Technical Committee 104 Physical Modelling in Geotechnics · Technical Committee 104 Physical Modelling in Geotechnics ... Ils incluent notamment des expériences de laboratoire menées

Technical Committee 104Physical Modelling in Geotechnics

Comité technique 104Modélisation physique en géotechnique

867

General Report for TC104 Physical Modelling in Geotechnics

Rapport général du TC104 Modélisation physique en géotechnique

White D.J., Gaudin C. Centre for Offshore Foundation Systems, University of Western Australia

Take W.A. Geo-Engineering Centre, Queen’s University, Canada

ABSTRACT: The session of the 18th ICSMGE on physical modelling in geotechnics, held by Technical Committee 104, provides anopportunity to reflect on the varied contributions that physical modelling makes to our profession – in both research and practice. The 27 papers contributed within this theme span many different applications of physical modelling. These range from simple small-scale laboratory tests to reveal mechanisms of soil-structure interaction – particularly where simulation by numerical methods is problematic – through to scaled-down versions of field constructions, performed outdoors, to validate the performance of newmonitoring technology. These examples show physical modelling in action across the entire spectrum of geotechnics, from the validation of fundamental analysis principles, to the assurance of construction technology. In all cases, physical models – defined here as physical idealisations of all or part of an envisioned geotechnical system – provide a more convenient method of gaining knowledge than observing or simulating the full geotechnical system.

RÉSUMÉ : La session de la 18e ICSMGE sur la modélisation physique en géotechnique est l’opportunité de faire le point sur lesdifférentes contributions que cette dernière a apporté à notre profession, dans le domaine de la recherche et des applications pratiques.Les 27 articles retenus dans les comptes rendus de la conférence couvrent un champ varié d’applications de la modélisation physique.Ils incluent notamment des expériences de laboratoire menées à échelle réduite pour révéler les mécanismes d’interaction sol-structure(notamment dans des cas ou la modélisation numérique peut s’avérer inadaptée), ainsi que des expériences de terrain réalisées sur desversions réduites d’ouvrage réels pour développer et valider de nouvelles techniques d’instrumentation et de suivi. Ces différentsexemples démontrent la variété des applications de la modélisation physique, de la validation de principes fondamentaux, àl’évaluation des performances de technologie de construction. Dans tous les cas, la modélisation physique, définie ici commel’idéalisation d’un system physique réel, permet d’accéder de manière rapide, économique et rigoureuse à une connaissance élargie duproblème étudié.

KEYWORDS: Physical modelling. Centrifuge modelling. Field monitoring.

1 INTRODUCTION

The papers submitted to the TC104-organised session on Physical Modelling in Geotechnics illustrate a range of applications of physical models to support geotechnical research and practice.

Physical modelling can serve a variety of purposes in geotechnical engineering. It can provide (Gaudin et al. 2010):

1. Insights into soil-structure interaction and geotechnical behaviour – qualitative and quantitative. Given the complexity of soil constitutive behaviour and the complex deformations and processes in some construction technologies, physical modelling provides a basis to assess fundamental modes of behaviour in controlled conditions.

2. Performance data to calibrate analytical or numerical models. Physical models use well-characterized soil and known boundary conditions, providing reliable performance data for a given idealized problem.

3. Specific performance data for design and operation.Physical modelling can be used directly to design geotechnical systems or tools. For example, site-specific soil and loading conditions can be replicated in a centrifuge model, to validate a foundation design. Or, the shape of a geotechnical tool – such as a novel form of ground improvement device or a proprietary anchoring system – can be optimised using performance data from physical models.

4. Soil characterisation data – through testing of samples using miniature versions of conventional in situ testing tools. Using physical models in this way can be more cost-effective than full scale in situ tests.

Physical modelling, in its various forms, contributes across the entire proceedings of the 18th ICSMGE, not only the 27 papers that were allocated to this session. Physical modelling is a technique that cuts across the traditional categories of ground engineering. The majority of the sessions at the ICSMGE are defined by type of geotechnical system – foundations, slopes, ground improvement and so forth. These are the traditional groupings that our profession forms, pooling expertise and specialist knowledge from experiences in design and construction.

The physical modelling community within geotechnical engineering pools a different type of specialist expertise. The TC104 activity focuses on the technology and principles that allow physical modelling facilities to flourish and provide capabilities that can be applied across all geotechnical topics.

When the ICSMGE was last held in Paris, in 1961, the ISSMFE did not have Technical Committees and there was no session dedicated to physical modelling. However, physical modelling featured heavily in the presented papers. The Proceedings of the Paris conference feature classic physical modelling studies of retaining walls by Rowe and Schofield, as well as model pile tests in the field performed by De Beer, Kerisel and Meyerhof. Many papers show film photographs of

868

Proceedings of the 18th International Conference on Soil Mechanics and Geotechnical Engineering, Paris 2013

distorted layers of coloured sand, or time-lapse exposures that reveal particle trajectories. Scaling laws are never referred to, aside from commentary by Meyerhof on the effect of stress level on friction angle, and the resulting scale effect between the bearing capacity of small models and field scale shallow foundations.

Over the past 50 years, physical modelling technology has evolved through to two key developments. The development of the centrifuge in the 1970s allowed the realism of physical modelling to be enhanced, through the correct modelling of self-weight stresses. The subsequent development of miniaturised electronics and micro-computers has led to enhanced methods of data acquisition, control, and image analysis. The refinement of these techniques continues to yield dramatic improvements in the utility of physical modelling. More realistic simulations can be conducted, and more detailed observations can be gathered.

Reviews of recent technological developments within physical modelling are described by Mayne et al. (2009) and White (2008).

Meanwhile, continuing cross-disciplinary efforts have led to wider recognition of definitive scaling laws to allow small scale physical models to be related to field scale conditions. TC104 has overseen the cataloguing of scaling law research. The initial publication of a TC2 (as TC104 was previously known) Scaling Law Catalogue (Garnier et al. 2007) has been followed by continuing development of this resource.

Proper application of scaling laws is vital when interpreting physical model tests for purposes (2) and (3) given on the previous page. If physical model test data is to be correctly linked to a field scale prototype, or to a numerical simulation, correct account must be made of the influences of size and timescale effects.

Scaling laws are well-established and straightforward to adhere to for small scale modelling of many geotechnical problems, particularly in fine-grained saturated soils. It is therefore no coincidence that physical modelling is a well-accepted technology in both research and design practice in offshore geotechnics, where soft normally- or lightly over-consolidated soils predominate. Recent state-of-the-art review papers have summarised many such applications (Martin 2001, White 2008, Gaudin et al. 2010).

For purpose (1), listed above, similitude and scaling laws are less significant, since the physical model might be an abstract component of the full system, or precise scaling of particular conditions may not be important. In this categorisation, we are referring to studies that aim to uncover a building block of the geotechnical system behaviour. Perhaps the overall system is not being modelled, or perhaps there is established theory to be tested. In the words of the title of the first Schofield Lecture, the aim might simply be to “expect the unexpected” (Bolton 2013).

It is this exploratory nature that makes physical modelling an attractive tool for many researchers and educators in geotechnics. By observing geotechnical systems in action, an intuitive understanding of soil mechanics can be gained, complementing the study of theory. Physical modelling tools for geotechnical education range from the venerable Hele-Shaw cell (Hele-Shaw 1898) to modern miniature – but highly instrumented – geotechnical centrifuges.

The Hele-Shaw cell provides solutions to seepage flow according to Darcy’s Law as a consequence of the Navier-Stokes equations simplifying for narrow planar flow. Henry Darcy studied in Paris – the venue of this 18th ICSMGE – at the Ecole Polytechnique and then Ecole des Ponts et Chausées, before joining the Corps des Ponts et Chausées. His most famous public work was the water supply system for Dijon. Appendix D of his published account of the Dijon works contains his report on the classic physical model tests in which the linear relationship between head gradient and flow velocity was identified. This work was published in 1856 and two years later Darcy died of pneumonia here in Paris.

Undergraduates studying geotechnics and fluid mechanics have faced Hele-Shaw cells for almost a century. Falling head permeability tests of the form analysed by Darcy in his classic work are standard undergraduate geotechnical laboratory experiments. Some of today’s undergraduates also have access to more sophisticated apparatus. These include miniature experiments that provide detailed measurements and observations of geotechnical constructions – often via bench-top centrifuges. These facilities provide opportunities to apply the analysis tools taught in lectures, completing the learning cycle through practical experience (Wartman 2006, Kolb 1984).

Meanwhile, to inspire the next generation of geotechnical engineering students, physical models are the most commonly called-upon facility within university departments to enliven events for school students. Physical models show engineering systems in action in a way that is immediately comprehensible.

Figure 1. Darcy’s physical model for investigating flow in porous media (Darcy 1856)

The following review summarises and discusses many of the papers contributed to the 18th ICSMGE that fall under the TC104 session. The contributed papers have been divided into the physical modelling categories listed at the start of this report.2 INSIGHTS INTO SOIL-STRUCTURE INTERACTION AND GEOTECHNICAL BEHAVIOUR

2.1 Reinforced ground

Several papers have focused on the performance of ground improvement systems utilising reinforced columns – either of cemented material, sand or stone. These systems are difficult to analyse, due to the complexity of the construction process and

869

Technical Committee 104 / Comité technique 104

the resulting uncertainty in the stress and strength within the vicinity of the column.

Physical modelling is particularly appropriate to investigate ground improvement considering the large resources required to undertake field scale testing and the complexity of numerical modelling: there is complexity associated with modelling of the reinforcement soil interface and of the process of reinforcement installation or construction. In some instances, the research undertaken has led to the development of very sophisticated testing technology, associated with the generation of the reinforcement or mechanism visualisation, which have provided insights that could not have been gained from other investigation techniques.

Ground improvement by the use of reinforced columns can be optimised by minimising the length of columns. They may not need to penetrate entirely through the soft layer.

Physical model tests reported by Tekin and Ergun (2013) compared the settlement of surface foundations on clay, reinforced by sand columns of varying length. The experiments featured a novel miniature extensometer arrangement. This used an antenna to detect magnets buried at multiple elevations within the sand columns. The efficiency was shown to depend on the length of the columns relative to the zone of loaded soil beneath the surface foundation. Columns that were too short settled with the surrounding foundation soil. Columns that extended deeper than the breadth of the surface foundation showed significantly less settlement. The vertical strain field measured by the extensometers supported these observations.

The theme of sand columns partially penetrating through clay continues in the paper by Sadek and Lattouf (2013). They performed drained triaxial tests on models of sand columns with varying volume ratio, relative to the full clay sample. The sand columns did not extend to the base of the clay. Careful exhumation of the samples after testing allowed the failure mechanism to be identified. The samples were treated as a single soil element in the interpretation. By fitting a Mohr-Coulomb failure envelope to the ultimate loads, the ‘smeared’ strength of the composite element was determined. These physical models essentially simulate a building block of a larger network of ground improvement columns.

The previous two studies focused on the in-service behaviour of a reinforcing column placed within a pre-bored hole in clay. Gautray et al (2012) focus more closely on the stone column construction process, through centrifuge modelling. Their aim was to examine the changes in column geometry and the pore pressures in the surrounding soil, as the compaction process evolves. Their tests include the full process of lance insertion –expanding a cased hole in the clay – followed by a cyclic retraction process whilst granular material is filled into the hole. Their data show the loading of the surrounding soil as the lance is inserted. The changes in pore pressure are significant. They are precipitated both by the initial insertion of the lance, and also by the lateral expansion of the granular column by the vertically-oscillating lance, during extraction.

After completion of the model construction process, the granular columns were loaded by a model foundation. The exchange from a column installation tool to a model foundation was made possible using the independent tool table of the ETH Zurich drum centrifuge. The model foundation was pushed into the ground, over the stone column, and the bearing capacity of the reinforced ground was identified. In this case, analytical solutions based on previous studies were able to bracket the identified capacity.

Two further system for studying column-reinforced ground are reported. Houda et al. (2013) describe a modular experimental apparatus that has been developed to allow parametric studies, including cyclic loading events, to be imposed on improved ground – using combinations of columns and geosynthetics.

Takano et al (2013) describe a highly sophisticated system which allows grout columns to be constructed within a X-ray

CT scanner, providing data of the changes in density surrounding each grout bulbs, as the column is constructed in increments. The sample container is also instrumented to calculate any changes in earth pressure coefficient. Subsequent centrifuge model tests, using a shaking table, demonstrate that the increased earth pressure coefficient leads to a reduction in the tendency to liquefy.

An alternative ground improvement system – the use of a cellular geosynthetic – is explored by Xu and Wang (2013). They describe investigations into the bearing capacity of footings on saturated granular soil, with and without geocell reinforcement. This ground improvement technology is relatively new development, and is most suited to reclaimed or filled ground. The geocell is laid on the ground then covered by fill. The tests showed that the geocell reinforcement serves to provide tensile capacity within the composite material. This changes the bearing capacity by altering the failure mechanism. The settlement around a surface footing becomes more bowl-shaped, rather than involving a punching shear mechanism. The punching shear is prevented by the tensile action of the geocell, which also serves to prevent tension cracks from opening adjacent to the footing.

These studies into ground improvement reflect the increasing need for urban developments and transport corridors to utilise poor ground, requiring mitigation measures to limit settlements.

2.2 Shallow and deep foundations

Several papers report experimental studies into the behaviour mechanisms of shallow and deep foundations. These focus on interaction effects – between pile base resistance and tunnels (Williamson et al. 2013), between pile base resistance and water jetting (Shepley and Bolton 2013) and between foundation and soil stiffness (Arnold and Laue 2013).

While loading of shallow and deep foundations can be undertaken relatively easily at full scale, physical modelling offers the possibility to investigate complex installation and interaction processes as illustrated in the examples below. The development of miniaturised electronics and sophisticated computer controlled motion systems, has enabled a continuously increasing realism of the modelling, providing invaluable insights into problems related to soil-structure interaction.

The complex experimental arrangement presented by Williamson et al. (2013) allows three model piles to be independently loaded, whilst a model tunnel is ‘constructed’ (through the simulation of volume loss) in the ground beneath. The effects of tunnel construction on overlying piled foundations is of increasing relevance as new urban railways – such as London’s Crossrail project – burrow beneath existing buildings.

During these centrifuge experiments, the soil movements are watched intently by an array of cameras, allowing displacement fields to be generated through particle image velocimetry (PIV). These results allow the full soil deformation mechanism to be visualised – rather than gathering only boundary movements measured by instruments located at the edges of the model. The displacement field extending from the tunnel construction to disrupt the existing piles can then be observed.

Pile construction can also be disruptive to existing infrastructure, particularly if dynamic or percussive installation methods are used. An alternative is to install piles through a jacking method, and the limited capacity of hydraulic pile jacking machines can be countered by the use of water jetting to reduce the penetration resistance. Shepley and Bolton (2013) describe centrifuge experiments which investigated the effects of water jetting.

The interaction between water jetting and the penetration resistance in sands is complex. Jetting serves to locally raise the pore water pressure at the pile tip, whilst also potentially causing migration of fines. Both of these effects will ease

870


penetration, but the former is temporary whilst the latter may have an irreversible effect on the properties of the surrounding soil.

The model tests showed the variation in pile penetration resistance into saturated sand with jetting flow rate. The response is highly non-linear. There is only a modest reduction in resistance at low flow rates, until a significant fall in penetration resistance occurs. Full liquefaction appeared present during penetration at the highest flow rate, with negligible resistance encountered. Scaling of this behaviour is naturally very challenging, particularly if the response is due to both particulate (internal erosion) and continuum (effective stress reduction) effects.

Arnold and Laue (2013) describe an experimental study into the load distribution beneath surface foundations, and the influence of the relative stiffness between the foundation and the soil. Centrifuge model tests were performed with a vertical point load applied at the central point of the foundation. Two model foundations were used, with and without edge stiffening to represent building walls. The stiffer foundation shows a more even distribution of foundation-soil pressure. The same response is evident in field measurements using pressure cells built into building foundations.

2.3 Flow-induced migration through porous media

Truong et al (2013) describe a set of experiments using an apparatus which is somewhat similar to Darcy’s (Figure 1). They have studied clogging effects during one-dimensional flow using a 2.5m long cylinder, 0.18 m in diameter. Unlike Darcy, they used electrical transducers to record flow rates and pressure, mercury manometers being general outlawed in modern laboratories. The experimental results show the inadequacy of assuming Darcian flow if the pore fluid introduces fine particles that may accumulate and block pores. The paper presents an example of steady flow through the tube, with a Darcian distribution of pore pressure. On injection of a bentonite slurry, the pore rapidly clog downstream of the injection point, causing a sudden rise in the upstream pore pressure.

These results are highly relevant to seepage through dams or embankments, where careful control of drainage is important to assure stability. The experiments illustrate that flow regimes and pore pressure distributions can be quickly altered if fine particles are transported within the pore fluid. The intrinsic permeability within Darcy’s Law may be a material property, but fines migration can rapidly change the composition of a material, and therefore its properties.

Sarma and Sarma (2013) discuss the flow of cementitious material into the zone surrounding a bored pile. They report a detailed laboratory investigation which mimicked the bored pile construction process to evaluate the parameters controlling the thickness of the sol zone at a pile wall that is strengthened by the inflow of cementing products during construction. The distance of impregnation was identified by a novel staining method, in which the carbonated cementation products are highlighted. The subsequent measurements of impregnation depth are extremely detailed, and have been elegantly interpreted into a link between voids ratio and particle size.

2.4 Reactive effects on soil behaviour

Cardoso and Nogueira Santos (2013) describe a careful physical model of electrokinetically-enhanced consolidation behaviour. Their study is motivated by the potential use of electrokinetics to improve the efficiency of soft ground improvement, in combination with drains. They initially consider one-dimensional consolidation of kaolin clay, using a modified cell that allows a voltage to be applied across the sample. Improvements in the rate of consolidation are identified, with cvincreasing by a factor of 6, typically.

Subsequent tests use a larger Rowe cell, incorporating vertical drains that encourage radial and well as vertical flow. The electrical field is then applied between a central drain and the outer circumference of the sample. In this case the beneficial effect of the radial flow overshadows the benefit from the electrokinetic effect. The conclusion is that electrokinetics can enhance engineered consolidation of soft ground, but only in certain circumstances.

A similar study using a physical model of an element of cemented barrier cut-off wall is reported by Verástegui-Flores et al. (2013). Laboratory apparatus was modified to provide long term measurements of the permeability and small strain stiffness of cement-stabilised bentonite clay. This is a widely used material for barrier wall construction, but can suffer from deterioration through chemical attack. Two novel pieces of apparatus were developed to provide simple methods of measuring the shear wave velocity and, in the second apparatus, the permeability. The combined data of permeability and small strain stiffness over >250 days shows the effect of sulphate attack on the properties of the cement-stabilised soil. As the cement hydrates, the pores become blocked by cementation, which also raised the shear wave velocity. However, when sulphate is added, these processes are halted, based on the measurements. These new types of test allow the performance of specific stabilised soil mixes to be determined via simple laboratory tests that are more representative than conventional methods.

3 PERFORMANCE DATA TO CALIBRATE MODELS

3.1 Seismic soil-structure interaction

Several papers contributed to this 18th ICSMGE focus on the seismic response of slopes, and the estimation of pore pressure build-up and lateral spreading. The use of highly instrumented centrifuge models provides detailed evidence of the internal accelerations and pore pressures within the slope. The models are usually plane strain, with a transparent window allowing the soil movement to be observed.

Due to the inability to undertake seismic field experiments, physical modelling has been used extensively to investigate seismic soil-structure interaction, using both 1g shaking tables and on-board shakers in geotechnical centrifuges. The ability to recreate accurate dynamic loading conditions and to measure pore pressure generation and soil displacements have proven to be essential to provide data to calibrate models.

Haeri et al. (2013) report a centrifuge model test of a slope reinforced by a 3 3 pile group, with a surficial non-liquefiable layer. During shaking, the soil liquefies and slides downslope, applying passive loading to the pile. The maximum loads occur at the start of the event, after the motion begins but before the soil is fully liquefied. As the degree of liquefaction increases the passive load reduces, although the load from the unliquefied layer persists.

Back-analysis of the lateral pressures and the resulting internal bending moment in the piles shows that the Japanese Roads Authority (JRA) design code provides good predictions of the maximum pile loading. However, the detailed distribution of load within a pile group is not considered in this code. The experimental data shows that significant shielding effects are present. The most heavily loaded pile was actually the downslope pile: the high load came not from the passive pressure from the upslope soil, but from the loss of active support on the downstream pile, as the downslope soil failed and slid away.

Higo et al (2013) report a study of embankment stability under seismic loading, using similar experimental techniques in a geotechnical centrifuge. They used a compacted clayey silty sand with the same composition as a material used to reinforce river levees in the Kansai region. Their study is focused on the

871


stability of a slope that is under steady state seepage conditions, with water emerging from close to the toe with the upper part of the slope being unsaturated.

A numerical back-analysis with a kinematic hardening plasticity constitutive model is performed, replicating many features of the physical model observations. In particular, both the physical and numerical models highlight how earthquake-induced excess pore pressures can lead to a rise in the water table and a loss of suction and strength in an unsaturated slope.

Moving to steeper slopes, Aklik and Wu (2013) describe a study of geotextile-reinforced walls, standing at angles of up to 85. Model tests were performed using a geotechnical centrifuge to induce collapse of the slopes using the ‘gravity turn-on’ method. These simple tests explored the failure mechanisms within the slope and the embedded geotextile layers. A simple camera system was used to record frequent images as the slopes ‘grew in height’. The failure mechanisms were quantified through particle image velocimetry analysis. The failures were shown to occur above the toe of the slopes, and controlled by the spacing of the geotextile layers. This study illustrates how complex geotechnical systems can be investigated using simple rapid experiments in a compact geotechnical centrifuge, with image analysis providing detailed quantification of the soil failure mechanisms.

Dashti et al (2013) present a series of centrifuge model tests results exploring the performance of buried water reservoir structures, made from concrete. The test arrangement includes a novel transparent laminar shaking container which allows the internal deformation to be visualised. In addition, the end walls of the container have pressure pad sensors, to record the distribution of pressure during shaking.

The tests are performed with and without a buried structure, providing calibration data for 2D and 3D numerical models. Time histories from historic earthquakes are used as the input shaking motion, after filtering out frequency components that are irrelevant or would damage the centrifuge. The data from these tests is currently being used by the Los Angeles Department of Water and Power, to assess the seismic performance of existing and planned subsurface reservoirs.

3.2 Ground improvement

Two studies describe experimental work to determine benchmarking data for the performance of ground improvement techniques for enhanced seismic performance.

Bahadori et al. (2013) report a series of shaking table tests used to evaluate the performance of tyre chips as soil reinforcement to improve liquefaction resistance. The stiffness and damping properties of tyre-chip – sand mixtures are assessed through intensively instrumented physical model tests. A level ground surface was modelled, and the results from an array of accelerometers were used to derive stress-strain loops at different elevations within the soil. These loops allow the stiffness and damping ratio to be determined.

An alternative novel material to improve the seismic performance of structures is expanded polystyrene – known as geofoam. In the example application studied by Dave et al. (2013), a layer of geofoam is used at the rear of a retaining wall. Tests were performed on a 1m3 sample using the shaking table at IIT Bombay. A surcharge was applied at the top of the retained soil to mimic field scale stress conditions. Varying magnitude of shaking were applied, whilst the pressures and acceleration within the backfill and at the wall facia were recorded. This soft layer of geofoam served to reduce the ground accelerations felt at the wall, and lowers the lateral pressures generated within the backfill.

The final contribution on the topic of ground improvement, to calibrate new analysis methods, is a paper describing a new foundation system for embankments on soft soil. Detert et al. (2013) describe a hybrid structure comprising two parallel sheet pile walls connected by a tension membrane. The embankment

is constructed in top of the membrane, and undergoes reduced settlement due to the support from the membrane.

This system acts to reduce embankment settlements through a combined action. The sheet pile walls prevent spreading, whilst the membrane generates tensile forces when distorted, supporting the embankment and relieving the soft layer of load.

To validate this foundation system, parallel streams of research using complementary physical and numerical modelling have been undertaken. Physical modelling is particularly important since large deformations and a complex composite system are involved. However, once the mechanisms of behaviour are clarified, suitably calibrated numerical modelling is planned to allow the system to be optimised.

3.3 Vibration screening barriers

The final paper in this category is concerned with geotechnical barriers to insulate sensitive areas from vibrations created by railway traffic. Masoumi and Vanhonacker (2013) describe an experimental programme studying the transmission of vibrations through a bed of sand, 10 m3 in volume. Careful attention was given to bed uniformity and scaling laws, with the vibration frequency being scaled up to reflect the reduced scale of the experiment.

The impedance of the soil bed was first established by impact testing, to determine stiffness and damping parameters for subsequent numerical analysis. Vibration transmission tests were then performed using a line of accelerometers at the soil surface, and a surface foundation with a vibrating live load. Tests were performed with and without concrete isolation walls, which were shown to reduce the transmission of vibrations. The results were compared with complementary numerical results.

4 PERFORMANCE DATA FOR DESIGN AND OPERATION

4.1 Pipeline protection from anchor dragging

Offshore pipeline-soil interaction is a relatively new field of investigations, which has greatly benefited from the development of physical modelling techniques. The advantage in terms of resources and timescale compared to field testing is obvious. More importantly, the recent development of sophisticated motion control has enabled modelling of the full life cycle of a pipeline, from the 3D dynamic motion at the touch down zone during laying, to the on-bottom stability under storm loading and large ‘post-failure’ lateral sweeps under thermal buckling. The large database developed from physical modelling tests over the last five years has been used to develop analytical models now currently used in design (Gaudin & White 2009, White & Cathie 2011).

Pipelines that cross shipping lanes or lie in regions of intensive fishing often require protection from anchor dragging. If a ship anchor or fishing trawlgear snags a pipeline both may be damaged. Pipelines can be protected by burial, often with rockdump backfill.

To assess the required depth of burial, and the performance of the protection scheme, the interaction between a passing anchor and the backfill must be assessed. Physical modelling is commonly used for this purpose, since the large deformations and chain-anchor-backfill interactions must be properly accounted for.

A model testing arrangement that can be used to determine performance data for direct use in design is described by Bezuijen et al. (2013). An elongated centrifuge strongbox is used to allow significant lengths of anchor dragging to be simulated. A faithful reproduction of a ship anchor, manufactured by 3D printing, is attached to a miniature chain, and dragged over a model seabed. A pipeline is buried in a trench that has been back-filled with coarser material.

The paper compares results from tests performed at 1g and at 80g in the Delft centrifuge. The results can be compared using

872


well-established scaling laws, although a limitation is the assumption that the sandy soil behaves as a purely frictional material, with a constant angle of friction. The results show that such an approach does not provide similitude between the two tests. The 1g tests indicate higher scaled resistance, but reduced anchor penetration into the seabed – which is unconservative. This discrepancy could be due to the higher operative friction angle in the sand and gravel at 1g. Correct scaling is shown to be important when gathering design data for anchor-pipeline interaction.

4.2 Open-pit mine stability

The stability of open-pit mines is investigated through physical modelling by Pipatsonga et al. (2013). This study was prompted by stability concerns at the Meo Moh mine in Thailand. A series of model tests was performed to investigates the mechanism by which arching across a steep cut slope provides additional stability, compared to the simple planar case.

The investigation comprised physical model tests performed at 1g and in a centrifuge using the gravity turn-on method, to identify different failure mechanisms. Slopes with supports of different geometry were constructed. Failure mechanisms included sliding and buckling of the undercut slope. The critical conditions for failure depended on the strength and stiffness of the buttressing supports.

These observations were used to calibrate simple design expressions for the conditions for failure. The results were immediately applied to the Meo Moh mine in Thailand, using site-specific characterisations data to provide input soil properties. The improved design of the open pit supports resulted in reduced requirements for excavation, transportation and dumping.

4.3 Performance of dike monitoring systems

Physical modelling on an extremely large scale is reported by Keolewijn et al. (2013). They describe a set of field experiments which were focused both on understanding the geotechnical performance of the structures, and also on the performance of the monitoring systems. A set of 3.5 m high dikes were constructed at a soft ground test site in the north east of the Netherlands. The study trialled different proprietary systems for monitoring dike performance and providing data that can be used to guide flood protection activities. Real time monitoring of dike performance is important for safety and maintenance programming in the Netherlands. New technology for remote sensing and data transmission allow dike networks to be monitored continuously from a central control location.

Various forms of instrumentation, including piezometers, inclinometers, fibre optic strain gauges, synthetic aperture radar and thermal cameras were used to monitor the behaviour of the three test dikes during controlled impounding on one side. The dikes were constructed in different ways, to encourage different failure mechanisms, including basal piping, internal erosion and overtopping.

The monitoring systems used the data streams in different ways to assess the dike status and predict the onset of instability. In some cases, the data was linked in real time to finite element simulations. These physical model tests have proven successful in demonstrating the potential of these dike monitoring systems, some of which are now in use in other countries worldwide.

5 CLOSING COMMENTS

The papers submitted to the TC104 session, and the physical modelling papers to the wider ICSMGE program provide contributions across the whole realm of geotechnical engineering. This report describes highlights, and provides a

broader commentary on the role that physical modelling plays in advancing research and practice in geotechnics.

6 ACKNOWLEDGEMENTS

Support from the ARC Future Fellowships program and from Shell Australia is acknowledged by the first author.

7 REFERENCES

Aklik P. & Wu W. 2013. Centrifuge model tests on foundation on geosynthetic reinforced slope. 18th ICSMGE Paris (this proceedings)

Arnold A. & Laue J. 2013. Loading behaviour of flexible raft foundations in full scale and centrifuge models. 18th ICSMGE Paris (this proceedings)

Bahadori H. & Manafi S. 2013. Investigation on the dynamic properties of saturated sand-tire chips mixture by shaking table. 18th ICSMGE Paris (this proceedings)

Bezuijen A., Zwaan R. & van Lottum, H 2013. The influence of the g-level for anchor tests in sand. 18th ICSMGE Paris (this proceedings)

Bolton M.D. 2013. Centrifuge modelling: Expect the unexpected (the 1st Schofield Lecture). 18th ICSMGE Paris (this proceedings).

Cardoso R. & Nogueira Santos J. 2013. An experimental study on the consolidation of soft clayey soils using electrochemical methods. 18th ICSMGE Paris (this proceedings)

Darcy H. 1856. Les fontaines publiques de la ville de Dijon. Dalmont, Paris.

Dashti S., Hushmand A., Ghayoomi M., McCartney J.S., Zhang M. Hushmand B., Mokarram N., Bastani A. Davis C., Lee Y., Hu J. 2013. Centrifuge modeling of seismic soil-structure-interaction and lateral earth pressures for large near-surface underground structures. 18th ICSMGE Paris (this proceedings)

Dave T. N., Dasaka S. M., Khan N. & Murali Krishna, A. 2013. Evaluation of seismic earth pressure reduction using EPS Geofoam. 18th ICSMGE Paris (this proceedings)

Detert O., Schanz T., Alexiew D. & König D. 2013. Analysis of an adaptive foundation system for embankments on soft soils by means of physical and numerical modelling. 18th ICSMGE Paris (this proceedings)

Garnier, J., Gaudin, C., Springman, S. M., Culligan, P. J., Goodings, D., Konig, D., Kutter, B., Phillips, R., Randolph, M. F., and Thorel, L. 2007. Catalogue of scaling laws and similitude questions in geotechnical centrifuge modelling. International Journal of Physical Modelling in Geotechnics, 7(3), 1-24.

Gaudin C. & White D.J. 2009. New centrifuge modelling techniques for investigating seabed pipeline behaviour. Proc. XVIIth Int. Conf. on Soil Mechanics & Geotechnical Engineering. Alexandria, 1, 448-451.

Gaudin, C., Clukey, E.C., Garnier, J. & Phillips, R. 2011. New frontiers for centrifuge modelling in offshore geotechnics, Frontiers in Offshore Geotechnics II, The Netherlands, CD, 155-188

Gautray, J., Laue J., Springman S.M. & Almeida M. 2013. Development of pore water pressure around a stone column. 18th

ICSMGE Paris (this proceedings) Haeri, S.M., Kavand A., Asefzadeh A. & Rahmani, I. 2013. Large scale

1-g shake table model test on the response of a stiff pile group to liquefaction induced lateral spreading. 18th ICSMGE Paris (this proceedings)

Hele-Shaw H. S., 1898. The flow of water, Nature 58 no. 1489, 33–36 Higo Y., Oka F., Kimoto S., Kinugawa T., Lee C.-W. & Doi T. 2013.

Dynamic centrifugal model test for unsaturated embankments considering seepage flow and the numerical analysis. 18th ICSMGE Paris (this proceedings)

Houda, M. Jenck O. Emeriault F. & Briançon L. 2013. Développement d’un modèle réduit tridimensionnel du renforcement des sols par inclusions rigides. 18th ICSMGE Paris (this proceedings)

Koelewijn A.R., de Vries G. & van Lottum, H. 2013. Full-scale field validation of innovative dike monitoring systems. 18th ICSMGE Paris (this proceedings)

Kolb, David A. 1984. Experiential Learning: Experience as the Source of Learning and Development. Prentice-Hall, Inc., Englewood Cliffs, N.J

Martin, C.M. 2001. Impact of centrifuge modelling on offshore foundation design. Proc. Int. Symp. "Constitutive and centrifuge geotechnical modelling: two extremes", Balkema, pp 135-153.

873


Masoumi H. & Vanhonacker P. 2013.Physical modeling of the vibration mitigation by an isolating screen. 18th ICSMGE Paris (this proceedings)

Mayne, P.W., Coop, M.R., Springman, S., Huang, A-B., and Zornberg, J. 2009. State-of-the-Art Paper (SOA-1): GeoMaterial Behavior and Testing. Proc. 17th ISCMGE, Alexandria, Egypt, Millpress/IOS Press Rotterdam, 2777-2872.

Najjar S. Sadek S. & Lattouf B. 2013. The drained strength of soft clays with partially penetrating sand columns at different area replacement ratios. 18th ICSMGE Paris (this proceedings)

Pipatpongsa T, Khosravi M.H. & Takemura J. 2013. Physical modeling of arch action in undercut slopes with actual engineering practice to Mae Moh open-pit mine of Thailand. 18th ICSMGE Paris (this proceedings)

Sarma, D. & Sarma, M.D. 2013. Critical state modelling of soil-structure interface for advanced design. 18th ICSMGE Paris (this proceedings)

Shepley P. & Bolton M.D. 2013. Water injection aided pile jacking centrifuge experiments in sand. 18th ICSMGE Paris (this proceedings)

Takano D., Morikawa Y., Nishimura S. & Takehana K. 2013. Experimental study on compaction grouting method for liquefiable soil using centrifuge test and X-ray tomography. 18th ICSMGE Paris (this proceedings)

Tekin M. & Ergun M.U. 2013. A model study of strains under footings supported by floating and end-bearing granular columns. 18th

ICSMGE Paris (this proceedings) Truong Q.Q., Dupla J.-C., Canou J., Chevalier C. & Chopin M. 2013.

Modélisation physique du blocage d’un écoulement d’eau dans un sol par injection d’un produit de colmatage. 18th ICSMGE Paris (this proceedings)Wartman J. 2006. Geotechnical physical modeling for education: Learning theory approach. Journal of Professional Issues in Engineering Education and Practice, Vol. 132, No. 4, October 2006, 288-296,

Verástegui-Flores R.D., Di Emidio G. & Bezuijen A. 2013. Hydraulic conductivity and small-strain stiffness of a cement-bentonite sample exposed to sulphates. 18th ICSMGE Paris (this proceedings)

White D.J. 2008. Contributions to Géotechnique 1948-2008: Physical Modelling. Géotechnique, 58(5)413-421

White D.J., & Cathie D.N. 2011. Geotechnics for subsea pipelines. Proc. 2nd Int. Symp. on Frontiers in Offshore Geotechnics. Perth, 87-123

Williamson M.G., Elshafie M.Z.E.B. & R.J. Mair 2013. Centrifuge modelling of bored piles in sands. 18th ICSMGE Paris (this proceedings)

Xu Y. & Wang J.P. 2013. Stability and performance of ground improvement using geocell mattresses under extreme weather. 18th

ICSMGE Paris (this proceedings)

875

Centrifuge model tests on foundation on geosynthetic reinforced slope.

Essais en centrifugeuse d'une fondation sur une pente renforcée par géosynthétique

P. Aklik, W. Wu Institute of Geotechnical Engineering, University of Natural Resources and Life Sciences, Vienna, Austria.

ABSTRACT: Centrifuge modelling is a powerful tool for physical modelling of reinforced slopes and offers the advantage to observethe failure mechanisms of the slopes. In order to replicate the gravity induced stresses of a prototype structure in a geometrically 1/N reduced model, it is necessary to test the model in a gravitational field N times larger than that of the prototype structure. In this paper,a series of model tests in a geotechnical centrifuge on reinforced slopes is presented. The geotextile reinforced slopes have the same height of 270 mm and is built with soil layers of the same properties. Photographs with high resolution are taken in short time intervals through the glass wall during flight and the soil deformations of geotextile reinforced slopes loaded with a footing are evaluated with Particle Image Velocimetry (PIV). The experimental results of reinforced slopes are presented. The test data provideinteresting insight into the failure mechanisms and the progressive failure characteristics of geo-synthetic reinforced slopes. RÉSUMÉ : La modélisation en centrifugeuse est un outil puissant pour la modélisation physique des pentes renforcées et offrel'avantage d'observer les mécanismes de rupture des pentes. Pour reproduire les contraintes induites par la gravité d'une structure prototype sur un modèle réduit à l'échelle 1/N, il est nécessaire de tester le modèle dans un champ de gravitation N fois plus grand quecelui de la structure prototype. Dans cet article, une série d'essais sur modèle dans une centrifugeuse géotechnique sur les pentesrenforcées est présenté. Les pentes renforcées de géotextiles ont la même hauteur de 270 mm et sont construites avec des couches desol de mêmes propriétés. Des photographies haute résolution sont prises à des intervalles de temps courts à travers la paroi de verrependant le vol et les déformations du sol de pentes renforcées de géotextiles chargées par une semelle filante sont évaluées envélocimétrie par images de particules (PIV). Les résultats expérimentaux de pentes renforcées sont présentés. Les données d'essaidonnent un aperçu intéressant sur les mécanismes de rupture et les caractéristiques de rupture progressive de pentes renforcées.

KEYWORDS: centrifuge, reinforced slope, foundation, PIV (Particle Image Velocimetry).

1 INTRODUCTION.

A wide range of geotechnical problems can be investigated using physical modeling techniques. Centrifuge modelling has become a powerful technique in geotechnical engineering for studying the stability of prototype slopes. In order to replicate the gravity induced stresses of a prototype structure in a geometrically 1/N reduced model, it is necessary to test the model in a gravitational field N times larger than that of prototype structure (Viswanadham and König, 2009). Substantial research demonstrated the effectiveness of centrifuge modelling for studying the behaviours of geosynthetic reinforced walls and slopes, as reported by Porbaha and Goodings (1994, 1996), Zornberg et al. (1997; 1998a,b), Zornberg and Arriaga (2003), Viswanadham and Mahajan (2007), Chen et al. (2007) and Viswanadham and König (2004, 2009).

Foundations are sometimes built on slopes or near the edges of slopes. Knowledge of the treatment of reinforced slopes loaded with a surface footing is of practical importance to geotechnical engineers. Although there are several research studies on reinforced level ground, investigations of footings on reinforced slopes are rather limited (Selvadurai & Gnanendran, 1989; Omar et al., 1993; Huang et al., 1994; Lee & Manjunath, 2000; Yoo, 2001; El Sawwaf, 2007; Alamshahi & Hataf, 2009).

In this paper, a series of reinforced slope models with a slope inclination of about 65, 75 and 85 degrees were tested in a geotechnical centrifuge. The aim is to investigate the effect of the foundation on the geotextile reinforced slopes. Moreover, a technique called Particle Image Velocimetry (PIV) is used in

this research to reveal the failure mechanisms of the geotextile reinforced slopes. The experimental results provide reproducible database for rational design of geosynthetic reinforced slopes.

2. MODEL DESIGN

2.1. Centrifuge

The geotechnical centrifuge at the Institute of Geotechnical Engineering, University of Natural Resources and Life Sciences (BOKU) in Vienna was manufactured by Trio-Tech, USA and was put into operation in 1990 with partial financial support from the Austrian Science Foundation (Trio-Tech 1988). The beam centrifuge has the following components: a swinging basket, a balancing counterweight, a DC motor and aerodynamic enclosure. It is equipped with 56 electrical slip rings for process control and data acquisition. By using the dual platforms, two models can be tested at the same time. However, it is usual to have only one swinging basket carrying a model, while a balance weight is loaded on the other platform. The centrifuge has been used to investigate various problems in geotechnical engineering, such as retaining wall, shallow foundation and pile foundation. Recent development in digital image processing offers excellent possibilities to study strength of geosynthetic reinforced slopes. The technical specifications of the centrifuge are listed in Table 1 and illustrated in Figure 1.

Centrifuge model tests on foundation on geosynthetic reinforced slope

Essais en centrifugeuse d'une fondation sur une pente renforcée par géosynthétique

Aklik P., Wu W. Institute of Geotechnical Engineering, University of Natural Resources and Life Sciences, Vienna, Austria.

876


Table 1. Technical specifications of the centrifuge.

Diameter [m] 3.0

Radius of the swinging basket [m] 1.3

Maximum radial acceleration [g] 200

Maximum model weight [kg] 90

Maximum model height [cm] 56

Figure 1. Photo of the centrifuge and its swinging basket.

2.2. Model box

The model box (Figure 2) has the dimensions of 440mm*400mm*155mm in depth. A transparent Plexiglas plate with a thickness of 30mm was used on one side of the box to enable digital images to be taken during testing. The other walls of the box were aluminum plates with a thickness of 15mm. The box is sufficiently rigid to maintain plane strain conditions in the model.

Figure 2. Geotextile reinforced slope model with a digital camera in the front and LED lights on the left and right sides.

2.3. Slope model and model textiles

Reinforced slope models have a slope inclination of about 65, 75, and 85 degrees. The geotextile reinforced slopes had the same height of 270mm and was built on a soil layer of the same properties. The slope models were loaded with a surcharge of the same soil on the top of the slope. Due to the inherent symmetry of the slope, only half of the slope was modelled.

2.4. Soil

The soil used in the experiments was uniform coarse sand (Table 2), Standard Sand II (DIN 1164/58). The sand was not compacted but each layer had the same weight for all three models.

T able 2. Properties of soil

Specific weight ρs [g/cm³] 2.644

Density range ρmin, ρmax [g/cm³] 1.44 – 1.65

Void ratio emin, emax 0.607 – 0.844

Coefficient of uniformity 1.4

Friction angle φ [°] 34

Cohesion c [kN/m²] 0

2.5. Instrumentation

The displacement of the geotextile reinforced slope models was measured by PIV (White et al. 2001; 2003). For this purpose, a 14.7 MP Canon G10 digital camera was used to obtain high resolution digital images of the sand grains behind the Plexiglas wall. Black dots surrounded by white circles were applied to the Plexiglas as can be seen in Fig. 2, and were used as reference points for monitoring displacements within the soil. Two panels of 33 LED lights were used on both sides of the model box for lighting the centrifuge during testing. A laptop computer was mounted close the rotating axis of the centrifuge and connected to the main computer in the control room to save the photos during centrifuge testing.

2.6. Method

The soil displacement analysis was carried out with GeoPIV8 software, developed by White & Take (2002). The first image is divided into a grid of test patches. Each test patch consists of a sample of the image matrix of size 20 * 20 pixels and the images were captured in 6 s intervals until the failure of the model. The recorded photographs are used to reveal the failure mechanisms of the slope after testing.

3. RESULTS

The slope deformations before and after the slope failures are evaluated with PIV analysis. The shear strain in the model slopes with inclinations of 65, 75, and 85 degrees are shown in Figures 3-5.

In Figures 3a and 3b, the slope has an inclination of 65 degrees. The failure surfaces can be clearly observed (Fig. 3b). The failure surface does not pass through the toe as is often observed in unreinforced slopes but emerges from the lower part of the slope.

Figures 4a and 4b show the strain distribution in a steeper slope with an inclination of 75 degrees. When compared with the slope having an inclination of 65 degrees, the shear strains especially in front of the shear surface are more pronounced than in the previous slope (Fig. 4b).

Steeper slope (Figs. 5a and 5b) shows larger shear strain than flatter slopes. Larger shear strain is observed not only in the top of the slope but also along the whole surface of the slope (Fig. 5b).

877


(a) (b)

Figure 3. (a) Beginning of slope failure of geotextile reinforced slope with a slope inclination of 65 degrees, (b) Failure surfaces of geotextile reinforced slope with a slope inclination of 65 degrees.

(a) (b)


(a) (b)


878


An important parameter in geotextile reinforced slopes is the vertical spacing between the reinforcement layers. There are 6 layers in the slope having 65, 7 layers in the slope having 75, and 8 layers in the slope having 85 slope inclinations. The slope failure is induced by breakage rather than pull-out of the reinforcement. The spacing between adjacent reinforcements is not investigated in the present study. This will be studied later.

4. CONCLUSIONS

The failure mechanisms of geosynthetic reinforced slopes are investigated in a geotechnical centrifuge. The failure surfaces emerge from the lower part of the slopes rather than from the slope toes. Slope failure is mainly dictated by the tensile strength of geotextile when geotextile is intersected by the failure surface. PIV is an efficient tool to instrument the soil deformation of model slopes in geotechnical centrifuge.

5. ACKNOWLEDGEMENTS

The work of the first author is supported by the Otto Pregl Foundation for Geotechnical Fundamental Research.

6. REFERENCES

Alamshahi S., Hataf, N., 2009, Bearing capacity of strip footings on sand slopes reinforced with geogrid and grid-anchor, Geotextiles and Geomembranes, Vol 27, 217-226.

Chen, H.-T., Hung, W.-Y., Chang, C.-C., Chen, Y.-J. & Lee, C.-J. 2007. Centrifuge modelling test of a geotextile-reinforced wall with a very wet clayey backfill. Geotextiles and Geomembranes 25 (6), 346–359.

El Sawwaf, M.A., 2007, Behavior of strip footing on geogrid-reinforced sand over a soft clay slope, Geotextiles and Geomembranes 25, 50-60.

Huang, C., Tatsuoka, F., Sato, Y., 1994, Failure mechanisms of reinforced sand slopes loaded with a footing, Soils and Foundations 24 (2), 27-40.

Lee, K.M., Manjunath, V.R., 2000, Experimental and numerical studies of geosynthetic-reinforced sand slopes loaded with a footing, Canadian Geotechnical Journal 37, 828-842.

Porbaha, A. & Goodings, D.J. 1994. Geotextile reinforced cohesive slopes on weak foundations. Proc. International Conference Centrifuge 94, Singapore, 623-628.

Porbaha, A. & Goodings, D.J. 1996. Centrifuge modeling of geotextile-reinforced cohesive soil retaining walls. Journal of Geotechnical and Geoenvironmental Engineering, Vol.122, No.10, 840-848.

Selvadurai, A., Gnanendran, C., 1989, An experimental study of a footing located on a sloped fill: influence of a soil reinforcement layer, Canadian Geotechnical Journal, 26 (3), 467-473.

Take, W.A. 2003. The influence of seasonal moisture cycles on clay slopes. PhD dissertation, University of Cambridge Engineering Department, UK.

Trio-Tech 1988. Technical proposal for a geophysical test centrifuge model 1231. Trio-Tech International, San Francisco, CA.

Viswanadham, B.V.S. & Konig, D. 2004. Studies on scaling and instrumentation of a geogrid. Geotextiles and Geomembranes, Vol 22, 5, 307-328.

Viswanadham, B.V.S. & Konig, D. 2009. Centrifuge modeling of geotextile-reinforced slopes subjected to differential settlements. Geotextiles and Geomembranes, Vol 27, 77 - 88.

White, D.J. 2002. An investigation into the behaviour of pressed-in piles. PhD dissertation, University of Cambridge Engineering Department, UK.

White, D. & Take, A. 2002. GeoPIV: Particle Image Velocimetry (PIV) Software for use in Geotechnical Testing. Cambridge, UK.

White, D.J., Take, W.A. & Bolton, M.D. 2001. Measuring soil deformation in geotechnical models digital images and PIV analysis. 10th International Conference on Computer Methods and Advances in Geomechanics, Tuscan, Arizona, 997-1002.

White, D.J., Take, W.A & Bolton, M.D. 2003. Soil deformation measurement using particle image velocimetry (PIV) and photogrammetry. Géotechnique Vol.53, No.7, 619-631.

Yoo, C., 2001, Laboratory investigation of bearing capacity behavior of strip footing on geogrid-reinforced sand slope, Geotextiles and Geomembranes, Vol 19, 279 - 298.

Zornberg, J.G. & Arriaga, F. 2003. Strain distribution within geosynthetic-reinforced slopes. Journal of Geotechnical and Geoenvironmental Engineering, Vol.129, No.1, 32-45.

Zornberg, J.G., Mitchell, J.K. & Sitar, N. 1997. Testing of reinforced slopes in a geotechnical centrifuge. Geotechnical Testing Journal, Vol.20, No.4, 470-480.

Zornberg, J.G., Sitar, N. & Mitchell, J.K. 1998. Performance of geosynthetic reinforced slopes at failure. Journal of Geotechnical and Geoenvironmental Engineering, Vol.124, No.8, 670-683.

Zornberg, J.G., Sitar, N. & Mitchell, J.K. 1998. Limit equilibrium as basis for design of geosynthetic reinforced slopes. Journal of Geotechnical and Geoenvironmental Engineering, Vol.124, No.8, 684-698.

879

Loading behaviour of flexible raft foundations in full scale and centrifuge models

Comportement de radiers flexibles dans des essais grandeur nature et en centrifugeuse

Arnold A. Dr. Vollenweider Ltd., Zurich & Lucerne University of Applied Sciences and Arts, Lucerne

Laue J. Institute for Geotechnical Engineering, ETH Zurich

ABSTRACT: Flexible rafts are commonly used foundation systems for different kinds of constructions. A raft is easy to build and to design even though the dimensioning is not straight forward. Two buildings were equipped to measure the stress distribution between raft foundations and the adjacent soil, and to measure the deformation of the load-carrying column on top of the foundation in order to know the load extent. To compare the full scale measurements with another model, centrifuge tests have been conducted in a drumcentrifuge at ETH Zurich. The loading behaviour of different raft foundations has been studied on these two models. This contribution gives a short summary of the comparison between the measurements gained from full scale and the model tests in the centrifuge. The conditions in the centrifuge provide an ideal stress distribution between raft and soil, while different influences on a raft foundation in full scale such as the geometry of the load bringing structure and loading sequences influence the stress distribution in the real world.

RÉSUMÉ : Les fondations flexibles sur radier sont un système de fondations utilisé communément pour différents types deconstruction. Un radier est facile à construire et à dimensionner, même si le dimensionnement n'est pas immédiat. Deux bâtiments ontété instrumentés pour mesurer la distribution des contraintes entre les fondations sur radier et le sol adjacent ainsi que lesdéformations de la colonne porteuse située sur la fondation afin de connaitre l'importance de la charge. Des essais en centrifugeuse ontété réalisés dans la centrifugeuse tambour à l'ETH Zürich afin de comparer les mesures grandeur nature à un autre modèle. Lecomportement sous charge de différentes fondations sur radier a été étudié pour ces deux modèles. Cette contribution donne un courtrésumé de la comparaison entre les mesures obtenues grâce aux essais grandeur nature et aux essais dans la centrifugeuse: lesconditions dans la centrifugeuse fournissent une distribution des contraintes idéale entre le radier et le sol, alors que les différentesinfluences sur une fondation sur radier, telles que la géométrie de la structure transmettant la charge et les séquences de chargementont un effet sur la distribution des contrainte dans une situation grandeur nature.

KEYWORDS: raft foundations, loading behaviour, system-stiffness, centrifuge tests, full scale measurements

1 INTRODUCTION

Raft foundations are frequently used systems to distribute loads of different structures into the ground. They are cheap and fast in construction using simple design assumptions. They provide a robust system which is not sensitive in terms of settlements – especially for overconsolidated clays and coarse grained soils. One may use piled foundations for normally consolidated fine grained soils to avoid unacceptable settlements.

been studied in model tests and in full scale to improve the analytical approach by means of investigating the changing stress distribution due to stiffness variation in soil and structure.

1.1 Analytical models

A short summary of the different analytical models is given here. The simplest model to obtain a stress distribution between foundation and soil is to focus purely on the vertical- and on the momentum equilibrium of force. The approach given in figure 1 does not care about the deformation, which must be identical on the foundation plate as well as in the soil. Since the deformation of the foundation system is not regarded, changes in soil- and

structure stiffness are neglected with this method. Thus, those models provide only a preliminary distribution of the stresses.

Figure 1: Stress distribution between foundation and soil fulfilling the vertical- and momentum equilibrium (Kany and El Gendy 1996).

Another method deals with a coefficient of subgrade reaction, based on the approach after Winkler (1867) and Zimmermann (1888). As given in figure 2 each spring is independent of the other springs, which results in an unrealistic distribution of settlements especially at the corners of the foundation.

Figure 2: Independent springs on the approach of coefficient of subgrade reaction (Kany and El Gendy 1996).

A third approach is based on the linear-elastic behaviour of soils after Boussinesq (1885) which has been developed to an

Even though raft foundations are easy to build, the dimensioning of such structures is not straight forward and partially to simplistic. The analytical approaches mostcommonly used base on equilibrium and linear-elasticbehaviour of soils, which usually provides only an ideal shape of the stress distribution acting on a foundation.

The stress distribution between raft and soil has therefore

880


approach for practical purposes by Kany (1974). This approach provides realistic settlements also at the edges of foundations. But it is not able to describe more complex soil behaviour such as hardening or softening (Muir Wood 1990).

een flexible and stiff behaviour of the foundation system.

s = 1/12 · E /E ·(d/L)3 (1)

tructu

: Foundation length [m]

din alculatedbehaviour (e.g. Leussink et al. 1966).

whole centrifuge test program can be found in Arnold (2012).

pringman et al. 2002). The test setup is given infigure 3 and 4.

Figure 3: Setup of the centrifuge test (Laue and Arnold 2008).

me of the test setup in the drum centrifuge (Laue and

no pressure. For these conditions, a flexible behaviour can be observed with maximum str s distributed near to the column.

um plate at 50g (Arnold and Laue 2009); right hand side: Resolution of the tekscan measurements given in kPa (Arnold and Laue

aves stiffer as four unloaded walls are placed on top of all sides. Details on this test can be found in Arnold and Laue (2009).

ted at 50g. Protoype load: 10925 kN. Prototype ttlement: Approx. 220 mm (Arnold and Laue 2009). The resolution is

eq

ituated in the area of the load-bringing column and the distribution is more uniform over the whole area

1.2 System-stiffness after DIN-code

The DIN-code 4018 (1981) defines a system-stiffness (eq.1), which allows distinguishing betw

K b s

Ks: System stiffness [-] Eb: Stiffness of the foundation s re [N/m2]

g on the c

Es: Stiffness of the soil [N/m2]d: Foundation thickness [m]L

The behaviour of the foundation is distinguished (Meyerhof, 1979) depending on the value of Ks with Ks = 0 representing flexible, 0.001 < Ks < 0.01semi-flexible, 0.01 < Ks < 0.1semi-stiff and Ks > 0.1 stiff behaviour. This allows choosing the stress distribution for design depen

2 CENTRIFUGE MODELLING

Details about the centrifuge modelling can be found in Schofield (1980) and Laue (2002). The centrifuge tests, which are presented in this contribution, have been conducted in the drum centrifuge at ETH Zurich (Springman et al. 2001). Detailed information about the

2.1 Centrifuge test on a flexible raft foundation

80 Centrifuge tests were conducted for studying the loading behaviour of flexible raft foundations (Arnold 2012). The stress distribution between raft and soil was measured with tactile pressure pads (S

Figure 4: ScheArnold 2008).

The loading of these tests was conducted on a 4 mm thick square aluminium plate as foundation with a side length of 11.2 cm under an enhanced g-level of 50. The model foundation represents a prototype foundation with a side length of 5.6 m and a thickness of 20 cm. Figure 5 shows the measured stress distribution for a load of 4.25 kN (equivalent to a prototype load of 10625 kN) and a settlement of 5 mm (equivalent to a settlement of 250 mm at prototype scale). The white areas show the highest pressure, black areas show

esse

Figure 5: Left hand side: Stress distribution under a 4 mm thick alumini

2009).

2.2 Centrifuge test on a stiffened raft foundation (4 unloaded walls)

Figure 6 shows the stress distribution under a foundation stiffened by four unloaded walls. The stiffened foundation has also a thickness of 4 mm but beh

Figure 6: Stress distribution under a 4 mm thick aluminium plate stiffened with 4 unloaded walls situated on the four edges of the foundation. Test conducse

uivalent to Figure 5.

The stiffer stress distribution can be seen in Figure 6. Less clear peak pressure is s

of the foundation slab.

3 FULL SCALE MEASUREMENTS

Detailed information about the full scale measurements is given in Arnold and Laue (2010) and Arnold (2012). Two different buildings were equipped with oil filled pressure plates manufactured by Gloetzl (Schmidt 1991) to gain some information about the load extent on the raft and the stress distribution between raft and adjacent soil. One building is

881


situated in central Switzerland (Alpnach) and the other one in the northern part of Switzerland (Merenschwand). Two foundations of a supporting girder (4 storey-building) were controlled in Alpnach. In Merenschwand two foundations of an

easured.

geodetic measurement provided information of the settlements. One of the pressure pad used in Alpnach is shown in figure 7.

Figure 7: Pressure pad embedded in a layer of fine sand (Arnold and

igure 8. The darker shaded areas in the

mbers 3, 7, 8, 9 at the center part of the rafts and 1, 2, 5, 6 on the foundation edges

measurement points (Arnold 2012).

Th

dle field by

1 and 3 show even tension at lower rates of loads.

(Arnold, 2012).

mns

The arrangement and the dimensions of the campaign in

t Nr. 1 and 7 & 13 at Raft Nr. 2: Pressure pads to me

loads are distributed near the columns (Pads 7 and 17) while

earth-fill supporting roof were m

3.1 Measurement equipment

The measurement equipment consisted of pressure pads which are able to measure the pressure at a reduced area, of strain gage devices and of displacement transducers to measure the deformation of the load bringing columns. Additional

Laue 2010).

4 MEASURING CAMPAIGN IN ALPNACH

The arrangement of the measurement system at the location in Alpnach is given in fmiddle of the foundation indicate the columns (cross-section are : 1.0 m · 0.30 m). a

Figure 8: Sketch of the measurement systems in Alpnach: D1 – D4 are strain gages to measure the deformations of the columns. Numbers 4, 5, 10, 14 correlates to the pressure pads under the rafts. Nu

indicate geodetic

4.1 Results

e measurements of the pressure pads in Alpnach are given in figure 9.

The measurements in Alpnach show the high dependency of the static system on the development of pressures under a foundation. Higher stresses are measured with the pressure pads 4 and 5, which are positioned at locations towards the inward field between the two foundations while values measured with the other two pads remain smaller. This can be explained using the measured the strains on both sides of the column (Figure 10). Both columns show bending towards the mid

higher compression on strain gauges 2 and 4 while strain gauges

Figure 9: Measurements of the pressure pads 4, 5, 10 and 14 at Alpnach

Figure 10: Measurements of the strains in the two investigated coluat Alpnach (Arnold, 2012).

5 MEASURING CAMPAIGN IN MERENSCHWAND

Merenschwand are given in Figure 11 while results of the measurements with the pressure pads are shown in Figure 12.

Figure 11: Sketch of the measurement systems in Merenschwand: D1 – D4: Displacement transducers to measure the deformations of the col mns; 6 & 17 at Rafu

asure the pressure between foundation and soil; 1 – 4 at Raft Nr. 1 and 5 – 8 (7 at foundation edge) at Raft Nr. 2: geodetic measurement points (Arnold 2012).

Higher loads are introduced into the ground under raft 2 than under raft 1. Even though the strain measurements in the column showed small bending of the columns, the stress distributions anticipates a more expected behaviour here. Higher

Raft Nr. 1 Raft Nr. 2

Raft Nr. 1 Raft Nr. 2

882


lower stresses are distributed in the outer areas. The measurements under these foundations with a thickness of 0.40m (raft 1) respectively 0.50 m (raft 2) indicate a flexible behaviour of this particular footing.

Figure 12: Measurements of the pressure pads 6, 7, 13 and 17 at

and the results of the centrifuge model tests. A flexible behaviour (as expected by the definition of DIN

d.

eld for future research, where the interaction of the whole building-structure with the soil should be

are also most grateful to Pierre Lehmann and Sandra Kümin (CES Ltd.)

Huwiler & Portmann Ltd.) who made the

Arn

Arn

Arn

Kan

Lau

Lau

Sch

Spr

Modelling in Geotechnical Engineering, Balkema, 113-118.

inkler, E. 1867. Die Lehre von der Elastizität und Festigkeit.Domenicus. Prag.

Zimmermann, H. 1888. Die Berechnung des Eisenbahn-Oberbaues.Verlag Wilhelm Ernst & Sohn. Berlin.

Merenschwand (Arnold, 2012).

6 COMPARISON OF THE RESULTS

There is a clear difference between the two measured foundation systems (Alpnach & Merenschwand) concerning the stress distribution between foundation and adjacent soil. The bending moment in the girder originating from the loads of the 4 storey building is dominating the stress distribution at the building in Alpnach. It is passed from the supporting girder to the column and further down to the foundation. This bending moment can be verified by the measurements of the strain gages and allow the construction process to be followed. The moment clearly dominates the stress distribution while a stiff behaviour can be noticed.

Little bending moment is passed from the earth-fill supporting roof to the foundation at the building in Merenschwand. Therefore the stress distribution here can be more easily compared to the ideal situation assumed with some of the simplified models

4018) can be identifie

7 CONCLUSION

The full scale measurements show a clear influence of the loading situation to the stress distribution between raft and adjacent soil. Bending moments are passed from roofs via columns and walls to the foundations where they influence the soil-structure interaction by changing the stress distribution between structure and adjacent soil. The so found influence of the loading situation to the stress distribution could not be shown in the centrifuge tests where the ideal test conditions without bending moments have been studied. The system-stiffness equation is only valid for this type of “ideal” loading situations, where bending moments in the structure do not play a significant role. Bending moments among other parameters like e.g. inhomogeneous design of a foundation do influence this stress distribution. Thus a single value of system stiffness for the whole foundation can be misleading as the stress dependency of the modulus of the ground is not taken into account but will have for rafts an influence on the design. This opens a new fi

investigated.

8 ACKNOWLEDGEMENTS

The authors are most grateful to the ETH-Grant (TH21 07-2) which made the presented investigation possible. A special thank goes to Markus Iten who conducted all the centrifuge tests and to Ernst Bleiker who was responsible of the measurement systems in Alpnach and Merenschwand. The authors

and Sepp Portmann (full scale measurements possible by providing the construction sites for the installation of the measurement systems.

9 REFERENCES

old, A. 2012. Tragverhalten von nicht starren Flachfundationen unter Berücksichtigung der lokalen Steifigkeitsverhältnisse. vdf. Zürich. old, A. and Laue, J. 2010. Physical modelling and full scale measurements on soil-structure interaction of flexible raft foundations. 11th International Conference on Geotechnical Challenges in Urban Regeneration, London. old, A., Laue, J., Espinosa, T. and Springman, S.M. 2010. Centrifuge modelling of the behaviour of flexible raft foundations on clay and sand. International Conference on Physical Modelling in Geotechnics, Zurich, CRC Press, 679-684. old, A. and Laue, J. 2009. Influence of unloaded walls on the strArn ess distribution under a raft foundation. 17th International Conference on Soil Mechanics and Geotechnical Engineering. Alexandria, Egypt. IOS Press, 1124-1127. ssinesq, IBou . 1885. Applications des potentiels à l’étude de l’équilibre et du mouvement des solides élastiques. Gauthier-Villars, Paris. . 1981. DIN 4018 – Berechnung der SohldruckverteDIN ilung unter Flächengründungen. Beuth Verlag GmbH. Berlin. y, M. (1974). Berechnung von Flächengründungen. Ernst & Sohn. Berlin.

Kany, M. and El Gendy, M. 1996. Sicherheitsüberlegungen bei Flächengründungen. Fraunhofer IRB Verlag. Stuttgart. e, J. (2002). Centrifuge Technology. Workshop on constitutive and centrifuge modelling: two extremes. Monte Verità. Balkema, Rotterdam, 75-105. e, J. and Arnold, A. 2008. Physical Modelling of Soil-Structure Interaction of Flexible Raft Foundations. IN Proc. 2nd BGA Int. Conf. on Foundations. Dundee Scotland. Balkema. Rotterdam. 1569-1580.

Leussink, H., Blinde, A. and Abel, P.G. 1966. Versuche über die Sohldruckverteilung unter starren Gründungskörpern auf kohäsionslosem Sand. TU Karlsruhe.

yerhof, G.G. 1979. Generalbericht: SoMe il-Structure Interaction and Foundations. 6th Panam. Conference. Lima. Peru.

ir Wood, D. 1990. Soil behavioMu ur and critical state soil mechanics.Cambrigde University Press.

midt, H. 1991. Erddruckfragen bei Tunnelbauwerken der offenen Bauweise. Bauingenieur 66, 517-527.

Schofield, A.N. 1980. Cambridge geotechnical centrifuge operations. Géotechnique 30(2), 227-268.

ingman, S.M., Laue, J., Boyle, R., White, J. and Zweidler, A. 2001. The ETH Zurich Geotechnical Drum Centrifuge. International Journal of Physical Modelling in Geotechnics 1(1), 59-70.

ingman, S.M., NateSpr r, P., Chikatamarla, R. and Laue, J. 2002. Use of flexible tactile pressure sensors in geotechnical centrifuges. International Conference of Physical

W

883

Investigation on the dynamic properties of saturated sand-tire chips mixture by shaking table

Étude des propriétés dynamiques d'un mélange de sable saturé et de chute de pneumatiques sur table vibrante

Bahadori H., Manafi S. Urmia University, Iran

ABSTRACT: Liquefaction of saturated sands is one of the most important topics in geotechnical engineering. Reinforcing saturated sands is one of the solutions to mitigate liquefaction potential. Scrap derived recycle materials (such as tire chips and tire shreds) are some kinds of reinforcing materials. In addition to mitigation effects, the reinforcing materials cause an improvement in dynamic properties of the soils. A series of one dimensional 1-g shaking table model tests were conducted on sand and sand mixed tire chips. Firoozkuh sand No. 161 is used in this study. Four different percentages of sand-tire chips were tested in this research. Shear modulus and damping ratio degradation curves are presented in the hysteresis loops too. Results show that damping ratio increases with increasing tire chips content in mixture. Also at the certain shear strain amplitude, shear modulus of reinforced soil decreases with increasing percentage of tire chips. RÉSUMÉ : La liquéfaction des sables saturés est l’un des sujets les plus importants en géotechnique. Le renforcement des sables saturés est une solution pour atténuer le potentiel de liquéfaction. Des matériaux recyclés (tels que des copeaux et des lambeaux de pneus) sont utilisés comme matériaux de renforcement. Outre l’effet d'atténuation, les matériaux de renforcement amènent une amélioration des propriétés dynamiques des sols. Une série d'essais de chargement sur table vibrante 1D ont été réalisés à 1g sur des mélanges de sable et de copeaux de pneus. Le sable Firoozkuh n ° 161 est utilisé dans cette étude. Quatre différents pourcentages de copeaux de pneus ont été testés. Le module de cisaillement et l'amortissement sont tirés des cycles d'hystérésis. Les résultats montrent que le taux d'amortissement augmente avec la teneur en copeaux de pneus contenu dans le mélange. Également, à une certaine amplitude de déformation de cisaillement, le module de cisaillement de sol renforcé diminue avec l’augmentation du pourcentage de copeaux de pneus.

KEYWORDS: Tire chips, liquefaction, shaking table

1 INTRODUCTION

When saturated clean sand deposit is subjected to seismic loading, the pore water pressure gradually increases until liquefaction happens and settlement occurs during and after an earthquake. The mentioned problem is attributed to rearrangement of grains and redistribution of voids within the soils. Over the years many methods have been presented to increase liquefaction resistance. However the main methods utilized in liquefaction mitigation are classified as densification, solidification, drainage and reinforcement techniques (Ghiassian and Ghazi 2009). Utilizing tire chips in soils is a kind of soil reinforcement which has a wide range of application. Many research works have been performed to achieve fundamental engineering properties of soil- rubber mixture, such as compaction characteristics, permeability, compressibility, modulus of elasticity, and Poisson’s ratio. Feng and Sutter (2000) conducted a series of resonant column test to obtain shear modulus and damping ratio of sand reinforced with rubber. They expressed that shear modulus and damping ratio of the mixtures is strongly influenced by the percentage of the rubber inclusion. Few studies have been performed on the effect of adding tire chips in mitigating the liquefaction potential of sand. Hyodo et al. (2007) carried out undrained cyclic triaxial tests on

sand samples reinforced with tire chips. They found out that tire chips control build-up of excess pore water pressure of the mixture during shear. They obtained that for sand fraction (i.e. sand volume /tire chips volume) lower than 50 percent, liquefaction does not occur at all. Studies on liquefaction resistance of reinforced soils with tire chips have been so far limited to almost element tests. In this paper a series of 1g shaking table tests were carried out to investigate on the effect of tire chips -sand mixture in reducing liquefaction potential, settlements after earthquake and pore water generation.

2 SHAKING TABLE AND MATERIALS

2.1. Model configuration and instrumentation

Figure 1 illustrates shaking table and its instruments. Container is made of Plexiglas with inner dimensions of 200×50×70 cm. At bottom of the container a void chamber is made by using a number 200 sieve (Bahadori and Motamedi 2011). A plastic plate was rigidly fixed at the center of container to separate reinforced and unreinforced parts from each other and waterproofing carefully. So two models (reinforced and unreinforced) can be tested at once with the same input acceleration (Uchimura et al. 2007).

884


Figure 1.General view of shaking table model

2. 2. Materials

The tire chips used in this study was made from discarded tires. The particles shape was very irregular and angular. The tire chips particles have negligible water absorption, and very small volumetric compression due to isotropic pressure. Table 1 demonstrates physical properties of tire chips. Table 1. Physical properties of tire chips material

Material D10 (mm)

D50 (mm)

Cc Cu Gs

Tire chips

2.1 3.9 0.99 2.05 1.16

Firoozkuh No.161 sand was used for the mixture in

reinforced side, and pure sand in unreinforced side. Table 2 demonstrates physical properties of sand. Table 2.physical properties of sand material Material Gs emax emin Cc Cu D50

(mm) Sand 1.16 0.874 0.548 0.97 2.58 0.3

2.3. Experimental procedure

Uchimura et al. (2007) presented following relation to calculate mixture ratio of tire chips that were evaluated by the dry weight of the tire chips relative to the total mixture material:

TCr

S TC

MTCM M

(1)

( : Tire chips content, rTC TCM : Weight of tire chips ,

SM : Weight of Firoozkuh sand). In this study 4 mixture ratio (TCr=10%, TCr=20%,

TCr=30% and TCr=40%) were selected. Maximum mixture ratio was limited to 40 percent, because if tire chips content were higher, the sand could not fill the entire voids among tire chips particles and the model became non-uniform.

Relative density of tire chips-sand mixture was calculated by following relation:

max

max min

(( )

sr

e eDe e

) (2)

( se

mine: sand void ratio, : maximum void ratio of sand

, : minimum void ratio of sand). maxe

Where se can be calculated as: Total s Tc

ss

V V VeV (3)

( : Total volume of mixture, TotalV sV : Volume of sand particles , :Volume of Tire chips particles) TcV

Both of unreinforced (pure sand) and reinforced (tire chips-sand mixture) models were prepared by wet tamping method, in which soil is mixed with 5% water.

Each model (reinforced or unreinforced) was prepared in six layers. The required weight for each layer was considered based on the desired density (equivalent value of the maximum void ratio or zero relative density) and exact volume of the layer. Each portion was placed into the model container and then tamped with light trowel to reach desired level. Carbon dioxide (CO2) was allowed to pass through the specimen at a low pressure in order to replace the air that trapped in the pores of the specimen. Then water was allowed to flow upward through the bottom of the container at low pressures in order to flush out the CO2 that cause increasing the final degree of saturation. Vibration with approximate uniform amplitude and 2.1 Hz freq was manually applied to the container (the shaking table was designed to vibrate at around 2 Hz frequencies).

3 TEST RESULTS AND DISCUSSION

3.1. Time history of acceleration

Figure 2 is a typical plot of time history of base acceleration measured by accelerometers (a5). It is noticeable that in all models base acceleration was continued for 23 second. Results indicated that acceleration within the soil tends to be increased towards the soil surface. On the other hand, after initial liquefaction (that occurred at un-reinforced models and also reinforced model with TCr=10%), acceleration is decreased due to the increase in excess pore water Pressure.

(d) Figure 2.Typical Time History of Base Acceleration

3.2. Shear stress-strain relationship

From the original shear beam equation, shear stress τ at any depth z may be written as the integration of density (ρ) times acceleration (ü) through higher levels (Eq.4).

0

z

udz (4)

A linear fit is recommended between adjacent pairs of instruments, which may be extrapolated from the top pair to surface (Eq. (5)).

885

Technical Committee 104 / Comité technique 104 Proceedings of the 18th International Conference on Soil Mechanics and Geotechnical Engineering, Paris 2013

2 11 2 1

2 1

( )( ) ( )u uu z u z zz z

(5)

Shear stress is evaluated using Zeghal

and Elgamal’s expression with the inobtained from Eq. (5) with z=0:

To calculate the damping ratio in each cycle of motion (Eq. 9), first, dissipated energy

terpolated surface acceleration

1( ) ( (0) ( ))2

z z u u z (6)

If only two instrument

s are present in a given soil layer, a simple firscalculate shear strain:

t order approximation must be applied to

2 1

2 1

( )( )u uz z

(7)

This applies for any point between instruments 1 and 2, and as such is more appropriate for the midpoi

nt. Figure 3 shows Stress–strain loops at P1 and P2 in reinforced and unreinforced sides of test with TC =40%.

r

Figure 3. Shear stress-strain loops for unreinforced and reinforced models with TCr=40%

It is clear that the hysteresis loops in unreinforced model tends to become progressively flatter and narrower as the sample begins to liquefy and display a clear reduction in stiffness. One reason for the good performance of sand-tire chips mixture in reducing generation of excess pore water pressure and increasing liquefaction resistance is high permeability of reinforced mixture, as compared with the pure sand. Another reason is probably that the stiffness of tire chip particles is less than that of sand grains, consequently allowing some volume compression under developed excess pore water pressure. Thus the volume compression of tire hip produces a situation similar to drc ainage or

xcess pore

ed from the ratio of the difference in maximum and minimum stress and strain developed in desired lo

dewatering which decreases the extent of ewater pressure (Towhata, 2008).

3.3. Shear modulus and damping ratio

Shear modulus is obtain

op.

(8)

( ) and absorbed energy ( ) must be calculated (Fig. 4).

(9)

Figure 4. Definition of Damping ratio and Shear Modulus Shear modules and damping ratios at 0.3m depth for all reinforced and unreinforced models are shown in Fig. 5.

(a)

(b)

(c)

(d)

Figure 5. Shear modulus of Reinforced and Unreinforced models: a) TCr=10% b) TCr=20% c) TCr=30% d) TCr=40% As it is depicted in Fig. 11, the shear modulus curve of reinforced model is placed over the unreinforced one.

886

Proceedings of the 18th International Conference on Soil Mechanics and Geotechnical Engineering, Paris 2013 Proceedings of the 18th International Conference on Soil Mechanics and Geotechnical Engineering, Paris 2013

n curves of reinforced odels shift to left side of strain axis and consequently

obtained shear strain is reduced.

Also, increasing content ratio of tire chips in mixture causes shear modulus degradatiom

Figure 6. Damping ratio of Reinforced and Unreinforced model with TCr=40%.

Damping ratio curve of reinforced and un-reinforced model with TCr=40% are presented in Fig. 6. At all models, the scattering of damping ratio values is high. As a general result, value of damping ratio is reduced by increasing strain amplitude. This process is similar to conclusions of Brennan (2004), which explained mentioned phenomena happens when the soil particles lose contact each other due to increase of pore water pressure and consequently cause frictional energy of soil skeleton reduced and since damping ratio is the ratio of dissipated energy to absorbed energy therefore damping

alue of the 10-2, is

cr

atio of Reinforced Models

build-up of excess

-up of

mixture due to

decreasing excess pore water generation. increased with increasing tire chips mixture

5Ghi

so

Baha

Uch

Tow

ab halandarzadeh, A., Fakher, A. 2009. Experimental study on seismic deformation modes of reinforced-soil walls. j. Geotextiles and Geomembranes., 27(9), 121-136.

3) Maximum shear modulus of reinforced soil increasedwith increasing tire chips content in

4) Mean damping ratio ischips content in sand-tire

REFERENCES assian, H. and Ghazi, F. 2009. Liquefaction analysis of fine sand reinforced with carpet waste fibers under triaxial tests, 2nd International Conference on New Developments in Soil Mechanics and Geotechnical Engineering, Near East University, Nicosia, North Cyprus, pp. 448-455.

Zheng-Yi, F. and Sutter, G. Dynamic Properties of Granulated Rubber/Sand Mixtures, Geotechnical Testing Journal, GTJODJ, Vol. 23, No. 3, September 2000, pp.338-344.

Hyodo, M., Yamada, S., Okamoto, M. 2007, Undrained cyclic hear properties of Tire chips–sand mixture, Int. workshop n scrap Tire derived geomaterials, Yokosuka, Japan. dori, H. and Motamedi, H. 2011, An investigation on the effects of geogrid and geogrid- geomembrane geocomposite on the reduction of settlement due to liquefaction”, 6th International Conference on Seismology and Earthquake Engineering, Tehran, Iran

imura, M. ,Chi, N.A. ,Nirmalan ,S. ,Sato ,T. ,Mediani , M., Towhata , I. 2007. Shaking table test on the effect of tire chips and sand mixture in increasing liquefaction resistance and mitigation uplift of pipe, International workshop on scrap tire derived geomaterials, Yokosuka, Japan.

Zeghal, M. and Elgamal, A.W. 1994. Analysis of site ratio is reduced. liquefaction using earthquake records. J. Geotech. Eng., 120(6), 996–1017. hata, I. 2008. Geotechnical earthquake engineering,

3.3.1. Mean damping ratio

Due to observing relatively irregular and non-uniform trends of damping ratios versus shear strain that has occurred because of various reasons such as high shear strain amplitude and sudden increase in pore water pressure, a new parameter is defined as the mean value of damping ratio by Sabermahani et al. (2009) to compare the values of damping ratio of reinforced models with each other. The values of mean damping ratio versus tire chips content ratio of reinforced model tests are plotted in Fig. 7. Results show that v

Springer First Edition. Brennan, A. J. 2004. Vertical drains as a countermeasure to

earthquake-induced soil liquefaction. PhD thesis, Univ. of Cambridge, Cambridge, U.K. ermahani, M., GS

mean damping ratio at the shear strain range of eased with increase in tire chips content. in

Fig.7 .Mean damping r

4 CONCLUSIONS

1) It seems tire chips can control the pore water pressure of the mixture during earthquake and increase liquefaction resistance. 2) Although unreinforced sand shows reducing in stiffness during earthquake due to rapid buildexcess pore water pressure, no sign of losing in stiffness was observed in reinforced sand with tire chips.

887

The influence of the g-level for anchor tests in sand

L’influence du niveau de g pour les tests d’ancrage en sable

Bezuijen A. Ghent University Ghent, Belgium/Deltares, Delft, Netherlands

Zwaan R., Lottum van H. Deltares, Delft, Netherlands

ABSTRACT: Physical model tests in geotechnics are quite often performed in a centrifuge, because then the stresses are the same inmodel and prototype, leading to comparable stress-strain behaviour. However, in theory for a pure friction material as sand, it shouldbe possible to get the same results in a reduced stress 1-g model as in an N-g model. This was checked in a series of anchor pulling tests. The anchor was pulled through a sand bed and a gravel berm. Tests were run with the same set-up at 80-g and at 1-g. The pulling force was measured as a function of time. Results show that there is a clear distinction between the 1-g and 80-g tests. The pulling force was relatively higher in the 1-g tests. This means that also for a pure friction material, stresses has to be the same in model and prototype.

RÉSUMÉ : Des essais sur modèles physiques en géotechnique sont souvent effectués en centrifugeuse, parce que les contraintes sontles mêmes dans le modèle et le prototype, ce qui offre un comportement contrainte-déformation comparable. Cependant, en théorie,pour un matériau purement frottant comme du sable, il devrait être possible d'obtenir les mêmes résultats dans un modèle 1-g auxcontraintes réduites, comme dans un modèle à N-g. Ceci a été vérifié dans une série de tests de traction d'ancre. L'ancre a été tirée à travers un lit de sable et une berme. Le tests à 80-g et à 1-g ont été effectués d’un arrangement identique. La force de traction a été mesurée en fonction du temps. Les résultats montrent qu'il y a une distinction claire entre les tests 1-g et les tests 80-g. La force de traction est relativement plus élevée dans les essais 1-g. Cela signifie que pour un matériau purement frottant, il faut que lescontraintes soient identiques dans le modèle et le prototype.

The influence of the g-level for anchor tests in sand

L’influence du niveau de g pour les tests d’ancrage en sable

A. Bezuijen Ghent University Ghent, Belgium/Deltares, Delft, Netherlands

R. Zwaan, H. van Lottum Deltares, Delft, Netherlands

ABSTRACT: Physical model tests in geotechnics are quite often performed in a centrifuge, because then the stresses are the same inmodel and prototype, leading to comparable stress-strain behaviour. However, in theory for a pure friction material as sand, it should be possible to get the same results in a reduced stress 1-g model as in an N-g model. This was checked in a series of anchor pulling tests. The anchor was pulled through a sand bed and a gravel berm. Tests were run with the same set-up at 80-g and at 1-g. The pulling force was measured as a function of time. Results show that there is a clear distinction between the 1-g and 80-g tests. The pulling force was relatively higher in the 1-g tests. This means that also for a pure friction material, stresses has to be the same in model and prototype.

RÉSUMÉ : Des essais sur modèles physiques en géotechnique sont souvent effectuésen centrifugeuse, parce que les contraintes sontles mêmes dans le modèle et le prototype, ce qui offre un comportement contrainte-déformation comparable. Cependant, en théorie,pour un matériau purement frottant comme du sable, il devrait être possible d'obtenir les mêmes résultats dans un modèle 1-g auxconstraintes réduites, comme dans un modèle à N-g. Ceci a été vérifié dans une série de tests de traction d'ancre. L'ancre a été tirée à travers un lit de sable et une berme. Le tests à 80-g et à 1-g ont été effectués d’un arrangement identique. La force de traction a été mesurée en fonction du temps. Les résultats montrent qu'il y a une distinction claire entre les tests 1-g et les tests 80-g. La force de traction est relativement plus élevée dans les essais 1-g. Cela signifie que pour un matériau purement frottant, il faut que lescontraintes soient identiques dans le modèle et le prototype.

KEYWORDS: centrifuge tests, scaling, anchor tests, friction material.

1 INTRODUCTION

Dragging anchors can be a real threat for pipe lines located at the sea bottom. With the number of pipelines and cables increasing as well as the number and size of the ships, it can be expected that this threat will increase in the future.

Pipelines and cables that cross shipping lanes are usually protected by gravel berms. The berm has to be stable against the chain of the anchor and the anchor itself. Some damage to the berms is allowed, but the pipeline and cable has to be protected, even for the heaviest anchors that can be expected. These berms are designed by experience and traditionally tested using large scale (scale around 1:5) model tests. Some first attempts have been made to simulate the process numerically using the so-called ‘rigid body technique’, see the visualisation of a numerical result in Figure 1. This is a promising path, see also leQin (2010), but up to now not ready to be used in a design.

Figure 1.Visualisation of numerical simulation of an anchor passing a berm using 'rigid body dynamics' (Bezuijen, 2011).

To avoid the relatively expensive large scale model tests, it is also possible to use a centrifuge model. The advantage of a centrifuge model is that a much smaller model is possible and still the stresses are the same in model and prototype. For a 1-g

scale model the stresses in the model will always be smaller than in the prototype, see Table 1.

However, in theory for a pure friction material as sand, it should be possible to get the same results in a reduced stress 1-g model as in an N-g model. This was checked in a series of anchor pulling tests. The anchor was pulled through a sand bed and a gravel berm. Tests were run with the same set-up at 80-g and at 1-g. The pulling force was measured as a function of time.

This paper presents the scaling rules, the set-up and results of the 1-g and 80-g tests will be described in the paper.

2 SCALING

2.1 N-g scaling

In a centrifuge model the length is N times smaller than in the prototype and the acceleration N times higher. The scaling relations the relevant parameters are presented in Table 1. As usual in centrifuge modelling the sand is not scaled from prototype to the model, because the sand grains are much smaller than the dimensions of the anchor, but the gravel material is scaled and N-times smaller in the model compared to prototype. It is difficult to fulfil the scaling rule for the velocity. It is necessary that the velocity is the same in model en prototype when dynamic scaling is assumed, but the velocity has to be even N times higher in the model compared to prototype when consolidation is the dominant mechanism. Since ships dragging anchors can still have a velocity of several metres per second, it is rather difficult, even to achieve the ‘dynamic’ scaling rule. In our tests an anchor velocity of 100 mm/min = 0.00167 m/s is used (for higher velocities it would be difficult to control and monitor the process during the test). This velocity will create a drained behaviour of the sand in the model while a partly drained behaviour in prototype is expected (see Van Lottum et al, 2010) and a drained behaviour in the gravel for both model

888


and prototype conditions. Further it is assumed that dynamic forces are limited during the anchor dragging.

Table 1. General scaling laws with scaling factor N.

Parameter scaling law model/prototype Ng-model 1g-model

Unit

Length 1/N 1/N mMass 1/N3 1/N3 kgForce 1/N2 1/N3 NStressTime (dynamic) Time (consolidation)

11/N1/N2

1/N1/√N1/N2

kPass

Velocity (dynamic) Velocity (consolidation)

1N

1/√NN

ms-1

ms-1

2.2 1-g conditions

The scaling in 1-g conditions is also presented in Table 1. It appears from the table that the stresses will be N times lower in the model compared to prototype. This means that also the strength of the soils has to be N times lower. For a soil with an undrained shear strength, as clay, this is difficult to achieve. However, for a pure friction material this is rather easy, because the N-times lower stress results automatically in a lower strength, assuming that the friction angle remains constant for the various stress levels.

Using dynamic scaling, the same scaling law for the velocity (Froude scaling) as in 1-g hydraulic modelling tests is found. However, when consolidation is dominant, again the velocity in the model has to be N times higher than in the model. As in the centrifuge model, it is assumed that the anchor will behave drained in both the sand and the gravel layer.

2.3 Conclusions scaling

The scaling laws cannot be fulfilled completely with respect to the prototype. However, assuming that consolidation is more important than dynamic forces, the error made because of assuming undrained behaviour in the sand and drained behaviour in the gravel is exactly the same in both models. This makes a good comparison possible between the 1-g and N-g 1:80 g models.

3 TESTS PERFORMED

3.1 Test set up centrifuge tests

Tests were run at 80 g in a specially developed container of L x W x H: 1.80 x 0.5 x 0.5 m, see Figure 2 and Figure 3. The length was necessary since a berm can be damaged not only by an anchor, but also by the anchor chain that removes stones on the berm before the anchor reaches the berm. The container is placed on a water reservoir, so that the water level can be changed during the test (by adding water from the reservoir or vice versa). This is of importance for such a long container, since during spinning up and spinning down, water movements in the container can destroy the soil model (sand and anchor berm). Therefore the water level was increased after spinning up and decreased before spinning down.

A pulley system was constructed on top of the container, see Figure 4, to be able to drag the anchor over the full length of the container using a hydraulic plunger with a stroke of 0.5 m. As usual in the Geo-Centrifuge of Deltares tests were performed under reduced air pressure conditions of 50-60 mbar. More details on the set up can be found in Van Lottum et al. (2010).

The anchor used in the tests was an AC-14 anchor. The model is shown in Figure 5. The model anchor and anchor chain were made of stainless steel using a 3-D print technique and cast with the so called lost wax method. The anchor and chain is

printed in wax, which is replaced by stainless steel. By this technique an accurate scaled copy of the original was obtained.

Water reservoir Assembly plate

Hydraulic actuator

Valves for water su

AnchorPulley system

Cameras

pply

Figure 2. Anchor dragging test setup on assembly plate

32 cm125 cm

ChainAnchor Dyneema Rope

Pulley system

Force transducer

Actuator rodHydraulic actuator

10 cm

18 cm

Figure 3. Sketch set-up

Figure 4. Pulley system in test set up.

30 mm46 mm

52 mm

Figure 5. Model AC-14 anchor.

The soil model consists of a homogeneous sand layer of Baskarp sand (d50= 135 m) with a relative density of 65 – 75% and a peek friction angle of 40 degrees. On the sand a pipe line of 13 mm diameter and a gravel berm was placed (d50=5.3 mm), see Figure 6. The porosity of the gravel was around 40% and the peek friction angle 48 degrees.

0

20

40

Y(m

m)

0 50 100 150 200 250 300X (mm)

Figure 6. Dimensions of model berm

889


Tests were performed under saturated conditions with 0.15 m of water on top of the sand bed. The anchor was pulled through the sand bed until it was on the top of the berm. During the test the displacement of the plunger and the force of the plunger were measured continuously. After the test the water table was lowered to create some capillary forces to keep the anchor in position during spinning down. Back at 1-g the position of the anchor was carefully measured, see Figure 7.

Figure 7. Carefully measuring the position of the anchor after a test.

3.2 1-g tests

The set-up for the 1-g tests was exactly the same as for the centrifuge test. The same soil preparation technique, container, plunger and pulley system were used only now the tests were run outside the centrifuge at normal 1 g conditions under atmospheric pressure. Measurements performed during the tests and after the tests were the same as in the centrifuge. Three tests were performed.

4 TEST RESULTS

4.1 Corrections on measurement data

The parameters of importance are the penetration and the displacement of the anchor and the pulling force on the anchor.The penetration was measured after the test. The other parameters were determined during the test from the displacement of the plunger and the force that was measured on the plunger. The cable used in the pulley system was 3 mm dyneema cable with a maximum pulling strength of 5 kN. In order to limit elongation during the test, the cable was pre-stressed with a force of 2.8 – 3.0 kN. However, there still was some elongation of the cable. Furthermore, there will be friction in the pulley system. A dummy test was performed to correct for the friction both at 1-g and 80-g. In this test the Dyneema cable was connected with a spring connected in the centrifuge and an extra force transducer was located between the spring and the cable. Such a transducer could not be placed between the anchor and the cable during the real tests because the dimensions of the transducer and the necessary electrical cables would influence the test results. In the tests, the force on the cable at the spring and the force on the plunger were measured. The results of the measurements are presented in Figure 8. Due to friction in the system, the results differ depending on the direction of movement. The movement from left to right in the plot is the movement during anchor pulling. It appears that, apart from very small puling forces at plunger displacements around -120 mm, during pulling the pulling force as measured in the cable with the force transducer near the spring is always about 0.75 times the force measured with the force transducer at the plunger (and divided by 5 to correct for the pully system). This is only possible when the friction in the system increases linearly with the pulling force. This correction was applied in Figure 9.

-0.30

-0.25

-0.20

-0.15

-0.10

-0.05

0.00

0.05

0.10

forc

ein

cabl

ean

dfri

ctio

n(k

N)

-1.4

-1.2

-1.0

-0.8

-0.6

-0.4

-0.2

0.0

0.2

0.4

Forc

eon

plun

ger,

blue

line

(kN

)

-130 -120 -110 -100 -90 -80 -70 -60displacement plunger (mm)

Force measured on cableForce on plunger (right Y- axis)friction

Figure 8: Comparison forces measured in a dummy test on the cable and on the plunger at 80g.

The correction for the elasticity of the cable was only performed for the 80-g tests. Due to the much smaller forces this was not necessary for the 1-g tests. The elasticity of the cable can be seen at the end of a test. When the anchor is pulled to its final position (on top of the berm) the pulling force is decreased retracting the plunger, while the anchor remains at the same position (controlled by the cameras). This allowed for higher pulling forces to measure is the elastic deformation of the cable. For low pulling forces there is an additional mechanism, the cables sag due to gravity. The last mechanism is only of importance for low pulling forces. Only the elastic relaxation is of importance during anchor pulling. Figure 9 shows the movement of the plunger as a function of pulling force duringrelaxation as measured in a test.

0.0

0.1

0.2

0.3

0.4

0.5

0.6

pulli

ngfo

rce

onpl

unge

r(kN

)

135 140 145 150 155 160displacement (mm)

Figure 9. Relaxation of cable and sagging at the end of a 80 g test. The slope of the steep vertical part of the measured plunger force is determined by the elastic strain. The flatter part at low pulling force is caused by sagging of the cables.

0.00

0.05

0.10

0.15

0.20

0.25

0.30

pulli

ngfo

rce

onan

chor

(kN

)

-600 -400 -200 0 200 400 600 800displacement (mm)

force corr.force not corr.

Figure 10. Pulling force and displacement with and without corrections on both force and displacement for a 80 g test.

Young’s modulus is about 4 kN/m, measured at the plunger, thus Young’s modulus of the cable is 4/25=0.16 kN/m.

The influence of the corrections for both the displacement and the force on the results are shown in Figure 10.

It is clear that the correction for the displacement hardly influences the results even at 80 g, but that the influence of the correction for the friction force is considerable.

890


4.2 Observations and results

In the 80-g tests the AC-14 anchor appeared to be a reasonable stable anchor. This means that pulling the anchor with the device described above, the anchor digs into the sand and does not rotate or rotated partly (up to 90 degrees). This was different for the 1-g test. In this test the anchor rotated 180 degrees around its pulling axis in front of the berm. In the model anchor the flukes were fixed (different from a real anchor). This means that when the anchor rotates, the flukes are pointing upwards and the anchor will not dig into the sand or the gravel berm. To avoid that the rotation of the anchor dominates all results the last test was performed with the anchor just in front of the berm and it was pulled over a short distance only.

The measured force displacements of both the 80-g tests and the 1-g tests are shown in Figure 11. The forces measured in the 80-g tests were divided by 80 to make them comparable with the results of the 1-g tests. Perfect scaling would mean that the 80-g test is 80 times higher, see Table 1. Thus dividing this force by 80 should result in the same value as the result of the 1-g test; Figure 11 shows that this is not the case. The force measured in the 1-g test is relatively higher.

0

1

2

3

4

5

6

7

8

Forc

e(N

)sca

led

to1g

-800 -600 -400 -200 0 200 400 600 800Displacement (mm)

Test ATest BTest CTest 1 80gTest 2 80 g

80-g tests

1-g tests

Figure 11. Force versus displacement for 1-g and 80-g (scaled, see text) tests.

Due to the rotation of the anchor just in front of the berm in 2 of the 3 tests, there is only one measurement of the maximum penetration of the anchor in the berm. This was on average 25.4 mm for the 80-g tests and 21.8 mm for the 1-g test. The difference is visible on the pictures taken after the test. After a 80-g test, Figure 12 the anchor flukes are completely in the berm (one fluke is visible in the picture but this is because the gravel is taken away for the measurement of the position of the fluke, the fluke in the upper part of the picture shows the original situation). Figure 13 shows that after the last 1-g test the flukes do not completely penetrate into the berm.

Figure 12. Position of anchor at the end of an 80-g test.

Figure 13. Position of anchor at the end of last 1-g test.

5 DISCUSSION

All results indicate that the soil and berm at the low stress levels of a 1-g test behave relatively stronger and stiffer than at the original stress level that is present during an 80-g test. If the stresses are not properly scaled, but lower than in reality; the soil behavior in a model test is stiffer and stronger than in the prototype. This means that also for purely friction materials as tested here, the proper representation of the stress-state is important. To test the protection efficiency of a berm against anchor dragging, 1:5 scale tests at 1-g are quite common. Looking at the results of this research, it is very likely that the results of these 1:5 scale model tests underestimate the penetration depth of the anchor in prototype, which is the primary objective of these tests, because that determines whether or not a pipe line is sufficiently protected. At a scale 1:5 the error will be smaller than at the scale 1:80 tested here, but can still be of importance. 6 CONCLUSIONS

Comparing the results of anchor tests at a scale 1:80 at the original stress level in a centrifuge with the results of a further identical 1:80 test at 1-g with thus a reduced stress level, led to the following conclusions: - The drag forces at 1-g are higher than 1/80 of the drag forces at 80-g . - The stability of the anchor is less during the 1-g tests. The penetration depth is lower in a 1-g test (only one test result) - Consequently the results indicate that in general a 1-g scale model test underestimates the penetration depth of the anchor and therefore overestimates the protection efficiency of the berm.7 ACKNOWLEDGEMENTS

The authors want to acknowledge Deltares for providing the possibility to perform the 1-g tests and Thijs van Dijk, Frans Kop, Jennifer Rietdijk and Ferry Schenkeveld for their ontribution to these tests. c

8 REFERENCES

Bezuijen A. (2011), Rigid body calculations (personal communication). LeQin Wang, HongKiat Chia. (2010) Optimization study of pipeline

rock armour protection design based on finit element analyses.Proceedings of the ASME 2010 29th International Conference on Ocean, Offshore and Arctic Engineering OMAE2010 © 2010 ASME.

Van Lottum H., Luger H.J., Bezuijen A. (2010) Centrifuge anchor dragging tests in sand and clay. Proc. Physical Modelling in Geotechnics – Springman, Laue & Seward (eds)© 2010 Taylor & Francis Group, London, ISBN 978-0-415-59288-8 1063-1068.

891

An experimental study on the consolidation of soft clayey soils using electrochemical methods

Étude expérimentale de la consolidation des argiles molles avec des méthodes électrochimiques

Cardoso R., Nogueira Santos J. ICIST, Instituto Superior Técnico, Lisbon Technical University, Portugal

ABSTRACT: An experimental study was performed where the acceleration of the consolidation using electroosmosis of normally consolidated saturated white Kaolin was investigated. The speed of consolidation was measured through the consolidation coefficientcomputed using the results of oedometer tests in which DC voltage was applied during the entire loading period. The inclusion ofdrains and electrodes in the drains was also studied in an experimental apparatus developed for this purpose. The results are interpreted and data is analyzed considering the application of this technique in practice.

RÉSUMÉ : Une étude expérimentale a été réalisée où la consolidation est accélérée en utilisant l'électro-osmose dans du Kaolin blancsaturé normalement consolidé. La vitesse de consolidation a été mesurée par le coefficient de consolidation calculé en utilisant lesrésultats des essais oedométriques dans lesquels la tension continue était appliquée au cours du chargement. L'inclusion des drains etdes électrodes dans les drains a été également étudiée dans un appareil expérimental développé à cet effet. Les résultats sontinterprétés et les données sont analysées en examinant l'utilisation de cette technique pour la pratique dans plusieurs conditionsnaturelles.

KEYWORDS: Electrokinetic, consolidation, water percolation, electrical technique, electroosmosis.

1 INTRODUCTION

Soil treatment with binders or other cementing materials, or the adoption of speed up consolidation techniques are necessary to treat soft clayey soils to make them appropriate to be used as foundation soils. Several techniques are available nowadays, such as pre-charge embankments, radial drainage, drainage with vacuum, electrochemical methods, etc, which have specific advantages/ disadvantages, variable cost and different implementation challenges. This paper focuses on the use of electrochemical methods, such as electroosmosis. Some field cases are described in the literature (see Glendinning et al., 2005, for example) and concern stabilization of slopes, excavations and embankments. Several aspects must be analyzed concerning the use of dewatering techniques based on electrochemical effects, mainly related with design, cost/efficiency and possible collateral effects.

The working principle of electrochemical methods is to apply a difference in electrical potential between electrodes placed in the soil to move the positive ions (cations) towards the negative electrode (cathode) and the negative ions (anions) towards the positive electrode (anode). These techniques are used to drain saturated soils and therefore to improve their mechanical properties, as well as for decontamination purposes. For the particular case of clay minerals, due to their negatively charged surface, they attract positive ions and immobilize them on the double layer to neutralize the electrical forces involved. Thus, a movement of cations will occur, which will carry the pore water in the same direction. The water is carried in this flow of ions in a viscous manner and, if water is drained, this will results in dewatering from the positive electrode (anode) to the negative electrode (cathode).

The work presented in this paper describes an experimental study where normally consolidated saturated white Kaolin specimens were investigated in order to understand the advantages of using electroosmosis to accelerate consolidation when compared with adopting a mesh of drains to ensure radial drainage. The last method was also included in the study because radial drainage is one of the most used methods for accelerating the consolidation of clayey soils.

The speed of consolidation was measured through the consolidation coefficient and the consolidation time necessary for a given settlement to occur. The electroosmosis method was

studied in tests performed in oedometer conditions, in which DC voltage influence on settlements was analyzed. The study of radial drainage was performed in an experimental apparatus developed for this purpose. The inclusion of electrodes in the drains was also studied.

2 ELECTROOSMOSIS

As a technique for accelerating consolidation, electroosmosis can be studied has if it was a case of pore pressure increment, which can not cause undrained failure in the soil. Esrig (1968) studied the different types of pore pressures that can be developed in the soil mass as function of the drainage conditions. The more usual condition on consolidation by electroosmosis is assuming that drainage is performed only through the cathode electrode. Water flow qh is given by Equation 1, where ke is the coefficient of permeability measured when water percolates only due to electroosmotic effects, kh is the coefficient of hydraulic permeability, u is the excess of pore water pressure caused by electroosmosis (and/or by the increment of vertical stress) and Vis the voltage. This equation shows the proportionality between the voltage in the soil and the pore water pressure developed and is used to find the distribution of the pore water pressure.

xVk

xu

kkq e

e

hh

(1)

Equation 2 comes from introducing Equation 1 in Darcy’s equation and both in the equation that governs the one-dimensional consolidation (Terzaghi et al. 1996). In Equation 2, mv is the compressibility index of the soil, w is the volumetric weight of water, t is time and cv is the coefficient of consolidation. Distance x is measured along the flow path, one-dimensional in this case.

tu

cxV

kk

xu

vw

h

e

1

2

2

2

2

(2)

The solution of Equation 2 is Equation 3, which is used to compute the excess of pore water pressure developed during the

892


electroosmosis. In this equation, Vm is the maximum voltage applied to the soil and Tv is the time factor, which depends on the distance L between the electrodes and on the time t, and is given by Equation 4.

vn

n

wh

mew

h

e TABsenAk

VkxVkku 2

02 exp)1(2)(

(3)

with2

21

nA and

L

xnB

21

2LtcT v

v (4)

According to Mitchell and Soga (2005), the solution of Equation 3 is given by Equation 5 (parameter A given in Eq. 3) (Mitchell and Soga 2005), where U is the average degree of consolidation. These authors present some abacus with the solution for several cases.

(5) v

n

n

TAA

U 2

02 exp)1(41

In case of radial flow occurring simultaneously, Equation 2 can be converted into Equation 6, where r is the distance measured in the horizontal direction, cr is the coefficient of consolidation in this direction, x here is measured along the vertical direction, as well as cv, and the other parameters were already explained. Mitchell and Soga (2005) also found the solution for this equation. The case voltage V=0V can also be found by solving this equation, by correcting radius r to consider each drain.

tu

xuc

rV

kk

ru

rrV

kk

ruc vw

h

ew

h

er

2

2

2

2

2

2

2

2 1 (6)

3 SOIL, EQUIPMENT AND TESTS PERFORMED

The material used in the tests is a commercial white Kaolin (wL=75%, IP=40%, classified as CH). Reconstituted specimens were prepared with water content equal to 1.5 wL and were normally consolidated for a maximum stress of 12 kPa. The electrical resistivity of the saturated soil for different water contents (and therefore void ratios) was also measured in order to confirm that this parameter does not changes significantly during the performance of the oedometer tests. Further details can be found in Nogueira Santos (2012).

Some calibration tests were performed first to ensure that the oedometer equipment was isolated from the electrical system. This motivated the adoption of a PVC ring instead of a stainless steel ring, because PVC is an electrical insulator material.

A commercial 9V battery cell was adopted to apply the electrical flow to the soil. Later, a modified mobile phone battery charger was used, which is shown in Figure 1. This source has a DC voltage of 6.39V and an intensity of 0.71A and was chosen because the batteries were not able to keep constant voltage for long periods of time.

Figure 1. Modified mobile phone battery charger

Two different types of tests were performed where several different cases were tested. The specimens of the first type were tested in a normal oedometer cell adapted to apply an electrical field to the soil. Tests were performed with and without the application of electrical DC voltage and two different voltages

were tested: 6.35V and 9V. The oedometric cell used was modified to include four silver electrodes (square plates) in the top and in the bottom porous stones, as shown in Figure 2.

Figure 2. Silver electrodes on the porous stone

For the second type of tests a new consolidation cell was developed to include vertical drains. The spacing of the drains was designed so that radial flow would occur instead of vertical flow. This cell (120mm diameter and 70mm high) is made of acrylic and is shown in Figure 3. The top load plate of the cell was drilled to allow the inclusion of the drains and the settlement of the soil without interference. A geosynthetic material was placed between the specimen and the load plate to enable drainage from the top. The drains introduced allowed drainage by hydraulic gradients generated by the increment of vertical stress, or drainage generated by this mechanical action as well as with the application of an electrical field. The radius for the volume of soil surrounding each drain is 14mm, which allows considering that drainage occurs mainly in the radial direction. For the last case, medical needles were used as electrodes, placed inside the drains. The drains considered are the needles cases filled with fine sand shown in Figure3.

Figure 3. Apparatus for the radial flow test and detail of the drains.

Electrodes corrosion and the formation of an oxide were detected during the electroosmotic one-dimensional tests, as well as the formation of gas bubbles. Figure 4 shows some photographs of the gas formation (a) and the electrode corrosion (b). The silver oxide produced in test EO3 is shown in Figure 5. Only electrodes corrosion was observed in the electroosmotic radial flow tests.

(a) (b)Figure 4 – (a) Gas formation (b) Electrode Corrosion.

893


Figure 5. Silver oxide in the top of the specimen.

4 ONE-DIMENSIONAL CONSOLIDATION TESTS WITH ELECTROOMOSIS

Several oedometric tests were performed in order to determine the influence of the electric current in soil consolidation speed. Besides the reference test EC1, where there was no application of electrical current, the cases with electrical current allowed to study the influence of increasing voltage (EO1 with DC voltage 9V and EO2 with DC voltage 6.35V) and the influence of applying a reversible current flow (EO3 with DC voltage 6.35V). The stress path adopted was the same in all cases, consisting in increasing the vertical stress each 24h: 12kPa-25kPa-50kPa-100kPa-200kPa-400kPa-800kPa-1600kPa-400kPa-12kPa.

The influence of electroosmosis in consolidation was studied by comparing the results of tests EC1 with those of EO1 (9.0V) and EO2 (6.35V). Figure 6 shows the plot effective stress vsvoid ratio, where it can be seen that the electrical treatment applied for 24h slightly increases the magnitude of the settlements, however the compressibility characteristics of the soil are not much affected (EC1 (0V): Cc=0.470 and Cs=0.128; EO2 (6.35V): Cc=0.541 and Cs=0.146; EO1 (9.0V): Cc=1.135 and Cs=0.133).

The comparison of the results found in tests EO2 and EO1 allows understanding the influence of the applied voltage. As it can be seen in Figure 6, the use of a higher voltage for the same period of time (24h) increases the magnitude of the settlements. This is because the volume of water extracted increases.

The efficiency of the consolidation process must be measured in terms of the settlements rate instead of total settlements, in particular because, in practice, electroosmosis is applied during short periods of time because its effects are visible faster and also for reducing energy costs. This rate is measured through the coefficient of consolidation cv. The main differences between the three tests are illustrated in Figure 7. As shown in this figure, this coefficient does not depend on the applied stress and is higher for the tests where the electroosmotic treatment was applied. This coefficient increases one order of magnitude when the electrical current is applied, which proves that the current accelerates in a significant manner the consolidation process.

0.40

0.60

0.80

1.00

1.20

1.40

1.60

1.80

2.00

1.00 10.00 100.00 1000.00 10000.00

vertical stress (kPa)

void

ratio

EC1 (0V)

EO2 (6.35V)

EO1 (9V)

Figure 6. Effective stress vs void ratio curve for EC1, EO1 and EO2.

0.01

0.10

1.00

1.00 10.00 100.00 1000.00 10000.00

vertical stress (kPa)

cv x

10-4

(m2 /s

)

EC1 (0V)

EO2 (6.35V)

EO2 (9V)

Figure 7. Coefficient of consolidation as function of the applied stress.

The comparison of the coefficients of consolidation measured in all tests presented in Figure 7 shows that the values are very similar for both tests where voltage was applied, which indicates that the increment of voltage had no significant effect in the time necessary to consolidate the soil for high consolidation degrees. This may be due to the fact that the two voltages are very similar. Further analysis on the settlements measured in these two cases, and considering the similar consolidation coefficients measured, showed that secondary settlements increase with voltage, and for this reason there is a larger reduction in void ratio for 9.0V.

Figure 8 presents the evolution of the settlements in time under the vertical stress of 50kPa (load increment of 25kPa) for the tests EC1 and EO2 (6.35V). When the two curves are compared it can be seen that the settlements measured in EC1 after 24h (at the end of the consolidation) were reached in test EO2 about 4h after the application of the load increment. It can be seen also that settlements increase in time with constant rate when electroosomosis is applied, which indicates important secondary consolidation.

To conclude, 4h was the time needed in each load step of EO2 to achieve the settlement measured in EC1 at the end of consolidation in this test. This means that consolidation using electroosmosis was about 6 times faster than in the other cases if any type of treatment is adopted. Similar values were found for each loading increment, as well as for test EO3(6.35V) where reversible current was applied for load increments with the duration of 24h each. This last test had the advantage of reducing the formation of silver oxide.

894


-0.8

-0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

00 4 8 12 16 20 24

time (h)

Ver

tical

dis

plac

emen

ts (m

m)

EC1 (0V)

EO2 (6.35V)

Figure 8. Evolution in time of the vertical displacements under the vertical stress of 50kPa in tests EC1 and EO2.

5 RADIAL CONSOLIDATION TESTS

As mentioned before, a similar study was made for the radial flow tests EOR and ER. Figure 9 is similar to Figure 8 but presents the vertical deformations measured in the tests EC1 and ER, when the vertical stress of 50kPa is applied (load increment of 25kPa). Deformations are shown instead of vertical displacements to account with the different heights of the specimens. The curve from ER test was corrected to account with the stiffness of the equipment (including drains, top geotextile and load plate). When the two curves are compared it can be seen that the deformations measured in EC1 after 24h (at the end of the consolidation) were reached in test ER about 2hours after load increment (12 times faster than if there would be no drainage). This increment is larger than the one observed when tests EC1 and EO2 are compared, which indicates that the efficiency of radial drainage in accelerating consolidation is higher than that of electroosmosis.

-0.035

-0.030

-0.025

-0.020

-0.015

-0.010

-0.005

0.0000 1 2 3 4 5 6 7 8

time (h)

Ver

tical

def

orm

atio

ns

EC1 (0V)

ER (0V)

Figure 9. Evolution in time of the vertical deformations under the vertical stress of 50kPa in tests EC1 and ER.

There is no significant difference in the results found for the two tests performed with drainage in the radial direction. For the test where electrical current was applied the anode was in the exterior circle and the cathode was in the interior circle. The comparison of the values measured for the coefficient of consolidation in the radial direction (1x10-4m2/s, independently from the use or not of electrical current) (Nogueira Santos, 2012) and those measured in the oedometric cells presented in Figure 7 shown that the main mechanism of water drainage in these tests is through the drains because the values in radial direction are about one order of magnitude higher than those in

the vertical direction. The values of the coefficient of consolidation in the radial direction slightly increase when electrososmosis occurs but the increments are not significant. This may be explained by the fact that the drains, both in anode and cathode, were left open and water could flow also as in the test where electricity was not used.

The small difference between the cases using drains with and without voltage shows that the inclusion of electrical current brings no evident earning in time savings for the particular case of radial drainage. The application of electrical current may eventually reduce the number of drains but it will affect their radial efficiency and this must be studied in the future.

6 CONCLUSIONS

The study presented confirms that electroosmosis accelerates the one-dimensional consolidation of clayey soils because the consolidation coefficient cv increases about one order of magnitude. This is a very significant improvement.

The studies where an electrical field was applied for 24h indicated that the settlement obtained can be larger than if there was no electrical current. For this reason the duration in time of the treatment must be controlled. For the one-dimensional consolidation tests performed, the use of electrical current allowed reducing the consolidation period in a factor of 6. Regarding the value of voltage applied to the soil, higher voltages can result in larger settlements if they are applied the same period of time. It can be deduced that the time during which voltage is applied can be reduced if high voltages are applied, as long as they are safe in the field.

Accordingly with the results of the tests with radial consolidation, considering the overlap between radial and vertical directions taken as a hypothesis for Esrig (1968), if the drains mesh is well designed, the inclusion of radial drainage direction has the most important role. Eventually, drainage in the vertical direction may not be considered.

The inclusion of radial drains without any electrical current allowed reducing the consolidation period by a factor of 16, which is 2.6 times larger than the factor found for the one-dimensional consolidation test with electroosmosis. Comparing the two techniques for accelerating consolidation, although the good results obtained when electrical current is used, the reduction achieved may not compensate the cost increment due to the energy spent with the process, as well as the need of specialists to install and control the technique. The installation of drains is proven to be efficient and economical, which explains the use of this technique in practical and current cases. Nevertheless, it is believed that the use of electroosmosis can be advantageous in Environmental Geotechnics problems and for this reason studies like the one presented will be useful to understand better the potentiality of this technique.

7 REFERENCES

Esrig, M. 1968. Pore pressures, consolidation, and electrokinetics. Journal of the Soil Mechanics and Foundation Division. ASCE, vol 94 SM(4), pp. 899-921.

Glendinning S., Jones C.J. and Pugh R.C. 2005. Reinforced Soil Using Cohesive Fill and Electrokinetic Geosynthetics. International Journal of Geomechanics, vol 5(2), pp.139-146.

Mitchell, J. and Soga, K. 2005. Fundamentals of Soil Behavior, 3rd

Edition. John Wiley and Sons. Nogueira Santos, J. 2012. Study on the use of electrical techniques for

accelerating the consolidation of clayey soils. MSc Thesis, Instituto Superior Técnico, Universidade Técnica de Lisboa (in Portuguese).

Terzaghi, K., Peck, R.B. and Mesri, G. 1996. Soil Mechanics in Engineering Practice, 3rd Edition. John Wiley and Sons, Inc.

895

Variation of Friction Angle and Dilatancy For Anisotropic Cohesionless Soils

Variations de l’angle de Frottement et de la Dilatance pour les Sols Anisotropes Sans Cohésion

Cinicioglu O., Abadkon A., Altunbas A., Abzal M. Bogazici University, Istanbul, Turkey

ABSTRACT: The goal of this paper is to investigate and quantify the variation of peak friction and dilatancy angles of anisotropic cohesionless soils as functions of the in-situ state of the soil. In this context, in-situ state of the soil is used as a broad term that encompasses the combined effects of the stress state, volumetric state, and stress history of the soil prior to any shearing. Accordingly,the parameters that define the in-situ state of soil are in-situ confining pressure (pi), relative density (ID) and overconsolidation ratio (OCR), respectively. In order to quantify the influences of these parameters on the peak friction angle and dilatancy angle, a specialtesting program was designed that employs mainly CKoD triaxial tests. These tests were conducted on reconstituted sand samples at different pi-ID-OCR combinations. Analyzing the obtained results, two new functions are proposed that allow the calculation of thepeak friction angle and dilatancy angle of anisotropic cohesionless soils. The greatest advantage of the proposed functions is that they use directly measurable or calculable parameters as input. Finally, using similar test data collected from literature, the proposedempirical equations are validated.

RÉSUMÉ : Variations de l’angle de Frottement et de la Dilatance pour les Sols Anisotropes Sans Cohésion RÉSUMÉ : Le but decette étude est de chercher et de quantifier les variations des angles de frottement maximum et de dilatance de sols anisotropes sanscohésion comme des fonctions de l’état in-situ du sol. Dans ce contexte, l’état in-situ du sol est utilisé comme un terme général qui entoure les effets combinés de l’état de contrainte, l’état volumétrique, et l’histoire des contraintes du sol avant tout cisaillement. Par conséquent, les paramètres qui définissent l’état in-situ du sol sont la pression de confinement, la densité relative et le taux desurconsolidation, respectivement. Afin de quantifier les influences de ces paramètres sur l’angle de frottement maximum et l’angle de dilatance, un programme d’essai spécial a été conçu qui emploie principalement des essais triaxiaux. Ces essais ont été effectués surdes échantillons de sable reconstituées selon différentes combinaisons. L’analyse des résultats obtenus conduit à deux nouvelles fonctions qui permettent le calcul de l’angle de frottement maximum et de dilatance de sols anisotropiques sans cohésion.

KEYWORDS: dilatancy, friction angle, sand, Ko-consolidation, granular material

1 INTRODUCTION

Dilatancy is a property that is unique to granular materials. However, for soils, manifestations of dilatancy depends on grain size and shape; In case of fine-grained soils, we can describe dilatancy as latent dilatancy since dilatant behavior manifests itself as a change in the pore water pressure. Though, in case of coarse-grained soils, dilatancy is physically evident and can be directly measured by conducting simple soil tests. Even though for both fine and coarse-grained soils dilatancy influences strength, only for coarse-grained soils it has an effect on the formation of shear planes, thus controlling the geometry of failure mechanisms. Due to this fact, dilatant behavior of coarse-grained soils draws much attention from the academia (Taylor 1948, Rowe 1962, De Josselin de Jong 1976, Bolton 1986, Schanz and Vermeer 1996, Chakraborty and Salgado 2010). Even in the face of this ever-continuing scientific interest in dilatancy, a practical function that renders the quantification of dilatant behavior is yet to emerge. There are milestone works towards understanding dilatant behavior as listed in the references; however the proposed functions are either impractical or conceptual. For example, one of the well-known functions for calculating dilatancy () was proposed by Bolton (1986):

(1)

dv and d1 in Eq. (1) corresponds to the increments of volumetric strain and major principal strain, respectively. ID is the relative density ranging from 0 to 1 and pf is the corresponding mean effective stress at failure. Q and R are empirical fitting parameters whose units are dependent on the unit used for pf. Accordingly, IR is defined as the relative density index which yields pf dependent magnitude of ID. Later

Schanz and Vermeer (1996), relying on experimental results, improved Eq. 1:

(2) Recently Chakraborty and Salgado (2010) studied the values

of the fitting parameters Q and R, especially for low confinement conditions. However, it is clear that the variables of Eq. 1 and Eq. 2 are defined for the moment of soil failure and this approach significantly reduces the practicality of the proposed equations. Hence, the goal of this study is to calculate dilatancy angle using parameters that correspond to the in-situ state of the soil. Previous studies have shown that dilatant behavior is affected by the confinement and compactness of the soil. Accordingly, confinement is defined by confining pressure (p) and compactness is defined by the relative density of the soil (ID), as is the case in Eq. 1 and Eq. 2. In addition to the confinement and relative density, Vaid and Sasitharan (1991) showed that stress path affects the dilatant behavior. That is why, in this research the most ubiquitous stress path in nature is chosen for sample preparation which is the Ko consolidation. Even though stress path followed during sample preparation stage is confined to Ko consolidation, the influence of stress history is investigated by considering the overconsolidation ratio (OCR) as a third variable. Since all these can be achieved during a triaxial test, the tests conducted were Ko-consolidated and drained triaxial tests (CKoD). In order to achieve different OCRs, the samples were unloaded under Ko conditions.

In the remainder of this paper, the results of the tests conducted are presented followed by the construction of the dilatancy equations. Following, the proposed equations are validated using data collected from the literature.

896


2 EXPERIMENTAL STUDY

The experimental approach in this study is to conduct sufficient number CKoD tests at different pi-ID combinations so that it would be possible to define the pi- relationship for every 5% change in ID. This is achieved for an ID range within 0.35 to 0.95 by conducting 80 CKoD tests. It is important to emphasize that pi and ID are the in-situ (before shearing) mean effective stress and relative density values. The sand used in these tests is local sand called Silivri Sand. In order to have the same grain size distribution in all tests, this sand was sieved and prepared with the standard grain size distribution of Ottawa sand (Table 1). Table 1. Properties of the test sand.

Sand Gs Cu Cc emax emin Silivri sand with Ottawa

distribution (SP) 2.67 2.16 1.45 0.96 0.56

Samples were prepared by dry pluviation. Several tests were

conducted at different OCRs (1,2,4,8) to consider the influence of unloading on dilation. Overconsolidated samples were unloaded under Ko conditions.

3 TEST RESULTS

4 DILATANCY AS A FUNCTION OF PI AND ID

Dilatancy angle is calculated from the test results using the relationship proposed by Schanz and Vermeer (1996).

(3) The relationship given in Eq. 3 is preferred since it is

specifically developed for triaxial testing conditions. As the goal is to investigate the uncoupled effects pi and ID on , test results are divided into several ID ranges. In other words, pi- relationships are defined separately for each 0.05 increment in ID (i.e. a single pi- relationship is defined for the tests with 0.65≤ID<0.70, and this pi- relationship is considered to be applicable for ID=0.675). The reason for choosing the ID increment to be 0.05 is because this much variation in ID is within the measurement margin of error. So for each ID range, the tangents of calculated values (tan) are plotted against the corresponding end of consolidation (in-situ) mean effective stresses that are normalized with the atmospheric pressure (pi/pa). The tan-(pi/pa) relationships obtained for three different ID ranges are shown in Figure 1 as examples. For all tan-(pi/pa), the relationships that yield the greatest coefficient of determination (R2) are used.

As it can be observed in Figure 1, tan-(pi/pa) can be considered to be approximately a linear relationship. Therefore, it is defined using line equation as follows:

(4) Here in Eq. 4, and are unitless fitting parameters. The

variations of and with ID are plotted in Figure 2. It can be seen that the value of is approximately constant and varies linearly with ID. However, in order to propose functions that would be applicable to all soils, and are defined as linear functions:

(5)

(6)

Figure 1. Tan-(pi/pa) relationships for two different ID ranges for Silivri sand.

Constants a, b, m, and n are fitting parameters. As a result, a general equation can be written with the form given below.

(7) However, for the soil tested, their values are given in Figure

2. According to Figure 2, a=0, b=-0.06, m=0.353, and n=0. Hence, the dilatancy equation for Silivri Sand with Ottawa distribution can be written as

(8) When the test results are analyzed considering the influence

of OCR, it is noticed that unloading has no effect on the dilatant behavior. Thus, OCR does not affect the proposed equations.

Figure 2. -ID and -ID relationships for Silivri sand.

5 INFLUENCE OF DILATANCY ON PEAK FRICTION ANGLE

Peak friction angle () is a function of critical state friction angle (crit) and dilatancy which can be defined as in Eq. 9.

(9) Here, the parameter r defines the proportion of dilatancy

contribution to the frictional strength of the material. Up until now, researchers defined parameter r as a soil dependent constant. However, in this study the influences of ID and p on parameter r are also investigated. The same method of

897


uncoupling the influences of ID and p is also used here. Accordingly, for each 0.05 increment of ID, corresponding r-p relationships are obtained. The results for 0.70≤ID<0.75 and 0.90≤ID<0.95 ranges are given in Figure 3 as examples.

Figure 3. r-(pi/pa) relationships for two different ID ranges of Silivri Sand with Ottawa grading.

As it can be observed in Figure 3, obtained r-(pi/pa)

relationships are approximately linear. Therefore, the relationships are defined using a line equation.

(10) Similar to Eq. 4, r and r are line-fitting parameters.

Variations of r and r with ID are given in Figure 4.

Figure 4. r-ID and r-ID relationships for Silivri sand.

imately linear. Th

The r-ID and r-ID relationships are approxerefore, they are defined as

(11)

Parameters ar, br, mr, and nr are line-fitting parameters. Combining Eq. 10, Eq. 11, and Eq. 12, the overall function for calculating r is obtained.

(13) For the Silivri sand with Ottawa grading, the parameters of

Eq. 13 are as follows: ar=-1.2, br=1.12, mr=1.03, and nr=-0.34. These values are obtained from Figure 4.

6 EVALUTION OF THE PROPOSED FUNCTIONS

The proposed equations (Eq. 8 and Eq. 13) were developed by investigating the results of the tests conducted on Silivri sand with Ottawa grading. Therefore, it is necessary to evaluate the proposed equations against data sets of different soils. However, it is very difficult to find a complete data set that provides sufficient number of pi-ID-- combinations. Fortunately, Vaid and Sasitharan (1992) conducted a broad triaxial testing program on Erksak sand. Erksak sand has Cu=1.8, emax=0.775, and emin=0.525. Evidently, it is more uniform than Silivri sand with Ottawa grading.

The goal of their research was to identify the effects of stress path and loading direction on the strength and dilatancy of sands. Accordingly the researchers conducted tests with 10 different stress paths. One of the stress paths is the same as the tests of this program; Conventional drained triaxial compression test on consolidated sand. However the samples were isotropically consolidated. But, the data set of this test provided an invaluable source against which to evaluate the proposed set of equations.

Vaid and Sasitharan (1992) conducted their tests at three different relative densities and under several different confining pressures. All relevant tests, except the tests with pi>2000kPa, are used for the evaluation of the equations. The reason for discarding the results of the tests with pi>2000kPa is to prevent the possible grain-crushing mechanism from altering the results. As a result, again for each ID, it is observed that tan-(pi/pa) relationships are approximately linear (Figure 5). At this point,

Figure 5. Tan-(pi/pa) relationships for two different ID values o

it is interesting to note that tan-(p /p ) relationships were even more line e considered.

f

For Erksak sand, the variations of and of Eq. 4 are ob

i aar when the tests with pi>2000kPa wer

Erksak sand.

(12)

tained from Figure 6. Clearly the relationships have the same form as in the case of Silivri sand with Ottawa grading.

898


igure 6. -ID and -ID relationships for Erksak sand.

Evidently, when the parameters of Erksak sand (a=0, b=-0.0

F

12, m=0.687, n=0) are inserted into Eq. 7, the following function is obtained.

(14) When Equations 8 and 14 are compa

for

uence of dilatancy on the frictional behavior is also inv

sand.

f the g(pi,ID) fun tion shown in Eq. 13 can also be defined for

r

In this paper, dilatancy angle and its influence on friction angle are quantified for cohesionless soils. This is achieved by analyzing the results of an extensive triaxial testing program on

onless soils. The results are arranged in a

red, it can be seen that

Ko-consolidated cohesiway that allows the observation of the uncoupled effects of the influential parameters; ID and pi. Moreover, it has been shown that OCR does not affect dilatant behavior. Even though the general form of the =f(pi,ID) function is given in Eq. 7, the present data suggests a simpler version as shown in Eq. 8:

(r.8) The data from Silivri sand with Ottawa grading and Erksak

sand, both support the Eq. 8 form of =f(pi,ID) function. Here,

both soils the same form of tan-(pi/pa) relationship is obtained.

The inflestigated. For the r parameter, the obtained r-ID and r-ID

relationships are given in Figure 7.

Figure 7. r-ID and r-ID relationships for Erksak

As it can be observed from Figure 7, the same form ocr=

E ksak sand. Of course, the line-fitting parameters are clearly different but this can be attributed to the differences in the grain shape, size and distribution between the two sands. Erksak sand is obviously more uniform compared to Silivri sand with Ottawa grading. The difference between emax and emin is greater for Silivri sand than it is for Erksak sand. It might be proposed that the uniformities of sands control the influence of dilatant behavior on strength, but this proposition requires further testing on different sands with varying uniformities.

7 CONCLUSIONS

b and m are soil dependent unitless constants. For now, there is not sufficient data to correlate the vgr

) for providing financial support to this project under TUBITAK

trength and dilatancy of sands. Géotechnique 36

gado R. 2010. Dilatancy and Shear Strength of

Mukavemet ve Genleşim Özellikleri. 14th National

De on based on

Row

526.

Tay 948. Fundamentals of Soil Mechanics. John Wiley and

alues of b and m to ain shape, grain size distribution, and mineralogy. However, it

is believed that, as the corresponding constants for different soils are obtained, it would be possible to link b and m to mineralogy, grain shape, and grain size distribution characteristics. Similarly, the influence of dilatancy angle on the peak friction angle of the soil is defined. This influence is again a function of pi and ID. As a result, peak friction angle can be calculated by using Eq. 9 and Eq. 13.

In order to obtain the constants for Eq. 8 and Eq. 13, it is sufficient to conduct 12 triaxial tests on clean cohesionless sands. The most important advantage of the proposed equations is that the dilatancy and peak friction angles are calculated using directly measurable and/or calculable soil parameters. This attribute significantly increases the practicality of the dilatancy and peak friction angle calculations. Once the required parameters are defined for a specific soil, it will be possible to calculate the variations in dilatancy and friction angle just by tracking the changes in stress state and volumetric state.

8 ACKNOWLEDGEMENTS

Authors would like to acknowledge the Scientific and Technological Research Council of Turkey (TUBITAK

Project 110M595.

9 REFERENCES

Abadkon A. 2012. Strength and Dilatancy of Anisotropic Cohesionless Soils. Ph.D. Thesis. Bogazici University, Istanbul, Turkey.

Bolton M. D. 1986. S(1), 65-78.

Chakraborty T. and SalSand at Low Confining Pressures. ASCE Journal of Geotechnical and Geoenvironmental Engineering 136 (3), 527-532.

Cinicioglu O. and Abadkon A. 2012. Anizotropik Kohezyonsuz Zeminlerin Congress on Soil Mech & Foundation Eng, October 4-5, Isparta, Turkey (in Turkish).

Josselin de Jong G. 1976. Rowe’s stress-dilatancy relatifriction. Géotechnique 26 (3), 527-534. e P. W. 1962. The stress-dilatancy relation for static equilibrium of an assembly of particles in contact. Proc. R. Soc. A, 269 (1339), 500-527.

Vaid Y. P. and Sasitharan S. 1992. The Strength and Dilatancy of Sand. Canadian Geotechnical Journal 29 (3), 522-

Schanz T. and Vermeer P. A. 1996. Angle of friction and dilatancy of sand. Géotechnique 46 (1), 145-151. lor D. W. 1Sons, New York.

899

Centrifuge Modeling of Seismic Soil-Structure-Interaction and Lateral Earth Pressures for Large Near-Surface Underground Structures

Modélisation en centrifugeuse de l'Interaction sol-structure sismique et des pressions de terre latérales pour les grands ouvrages souterrains proches de la surface

Dashti S., Hushmand A., Ghayoomi M., McCartney J.S., Zhang M. University of Colorado Boulder

Hushmand B., Mokarram N., Bastani A. Hushmand Associates, Inc.

Davis C., Lee Y., Hu J. Los Angeles Department of Water and Power

ABSTRACT: The Los Angeles Department of Water and Power (LADWP) is planning the construction of a new buried reservoir inSouthern California. The current state of practice for evaluating the seismic response of underground structures relies heavily on simplified procedures or numerical tools that have not been verified adequately against physical model studies or detailed casehistories. A series of eight centrifuge tests are currently being conducted at the University of Colorado Boulder (CU Boulder) toproduce well-documented model “case histories.” The data from these tests help better understand seismic soil-structure-interaction(SSI) and the distribution of lateral seismic earth pressures on the walls of a buried structure restrained at top and bottom. This paper provides a brief overview of a centrifuge physical modeling investigation into the influence of the relative stiffness of the underground structure and the characteristics of the input motion on the seismic response of buried structures.

RÉSUMÉ: Le "Los Angeles Department of Water and Power" (LADWP) prévoit la construction d'un nouveau réservoir enterré, unecentrale hydroélectrique, et une station de régulation de débit en Californie du Sud. L'état actuel de la pratique d'évaluation de la réponse sismique des structures souterraines repose en grande partie sur les procédures simplifiées ou des outils numériques qui n'ontpas été comparés de manière adéquate à des des modèles physiques ou des études de cas. Une série de huit essais en centrifugeusesont actuellement en cours à l'Université du Colorado, Boulder pour produire des données complètes de prototypes. Les données deces essais aident à mieux comprendre l'interaction sol-structure sismique et la distribution des pressions des terres latérales sismiques sur les murs d'une structure enterrée maintenue en haut et en bas. Ce document donne un aperçu d'une modélisation physique encentrifugeuse de l'influence de la rigidité relative de la structure souterraine et des caractéristiques du signal source sur la réponse sismique de la structure enterrée.

KEYWORDS: Physical modeling ; Centrifuge modeling ; Seismic soil structure interaction; Underground structures.

1 INTRODUCTION

In order to better understand the seismic response of buriedwater reservoirs, a series of centrifuge tests are being performed on scale-model underground structures in a new, transparent flexible shear beam (FSB) type container developed by Ghayoomi et al. (2012a,b). The data from these tests serve two important purposes: 1) to better understand seismic soil-structure-interaction (SSI) and the distribution of lateral seismic earth pressures on the walls of a buried structure restrained at top and bottom; 2) to calibrate and improve numerical models. Specifically, the goal of the tests is to provide validation data for two-dimensional (2-D) and 3-D finite element analyses of the dynamic response of equivalent model underground structures with a range of stiffnesses.

In addition to describing the testing program on buried structures, we briefly discuss the results from a preliminary centrifuge test performed on a free-field soil specimen with no structure. The goal was to initially investigate the dynamic response of uniform dry sand and simultaneously evaluate the performance of the newly designed container in simulating 1-D conditions with minimum boundary effects. The next experiments, which are currently underway, evaluate the seismic response of three different tunnel structures with varying stiffnesses and soil conditions. Accelerations, displacements, and axial strains as well as the distribution of lateral earth pressures on the restrained walls are being measured during a suite of input earthquake and sinusoidal motions in flight. The influence of the relative stiffness of the underground structure to soil and the characteristics of the input

motion (i.e., amplitude, frequency content, and duration) on the seismic response of the buried structures are being studied. The insight gained from this investigation is aimed at improving the design and safety of the Los Angeles reservoirs and similar buried water storage structures in seismically active areas.

2 RESEARCH PROGRAM

2.1 Background

In order to comply with new water quality regulations in California, the Los Angeles Department of Water and Power (LADWP) is planning to cover or bypass each of its open reservoirs and replace them with buried reinforced concrete reservoirs. The proposed buried Headworks Reservoir includes 35 to 40-foot high walls that will be buried and restrained against rotational movement at the bottom and top by the reservoir floor and roof. The current state of practice for evaluating the seismic response of underground structures relies heavily on simplified procedures or numerical tools that have not been verified adequately against physical model studies or case histories, leading to significant uncertainties. Hence, a series of dynamic centrifuge tests were planned to evaluate seismic lateral earth pressures on a range of reduced scale underground structures.

2.2 Experimental Plan

A series of eight centrifuge experiments were planned to investigate the seismic response of relatively stiff buried

900


structures restrained at the top and bottom in medium-dense, dry Sand, at a spin acceleration of 60g. The testing plan for the first phase of the investigation is summarized in Table 1.

Table 1. Centrifuge Testing Plan (First Phase)

Test# Structure Model Soil

Type

SoilRelativeDensity

(Dr)

Soil Cover on Tunnel

1 None2

Simple Equivalent Prototype (SEP) 1.5 m

3 None

4

SEP (model fixed to the container

base) 1.5 m 5 None6 Stiff SEP 1.5 m 7 None8 Flexible SEP

NevadaSand 60%

1.5 m

2.2.1 Model Container A transparent FSB-type container (Fig. 1) was developed by Ghayoomi et al. 2012a,b) to enable better visualization of the response of underground structures . The container consists of a stack of transparent, rigid frames separated by soft rubber, bonded together with high strength epoxy. The combination of rigid frames and soft rubber provides a soft and flexible lateral deflection response during 1D horizontal shaking (Divis et al. 1996). In this manner, the container does not contribute additional stiffness to the soil layer.

To characterize the response of the empty container, it was placed on a dynamic shaking table mounted on the centrifuge platform and spun up to 60g of gravitational acceleration. Next, a series of sine-sweep motions were applied to the base of the container in flight. The frequency response function of the container was calculated using the power spectral ratios of the accelerations measured using accelerometers mounted horizontally on each of the frames. The fundamental natural frequency of the empty container was 40 Hz at a centrifugal acceleration of 60g (Figure 2), confirming the soft response.

Figure 1. Picture of the transparent FSB container at CU Boulder

101 1020

1

2

3

4

Frequency (Hz)

Tran

sfer

Fun

ctio

n

Frame 4Frame 3Frame 2

Figure 2. Frequency response of the FSB container (different acrylic frames with respect to the base) at 60g in model-scale.

2.2.2 Characterizing Soil Properties Nevada Sand was chosen for use in the testing program, as it is a well-characterized, uniform, fine, angular sand. A relative

density (Dr) of 60% was selected for testing as it is expected to show a slight dilative response during shaking which may provide a worst-case scenario for seismic lateral earth pressures. Table 2 summarizes the properties of Nevada Sand (Ghayoomi et al. 2012a,b).

Table 2. Properties of Nevada Sand (Ghayoomi et al. 2012b) Specific Gravity 2.65 (assumed)

Maximum Dry Unit Weight 16.39 kN/m3

Minimum Void Ratio 0.586

Minimum Dry Unit Weight 14.00 kN/m3

Maximum Void Ratio 0.852

2.2.3 Selection and Calibration of Ground Motions A suite of earthquake ground motions was selected for design based on the expected seismic hazard at the project site. The selected records included scaled versions of the following motions: 1) the Izmit Earthquake recorded at the Istanbul station (far field); 2) the Northridge Earthquake recorded at the Sylmar station (near field); 3) the Loma Prieta Earthquake recorded at the LGPC station (near field). In addition to earthquake records, sine-sweeps (with amplitude = 0.3g) were selected at frequencies ranging from 0.5 Hz to 7 Hz in the prototype scale. The goal was to evaluate the response of the soil-structure system under a range of motions with different characteristics.

The “desired” ground motions were converted into “target” ground motions that are safe to use in the centrifuge by filtering out frequencies that are beyond the capability of the shake table and are potentially damaging to the centrifuge (e.g., Mason et al. 2010). In this case, frequencies less than 0.1 Hz and greater than 15 Hz were filtered using an eighth-order band-pass Butterworth filter. The target motions were converted to model scale units for both time and acceleration values (e.g., accelerations multiplied by 60 and time values divided by 60), to covert the “target” motion to the “command” signal.

The “achieved” motion by the shake table may not be the same as the “command” motion because of the nonlinear response of the overall system. The shake table tends to damp out the higher frequency signals and amplify the lower frequencies. A frequency-domain transfer function was applied to the “command” signal iteratively in order to better match the “achieved” motion with the “target”. Particular attention was given to the Arias-Intensity time history of the “target” motion, roughly quantifying the energy of the ground motion as well as the 5%-damped spectral accelerations. Figure 3 compares an example of “achieved” and “target” base motions during the Northridge event with a scaled prototype PGA of 0.3g.

(a) (b)

(c) (d) Figure 3. Comparison of the “achieved” and “target” motions during the

Northridge event (scaled PGA = 0.3g) in prototype scale.

2.2.4 Design of Equivalent Model Underground Structures Three simple equivalent model underground structures were designed and constructed (e.g., Figure 4), to simulate prototype

901


structures with a range of expected dynamic properties (e.g., mass and stiffness). The first mode frequency of each structural model was measured in a 1-g shaking table test as shown in Figure 5a. The frequency values were in good agreement with the numerical estimates obtained using SAP and Abaqus. The quality of the weld between the walls of models was observed to be a key parameter in obtaining a good match between numerical and experimental values of the resonant frequencies.

(a) (b)

(c)Figure 4. Dimensions of three model structures in model scale: (a) SEP

Structure; (b) Stiff SEP; (c) Flexible SEP.

(a) (b) Figure 5. Baseline structure: a) 1-g shake table testing; b) Tekscan

pressure sensor placement on the tunnel wall.

2.2.5 Instrumentation Challenges Horizontal LVDTs were mounted on a rack attached to the stationary centrifuge platform and the light-weight cores were attached to the container frames. Vertical LVDTs were attached to a rack mounted on the top of the container. Permanent racking displacement of the tunnel structures was assumed to be small due to the high stiffness of these models. Hence, accelerometers were judged to provide a reasonable means for estimating transient racking deformations for each underground structure. Visual monitoring of the structures through the transparent walls of the container provides another means for the verification of racking behavior during shaking.

Tactile pressure sensors from Tekscan, Inc. were used in this study to measure dynamic earth pressures. They are flexible, thin sheets capable of measuring normal stresses applied with a matrix of sensels. This flexible sensor permits measurement of 2–D stress distributions on a surface with minimum deflection. Previous commercially-available tactile sensors were not reliable in capturing the full amplitude content of dynamic signals under the high-frequency environment of the centrifuge. This is in part due to signal aliasing and the sensor’s own frequency response (filtering effect). The sensor model used in this study (9500) has a sampling rate up to 4,000 Hz, which is rapid enough to avoid signal aliasing. The frequency response of each sensors was then characterized in dynamic tests using a load frame (Dashti et al. 2012). The frequency

recover the original pressure time histories. The response and accuracy of these tactile pressure sensors are affected by the presence of shear (Palmer et al. 2009). Shear was minimized by incorporating a Teflon-Teflon interface between the sensor and soil (Figure 5b) as recommended by Palmer et al. (2009)

response of these sensors was used as a transfer function to

PRELIMINARY FREE-FIELD TEST

e) was prepared and

3.1 Test Setup and Instrumentation

ive density of 60% was

3

A free-field soil model (with no structurtested at 60g of spin acceleration, as the baseline experiment to investigate the dynamic response of dry Nevada Sand and the performance of the container when filled with sand.

A layer of Nevada sand with a relatprepared by dry pluviation in the FSB container. The sand was placed atop a 5 mm-thick layer of gravel, which is intended to provide a no-slip boundary at the base of the soil profile. The dimensions of the sand specimen were: 700 mm long, 305 mm wide, and 336 mm high in the model scale. The instrumentation layout within the sand layer, including LVDTs and accelerometers, is shown in Figure 6.

Figure 6: Instrumentation layout in preliminary free-field test

3.2 Test Results

n example array of acceleration recordings

4

4.1 Test Setup and Instrumentation

EP model structure was

4.2 Test Results

Table 3 summarizes the sequence and PGA’s of the achieved base motions during Test 1. Figure 9 compares the acceleration

(dimensions in prototype scale).

Figure 7 presents awithin the soil column and a comparison of Arias Intensity-time histories recorded by the accelerometers in the center of the soil profile and near the boundary of the container. The comparisons show little difference between the two arrays, indicating minimum boundary effects in this container. The recorded settlement time histories at two locations were also consistent. The settlement measurements indicated little densification during the application of sine-sweeps, and considerable densification during each broad-band earthquake motion. Hence, the change in soil relative densities after each event must be incorporated into the numerical models.

CENTRIFUGE TESTING OF SOIL-STRUCTURE SYSTEM

A preliminary test on a trial flexible Sperformed to evaluate the proposed model instrumentation and response. The model was instrumented with accelerometers, LVDTs, strain gauges, and pressure transducers as shown in Figure 8. Accelerometers were placed away from, adjacent to, and on the structure to evaluate soil-structure-interaction effects. LVDTs were used to measure settlements at key locations. Strain gauges were placed on both walls to measure moment distributions and to indirectly calculate dynamic earth pressures.

902


records at the same elevation on the structure and in the free-field during the Izmit event, showing an amplification of movement near on the roof of the structure.

Figure 10 presents the recorded settlements at various locations with respect to the structure, showing larger settlements in the free-field, which decreased towards the structure. This settlement pattern was expected due to the smaller weight of the tunnel compared to the adjacent soil. A larger settlement of the surrounding soil compared to the tunnel led to an overall decrease in permanent lateral earth pressures on the walls after each shaking event. These results are currently being studied in combination with strain distributions and direct pressure measurements for different underground structures and base motions for Test 1 and the subsequent tests.

(a) (b) Figure 7: Measured acceleration recordings in the free-field test

compared in the middle and near the container boundary.

Figure 8. Instrumentation layout in Test-1 (prototype scale)

Table 3. Achieved Motions in Test-1

No. Ground Motion Achieved

PGA1 Izmit - Istanbul 0.3 2 Northridge - Sylmar 0.3 3 Northridge - Sylmar 0.8 4 Northridge - Sylmar 1.1 5 Loma - LGPC 1.0

5 CONCLUSIO

ynamic centrifuge were co ted on scalse-crete water reservoirs currently

uthern California. The goal of testing was

N

Dm

experiments nducodel buried reinforced con

being designed in soto verify 2-D and 3-D numerical models of equivalent underground structures restrained at the top and bottom. The data from these experiments will help evaluate the effects of seismic soil-structure-interaction (SSI) on the distribution of accelerations and lateral earth pressures on underground structures with different stiffnesses, soil conditions, and input ground motion characteristics. This paper includes a discussion of the centrifuge testing plan for evaluating the seismic response of buried structures, including the container characterization,

development of scale model structures, instrumentation challenges and preliminary results.

0 50 100-0.5

0

0.5

AC

C7

max = -0.22

0 50 100-0.5

0

0.5

AC

C 2

max = -0.23

0 50 100-0.5

0

0.5

AC

C8

time (s)

max = -0.25

0 50 100-0.5

0

0.5

AC

C 3

max = -0.22

0 50 100-0.5

0

0.5

AC

C9

max = -0.340 50 100

-0.5

0

0.5

AC

C 4

max = -0.27

Time (s) Time (s) Figure 9. Acceleration time histories (in prototype “g”) recorded in the

free-field and on the structure during the Izmit event in Test-1.

0 10 20 30 40 50 60 70 80 90 100-10

-8

-6

-4

-2

2

0

Time (s)

Settl

emen

t (cm

)

LVDT 1LVDT2LVDT3LVDT4LVDT5

Figure 10. Settlement recorded at various locations with respect to the structure in Test-1 during the Izmit event.

6 ACKNOWLEDGEMENTS

The authors w epartment ofWater and Power for the financial support of this project and the

versity of Colorado Boulder.

lg,” 15th World Conf. on EQ Eng., Lisbon.

Divis, C.J., Kutter, B.L., Idriss, I.M., Goto, Y., and Matsuda, T.

G

Zurich, Switzerland.

ould like to thank the Los Angeles D

centrifuge facility staff at the Uni

7 REFERENCES

Dashti, S., Gillis, K., Ghayoomi, M., and Hashash, Y. (2012). “Sensing of Lateral Seismic Earth Pressures in GeotechnicaCentrifuge Modelin

(1996). “Uniformity of Specimen and Response of Liquefiable Sand Model in Large Centrifuge Shaker,” 6th Japan-US Workshop on Earthquake Resistant Design of LifelineFacilities and Countermeasures against Soil Liq. pp. 259-274. hayoomi, M., Dashti, S., McCartney, J.S. (2012a). “Performance of a Transparent, Flexible Shear Beam-Type Container in Dynamic Centrifuge Modeling of Geotechnical Systems,” J. of Soil Dyn. and EQ Eng. (under review).

Ghayoomi, M., Dashti, S., McCartney, J.S. (2012b). “Effect of Boundary Conditions on the Performance of a Transparent Flexible Shear Beam-Type Container,” 2nd Int. Conf. on Perf.-Based Design EQ Geotech. Eng., Taormina, Italy.

Palmer, M.C., O’Rourke, T.D., Olson, N.A., Abdoun, T., Ha, D., O’Rourke, M.J. (2009). “Tactile Pressure Sensors for Soil-Structure Interaction Assessment.” J. of Geotech. and Geotech. Eng., ASCE, 1638-1645

Mason, H.B., Bray, J.D., Kutter, B.L., Wilson, D.W., and Choy, B.Y. (2010). “Earthquake motion selection and calibration for use in a geotechnical centrifuge.” 7th Int. Conf. on Physical Modeling in Geotechnics.

903

Evaluation of Seismic Earth Pressure Reduction using EPS Geofoam

Evaluation de la réduction de la poussée sismique en utilisant du Polystyrène Expansé

Dave T.N., Dasaka S.M., Khan N. Indian Institute of Technology Bombay, Mumbai, India

Murali Krishna A. Indian Institute of Technology Guwahati, India

ABSTRACT: Retaining structures are designed to withstand lateral pressures due to backfill, surcharge load from adjacent structures and traffic and earthquake loads. The cost of these structures is directly proportional to the earth pressures they are subjected to. Several techniques have been tried in the literature to minimize the earth pressure exerted on retaining walls. Among them, use ofgeofoam as a compressible inclusion placed at the wall-backfill interface, is found to be a simple and effective solution, based on preliminary studies. However, behaviour of EPS geofoam and its influence on the earth pressure reduction under seismic loadingconditions are not well understood, and need to be investigated further. In the present study, small scale physical model tests were performed on an instrumented retaining wall subjected to 1-D shaking, to evaluate earth pressures on the wall and to assess effectiveness of EPS geofoam to reduce seismic earth pressures. Firstly, static surcharge loading was applied in order to evaluate magnitude and distribution of earth pressure. Further, under maintained surcharge, a seismic load in the form of a stepped sinusoidalwave from 0 to 0.7 g was applied in increments of 0.045 g, each increment being applied for 5 seconds at 3 Hz frequency. The experimental results indicate that the earth pressures under the influence of a seismic load show an increase of the order of 23%. Moreover, by using the geofoam as a seismic buffer, it was observed that the total seismic force on the retaining wall reduced byabout 23% with a corresponding reduction in maximum lateral thrust by 27%.

RÉSUMÉ : Les structures de soutènement sont conçues pour résister à des pressions latérales dues au remblai, à la surcharge de structures adjacentes, au traffic et aux charges sismiques. Plusieurs études ont été réalisées dans la littérature minimiser la pression desterres sur des murs de soutènement. Dans la présente étude, des expérimentations ont été exécutées sur un mur de soutènement instrumenté pour évaluer la pression des terres et l'efficacité du Polystyrène, sous sollicitation sismique générée par une table vibrante1D. Premièrement, une surcharge statique était appliquée afin d'évaluer la distribution de la pression des terres. Puis, sous la surcharge maintenue, une charge sismique sous forme de vague sinusoïdale de 0 à 0,7 g était appliquée par paliers de 0,045 g, chaqueaugmentation étant appliquée pendant 5 secondes à 3 hertz de fréquence. Les résultats expérimentaux indiquent que les pressions des terres, sous l'influence d'une charge sismique montrent une augmentation de l'ordre de 23%. De plus, avec le polystyrène commeamortisseur sismique, on a observé que la force sismique totale sur le mur de soutènement diminue d'environ 23% avec une réduction de la poussée latérale maximum de 27%.

KEYWORDS: seismic load, earth pressure reduction, geofoam, shake table

1 INTRODUCTION

Earth-retaining structures are integral part of many infrastructure projects, and underground urban construction to retain soil on one of its sides. Rigid retaining walls are commonly found in basements, bridge abutments, box culverts etc. and they cannot be entirely replaced by reinforced soil walls. Lateral pressure acting on rigid retaining walls due to backfill, surcharge load from adjacent structures and loads due to traffic and natural calamities like earthquake etc. decides their sectional dimensions. Intensive earthquake loading, which impose larger forces compared to that of static active or at-rest conditions. The geotechnical profession has been constantly working for a viable solution to reduce the earth pressures exerted on retaining walls, which would eventually reduce the construction cost of the wall, and post-construction maintenance cost. A technique of placing a compressible inclusion at the soil-wall interface has come into existence to minimize earth pressures on retaining walls. Previous research studies indicate that provision of a compressible inclusion behind a rigid non-yielding/limited yielding or yielding wall would contribute to the economical design of the wall by imparting controlled yielding in the backfill material. Deformations in a retained soil mass mobilize a greater portion of the available shear strength

of the material and decrease the unbalanced lateral forces acting on the retaining structure.

2 REVIEW OF LITERATURE

Among all the methods, provision of a compressible inclusion in the form of Expanded Polystyrene (EPS) geofoam at the wall-backfill interface proved successful because of ease in construction and predictable stress-strain characteristics of the inclusion. In the past, studies were conducted with materials such as glass-fiber insulation (Rehnman & Broms, 1972) and cardboard (Edgar et al., 1989) for similar applications. However, they were not successful, as their stress-strain behavior was unpredictable and uncontrollable. On the other hand, Expanded Polystyrene (EPS) geofoam is considered as a suitable material as it fulfills the required stress-strain behavior and has smaller stiffness than any other geofoam materials. Additionally, Horvath (1997) documented 30 years of proven durability of EPS geofoam in several geotechnical applications.

A field study on reduction in lateral earth pressure behind rigid wall by using compressible geo-inclusion has been reported by Partos and Kazaniwsky (1987). Using instrumented model studies, McGown et al. (1988) demonstrated significant reduction in lateral earth pressure even below active earth pressure, when soil was allowed to yield in a controlled manner.

904


Karpurapu and Bathurst (1992) used a non-linear finite element analysis to simulate the controlled yielding concept for static load and concluded that compressible inclusion with t=0.01h (t – thickness of compressible inclusion, h – height of the wall) would provide active stress conditions in the backfill, if the stiffness of the compressible inclusion is sufficiently small.

Experimental investigations of the concept of reduction of seismic load on the retaining wall in the presence of geofoam inclusion were performed by several researchers on reduced scale models tested on shaking table (Hazarika et al. 2002, Bathurst et al. 2006, Zarnani and Bathurst 2007). Hazarika et al. (2002) showed reduction in the peak lateral loads in the range of 30% to 60% compared to that on an identical structure but with no compressible inclusion. Zarnani and Bathurst (2007) noticed that the magnitude of dynamic lateral earth force was reduced with decreasing geofoam modulus. Horvath (2010) highlighted compressive stiffness as the single most important behavioural characteristic of any compressible inclusion influencing the reduction. Athanasopoulos–Zekkos et al. (2012) observed that EPS of 20 kg/m3 density and relative thickness (t/h) of 15% to 20% can reduce the seismic pressure by up to 20%, and the seismic displacement of the wall by up to 50%, depending on shaking intensity and height of wall.

The available literature highlighted that with the use of EPS geofoam, the earth pressures on the rigid retaining walls can even be reduced below the active earth pressures. However, behaviour of EPS geofoam and its influence on the earth pressure reduction under seismic loading conditions are not well understood, especially in the presence of realistic surcharge loads, and need to be investigated further. Hence, the present study is aimed at evaluation of earth pressure under combined surcharge and seismic loading and to assess effectiveness of EPS geofoam, through experimental investigations on small scale models tested on 1-D shaking table facility.

3 EXPERIMENTAL PROGRAM

The physical tests described in this paper were carried out on 1.2 m × 1.2 m shaking table located at the Indian Institute of Technology Bombay. The table has 10 kN payload capacity and is driven by a 100 kN capacity Schenk hydraulic actuator with ancillary controller and PC software. The table was driven in the horizontal direction only, as it is noted that the horizontal component of seismic induced dynamic earth loading is typically the most important loading for the application under investigation. The table can excite the rated payload at frequencies up to 50 Hz and ± 5g. The maximum displacement of the table is ±125 mm. The instrumented retaining wall models were built in a stiff strong box (1.2 m long 0.31 m wide and 0.7 m high) and bolted to the steel platform of the shaking table. Detailed diagram and pictorial view of experimental set up are illustrated in Figs. 1-2. The model retaining wall was placed at a distance of 0.10 m from one of the ends, allowing 1.1 m as backfill length behind retaining wall. A 15 mm thick stainless steel plate was used as a model retaining wall and was instrumented with 7 diaphragm type earth pressure cells, attached flush with the surface of the wall. The wall was restrained laterally using three universal load cells rigidly connected to the other side of the retaining wall at 125, 325 and 555 mm elevations. One side of strong box was made-up of Plexiglas and other sides of stainless steel. The inside surface of the Plexiglas is covered by 120 mm wide and 60 µm thick greased polyethylene sheet with 10 mm overlap with each other. The combination of friction-reducing membrane and rigid lateral bracing was adopted to ensure that the test models were subjected to plane strain boundary conditions. A plywood sheet was bolted to the bottom of strong box, and a layer of sand was epoxied to the top surface of plywood to create a rough surface, so as to simulate backfill continuity in vertical direction.

A series of experiments were carried out without geofoam and with geofoam inclusion at wall-backfill interface. In all experiments, the sand was backfilled at 68% relative density using portable travelling pluviator (Dave and Dasaka, 2012) and top surface was manually leveled. The actual relative densities achieved in each test during the backfilling were monitored by collecting samples in small cups of known volume placed at different locations. Previous studies of the authors highlighted that EPS panel of density of 10D (10 kg/m3) and 75 mm thickness (t/H = 0.125) helps in maximum reduction in earth pressure by mobilization of its elastic compression. Hence, EPS panel of 10 kg/m3 density and dimensions of 700 mm x 300 mm and 75 mm thickness, prepared using hot-wire cutter, was pasted to retaining wall using ABRO tape to have proper contact of EPS panel with retaining wall during the test. Uniaxial compression tests were carried out on EPS samples at an axial strain rate of 10%/minute, and yield strength of the EPS geofoam was found as 29.3 kPa, as shown in Fig. 3.

Figure 1. Detailed diagram of experimental setup

Figure 2. Picotrial view of experimental setup

To apply uniformly distributed surcharge on the backfill, a rubber bellow was placed over an 8 mm thick rubber sheet laying on the surface of the backfill. Specially designed neoprene rubber bellow of 250 kPa capacity with non-return pneumatic valve was connected to a compressor to apply regulated pressure. A steel plate of 10 mm thickness with attachments to measure surface settlement was placed between rubber bellow and rubber sheet and a steel plate of 10 mm thickness was placed on the rubber bellow such that when inflated with compressed air, the plate moved upwards to mobilize reaction from frame, which was rigidly connected to the tank, thereby transferring pressure to the sand fill. Three LVDTs were used to measure vertical settlement at top of the backfill at 150 mm, 450 mm and 750 mm from retaining wall. The LVDTs were firmly mounted on the reaction frame with magnetic stand and were rested on angles welded on steel plate. Four accelerometers (PCB Piezotronics) were used to obtain acceleration-time excitation history. Out of these, three

905


were embedded in backfill at 100 mm, 300 mm and 500 mm from bottom and one accelerometer was mounted directly on the shaking table to record the input base acceleration–time excitation history, as shown in Fig. 1. The accelerometers were attached to mounting blocks before placing them at desired locations, to ensure that the devices remained level and moved in phase with the surrounding sand during shaking, as shown in Fig. 4.

Figure 3. Stress-strain behavior of 10D EPS geofoam

Figure 4. Positioning of Accelerometer in the backfill

The instruments were monitored by a separate high speed data acquisition system (MGC plus – HBM Inc. and Catman professional software). Data from a total of 17 instruments were recorded at a speed of about 100 Hz in order to prevent aliasing and to capture peak response values. After the model preparation was completed, surcharge pressure was applied in increments of 10 kPa up to 50 kPa and corresponding magnitude and distribution of earth pressure were monitored. Further, under maintained surcharge pressure, models were excited using a displacement–time history selected to match a target stepped-amplitude sinusoidal accelerogram with a frequency of 3Hz as shown in Fig. 5. The acceleration record was stepped in 0.045 g increments and each amplitude increment was held for 5 s. The maximum base acceleration was 0.7 g. The above frequency was adopted, as frequencies of 2–3 Hz are representative of typical predominant frequencies of medium to high frequency earthquakes (Bathurst and Hatami 1998) and fall within the expected earthquake parameters for North American seismic design (AASHTO, 2002). This simple

base excitation record is more aggressive than an equivalent true earthquake record with the same predominant frequency and amplitude (Bathurst and Hatami 1998, Matsuo et al. 1998). The models were only excited in the horizontal cross-plane direction to be consistent with the critical orientation typically assumed for seismic design of earth retaining walls (AASHTO 2002).

Figure 5. Stepped-amplitude sinusoidal excitation input

4 RESULTS AND DISCUSSION

Experimental evaluation of earth pressure under combined static surcharge and seismic acceleration was carried out for model tests without and with geofoam inclusion. In this paper, results of model tests with 10D geofoam are compared with experiments without geofoam. For the sake of brevity, earth pressure results corresponding to the maximum surcharge load of 50 kPa and seismic loading are only presented here. Under static surcharge load, observed earth pressure distribution was approximately triangular in shape as shown in Fig. 6. However, just above the base of wall, lower earth pressures were observed, this may be due to arching of backfill soil. Experimental evaluation of seismic earth pressure on retaining wall by application of seismic acceleration revealed reduction in the earth pressure in top 1/3 portion of wall, while increase for remaining wall height as shown in Fig. 6.

Figure 6. Earth pressure distribution for experiments without geofoam inclusion

During seismic loading, top portion of the wall might have moved sufficiently to achieve active condition, showing

906


reduction in pressure in top portion of the wall; whereas, rest of the wall might not have undergone sufficient displacement, and hence resisted the seismic loading, causing increase in the pressure. The increase in total lateral thrust was negligible for 0.18 g (about 2.36%), however, after 0.36 g, increase in earth pressure was observed throughout the wall height. The total lateral thrust increased with increase in seismic acceleration and the maximum increase in total lateral thrust was observed to be of 23% at 0.7 g. Maximum increase in lateral thrust of 49.5% was observed at about 0.35h from bottom; however reduction in lateral thrust near the top was observed. The observed reduction may be due to sufficient lateral movement of retaining wall, and subsequent mobilization of backfill strength and reduction in effect of surcharge load due to wall movement as shown in Fig. 6.

Earth pressure distribution with geofoam inclusion is presented in Fig. 7. The measured total thrust under 50 kPa surcharge pressure was 23.2% less than that on wall without geofoam inclusion. Reduction in total lateral thrust under surcharge loading is attributed to compression of geofoam and associated backfill strength mobilization which resulted in settlement of backfill. As during surcharge load application phase, compression of geofoam had reached its elastic limit, hence further reduction in earth pressure was negligible during seismic loading phase.

Figure 7. Earth pressure distribution for experiments with geofoam inclusion

Maximum reduction in total lateral thrust under combined loading was 26.9% corresponding to applied seismic acceleration of 0.36 g. At the seismic acceleration of 0.7 g, the reduction in maximum total lateral thrust was about 23%. Experiments with geofoam inclusion showed 54% increase in maximum lateral thrust under seismic loading, though it was 9.75% lower than the corresponding lateral thrust in the absence of geofoam inclusion. The maximum lateral thrust was reduced by 54% due to geofoam inclusion at location h/3 from base of wall. Though, provision of EPS geofoam at backfill-wall interface showed significant reduction in static and seismic loads, due to small scale model studies and associated boundary conditions, the reduction in magnitude of earth pressure was less than that noted from numerical study on a 6 m high wall carried out by the authors.

5 CONCLUSIONS

Following are the salient conclusions derived from the present studies: Increase in total lateral thrust was found negligible up to

0.18 g seismic acceleration. However, after 0.36 g,

increase in earth pressure and total lateral thrust were observed throughout the wall height.

Increase in total lateral thrust was observed to be around 23% at 0.7 g with maximum increase of 49.5% at 0.35h from bottom of the wall.

Provision of EPS geofoam as compressible inclusion at backfill-retaining wall interface reduced the earth pressure under static surcharge loading and combined surcharge and seismic loading by 23.2% and 23%, respectively.

Maximum reduction in total lateral thrust was found to be 26.9% at 0.36 g seismic acceleration.

6 ACKNOWLEDGEMENT

The work reported in this paper is substantially supported by the Department of Science and Technology (DST), India (Project No. SR/FTP/ETA-69/2008).

7 REFERENCES

Rehman, S. E. and Broms, B. B. 1972. Lateral pressures on basement walls: Results from full scale tests. In Proceedings of the fifth Europian Conference on Soil Mechanics and Foundation Engineering, 189-197.

Edgar, T. V., Pucket, J. A. and D’Spain, R. B. 1989. Effects of geotextiles on lateral pressure and deformation in highway embankments. Geotextiles and Geomembranes 8(4), 275-292.

Horvath, J. S. 1997. The compressible inclusion function of EPS geofoam. Geotextiles and Geomembranes 15(1-3), 77-120.

Partos, A. M. And Kazaniwski, P. M. (1987) Geoboard reduces lateral earth pressures. In Proceedings of Geosynthetics –’87. 628-639.

McGown, A., Andraws, K. Z. and Murry, R. T. 1988. Controlled yielding of the lateral boundaries of soil retaining structures. In Geosynthetics for soil improvement, ed – R. D. Holtz, 193-210, ASCE.

Karpurapu, R. and Bathurst, R. J. 1992. Numerical investigation of controlled yielding of soil retaining structures, Geotextiles and Geomembranes 11(2), 115-131.

Hazarika, H., Okuzono, S. and Matsuo, Y. 2002. Compressible geo-inclusion as seismic earth pressure reduction technique. In Proceedings of 13th International offshore and polar engineering conference, Honolulu – USA. 1244-1249.

Bathurst, R, Zarnani, S. and Gaskin, A. 2006. Shaking table testing of geofoam seismic buffers, Soil Dynamics and Earthquake Engineering 27, 324-332.

Zarnani, S. and Bathurst, R. J. 2007. Experimental investigation of EPS geofoam seismic buffers using shaking table tests, Geosynthetics International 14(3), 165-177.

Horvath, J. S. 2010. Lateral pressure reduction on earth-retaining structures using geofoams: Correcting some misunderstandings, In Proceedings of ER2010: Earth Retention Conference 3, ASCE.

Athanasopoulos–Zekkos, A., Lamote, K. and Athanasopoulos, G. A. 2012. Use of EPS geofoam compressible inclusions for reducing the earthquake effects on yielding earth retaining structures. Soil Dynamics and Earthquake Engineering 41, 59-71.

Dave, T. N. and Dasaka, S. M. 2012. Assessment of portable traveling pluviator to prepare reconstituted sand specimens. Geomechanics and Engineering 4(2), 79-90.

Bathurst, R. J. and Hatami, K. 1998. Seismic response analysis of a geosynthetic reinforced soil retaining wall. Geosynthetics International 5(1-2), 127-66.

Matsuo, O., Tsutsumi, T., Yokoyama, K. and Saito, Y. 1998. Shaking table tests and analysis of geosynthetic-reinforced soil retaining walls. Geosynthetics International 5(1-2), 97-126.

AASHTO 2002. Standard Specification for highway bridges. American Association of State Highway and Transportation Officials. Washington DC, USA.

907

Analysis of an adaptive foundation system for embankments on soft soils by means of physical and numerical modelling

Analyse d'un système de fondation adaptatif pour les remblais sur sols compressibles par modélisation physique et numérique

Detert O., Alexiew D. HUESKER Synthetic GmbH & Ruhr-Universität Bochum, Germany

Schanz T., König D. Ruhr-Universität Bochum, Germany

ABSTRACT: A new innovative foundation system for embankments on soft soil is currently being analyzed at the Ruhr-UniversitätBochum, Germany, in cooperation with the company HUESKER Synthetic GmbH. The system consists of two parallel vertical walls (e.g. sheet pile walls), which are installed into the soft subsoil and connected at their top via a horizontal tension membrane (e.g. geotextile). The embankment is then constructed on top of this tension membrane. The aim of this research project is to demonstratethe applicability of the system and to develop an analytical calculation algorithm for serviceability and ultimate limit state of the system. To study the complex interactive system behaviour a strategy is followed using geotechnical centrifuge technology andnumerical modelling. Measurement data from the centrifuge wall tests will be used for the validation and calibration of the numerical model.

RÉSUMÉ : Un système de fondation pour les remblais sur sol mou, innovant est actuellement analysé à l’université de la Ruhr àBochum, Allemagne, en coopération avec l’entreprise HUESKER Synthetic GmbH. Le système consiste en deux palplanchesverticales et parallèles, ancrées dans le sol mou et reliées par une membrane horizontale résistante à la traction. Le remblais est ensuiteconstruit par-dessus cette membrane. Le but de ce projet de recherche est de démontrer la pertinence du système et de développer un algorithme analytique de calcul pour les états limites de service et ultimes. Pour étudier le comportement complexe du remblai, oncomparera un modèle en centrifugeuse et une modélisation numérique. Les résultats des mesures des essais seront utilisés pour lavalidation et la calibration du modèle numérique.

KEYWORDS: soft soils, embankment, foundation system, geotextile, self-regulating, centrifuge tests, numerical modelling

1 INTRODUCTION

The construction of embankments on soft soils (e.g. for transportation, as break-waters or stockpiles) is a challenge due to their low shear strength, low permeability, high compressibility and high water content. The surcharge by the embankment can not only result in a local or total loss of stability (failure, see Figures 1 and 2) but also in unacceptable settlements or horizontal deformations, which could endanger structures nearby. The subsoil below the middle of the embankment will be loaded under approximately oedometric conditions, whereas the subsoil below the embankment shoulders will experience compressive as well as shear stresses.

Figure 1. Slope stability failure of an embankment directly founded on soft soil and numerical illustrated shear failure zone (foto: HUESKER)

To overcome these issues different solutions such as a phased construction of the embankment with a basal reinforcement, prefabricated vertical drains and consolidation phases or the use of (geotextile encased) granular columns or rigid inclusions, e.g. prefabricated concrete piles, with a horizontal geotextile reinforcing layer, are available. Each system has its limitations however, which can be related to the thickness of the soft soil layer, height of the embankment, time and also economic, ecological or technical reasons.

Figure 2. Soil extrusion below an embankment (sand filled geotextile tube) with basal reinforcement (sand mattress), system sketch and numerical illustrated shear failure zone (foto: HUESKER)

A new self-regulating foundation system for the construction of embankments on soft soils is presented in this

908


paper, which is believed to be a more feasible and economical solution under certain boundary conditions. The new concept and its application areas are presented in the following sections.

2. NEW SELF-REGULATING FOUNDATION SYSTEM

2.1. Description and basic ideas

The new self-regulating foundation system consists of two vertical parallel walls (e.g. sheet pile walls) which are installed at a certain distance between each other into the soft soil and connected to each other by a horizontal tension membrane (e.g. geotextile). The tension membrane is assumed to cover the whole area in between the vertical walls. The vertical walls may end within the soft soil layer or reach further down into a firm layer. The soft soil beneath the embankment is therefore confined by the membrane on top and the vertical elements (Figure 3).

The embankment will be constructed above the tension membrane, which is connected to the walls. This surcharge generates vertical and horizontal pressures and corresponding strains in the soft soil. The horizontal thrust tries to move the walls outwards. At the same time tension forces are mobilized in the tension membrane: first due to settlements (deflection) beneath the embankment and second due to the outward movements of the vertical walls tensioning the connected membrane additionally.

Figure 3. Sketch of deformed tension membrane foundation system

The basic ideas of the system are on the one hand to confine the soft soil by the vertical and horizontal elements to prevent excessive lateral deformation or even extrusion of the soft soil. This confinement results also in reduced vertical deformation. On the other hand a self-regulating mechanism of the system takes place. Each load increment provokes an increased horizontal pressure on the vertical walls and therefore a further outward deformation. This deformation results in a larger strain of the tension membrane and a corresponding higher tensile force. Thus the later provides an increased resistance to the outward displacement tendency of the walls. Say, the system reacts to a higher surcharge with a higher lateral restrain. The foundation system not only ensures the global stability of the embankment but also “automatically” prevents or reduces deformations.

2.2. Overview on related systems

The use of a geotextile basal reinforcement is a well established and documented method for the construction of embankments on soft soils. Many authors have reported about research and cases studies, as e.g. Rowe and Li (2005). This will be the most economic solution, if there are no restrictions regarding the settlements, the horizontal “spreading”, the time for consolidation etc. Wager and Holtz (1976) used in the 1960’s very short sheet pile walls connected via tie-rods to capture spreading forces of

embankment on soft soils. The tie-rods and sheet pile walls acted like a basal reinforcement mechanism and were just placed at the base of the embankment, not being embedded into the soft ground. It is reported that several projects applied this method. This solution was not followed further when geotextile reinforcements became readily available, mainly for cost

sed deformation. Design approaches have not been me

ion or failure while ear

sign approaches or system dep

ribed foundation system for embankments on soft soils.

3. RESEARCH STRATEGY

orithm for serviceabilityand ultimate limit state of the system.

3.2

reasons.Harata et al. (2008) reported about the use of sheet pile

walls at the toe of embankments on soft soils to cut off the settlement depression. Due to the installation of the sheet pile walls into the ground a stress discontinuity between the embankment and the surrounding ground is generated, which leads to a reduction of the vertical deformation outside the embankment. In the design concept of Harata et al. only the equilibrium of the vertical forces is considered. Ochiai et al. (1991) studied in small scale laboratory tests different arrangements of two parallel sheet pile walls at the toes of the embankment, where the wall length and inclination were varied. Additionally in two of the tests the influence of a connection via tie-rods between the walls has been investigated. As a result of the tests the authors rated the different arrangements in respect of the deformation outside the embankment. The use of tie-rods led to decrea

ntioned.Adalier et al. (2003), Elgamal et al. (2002) and Tanaka et al.

(2000) reported about the use of tie-rod connected sheet pile walls beneath embankments on loose, saturated sandy foundation soils to prevent earthquake-induced liquefaction. Adalier et al. (2003), analyzed the behaviour with centrifuge tests and Elgamal et al. (2002) performed numerical simulations based on these results. Tanaka et al. (2000) performed shaking table tests and numerical simulations. All researchers confirmed the benefit of tie-rod connected sheet pile walls beneath the embankment with respect to deformat

thquake-induced liquefaction occurs. In both applications only single tie-rods are used, thus the

embankment weight has to be carried only by the subsoil. A restraining tensile force as with the membrane foundation system is not generated by the embankment weight. Long time consolidation processes are not relevant in the case of the liquefaction issue and of little relevance where a stress discontinuity is of interest. De

endencies are not addressed. Cofferdams do have a similar set-up but they are mainly

constructed above the existing ground level. The infill is a well draining granular material, which provides the stability of the system. Cofferdams are mostly loaded horizontally from one side, so the construction sequence as well as the interaction between the structural elements and soil are completely different to above desc

3.1. Aim of the research

The aim of this research project is to demonstrate the applicability of the system, the self-regulation mechanism and to develop an analytical calculation alg

. Theoretical system behaviour

The stress and strain of the different system components, vertical walls, tension membrane and soft soil, are strongly influenced by their interaction. Due to consolidation processes in the soft soil the interactions are time dependent. The stiffness of the soil as well as the total stress on the walls are changing with the consolidation from undrained conditions at the beginning of the embankment construction to drained conditions in the final state. The system behaviour depends on

909


many factors e.g. the distance between the vertical walls, their length and degree of fixation, the weight of the embankment, the thickness, stiffness and shear strength of the soft soil, the stiffness of the vertical walls and tension membrane and the rel

is as mentioned before time and deformation de

earth pressure is acting, which rep

ry be

ign of the vertical walls shall be done for

ent origins of tensile

educe but at the same time the

brane and therefore to tensile forces within the me

and reaches its

to squeeze out but is hindered by the walls and the

n the membrane and may lead to some increased deformation.

3.3.

bration of the numerical model measurement dat

ität in Bochum, Germany (Jessberger and Güttler, 1988).

sile stiffness of the horizontal e

determined by derivation of the bending

used for the validation and calibration of the numerical model.

ts can be

ation of the latter between each other. A key factor for the design of the system is the knowledge of

the earth pressure distribution along the sheet pile walls. This distribution

pendent. Due to the low permeability of mostly saturated soft soils the

total surcharge load from the embankment during and directly after construction is carried by the pore water pressure. At this moment and under assumed oedometric conditions the additional horizontal pressure equals the vertical pressure from the surcharge, which represents the upper limit regarding the horizontal loading on the vertical elements. With progressing consolidation the additional horizontal pressure decreases until the effective horizontal

resents a lower limit. Since the self-regulating system is not infinitly rigid and the

loading is not uniform but trapezoidal (Figure 3) the horizontal earth pressure will be in between these two limits at the ve

ginning and will decrease with time during consolidation. Due to the deformation of the walls the earth pressure

outside the walls will change from the at rest condition to passive earth pressure. Similarly inside the walls the earth pressure will change towards the active earth pressure. It has to be analyzed to which degree the passive and active earth pressure will be activated and how this is influenced by the length and bending stiffness of the vertical elements, the tensile stiffness of the membrane and the relation of the latter between each other. The above described behaviour leads to the conclusion that the maximum bending moment does occur immediately after the construction of the embankment when the total surcharge load from the embankment is carried by the pore water. This means the des

undrained conditions. The tensile forces in the tension membrane do also depend

on time and deformation. Three differforces have to be considered (Figure 4).

Figure 4: Acting forces in the tensile membrane (qualitative sketch) Connection forces: Due to the outward movements of the

vertical walls a tensile force is developed within the tension membrane.

Spreading forces: Due to the imbalance of the earth pressure in the region of the embankment shoulders spreading forces are generated. These spreading forces are mainly

captured by the geotextile. In case of the installation of the sheet pile walls with an excess length (hk> 0) above the ground level the spreading forces rconnection forces increase.

Membrane forces: The settlements of the soft soil due to the embankment weight leads to a geometrical elongation of the tension mem

mbrane. For the tension membrane the connection force activated by

the wall movements will be at its maximum during and immediately after construction of the embankment. The membrane force will increase during consolidation

maximum when consolidation comes to an end.Possibly a further influence on the tensile forces is the heave

which can occur in the region of the embankment shoulder when soil tends

membrane. The different mechanisms are all interacting and increase or

reduce the total tensile force in the membrane. Furthermore the creep of the tension membrane has to be taken into account, which will lead to reduced stresses i

System analysis

Due to the complex and time dependent interaction and the multitude of influencing parameters a comprehensive numerical parametric study is planned for the system analysis. For the validation and cali

a is required. A real scale field test would generate the most reasonable

data but boundary conditions are hard to control and consolidation takes a long time, which means the reproducibility of the tests would be very low. Small scale tests overcome these drawbacks, but they do not represent the realistic stress fields of the system. The centrifuge technique combines the advantage from field and small scale tests. Realistic stress fields can be generated, boundary conditions are well controlled and consolidation takes considerably less time due to the shorter drainage path. Due to these reasons the system is analyzed in the beam centrifuge Z1 at the Ruhr-Univers

3.3.1. Centrifuge tests By means of centrifuge tests the earth pressure distribution under varying relations between the bending stiffness of the vertical walls and the tenm mbrane will be analyzed. Therefore the vertical model walls are instrumented with strain gauges and measurements are taken frequently during the staged construction of the embankment and consolidation phases. The measurement data can be transferred into bending moments by conversion factors. The total earth pressure distribution can then bemoment distribution. A detailed description of the centrifuge test set-up and execution can be found in Detert et al. (2012). The results of the centrifuge tests are analyzed and

3.3.2. Numerical parameter study Numerical methods are a powerful tool in analyzing complex mechanism with varying parameters. The right choice of the soil model is very important for the numerical simulation. The soil model has to be capable of reproducing the significant soil mechanical processes occurring in the system as well as the load history of the construction steps and the centrifuge test procedure. The data obtained from the centrifuge tesused to confirm the choice of the right soil model.

910


From the centrifuge test a resultant earth pressure (sum of passive and active pressure) distribution is obtained. By means of numerical simulations it is possible to distinguish between the different time dependent earth pressure components out- and inside the system, as well as the pressure from the embankment

nsitivity of

. Based on the results of the numerical investigations an

eveloped.

aterial is required to reach

al elements

geotextile and the vertical

t

performance of geogrid anchors for sheet pile walls or similar. e

to the

r endanchored back into the embankment. Also more flexible connections are possible if large settlements are expected.

iour is required. It was concluded

tor is the connection of the tension membrane to the vertical walls, which

lized in different applications.

Har

Och

Ada

Tan

Wa

. In ’88,

tions of Geogrids as tie-back Anchors for Vertical Walls, 4th European

Geosynthetics Conference, Edinburgh, Scotland, UK

weight. It is also possible to identify and observe the different origins of the tensile forces in the membrane. With the numerical parametric study the sedifferent parameters and their impact on the stress and strain of the different system components can be analyzed

analytical design approach will be d

4. CONSTRUCTION ASPECTS

For the system construction well established techniques such as sheet pile wall installation can be used and the appropriate machinery is generally available worldwide. Site preparation for the equipment is reduced to two lateral “construction roads” for the installation of only the vertical wall elements, compared to full width working platforms which are required in the case of conventional soil improvement techniques such as e.g. vibro stone columns. The soft soil becomes part of the system and no soil disposal is necessary. Less fill mthe final height of the embankment due to the reduced settlement and lateral deformation. Depending on the soft soil conditions and/or the lifetime of the embankment it is possible to reclaim the verticwith little effort. It is also possible to have a partly open wall system by installing shorter sheet piles in-between. From the practical point of few a key element of the system will be the connection of the elements. Large forces are to be transferred between the tension membrane and the vertical walls. The use of geogrids as an anchor element has already been applied several times as shown in Detert et al. (2008). Differencase studies demonstrate the applicability and the good

A possible connection detail is shown in figure 5. Thconnection consists here of u-shaped steel rings welded onsheet pile wall and a steel pipe pushed through these rings.

Figure 5. Connection detail between sheet pile wall and geogrid anchor

The geogrid is wrapped around this steel pipe and the uppe

5. CONCLUSION

The paper describes the theoretical behaviour of a new foundation system for embankments on soft soils. The system

consists of two parallel vertical walls, which are installed into the soft subsoil and connected via a horizontal tension membrane. The embankment is constructed on top of this membrane. It was found that the understanding of the earth pressure distribution along the walls, which is time and deformation dependent, is a key factor for the design. Due to the complex interaction between the system components a strategy was developed to analyze the system. Numerical simulations offer a very efficient method to perform comprehensive parametric studies for analyzing the impact on the system behaviour of the different system components. For the validation and calibration of the numerical model measurement data from the system behavthat centrifuge tests are the most beneficial technique for gaining this required data. From a practical point of view it was found that the installation of the system is not difficult. A key fac

has been previously rea

6. REFERENCES

ata N, Otani J., Ochiai H., Onda K.. and Okuda Y. 2008. Countermeasures against settlement of embankment on soft ground with PFS (Partial Floatin Sheet-Pile) method. Geotechnics of soft soils: Focus on Ground Improvement: Proceedings of the Second International Workshop on Geotechnics of Soft Soils, Glasgow, Scotland, 3-5 September 2008 iai H., Hayashi S., Umezaki T. and Otani J. 1991. Model test on sheet-pile countermeasures for clay foundation under embankment, Developments in Geotechnical Aspects of Embankments, Excavations and Buried Structureslier K., Pamuk A. and Zimmie T.F. 2003. Seismic Rehabilitation of Coastal Dikes by Sheet-Pile Enclosures, The Thirteenth International Offshore and Polar Engineering Conference, Honolulu, Hawai, USA, May 25-30, 2003 aka H., Murata H., Kita H. and Okamoto M. 2000. Study of sheet pile wall method as a remediation against liquefaction, The Twelfth World Conference on Earthquake Engineering, Auckland, New Zealand, 2000

amal A., Parra E., Yang Z. and Adalier K. 2002. NElg umerical Analysis of Embankment Foundation Liquefaction Countermeasures, Jounral of Earthquake Engineering, Vol. 6. No. 4 (2002)

Rowe, R.K. and Li, A.L. 2005. Geosynthetic-reinforced embankments over soft foundations, No. 12, 2005, Geosynthetics International

ger, O. and Holtz, R.D. 1976. Reinforcing Embankments by short Sheet Piles and Tie Rods. New Horizons in Construction Materials, International Symposium, Lehigh University, Bethlehem, Pennsylvania, November 1-3, 1976

ert O., König, D. and Schanz T. 2012. CentrifugeDet modeling of an adaptive foundation system for embankments on soft soils,Proceedings of Eurofuge 2012, Deflt, Netherlands berger,H. and Güttler, UJess . 1988. Bochum geotechnical centrifugeJ. Cort (Ed.), Proc. Int. Conf. Geotech. Cent. Mod. - CentrifugeParis , 37 -44

Detert O, Wehrli E and Cejka, A. 2008. Innovative Applica

911

Reliability analysis of empirical predictive models for earthquake-induced sliding displacements of slopes

Analyse de fiabilité des modèles empiriques de prédiction des déplacements sismiques de pentes

Fotopoulou S., Pitilakis K. Department of Civil Engineering, Aristotle University, Thessaloniki, Greece

ABSTRACT: The goal of this study is twofold: (i) to identify the influence of the earthquake characteristics on the magnitude of theresidual co-seismic slope displacements of a typical slope using different predictive analytical models and (ii) to compare the results of the analytical models with an exact fully dynamic non-linear analysis. In particular, three analytical models were used to predict thepermanent slope displacements: the classical Newmark rigid block model, the decoupled Rathje and Antonakos model and thecoupled Bray and Travasarou sliding block model. In addition, 2 dimensional fully non-linear numerical analyses were performed using the code FLAC for idealized sand and clayey step-like slopes considering different real acceleration time histories as inputmotion. All three models predict displacements that are generally in good agreement with the numerical results for the sand slopecase. On the contrary, for the clay more flexible slope the correlation is not so good. However it is shown that the some crucialparameters, like the frequency content of the input motion, are not always appropriately captured in all analytical models.

RÉSUMÉ : L'objectif de cette étude est (i) d'identifier l'influence des caractéristiques du tremblement de terre sur l'ampleur desdéplacements co-sismiques résiduels d’une pente, en utilisant différents modèles analytiques et (ii) de comparer les déplacmentsanalytiques avec une analyse numérique plus élaborée. En particulier, trois modèles différents étaient utilisés pour estimer lesdéplacements permanents : le modèle de base de bloc rigide de Newmark, le modèle découplé de Rathje et Antonakos et le modèlecouplé de Bray et Travasarou. L’analyse numérique a été effectuée sur la même pente avec le code FLAC et pour les mêmesmatériaux de sol (sable et argile). Dans le cas de pente sableuse les déplacements calculés par les trois modèles analytiques sontgénéralement en relativement bon accord avec les résultats numériques. La comparaison est moins bonne pour la pente argileuse.Néanmoins il a été démontré que tous les modèles analytiques ne tiennent pas en compte proprement quelques paramètres importantscomme la fréquence du mouvement fort des sols.

KEYWORDS: co-seismic slope displacements, Newmark-type displacement models, non-linear dynamic numerical analysis.

1 INTRODUCTION

It is common practice in geotechnical earthquake engineering to assess the expected seismic performance of slopes and earth structures by estimating the potential for seismically induced permanent displacements using one of the available displacement-based analytical procedures. Considering that (total and/or differential) displacements ultimately govern the serviceability level of a slope after an earthquake, the use of such approaches is strongly recommended. Typically, two different approaches of increased complexity are proposed to assess permanent ground displacements in case of seismically triggered slides: Newmark-type displacement methods and advanced stress- strain dynamic methods.

The sliding-block analog proposed by Newmark (1965) still provides the conceptual basis on which all other displacement-based methods have been developed aiming to yield more accurate estimates of slope displacement. This has been accomplished by proposing more efficient ground motion intensity measures (e.g. Saygili and Rathje, 2008), improving the modeling of dynamic resistance of the slope characterized by its yield coefficient (e.g. Bray, 2007) and by analyzing the dynamic slope response more rigorously (e.g. Bray and Travasarou, 2007; Rathje and Antonakos, 2011). In terms of their assumptions to analyze the dynamic slope response, displacement based methods can be classified into three main types: rigid block, decoupled and coupled. A short description of the different types of Newmark-type displacement methods

as well as recommendations for the selection of the most appropriate ones is given in Jibson (2011).

Advanced stress-deformation analyses based on continuum (finite element, FE, finite difference, FDM) or discontinuum formulations usually incorporating complicated constitutive models, are becoming recently more and more attractive, as they can provide approximate solutions to problems which otherwise cannot be solved by conventional methods e.g. the complex geometry including topographic and basin effects, material anisotropy and non-linear behavior under seismic loading, in situ stresses, pore water pressure built-up, progressive failure of slopes due to strain localization. Several investigators have implemented continuum FE or FD codes to evaluate the residual ground displacements of slopes using elastoplastic constitutive models (e.g. Chugh and Stark, 2006; Lenti and Martino, 2012 etc.).

In this paper we study the accuracy of three different Newmark-type based models i.e. the conventional analytical Newmark (1965) rigid block approach, the Rathje and Antonakos (2011) decoupled model and Bray and Travasarou (2007) coupled model, classically used to estimate the expected co-seismic slope displacements, with a more refined numerical approach, considering different earthquake input motions scaled to different PGA values and compliance of the sliding surface. For the purpose of this comparative study we selected a typical configuration of a 30o inclined sand and clayey slope.

912


2 IMPLEMENTATION OF NEWMARK-TYPE PREDICTIVE MODELS

Τhe Newmark conventional analytical rigid block method is used to predict cumulative slope displacements obtained by integrating twice with respect to time the parts of an earthquake acceleration-time history that exceed the critical or yield acceleration, ac (ky·g) (e.g. threshold acceleration required to overcome shear soil resistance and initiate sliding). The second approach is a two-parameter vector (PGA, PGV) model proposed by Rathje and Antonakos (2011) applied herein to evaluate co-seismic slope displacements. This model is recommended for use in practice due to its ability to significantly reduce the variability in the displacement prediction. For flexible sliding, kmax (e.g. peak value of the average acceleration time history within the sliding mass) is used in lieu of PGA and k–velmax (e.g. peak value of the k-vel time history provided by numerical integration of the k-time history) is used to replace PGV. The third one is the Bray and Travasarou (2007) model. In this model cumulative displacements are calculated using the nonlinear fully coupled stick-slip deformable sliding block model proposed by Rathje and Bray (2000) to capture the dynamic response of the sliding mass. They use a single intensity parameter to characterize the equivalent seismic loading on the sliding mass, i.e. the ground motion’s spectral acceleration Sa at a degraded period equal to 1.5Ts, which was found to be the optimal one in terms of efficiency and sufficiency (Bray 2007).

The first goal is to study the influence of the earthquake characteristics and the dynamic response of the slope on the magnitude of the residual slope displacements using the aforementioned three predictive models. In this respect, permanent displacements as a function of the critical acceleration ratio (e.g. ky/kmax or ky/PGA) are computed using the three approaches considering different earthquake input motions and compliance of the sliding surface. Comparisons between the models allowed evaluating their reliability. Mean displacements were calculated using the Newmark rigid block model, as reference, whereas median values ±1 standard deviation and median and 16th - 84th percentiles were derived for the decoupled and coupled approximations respectively. Table 1. Parameters describing the characteristics of the ground motions nd the dynamic response of the sliding mass a

Earthquake record name

Valnerina1979- Cascia_L

Northridge 1994- Pacoima

Dam_LEarthquake code Cascia PacoimaMoment magnitude (Mw) 5.9 6.7

PGA (g) 0.15 0.41 Fundamental period Tp (sec) 0.23 0.48

Mean Period Tm (sec) 0.295 0.507 Scaled PGA (g) 0.3 0.7 0.3 0.7 PGV (cm/sec) 10.3 30.9 14.6 43.9 Natural period of the sliding mass Ts (sec) 0.16 0.032 0.16 0.032

Sa(1.5Ts)/PGAscaled 2.93 1.07 2.26 1.03 Ts/Tm 0.54 0.11 0.32 0.06 The seismic input consists of two real acceleration time

histories recorded at rock outcropping conditions and scaled at two levels of PGA, i.e. 0.3 and 0.7g. Table 1 presents the parameters describing some basic characteristics of the ground motions and the flexibility of the potential sliding surface. The displacements were computed for nearly rigid (Ts=0.032sec) and relatively flexible (Ts=0.16 sec) sliding masses. The derived (mean or median) permanent displacements for the three different predictive models and for the different considered

earthquake scenarios plotted as a function of the critical acceleration ratio, ky/kmax or ky/PGA, are illustrated in Figures 2a, 2b and 2c when considering the nearly rigid sliding surface. Moreover in Figures 3a and 3b we compared between them the three analytical models for the Pacoima 0.7g input motion for the nearly rigid and the relatively flexible sliding mass respectively.

(a)

(b)

(c)

Figure 2. Newmark (a), Rathje and Antonakos (b) and Bray and Travasarou (c) displacement versus ky/kmax considering a nearly rigid sliding mass for different acceleration time histories (cascia, pacoima) scaled at different levels of PGA (0.3g, 0.7g)

The results prove the important role of the amplitude and

frequency content of the earthquake as well as the compliance of the sliding surface on the magnitude of the computed displacements. As it should be expected, time histories scaled at 0.7g produce larger displacements compared to those scaled at 0.3g for the same critical acceleration ratios. For the Newmark and Rathje and Antonakos models the lower frequency input motion (Pacoima- fp=2.1Hz) generally yields larger displacements in relation to the higher frequency input motion (Cascia- fp=4.4Hz). For the Newmark model (see Fig. 2a) this trend becomes more pronounced with the increase of the critical

913


acceleration ratio, whereas in Rathje and Antonakos (see Fig. 2b) this trend does not seem to be influenced by the critical acceleration ratio. Contrary to the previous models it seems that the importance of the frequency content is not taken into account in the Bray and Travasarou coupled model, which predicts slightly larger displacements for the higher frequency input motion (see Fig. 2c). The latter model generally predicts larger displacements compared to Newmark rigid block and Rathje and Antonakos decoupled models. In particular, the difference in the displacement prediction is by far more noticeable for the flexible (Fig. 3b) compared to the nearly rigid (Fig. 3a) sliding mass. Displacements computed using Rathje and Antonakos predictive equations are closer to the Newmark rigid block model. The comparison is even better for the higher frequency input motion and for the lower level of shaking.

The

Figure 3. Comparison of the different Newmark-type models when considering a nearly rigid (a) and a relatively flexible (b) sliding mass for a certain earthquake scenario (Pacoima scaled at 0.7g)

3 COMPARISON WITH THE DYNAMIC NUMERICAL ANALYSIS

The second goal is to compare the Newmark-type analytical models with an a-priori more accurate numerical model. For this purpose a two- dimensional fully non-linear FLAC (Itasca, 2008) model has been used. The computed permanent horizontal displacements within the sliding mass for the two idealized step-like slopes, characterized by different flexibility of the potential sliding surface, are compared with the three Newmark-type models. The geometry of the finite slope is shown in Figure 4. The discretization allows for a maximum frequency of at least 10Hz to propagate through the grid without distortion. Free field absorbing boundaries are applied along the lateral boundaries

whereas quiet boundaries are applied along the bottom of the dynamic model to minimize the effect of artificially reflected waves. The soil materials are modeled using an elastoplastic constitutive model with the Mohr-Coulomb failure criterion, assuming a non-associated flow rule for shear failure. Two different soil types are selected for the surface deposits to represent relatively stiff frictional and cohesive materials. The mechanical properties for the soil materials and the elastic bedrock are presented in Table 2.

Figure 4. Slope configuration used for the numerical modeling

T able 2. Soil properties of the analyzed slopes

Relatively stiff soil Parameter

sand clayStiffsoil

Elasticbedrock

Dry density (kg/m3) 1800 1800 2000 2300

Poisson's ratio 0.3 0.3 0.3 0.3

Cohesion c (KPa) 0 10 50 -

Friction angle φ (degrees) 36 25.0 27 -

Shear wave Velocity Vs (m/sec) 250 250 500 850

T able 3. Selected outcropping records used for the dynamic analyses

Earthquake Record station Mw R(km) PGA(g) Valnerina, Italy 1979 Cascia 5.9 5.0 0.15

Parnitha, Athens 1999 Kypseli 6.0 10.0 0.12

Montenegro 1979 Hercegnovi Novi 6.9 60.0 0.26 Northridge, California 1994 Pacoima Dam 6.7 19.3 0.41

Campano Lucano, Italy 1980 Sturno 7.2 32.0 0.32

Duzce, Turkey 1999 Mudurno_000 7.2 33.8 0.12

Loma Prieta, California 1989 Gilroy1 6.9 28.6 0.44

The initial fundamental period of the sliding mass (Ts) is

estimated using the simplified expression: Ts = 4H/Vs, where H is the depth and Vs is the shear wave velocity of the potential sliding mass. The depth of the sliding surface is evaluated equal to 2m for the sandy slope and 10m for the clayey one by means of limit equilibrium pseudostatic analyses. The horizontal yield coefficient, ky, is computed via pseudostatic slope stability analysis equal to 0.16 and 0.15 for the 30o inclined sand and clayey slopes respectively.

The seismic input applied along the base of the dynamic model consists of a set of 7 real acceleration time histories recorded on rock outcrop (see Table 3) and scaled at PGA=0.7g. To derive the appropriate inputs for the Newmark-type methods that include the effect of soil conditions, and to allow a direct comparison with the numerical results, we computed the time histories at the depth of the sliding surfaces through a 1D non-linear site response analysis considering the same soil properties as in the 2D dynamic analysis. It is noticed that the 1D soil profile is located at the section that approximately corresponds

(a)

(b)

914


to the maximum slide mass thickness of the slope (Section A in Figure 4). The bottom of the sliding surface is taken be consistent to the estimated fundamental period of the sliding mass (Ts) that is different for the clay and sand slopes.

Table 4 presents the computed numerical horizontal displacements together with those calculated using the different Newmark-type displacement methods. The average difference (%) of the Newmark-type models in the median (or mean) displacement estimation compared to the numerical displacement is shown in Figure 5a for both sand and clay slopes. The dispersion of the corresponding differences is presented in Figure 5b. Table 4. Comparison between numerical, Newmark (1965), Rathje and Antonakos (2011) and Bray and Travasarou (2007) displacements for sand and clayey slope materials and for outcropping accelerograms caled at 0.7g s

Slope soil

material

Earthquake code

Computed horizontal

displacement (m)

Average Newmark

(m)

Rathje and Antonakos

Median (m)

Bray and Travasarou

Median (m)

cascia 0.6 0.64 0.40 0.60 kypseli 0.50 0.55 0.50 0.65

montenegro 0.90 0.70 0.37 0.42 pacoima 0.70 0.53 0.49 0.57

sturno 1.70 1.38 0.83 0.81 duzce 1.10 0.94 0.36 0.57

sand

gilroy 0.20 0.23 0.28 0.57 cascia 0.50 0.36 0.16 0.57 kypseli 0.45 0.28 0.14 0.53

montenegro 0.82 0.47 0.16 0.72 pacoima 0.62 0.35 0.19 0.79

sturno 1.40 0.90 0.25 0.71 duzce 0.85 0.48 0.16 1.16

clayey

gilroy 0.20 0.09 0.09 0.55

4 DISCUSSION- CONCLUSIONS

In general the Newmark-type analytical models predict comparable displacements, at least in the order of magnitude, with the exact numerical analysis. The comparison is generally better for the sand slope case, while for the clayey more flexible slope the divergences are amplified. In particular Bray and Travasarou model tend to predict generally larger displacements with respect to the numerical analysis, whereas Newmark and Rathje and Antonakos models underpredict the corresponding displacements.

Among the three methods, Bray and Travasarou model was found to present the minimum average predictive error (%) in relation to the numerical analysis for both sand and clay slope cases. This is in line with the inherent coupled stick-slip assumption adopted in the method that offers a conceptual improvement over the rigid block and decoupled approaches for modeling the physical mechanism of earthquake-induced landslide deformation. However, Bray and Travasarou model presents a very large dispersion in the median displacement estimation (up to 70% for both sandy and clayey slopes). Thus, the use of Sa(1.5 Ts) seems rather insufficient to fully describe the characteristics of the seismic loading (i.e. amplitude, frequency content and duration) for site-specific applications.

Newmark analytical approach shows the minimum dispersion in the displacement prediction (less than 10-20%) with respect the numerical analysis results compared to the Bray and Travasarou and Rathje and Antonakos models. This may be justified by the fact that Newmark analytical method uses the entire time history to characterize the seismic loading as opposed to the Bray and Travasarou and Rathje and Antonakos models that use one [Sa(1.5 Ts)] and two (PGA, PGV) intensity parameters respectively. As such, uncertainties associated to the selection of the ground motion intensity parameters are lower in the Newmark analytical approach.

Overall, the differences in the displacement prediction between the three models are larger for the clayey slope. Thus,

the compliance of the sliding surface in relation with the way that the frequency content of the input motion is taken or not into account may produce some important errors to the estimated earthquake-induced sliding displacements of slopes. It is suggested that a better framework is deemed necessary to account for the various uncertainties in the seismic displacements prediction.

(b)

(a)

Figure 5. (a) Average difference (%) and (b) dispersion of the predictive models in the median displacement estimation compared to the corresponding numerical displacement considering nearly rigid (sand slope) and flexible (clayey slope) sliding masses

5 REFERENCES

Bray J.D. 2007. Simplified seismic slope displacement procedures. Earthquake Ggeotechnical Eengineering. K. D. Pitilakis Editor, 2007 Springer.

Bray J.D. and Travasarou T. 2007. Simplified procedure for estimating earthquake-induced deviatoric slope displacements. Journal of Geotechnical and Geoenvironmental Engineering 133(4), 381-392.

Chugh A.K., Stark T.D. 2006. Permanent seismic deformation analysis of a landslide. Landslides 3(1), 2-12.

Itasca Consulting Group 2008. Inc. FLAC (Fast Lagrangian Analysis of Continua), ver. 6.0. Itasca Consulting Group, Inc., Minneapolis.

Jibson R.W. 2011. Methods for assessing the stability of slopes during earthquakes-A retrospective. Engineering Geology 122(1-2), 43-50.

Lenti L. and Martino S. 2012. The interaction of seismic waves with step-like slopes and its influence on landslide movements. Engineering Geology 126, 19–36.

Newmark N.M. 1965. Effects of earthquakes on dams and embankments. Geotechnique 15 (2), 139–159.

Rathje E.M. and Bray J.D. 2000. Nonlinear coupled seismic sliding analysis of earth structures. Journal of Geotechnical and Geoenvironmental Engineering 126(11), 1002–1014.

Rathje E.M. and Antonakos G. 2011. A unified model for predicting earthquake-induced sliding displacements of rigid and flexible slopes. Engineering Geology 122(1-2), 51-60.

Saygili G. and Rathje E.M. 2008. Empirical Predictive Models for Earthquake-Induced Sliding Displacements of Slopes. Journal of Geotechnical and Geoenvironmental Engineering 134(6), 790.

915

Development of pore water pressure around a stone column.

Développement des pressions interstitielles autour d’une colonne ballastée.

Gautray J., Laue J., Springman S.M. Institute for Geotechnical Engineering, ETH Zürich, Switzerland

Almeida M. Federal University of Rio de Janeiro, Rio de Janeiro, Brazil

ABSTRACT: The bearing capacity of model stone columns installed in soft soil is investigated in a series of centrifuge model teststhat are carefully instrumented to reveal the response of the ground during penetration of the tool and the cyclic compaction process during withdrawal. Pore pressures are measured at various distances from the column axes as well as at different depths, and the influence of the excess pore water pressure build up and dissipation around the column and the development of the load transfer mechanism are examined. The data are analysed and compared to theoretical solutions, both for the installation phase of the column as well as for subsequent loading with a stiff, circular foundation. These provide a set of high quality data for validating numerical methods. The measurements, and the associated analyses, will help to determine the transient load bearing capacity of stone columnsand the effects of accelerated pore pressure dissipation, which will contribute to improving the understanding and use of this mode of ground improvement.

RÉSUMÉ : Une recherche sur la capacité portante de modèles de colonnes ballastées installées dans un sol mou est réalisée dans unesérie d’essais en centrifugeuse instrumentés avec soin afin de mettre en exergue la réponse du sol durant la pénétration de l’outil ainsique le processus de compaction cyclique durant l’extraction. Les pressions interstitielles sont mesurées à différentes distances de l’axede la colonne ainsi qu’à différentes profondeurs et l’influence de la formation et de la dissipation des surpressions interstitielles autourde la colonne et le développement du mécanisme de transfert de charge sont examinés. Les données sont analysées et comparées avec des solutions théoriques pour la phase d’installation de la colonne et pour le chargement consécutif avec une fondation circulairerigide, fournissant une série de données de haute qualité pour la validation de méthodes numériques. Les mesures, ainsi que les analyses associées, aideront à déterminer la capacité portante de colonnes ballastées ainsi que les effets de la dissipation accélérée despressions interstitielles, ce qui contribuera à améliorer la compréhension et l’utilisation de ce mode d’amélioration des sols.

KEYWORDS: Ground improvement, stone columns, consolidation

1 INTRODUCTION

Stone columns have proven to be an efficient ground improvement technique. They increase the vertical stiffness and reduce the consolidation time, as radial drainage dominates the consolidation process (e.g. Hansbo, 1981).

This paper presents the results of a centrifuge test conducted in the ETH Zürich geotechnical drum centrifuge (Springman et al., 2001) at multiple earth’s gravity, n = 50. A stone column has been installed in a clay model (Weber, 2008) and is loaded with a circular footing. The pore pressures developing during installation and the loading phase were recorded and studied.

2 SOILS

2.1 Soft clay bed - Birmensdorf clay

Remoulded natural clay from the traffic interchange near to Birmensdorf was consolidated in a large oedometer and used as soft clay bed for the experiment. The main properties of this clay are summarised in Table 1.

2.2 Granular column – quartz sand

As tested by Weber (2008), quartz sand (fraction 0.5 – 1 mm) was used for constructing the sand columns (see Table 2).

Table 1: Properties of the reconstituted Birmensdorf clay (after Weber, 2008).

USCS classification CH Clay particle content from

sedimentation analysis < 2μm [%] 42

Liquid limit wl [%] 45-62 (av. 60) Plastic limit wp [%] 18-26 (av. 21)

Plasticity index Ip [%] 27-36 (av. 30) Critical state angle of friction ’cv [°] 24.5

Cohesion c’ [kN/m2] 0 Specific density ρs [g/cm3] 2.75 Medium grain size d50 [μm] 4

Water-saturated permeability k [m/s] for a void ratio of e = 1.10 [-] 1.5.10-9

2.3 Filling material - Perth sand

Perth sand was used in order to fill the gap between the clay model and the wall of the model container (see Fig. 1). Selected properties of this material can also be found in Table 2.

916


Table 2: Selected sand properties i) column (Weber, 2008) ii) Perth sand (Buchheister, 2009).

Origin Column Perth USCS classification SP SP Density ρs [g/cm3] 2.65 2.65

Critical state angle of friction ’cv [°] 37.0 30.5

Medium grain size d50 [mm] 0.75 0.23 Coefficient of uniformity [-] 1.4 1.79 Coefficient of gradation [-] 1.0 1.26

Grain shape semi-angular- slightly rounded -

3 SAMPLE PREPARATION

The clay was consolidated in a 250 mm diameter oedometer with incremental loading up to a total vertical stress of 200 kPa. The sample was removed from the oedometer container and the pore pressure transducers (PPTs) were installed. Their locations are shown in Fig. 1. The sample was then put into the centrifuge strongbox (diameter 400 mm) and the gap of 75 mm between container wall and clay model was filled with Perth sand by dry pluviation without compaction (Fig. 1), resulting in an axisymmetric sample. In this test, the boundary conditions cannot be considered to be oedometric with no radial strain, as the sand/clay interface is not rigid.

Fig. 1: (a) Plan view and (b) cross-section of the model set-up.

4 T-BAR TEST

A T-bar test, the location of which can be seen in Fig. 1, was conducted in order to determine the profile of the undrained shear strength in the soft clay. The T-bar (Fig. 2) has a length of 28 mm and a diameter of 7 mm. It was driven at a rate of 0.5 mm/s up to a depth of 140 mm, where a waiting time of one

minute was observed before the tool was pulled back out of the model.

The undrained shear strength was calculated with the following equation:

[ub

F ]s kPaL B N

(5)

where F is the force recorded for the T-bar penetration, L the length of the T-bar, B the width of the T-bar and Nb the T-bar factor, set in this case at 10.5 [-] (Stewart et al., 1994).

Fig. 2: T-bar (Weber, 2008).

(a)

Fig. 3: Profile of the undrained shear strength obtained with the T-Bar.

5 INSTALLATION OF THE STONE COLUMN

The test procedure and the corresponding results are presented at model scale with the exception of the forces, which are scaled to prototype scale. According to the scaling relationships, stresses are scaled by the factor 1, whereas time scaling from model to prototype differs by a factor n2 for diffusion processes and by n during inertia processes. Forces are scaled by a factor n2 (e.g. Schofield, 1980), n being the factor by which earth’s gravity is increased.

(b)

Standpipe

P4P7

Fig. 4: Evolution of the pore water pressures during the in-flight consolidation.

P1 – P3P5P6

P2

917


5.1 Consolidation

The first step in the centrifuge model test is to reconsolidate the clay model in-flight at an acceleration of 50 g, due to the increase in self-weight. Dissipation of the excess pore water pressures took approximately 13 h (Fig. 4).

5.2 Stone column installation

The stone column installation tool developed by Weber (Fig. 5; Weber, 2008) has been used to construct the stone column. It consists of a steel tube with an outer diameter of 10 mm and an inner diameter of 8 mm. A drawing pin was used to prevent the tube from blocking during first penetration.

Fig. 5: Stone column installation tool (Weber, 2008).

The column tool was inserted up to a depth of 120 mm in the centre of the clay model and the column was built with a 15/10 compaction regime (i.e. once the desired installation depth was reached, the tool was extracted by 15 mm before being inserted again 10 mm, compacting the sand in the column). This compaction process increases the stone column diameter to 12 mm (see Fig. 1), at least in the softest clay layer near the surface. The insertion was displacement-controlled (2 mm/s) and the driving force as well as the pore water pressures were recorded during this phase (Fig. 6). The development of pore water pressure over time is given in the top part of Fig. 6. The middle part of Fig. 6 shows the scaled driving force required to penetrate the installation tool and the bottom part shows the position of the tip of the installation tool under the surface. It can be seen that the strongest reaction of the PPTs is observed when the tip of the column tool reaches the depth of the sensors (marked by horizontal dashed lines in the Fig. 6c). This is observed both for the penetration phase and the compaction phase, respectively.

90% of the excess pore water pressures are dissipated after about 2300 s, which, when scaled by 502, corresponds to a prototype time of about 67 days. This is significantly shorter than the time needed for dissipation of excess pore pressures at the beginning of the test (see Fig. 4). Indeed, for a drainage path of 4 m (half of the prototype height of the model) and a one dimensional stiffness modulus ME = 1780 kPa, a coefficient of consolidation 7 22.67 10 /vc m

2 / 589 d c

s

s

is obtained, leading to a consolidation time at 90% excess pore water pressure dissipation of 90 90v vt T . This reduces the time by a factor of 8.8, which is consistent with a combined drainage condition, i.e., vertical plus radially outwards (to the surrounding Perth sand) and inwards (towards the stone column) resulting from the insertion of the granular column.

day

6 FOOTING LOADING

As the third step in the test, the newly built stone column was loaded with a 56 mm-diameter stiff aluminium footing, after the excess pore water pressures caused by the installation of the column had dissipated. The loading was displacement-controlled (v = 0.02 mm/s) and a maximum settlement of 17 mm at model scale was attained before the footing was

removed and the loading-induced excess pore water pressures were left to dissipate. The first jump in excess pore water pressures (Fig. 7 top) between 0 and 1000 s is due to a technical problem, which triggered an unexpected loading of the stone column. The actual loading can thus be studied after 1000 s.

The sensors P1 (top layer close to the column) and the sensors P4, P5 and P6 in a depth of 96 mm below ground surface react in parallel to the loading, albeit with different magnitudes of pore pressure change, while P2 and P3 exhibit a less sharp response. This confirms that the column takes a larger part of the load than the soft clay and the pressure distribution with depth is not building up as it would in a homogeneous medium. Secondly, an explanation was sought for the increase (P1 to P6) or drop (P7) in pore water pressures that can be seen at about 1500 s, which might have coincided with failure of the column. A bulge could be identified in the upper third of the column as the model was being dismantled. The cause might be attributed to the development of this ‘local compression failure zone’ in the sand column.

(a)

P7

P4-P6

P5P1-P2-P3

(b)

(c)

Fig. 6: (a) Pore water pressures, (b) driving force (c) depth of the tip of the installation tool during the sand column installation.

The total load applied on the footing can be formulated as (Adam, 2011):

v sc sc clay clay sc clayA A A A (1)

with the corresponding load on the stone column as (Adam, 2011):

' '1 sin 2 1 sinsc sc clay clay scc (2)

918


where A corresponds to the area and the indices “v” and “sc” to vertical and stone column, respectively. The load on the stone column when the jump was observed, for a global loading of 65 kPa, is then derived to be 234.2 sc kPa .

Bergado et al. (1994) suggest the bulging failure load (Muir Wood et al., 2000) on a stone column is calculated from:

' 'max , , 0 .2 1 sin 1 sinc p c u p c p c s sq z K s K q K (3)

where γc is the unit weight of clay, z is the depth, Kp,c is the clay coefficient of passive earth failure, su is the undrained shear strength, q0 is the overburden pressure and ’s is the angle of friction of the column material. A failure load of

can be obtained. max 269.6 q kPaHughes & Withers (1974) propose a different equation to

calculate the bulging failure load: '

max 4 1 sin 1 sinc u sq z s 's

kPa

(4) where the nomenclature used is the same as in the formula of Bergado et al. (1994). In this case, a failure load of

is obtained. max 209.8 q

Fig. 7: (a) Pore water pressures, (b) footing loading and (c) footing settlement during the footing loading.

It can be seen that the two theoretical solutions proposed bound the value calculated from the data. As a consequence, it may be concluded that local bulging failure in the stone columns triggered an additional radial loading of the soft clay and caused a small load decrement below the column (marked

by the pore water pressure drop observed at P7, which subsequently consolidated out exponentially).

The local bulging failure described by Muir Wood et al. (2000) replicates the case of an axially loaded cylindrical specimen in a triaxial apparatus, in which shear discontinuities are formed as sections of the cylindrical specimen spall outwards from the central axis (Desrues et al., 1996).

7 SUMMARY

Data from a physical model test in a geotechnical centrifuge is provided in this contribution to validate numerical processes to simulate the installation effects of stone columns and their subsequent use as ground improvement under a footing. Pore pressure transducers have been installed in the vicinity and under the stone column in this axisymmetric test. This measurement provides valuable insight in the behaviour of the stone columns. These measurements enable the identification of the bulging failure load of the column, which lay between boundaries given by two analytical solutions.

8 ACKNOWLEDGEMENTS

The fourth author was supported by a grant from the Brazilian Research Council CNPq and funds from the ETH Rectorate and the Department of Civil, Environmental and Geomatic Engineering, during a two month stay (January-February 2012) at ETH Zurich. This support is gratefully acknowledged.

(a)P7

P4

9 REFERENCES

Adam. D. 2011. Bodenverbesserung versus Hybridgründung und Tieffundierung – Vergleich der Gründungskonzepte von drei Projekten mit Tragweite für Europa. Institut für Geotechnik, ETH Zürich, Kolloquium, 17.11.2011, www.igt.ethz.ch.

Bergado, D.T., Chai, J.C., Alfaro, M.C. and Balasubramaniam, A.S. 1994. Improvement Techniques of Soft Ground in Subsiding and Lowland Environment. Balkema, Rotterdam.

Buchheister, J. 2009. Verflüssigungspotential von reinem und siltigem Sand unter multiaxialer Belastung. Institut für Geotechnik, ETH Zürich, Diss. Nr. 18312, VDF-Verlag, ETH Zürich

Desrues, J., Chambon, R., Mokni, M. and Mazerolle, F. 1996. Void ratio evolution inside shear bands in triaxial sand specimens studied by computed tomography. Géotechnique 46 (3), 529-546.

Hansbo, S. 1981. Consolidation of fine-grained soils by prefabricated drains. X ICSMFE, Stockholm, Sweden (3), 677-682.

Hughes, J.M.O. and Withers, N.J. 1974. Reinforcing of soft cohesive soils with stone columns. Ground Engineering 7 (3), 42-49.

Muir Wood, D., Hu, W. and Nash, D.F.T. 2000. Group effects in stone column foundations model tests. Géotechnique 50 (6), 689-698.

Schofield, A.N. 1980. Cambridge geotechnical centrifuge operations: 20th Rankine lecture. Géotechnique. 30 (2), 129-170.

Springman, S., Laue, J., Boyle, R., White, J. and Zweidler, A. 2001. The ETH Zurich Geotechnical Drum Centrifuge. International Journal of Physical Modelling in Geotechnics 1 (1), 59-70.

Stewart, D.P. and Randolph, M.F. 1991. A new site investigation tool for the centrifuge. Centrifuge ’91, H.Y. Ko and F.G. McLean (eds). Balkema, 531-537.

Weber, T.M. 2008. Modellierung der Baugrundverbesserung mit Schottersäulen. Institut für Geotechnik, ETH Zürich, Diss. Nr. 17321, VDF-Verlag, ETH Zürich.

P3P5P6

P2 P1

(b)

(c)

919

Large scale 1-g shake table model test on the response of a stiff pile group to liquefaction induced lateral spreading

Réponse d'un groupe de 3 × 3 pieux rigides sous l'action d'un écoulement latéral induit par liquéfaction étudié à grande échelle sur table vibrante

Haeri S.M., Kavand A., Asefzadeh A. Department of civil engineering, Sharif University of Technology, Tehran, Iran

Rahmani I. Marine transportation and technology division, Transportation research institute, Tehran, Iran

ABSTRACT: Evaluation of pile response to liquefaction induced lateral spreading is an important step towards resistant design of pilefoundations against this destructive phenomenon. This paper investigates the response of a stiff 3×3 pile group under liquefactioninduced lateral spreading using large scale 1-g shake table test. The model ground consisted of a 3-layer soil profile including a base non-liquefiable layer, a middle liquefiable layer and an upper non-liquefiable layer. Different parameters of the response of laterally spreading soil as well as those of the pile group including accelerations, pore water pressures, displacements and bending momentswere recorded during the shaking that are presented and discussed in the paper. In addition, distribution of lateral pressures due to lateral spreading on individual piles of the group is investigated in detail. The results show that lateral forces exerted by the laterallyspreading soil vary in the individual pile of the group both in transverse and longitudinal directions depending on the pile position within the group. It was also found that the magnitude of lateral pressures due to lateral spreading on the stiff 3×3 pile group of thisstudy are close to the values recommended by the design code.

RÉSUMÉ : L'évaluation de la réponse d'un pieu à l'écoulement latéral induit par liquéfaction est une étape importante vers laconception de fondations sur pieux contre ce phénomène destructeur. Cet article présent l’étude de la réponse d'un groupe de 3 × 3pieux rigides en vraie grandeur à 1 g sur table vibrante. Le modèle de sol est composé de 3 couches comprenant une couche centraleliquéfiable et des épontes non liquéfiables. Différents paramètres de la réponse du sol ainsi que ceux du groupe de pieux, y comprisles accélérations, les pressions interstitielles, les déplacements et les moments de flexion ont été enregistrés pendant l’essai et sontprésentés et discutés dans cet article. De plus, la distribution des pressions latérales de sol dues à l’écoulement sur tous les pieux du groupe est étudiée en détail. Les résultats montrent que les forces latérales exercées par le sol écoulé varient dans le pieu individuel dugroupe à la fois dans les directions transversale et longitudinale, selon la position du pieu à l'intérieur du groupe. Il a également été constaté que la valeur des pressions latérales dues au sol écoulé sur le groupe de pieux sont proches des valeurs recommandées par lecode de conception.

KEYWORDS: Pile group, liquefaction, lateral spreading, 1-g shake table test.

1 INTRODUCTION

Liquefaction-induced lateral spreading is commonly observed in gently sloping grounds or lands ending in free faces as a result of liquefaction in underlying saturated loose cohesionless deposits. In these deposits, earthquake-induced excess pore water pressures can cause a significant decrease in soil shear strength resulting in ground movement towards downslope or free face due to existing static shear forces (Kramer and Elgamal 2001). Lateral spreading can impose significant lateral pressures on pile foundations. During past earthquakes, several examples regarding severe damages to piles and structures supported on them due to lateral spreading have been documented, among which the cases in the 1964 Niigata (Hamada et al. 1986), the 1995 Kobe (Tokimatsu and Asaka 1998), and the 2010 Haiti earthquakes (Eberhard Marc et al. 2010) are the most important ones in this respect.

Although some experimental studies including shaking table, centrifuge and field tests (e.g. Haeri et al. 2012, Motamed and Towhata 2010, Abdoun et al. 2003, Ashford et al. 2006) have been conducted to evaluate the response of pile groups to lateral spreading, but different aspects of the soil-pile interaction in laterally spreading ground are not yet fully understood. For example, there are not still enough effective researches concerning variation of the value and pattern of the lateral pressures from the liquefied layer against different

individual piles of a group. Motamed and Towhata [10] recently showed that the lateral spreading force in an individual pile within a group varies depending on the pile position in the group. They conducted a series of 1-g shaking table tests on pile groups behind quay walls in a two-layer soil profile, including a non-liquefiable layer overlain by a top liquefiable layer, and showed that in a pile group rear-row piles which are closer to the quay wall sustain larger lateral pressures, while front-row piles sustain smaller values.

In this paper response of a stiff 3×3 pile group under liquefaction induced lateral spreading in a 3-layer soil profile including a base non-liquefiable layer, a middle liquefiable layer and an upper non-liquefiable layer is studied. For this purpose 1-g shake table physical modeling is utilized. Different parameters of the response of laterally spreading soil as well as those of the pile group such as accelerations, pore water pressures, bending moments and displacements were recorded during the test that are briefly discussed in the paper. The main focus of this paper is on distribution of lateral soil forces in individual piles of the group.

920


2 PHYSICAL MODEL

The experiment of this study was conducted using shaking table device of the earthquake research center at Sharif University of Technology (SUT).

In order to hold the physical model, a rigid box was used which had a length of 3.5 m, width of 1 m and height of 1.5 m. Figure 1 shows the schematic cross section and plan view of the physical model along with the layout of transducers. As seen, the soil profile consists of three distinct layers including a non-liquefiable crust with a thickness of 25 cm and relative density of about 60%, that is made of sand and clay (10% by weight of sand); a 1m thick middle liquefiable layer consisting of loose sand with relative density of about 15% and a lower non-liquefiable dense sand layer having 25 cm thickness and relative density of about 80%. All the soil layers have a slope of 7%. The sand used in physical model is standard Firoozkuh silica sand (No. 161) which has a uniform grain size distribution and is widely used in Iran for geotechnical physical modeling. Model piles of this study were initially designed as steel piles in prototype scale according to recommendations by JRA 2002 since representing a stiff pile comparing to concrete ones. Subsequently, mechanical and geometrical properties of the piles were calculated in model scale using similitude laws proposed by Iai et al. (2005). In this regard, the geometrical scale was selected as λ=8 (prototype/model). All model piles were made of aluminum pipes. Material properties of the model piles are summarized in Table 1. As sketched in Figure 1, various types of transducers were employed in different parts of the model including accelerometers and pore pressure transducers in the free field (far from the piles) to measure soil accelerations and excess pore water pressures; pore pressure transducers close to the piles to monitor build-up and dissipation of the excess pore pressures in the near field (close to the piles); displacement transducers (LVDTs) attached to the pile cap and also in free field to record pile and soil lateral displacements and finally strain gauges pasted along the piles to record bending moments. Base excitation was applied parallel to the model slope. The excitation was a sinusoidal acceleration record having amplitude of 0.3g and frequency of 3 Hz whose duration was 12 sec consisting of two rising and falling parts, each of duration of about 1.0 sec at the beginning and end of shaking.

T able 1. Mechanical and geometrical properties of pile foundations.

Material Height (m)

Outer/inner diameter (cm) I (cm4) EI

(kN.m2)

Aluminum 1.25 5.2/4.7 5.904 4.054

3 SUMMARY OF EXPERIMENTAL RESULTS

In this section a summary of the main measured data during the shaking table test is briefly presented and discussed.

3.1. Soil acceleration in free field

Sample soil acceleration time histories in the free field part of the model (soil far from the piles) are shown in Figure 2. As can be observed in this figure, the amplitude of acceleration records in liquefiable layer decreased dramatically at the beginning stages of shaking as the soil underwent liquefaction.

3.2. Excess pore water pressure records

Representative excess pore pressure time histories recorded in free field area are shown in Figure 3. The trends show that the soil liquefied after about 3 cycles of shaking since the middle layer composed of very loose sand. Drainage of excess pore

pressures or consolidation of the liquefied sand initiated from the lower depths (PWP1) and followed by pore pressure reduction in upper elevations (PWP2).

Figure 1. Plan view and cross section of the physical model.

-0.4

-0.2

0

0.2

0.4

Acc

eler

atio

n (g

)

ACC4 (surface)

-0.4

-0.2

0

0.2

0.4

Acc

eler

atio

n (g

)

ACC3 (65cm depth)

-0.4

-0.2

0

0.2

0.4

Acc

eler

atio

n (g

)

ACC2 (95cm depth)

-0.4

-0.2

0

0.2

0.4

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

Acc

eler

atio

n (g

)

Time (sec)

ACC1 (Base)

Figure 2. Sample acceleration time histories of soil in the free field.

-202468

101214

Exce

ss p

ore

pres

sure

(kPa

) PWP 2

-202468

101214

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 3

Exce

ss p

ore

pres

sure

(kPa

)

PWP 1

4Time (sec)

Figure 3. Sample excess pore water pressure records in free field.

3.3. Soil and pile group lateral displacement records

Figure 4, summarizes displacement records of the pile cap and soil at the free field. As seen, the soil started to move downward right after being liquefied. Unlike the free field soil displacement which kept increasing until the end of shaking,

921


pile cap displacement reached its maximum displacement a few seconds after the shaking and then bounced back gradually having a residual displacement of about 9 mm. The maximum ground surface displacement was about 5.0 cm while the maximum displacement of the cap was about 5.4 cm.

-10123456

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

Disp

lace

men

t (cm

)

Ground surface

-2-10123456

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

Disp

lace

men

t (cm

)

Time (sec) Figure 4. Time histories of ground surface and lateral pile cap displacements.

3.4. Pile bending moments

Figure 5 shows time histories of bending moments in instrumented individual piles of the group at some representative depths, i.e. at base of the liquefiable layer and near the connection of piles to the cap. As seen, after lateral spreading that occurred about t=1.5 sec, bending moments increases significantly. However, during liquefaction, the soil loses most of its shear resistance; hence it fails and gradually moves around the piles. This movement reduces the lateral pressure on the piles; therefore the piles bounce back towards upslope due to their rigidity as the shaking continued. Due to this elastic rebound, bending moments in piles descend as well. It should be noted that time histories of bending moments in all piles consist of a cyclic component due to dynamic soil pressures as well as a monotonic component from the kinematic lateral soil pressures during lateral spreading. An interesting observation is that maximum positive bending moments differ in individual piles of the group depending on their position within the group. 4 LATERAL PRESSURE OF LIQUEFIED SOIL ON THE

PILES

The lateral pressures exerted on the individual piles of the groups were back-calculated from the monotonic component of bending moment data using the method introduced by Brandenberg et al. (2010). Figure 6 shows profiles of the monotonic component of back-calculated lateral pressures of liquefied soil along with the lateral forces proposed by JRA (2002) code for design of pile groups against lateral spreading. This code recommends using 30% of the total overburden pressure to be applied to the outermost width of the pile group as lateral forces due to lateral spreading. In cases with a top non-liquefiable layer, it suggests that the passive pressure from non-liquefiable layer should be considered as well. For design applications, implementing JRA (2002), it is assumed that the total lateral force exerted on the pile group is equally distributed among the individual piles of the group.

According to Figure 6, at the early stages of shaking when the soil was not yet liquefied, induced pressures are negligible. But upon liquefaction and lateral spreading, magnitude of lateral pressures increased significantly. In all diagrams, an increase in applied lateral pressures is observed at upper elevations where the non-liquefiable crust exists. In fact, the non-liquefiable crust moved with the underlying liquefied layer towards the downslope during lateral spreading, exerting extra pressures on the piles. As seen in Figure 6, the magnitude of lateral pressures on pile 3 (the downslpe pile) in upper elevations are greater than

of the soil from the downslope side of pile 3 during lateral spreading resulting in lack of lateral support. The agreement between the magnitudes and patterns of back-calculated lateral pressures with the values recommended by JRA 2002 is reasonable except in pile 3 which shows significant difference with JRA 2002 values in terms of pressure magnitude and pattern.

-0.1

-0.05

0

0.05

0.1

0.15

0.2

0.25

Ben

ding

mom

ent (

kN.m

)

Pile P1

Ground surface

those of the other piles which can be attributed to the separation

-0.1

-0.05

0

0.05

0.1

0.15

0.2

0.25

Ben

ding

mom

ent (

kN.m

)

Pile P2

-0.1

-0.05

0

0.05

0.1

0.15

0.2

0.25

Ben

ding

mom

ent (

kN.m

)

Pile P3

-0.1

-0.05

0

0.05

0.1

0.15

0.2

0.25

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

Ben

ding

mom

ent (

kN.m

)

Time (Sec)

Pile P4

Figure 5. Time histories of bend tative ing moments in represenindividual piles of the group.

0

0.2

0.4

0.6

0.8

1

1.2

-2 -1 0 1 2 3 4 5 6 7 8 9

Dept

h (m

)

Pile P1

t=1 sect=2 sect=4 sect=8 sect=12 secJRA

0

0.2

0.4

0.6

0.8

1

1.2

-3 -2 -1 0 1 2 3 4 5 6 7 8 9

Dep

th (m

)

Pile P2


0

0.2

0.4

0.6

0.8

1

1.2

-3 -2 -1 0 1 2 3 4 5 6 7 8 9

Dept

h (m

)

Monotonic soil pressure (kPa)

Pile P4


0

0.2

0.4

0.6

0.8

1

1.2

-3 -2 -1 0 1 2 3 4 5 6 7 8 9

Dep

th (m

)

Monotonic soil pressure (kPa)

Pile P3


Figure 6. Profiles of lateral soil pressures on individual piles of

3.5. Total lateral forces exerted on individual piles

ces exerted

the group during lateral spreading.

Monotonic components of maximum total lateral foron the piles were calculated by integrating the lateral soil pressures along the piles. These total lateral forces were

Depth=125cm

Pile cap

Depth=124cm

922


0.20

9 0.45

9

1.72

3

1.39

7

1.93

2

1.85

6

0.0

0.5

1.0

1.5

2.0

2.5

3.0

Experiment JRA

Mon

oton

ic c

ompo

nent

of

max

imum

tota

l late

ral f

orce

(kN

) Crust layerLiq. layerTotal

separately evaluated for the liquefied layer and the non-liquefiable crust. The calculated forces are displayed in Figure 7. By comparing total lateral forces in different piles following findings can be itemized: - The amount of total lateral force in pile P2 (located in

.24 times that exerted

t among all

rces in pile P1 (the middle p

middle row) is less than the piles located in upslope and downslope rows, i.e. piles P1 and P3.

- Total lateral force on pile P1 is about 1on pile P2. This occurs due to the shadow effect. Since the upslope pile is directly pushed by the laterally spreading soil and acts as a barrier for pile downslope pile, P2.

- Total lateral force exerted on pile P3 is the largesthe other piles. Total lateral force on pile P3 is about 1.43 and 1.76 times those of piles P1 and P2, respectively. This can be described by the separation of soil from the downslope side of pile P3 during lateral spreading resulting in lack of lateral support.

- Comparing total lateral fo

Figure 8. Comparison between monotonic components of maximum total lateral forces in pile group of this experiment and JRA 2002 recommended values.

ile in upslope row) and P4 (the side pile in upslope row) shows that the side pile receives larger force than the middle pile by a factor of about 1.27. This phenomenon is called neighboring effect.

4. CONCLUSIONS

0.02

6

0.01

1 0.03

4

0.02

5

0.14

2

0.12

5

0.20

6

0.18

7

0.16

8

0.13

6

0.21

3

0.00

0.05

0.10

0.15

0.20

P1 P2 P3 P4

Mon

oton

ic c

ompo

nent

of m

axi

tota

l lat

eral

forc

e (k

N)

Pile

0.24

0

0.25

0.30

mum

Crust layerLiq. layerTotal

3.6. Total lateral force exerted on the pile group

estimated by

Findings from a large scale shake table test on a stiff 3×3 pile group are presented and disscussed. The results show that total lateral forces due to lateral spreading on the pile group can be well predicted by JRA 2002 design code. However, based on the experimental results, lateral forces exerted on individual piles of the group varries depending on the pile positions within the group which is not considered by JRA code. The shadow and neighboring effects are found to be responsible for such an observation. It is recommended that this variation be considered in design applications.

5. REFERENCES

Kramer S.L. and Elgamal, A. 2001. Modeling soil liquefaction hazards for performance-based earthquake engineering. PEER report 2001/13, Pacific Earthquake Engineering Research Center, College of Engineering, Univ. of California, Berkeley.

Figure 7. Comparison of maximum total lateral forces on

Hamada H., Yasuda S., Isoyama R. and Emoto K. 1986. Study on Liquefaction Induced Permanent Ground Displacements. Research report, Association for the Development of Earthquake Prediction, Japan. different piles of the group.

Tokimatsu K. and Asaka Y. 1998. Effects of Liquefaction-Induced Ground Displacements on Pile Performance in the 1995 Hyogoken-Nambu Earthquake. Special Issue of Soils and Foundations, pages 163–177.

Total lateral forces exerted on the group can beadding all lateral forces exerting on individual piles of the group. It should be noted that in this experiment, only one side pile in upslope row of the group was instrumented but it was assumed that the ratio of lateral forces on the side piles of other rows to those of their corresponding middle piles is the same as the ratio between piles P4 and P1. Total forces exerted on pile group of this experiment are compared with those recommended by JRA 2002 in Figure 8. According to this figure, total lateral force exerted on the pile group is about 1.04 times the values calculated using recommendations of JRA 2002. This difference in total lateral forces is found to be negligible. But if only the lateral forces from the liquefiable layer be considred the differences will be more. However, the trend observed for the non-liquefiable crust layer is completely different as the lateral forces suggested by JRA [2002] is about 2.2 times the experimental values. The reason is that passive pressure recommended by JRA 2002 does not seem to be mobilized in this experiment.

Eberhard Marc O., Baldridge S., Marshall J., Mooney W. and Rix J. 2010. USGS/EERI Advance Reconnaissance Team: TEAM REPORT V 1.1, The MW 7.0 Haiti Earthquake of January 12, 2010.

Haeri S. M., Kavand A., Rahmani I. and Torabi H. 2012. Response of a group of piles to liquefaction-induced lateral spreading by large scale shake table testing. Soil Dynamics and Earthquake Engineering 38, 25-45.

Motamed R. and Towhata I. 2010. Shaking table model tests on pile groups behind quay walls subjected to lateral spreading. Journal of Geotechnical and Geoenvironmental Engineering 136(3), 477-489.

Abdoun T., Dobry R., O’Rourke T. and Goh SH. 2003. Pile response to lateral spreads: centrifuge modeling. Journal of Geotechnical and Geoenvironmental Engineering 129(10), 869-678.

Ashford S. A., Juirnarongrit T., Sugano T. and Hamada M. 2006. Soil–pile response to blast-induced lateral spreading. I: Field Test. Journal of Geotechnical and Geoenvironmental Engineering 132(2), 152-162.

JRA. 2002. Seismic design specifications for highway bridges. Japan Road Association, English version, Prepared by Public Works Research Institute (PWRI) and Ministry of Land, Infrastructure and Transport, Tokyo, Japan.

Iai S., Tobita T. and Nakahara T. (2005. Generalized scaling relations for dynamic centrifuge tests. Geotechnique 55(5), 355-362.

Brandenberg S. J., Wilson D. W., and Rashid M. M. 2010. Weighted residual numerical differentiation algorithm applied to experimental bending moment data. Journal of Geotechnical and Geoenvironmental Engineering 136(6), 854-863.

923

Dynamic centrifugal model test for unsaturated embankments considering seepage flow and the numerical analysis

Expérimentation en centrifugeuse et modélisation numérique de la réponse aux séismes de remblais non saturés en prenant en compte l’écoulement

Higo Y., Oka F., Kimoto S., Kinugawa T. Department of Civil and Earth Resources Engineering, Kyoto University, Japan

Lee C.-W., Doi T. Former graduate student of Department of Civil and Earth Resources Engineering, Kyoto University, Japan

ABSTRACT: Earthquake-induced failure of unsaturated road embankments has taken place during the past earthquakes. It has been pointed out that water flow or higher water content of road embankments was a possible reason of the damage. In this study, dynamicresistance of unsaturated embankments with and without the seepage flow has been studied through the centrifugal model tests ofunsaturated embankment and their numerical simulations by a multi-phase coupled finite element method with an elasto-plastic constitutive model. During the tests, displacement, pore pressures, and acceleration of embankments have been fully monitored. Thedynamic behavior of unsaturated embankments with infiltration of pore water has been discussed on the basis of comparison betweenthe experimental and the numerical results. From the present study, we have found that the seepage flow and the high water content extensively affect the dynamic stability of unsaturated road embankments.

RÉSUMÉ: Les remblais routiers non saturés peuvent s’écrouler lors d’un séisme, la cause de cet écroulement étant sans doute à rechercher dans l’écoulement d’eau ou dans des grandes valeurs de teneur en eau au sein du remblai. Dans cet article, la résistance dynamique d’un remblai non saturé est étudiée, avec ou sans écoulement, par des essais en centrifugeuse et par leur modélisation numérique aux Eléments Finis, en utilisant une approche hydromécanique couplée et un modèle constitutif élasto-plastique pour le sol. Les déplacements, les pressions interstitielles et les accélérations ont été mesurés tout au long des essais. Les résultats expérimentaux et les résultats numériques ont été comparés. Les résultats de cette étude confirment que l’écoulement et la forte teneur en eau ont effectivement un rôle majeur pour la stabilité dynamique des remblais routiers non saturés.

KEYWORDS: unsaturated soil, embankment, earthquake, failure, dynamic centrifugal model test, numerical simulation. 1 INTRODUCTION

It is well known that the road embankment is still vulnerable against earthquakes. Earthquakes damaged road embankments during the extensive earthquakes such as the 2011 off the Pacific Coast of Tohoku Earthquake (M9.0). In particular, road embankments constructed on mountain/hill sides were severely damaged by the 2009 Suruga Bay Earthquake, the 2007 Noto Hanto Earthquake and the 2004 Niigata-ken Chuetsu Earthquake etc.

In the cases of the Noto Hanto Earthquake and the Niigata-ken Chuetsu Earthquake, seepage water flow or higher water content in the embankments was a possible reason of the damage of the road embankments. Increase in the water contents causes loss of the inter-particle force caused by suction and decrease in the skeleton stress. This suggests that the effect of the seepage water flow and the high water content in the embankments on the dynamic failure of road embankment has to be studied in detail. However, to the authors’ knowledge, there are a limited number of physical model studies on unsaturated embankments considering seepage water as well as numerical ones (e.g., Hayashi et al. 2002, Doi et al. 2010).

In this study, seismic resistance of unsaturated embankments with and without the seepage water has been studied through the centrifugal model tests and their numerical simulations by a multi-phase coupled finite element method with an elasto-plastic constitutive model (Oka et al. 2008, Oka et al. 2011). The dynamic behavior of unsaturated embankments with infiltration of the pore water has been discussed on the basis of comparison between the experimental and numerical results.

2 DYNAMIC CENTRIFUGAL MODEL TESTS

2.1 Soil used in the test

The soil used in the model tests is Yodogawa-levee sand, which has been used to fix the embankment of Yodo River in Kansai area. The physical properties of Yodogawa-levee sand are listed in Table 1.

Table 1. Physical properties of Yodogawa-levee sand Parameter Value

Sand content (%) 73.2 Silt content (%) 14.7

Clay content (%) 12.1 Dmax (mm)* 2.0 D50 (mm) 0.29 ρs (g/cm3) 2.661 wopt (%) 13.7

ρdmax (g/cm3) 1.861 k(m/sec)** 4.79×10-6

*Maximum diameter of sieved particle **Permeability when degree of compaction is 90%

2.2 Testing procedure

The model configuration and the sensor locations are illustrated in Fig. 1. Prior to preparing the specimen, test samples were mixed with water to set up the initial water contents of 15%. Then, model embankments were prepared by compacting method in eight layers; the base ground and the embankment were separated into three layers (thickness: 30mm, 15mm, 15mm), and five layers (as same thickness: 20mm), respectively. During the model construction, the accelerometers and the pore pressure transducers were embedded at the prescribed locations. The degree of compaction of all the cases

924


was set to be 90%. After construction of the embankment, laser displacement sensors were installed at the prescribed locations. In addition, the targets were inserted in each compacted layer, for quantifying the displacements between before and after the tests by PTV (Particle Tracking Velocimetry) technique. The centrifugal gravity used in this study was 50G.

The infiltration of water has been performed from the three slits of the right side wall (see Figure 1). The water of 2,000mL was prepared in the water supply tank attached on the model container. The water level of the tank was set to be 6.75m. The valves of the three slits installed on the wall with the height of 1.5m, 3m, and 4.5m from the bottom were opened and the water flowed into the model embankment and the base ground through the slits. In the present study, the water was used as the pore fluid, whereas viscous fluids with the 50 times viscosity of water, e.g., metolose solution, are often used for satisfying the similarity rule. This is because the matric suction of the metolose solution is smaller than that of the water. Note that the permeability of the embankment is 50 times in 50G field.

The centrifuge was spun up to an acceleration of 50G and seepage in 50G field was started. The water level was increased and reached close to the steady state, dynamic loads was applied to the model embankment. The tapered sine waves with a frequency of 1 Hz and an amplitude of approximately 400 gal were used as an input wave, and the duration of the wave was 30 seconds.

Figure 1. Model embankment and arrangement of sensors

2.3 Testing program

Two dynamic centrifugal model tests were performed. One has been conducted without infiltration (Case 1), and the other has been done with the seepage flow (Case 2). Average water content measured after the test was 12.7% in Case 1, which is similar to the optimum water content.

2.4 Test results

Figure 2 demonstrates a time profile of the pore water pressure at the base ground during the infiltration process and the water level estimated by the pore water pressure. All of the experimental results are expressed in a prototype scale hereinafter. It is seen that the pore water pressures were increased induced by the infiltration of water and reached almost steady state after 12 hour. The seepage area is shown in Figure 3 obtained in the other test with the same testing conditions as Case 1 other than the use of the yellow colored water. It is seen that the seepage area is wider than that estimated by the pore water pressures as shown in Figure 2. This indicates that the unsaturated seepage flow occurs in the specimen.

Figure 4 shows the distribution of displacement vector, and Table 2 shows the displacements at the toe of the slope and the crest. Displacement in Case 2 is much larger than Case 1 due to the effect of infiltration. This suggests that an increase of the pore water pressure by infiltration causes the decrease in suction and the skeleton stress of the embankment.

Time profiles of the excess pore water pressure measured under the crest (No.1) and in the base ground (No. 4) are shown in Figure 5. Note that the pore water pressure at the beginning of the loading is shifted to zero. While the excess pore water pressure levels in Case 1 are almost zero, those in Case 2 are rather high. In particular, it is possible that liquefaction occurs because the pore pressure of P4 at the base ground of Case 1 increases up to the initial vertical stress.

0 4 8 12 16 20

0

5

10

15

20

25 P 1 P 2 P 3

P3

P4

P2 P 4

Por

e pr

essu

re (

kPa)

T im e (hour )

P1

Table 2. Displacements after the tests Toe of the Slope Crest

Casex y y

1 -0.4 7.1 0.22 -213.0 24.0 -380.0

* x: horizontal displacement, y: vertical displacement (in prototype, unit: mm)

Figure 3. Seepage area indicated by yellow colored water

Figure 2 Time profile of pore water pressure and the water level estimated by the pore water pressure (Case 2)

50 cm

Figure 4. Distribution of the displacement vectors

(a) Case 1

50 cm

(b) Case 2

925


0 3 0 60 90

0

5

10

15

20

12 0

P 1 C ase 1 P 4 C ase 1 P 1 C ase 2 P 4 C ase 2

Por

e pr

ess

ure

(kP

a)

T im e (sec )

Onset of seismic loading

Figure 5. Excess pore water pressure-time profile

10

1

2

3

4

5

Mea

sure

-inp

ut

rati

o

Pe r io d (sec )

A c c e le r om et e r 1 (at t he c r est )

A c c e le r om et e r 4 ( in t he em bankm ent )

0.3

Case 2

Case 1

Figure 6. Acceleration response spectrum (measured-input ratio)

Fig. 6 shows the acceleration response spectrum, namely, measured-input ratio, calculated using the EMPR program developed by Sugito et al. (2000) at the crest and in the embankment. Damping parameter used in this analysis is 0.05. The higher amplification occurs in Case 2 than Case 1, and the predominant period is about 0.5sec. This is probably because of the reduction in the strength and the stiffness of the embankment due to the infiltration.

3 NUMERICAL SIMULATION OF THE DYNAMIC CENTRIFUGAL MODEL TESTS

3.1 Multi-phase coupled liquefaction-analysis method for elasto-plastic unsaturated soils

In the formulation of the dynamic coupled analysis, the simplified three-phase method is used in which the compressibility of air is assumed to be very high, whereas the soil particle and the pore water are incompressible as compared with the air (Oka et al. 2007, 2008, Kato et al. 2009), namely, the three-phase method can be simplified into the soil-water coupled two-phase mixture theory. A cyclic elasto-plastic model based on the non-linear kinematic hardening rule (Oka et al., 1999) was used in the analysis with a modified plastic dependency of the modulus, into which the non-linear hardening rule was incorporated. As the stress variable of the constitutive model, the skeleton stress is used in order to describe the mechanical behavior of unsaturated soils (e.g., Oka et al. 2007, 2008, Oka and Kimoto 2012). Skeleton stress tensor

is defined as follows: , (1)

where is the total stress tensor, is Kronecker’s delta, is the average fluid pressure, is the pore water pressure, is the pore air pressure, and is the degree of saturation. For the full description of the behavior of unsaturated soil, it is necessary to incorporate the suction in the constitutive model.

A u-p formulation for the liquefaction analysis (Oka et al., 2004) is adopted to solve the governing equations, in which the displacement of the solid phase and the pore pressure are used

as independent variables. In the present analysis, the finite element method was used for the spatial discretization of the equation of motion for the whole mixture, and the finite difference method was used for the discretization of the continuity equation for the water phase.

3.2 Analysis model and the boundary conditions

Figure 7 shows the analysis model of the embankment and the finite element meshes used in the analysis. The embankment is initially unsaturated with an initial suction of 3.21 kPa corresponding to the initial degree of saturation of 67.8%. The right boundary is partly drainage one in order to simulate the water supply with the hydrostatic pressure. After the simulation of the water infiltration for 24 hours, dynamic analysis with the same input wave as in the experiment has been performed. Material parameters used in this analysis are listed in Table 3. The permeability coefficients of the elements just close to the drainage boundary are 10 times lower than the other parts because the water is likely to flow between the soil and the wall of the model container. The bulk modulus of the pore fluid of 5,000 kPa is lower than that of the water in order to model the mixture of the pore water and the pore air since the unsaturated seepage flow was observed in the experiment.

Hydrostatic

pressure

6.75

Impermeable boundariesPermeable boundaries No flow (unsaturated state) /drainage boundary (fully saturated state)

x

y

Figure 7 Analysis model and the boundary conditions

Table 3 Mater al parameters i Initial void ratio 0.589

Compression index 0.0804 Swelling index 0.0001

Elastic shear modulus 4000 Permeability (m/s) 4.79×10

-5

Bulk modulus of pore fluid (kPa) 6000 Phase transformation stress ratio 1.270

Failure stress ratio 1.270 Kinematic hardening parameter 10000 Kinematic hardening parameter 150 Kinematic hardening parameter 50

Quasi-overconsolidation ratio 1.3 Anisotropy parameter 2000 Dilatancy coefficient 1.0 Dilatancy coefficient 2.0

Referential strain parameter 0.008 Referential strain parameter 0.08

van Genuchten’s parameter (1/m) 19.6 van Genuchten’s parameter 1.2

3.3 Simulation results

Figure 8 shows the time profile of the pore water pressure during the seepage process. Comparing with the test results shown in Figure 2, the pore water pressure level at P1, P2, and P3 of the simulation results are higher than those of the test results. Meanwhile, the seepage area shown in Figure 9 is almost the same as that observed in the test (Figure 3). Namely, the unsaturated seepage flow was observed in experiment, while the fully saturated flow is obtained by this analysis. In this analysis, the unsaturated seepage flow has been modeled by the reduction of the bulk modulus of the pore fluid.

926


Displacements-time profile at the toe and the crest obtained by the analysis are in good agreement with the test results as shown in Figure 10. Figure 11 shows the distribution of at the end of the dynamic loading, in which ( : plastic deviatoric strain increment). It can be seen that the several strain localization zones appear from the toe of the embankment to the crest. In addition, another strain localization zone can be seen in the base ground, which is consistent with the test results (Figure 4).

The distribution of pore water pressure is shown in Figure 12. It is seen that the pore water pressure increases in the seepage area. Figure 13 shows the skeleton stress decreasing ratio SSDR (defined as , : current mean skeleton stress, : initial mean skeleton stress) at the end of the dynamic loading. The higher SSDR is observed just below the toe of the embankment. This suggests that the decrease in the mean skeleton stress due to the increase in pore pressure induces the large deformation of the embankment with water infiltration.

4 CONCLUSIONS

Dynamic behaviors of unsaturated embankment considering seepage flow have been studied through the centrifugal model tests and their numerical simulation. For the seepage process, the unsaturated seepage flow has been observed in the experiment, and the seepage area of numerical simulation has been similar to that of experiment. It is found in the dynamic loading process that the infiltration of water into the unsaturated embankment has induced the large deformation in the seepage area due to the generation of pore water pressure in the embankment. The numerical results have provided that the increase in the pore pressure leads to the decrease in the mean skeleton stress in the seepage area, in particular, just below the toe of the embankment.

5 ACKNOWLEDGEMENTS

This research was supported in part by the National Institute for Land and Infrastructure Management, MLIT, Japan (Grant for research and development of technologies for improving the quality of road policy, No. 21-4, 2009-2012).

6 REFERENCES

Doi, T., Higo, Y., Oka, F., Kimura, M., Kimoto, S. & Lee, C.-W., Proceedings of 23rd KKCNN 2010, Taipei, pp. 303-306, 2010.

Hayashi, H., Nishikawa, J. & Egawa, T., Proc. Int. Conf. Physical Modelling in Geotechnics, ICPMG ’02, Phillips, R., et al., eds., pp. 483-488, 2002.

Kato, R., Oka, F., Kimoto, S., Kodaka, T. & Sunami, S., Journal of Geotechnical Engineering, JSCE, 65 (1), 226-240, 2009 (in Japanese).

Oka, F., Yashima, A., Tateishi, A., Taguchi, T. & Yamashita, S., Géotechnique, 49(5), 661-680, 1999.

Oka, F., Kodaka, T. & Kim Y.-S., Int. J. Numer. Anal. Meth. Geomech., 28 (2), 131-179, 2004.

Oka, F., Kodaka, T., Kimoto, S., Kato, R. & Sunami, S., Key Engineering Materials, 340-341, pp.1223-1230, 2007.

Oka, F., Kimoto, S., Kato, R., Sunami, S. & Kodaka, T., Proc. 12th Int. Conf. IACMAG, Singh, D.N. ed., 2029-2041, 2008.

Oka, F., Kimoto, S. and Kato, R., First International Conference on Geotechnique, Construction Material and Environment, Mie, Japan, pp.15-22, 2011.

Oka, F. and Kimoto, S., Computational modeling of multiphase geomaterials, CRC Press, Taylor and Francis Group, 2012.

Sugito, M., Furumoto, Y., & Sugiyama, T., 12th World Conference on Earthquake Engineering, 2111/4/A CD-ROM, Auckland, New Zealand, 2000.

.

0 4 8 12 16 20 24

0

10

20

30

40

50

P4

P3

P2

P 1 P 2 P 3 P 4

Por

e pr

essu

re (

kPa)

T im e (hour )

P1

Figure 8. Pore water pressure-time profile during the seepage process

Figure 9. Distribution of pore water pressure at 24 hours of the seepage process (unit: kPa)

0 5 10 15 20 25 30- 0.4

- 0.3

- 0.2

- 0.1

0.0

0.1

V er t ic al d isp. (C r es t , s im .) V er t ic al d isp. (C res t , exp.) V er t ic al d isp. (T oe, s im .) V er t ic al d isp. (T oe, exp.) Hor izont al d isp. (T oe, s im .) Hor izont al d isp. (T oe, exp .)

Dis

plac

emen

t (m

)

T im e (sec ) Figure 10. Displacements-time profile during the dynamic loading process

Figure 11 Distribution of at 30 seconds of the dynamic loading process (max: 0.240)

Figure 12 Distribution of the pore water pressure at 30 seconds of the dynamic loading process (Unit: kPa, max: 109.0kPa)

Figure 13 Distribution of SSDR at 30 seconds of the dynamic loading process

927

Développement d’un modèle réduit tridimensionnel du renforcement des sols par inclusions rigides

Development of a three-dimensional small scale model to simulate soil improvement by rigid piles

Houda M., Jenck O., Emeriault F. Laboratoire 3SR (Université Joseph Fourier, Grenoble INP, CNRS), Saint-Martin d’Hères, France

Briançon L. CNAM, Paris, France et Laboratoire 3SR, Saint-Martin d’Hères, France

Gotteland Ph. Fédération Nationale des Travaux Publics, DTR Recherche, Paris, France

RÉSUMÉ: Un modèle réduit tridimensionnel à l’échelle 1/10 a été développé en laboratoire, simulant le renforcement des sols parinclusions rigides verticales, mettant éventuellement en œuvre un renforcement supplémentaire par géosynthétique. Des sollicitationsmonotones ou cycliques peuvent être appliquées en surface, via une membrane sous pression. Des campagnes expérimentales ont étéentreprises sur ce modèle de laboratoire, portant sur l’analyse du comportement sous chargement cyclique et sur l’étude de l’impact d’un renforcement par géosynthétique en base du matelas granulaire. L’objectif de ces études préliminaires est la validation dudispositif expérimental. La modularité de ce modèle en termes de géométrie, de mise en œuvre de matériaux et d’instrumentationpermettra l’étude d’autres problématiques géotechniques.

ABSTRACT: A three-dimensional model in 1/10th scale was developed in the laboratory, simulating the soil reinforcement byvertical rigid inclusions, with the possibility to add reinforcement by geosynthetic. Monotonic or cyclic loading can be applied onsurface via a membrane under pressure. Experimental campaigns have been conducted on this model, studying the behaviour under cyclic loading and the impact of geosynthetic reinforcement in granular base of the mattress. The main objective of these studies was to validate the experimental device. The modularity of this model in terms of geometry, materials and instrumentation allow the study of other geotechnical problems.

MOTS-CLES: Modèle réduit de laboratoire, chargement cyclique, renforcement des sols, inclusions rigides, géosynthétique.

KEYWORDS: Laboratory Model, cyclic loading, soil reinforcement, rigid piles, geosynthetic.

1 INTRODUCTION.

L’étude du comportement des ouvrages renforcés par inclusions rigides est un sujet d’actualité en France : le projet national ASIRI a donné lieu en juillet 2012 à des recommandations de dimensionnement et de mise en œuvre (IREX 2012). Cependant ce projet a été limité en général à des cas de chargement statique (Chevalier et al. 2010) et ces recommandations ne traitent pas suffisamment des cas où la plateforme de transfert de charge est renforcée par des nappes géosynthétiques (Briançon et Simon 2011), applications cependant en plein essor.

Des modèles physiques en laboratoire ont été développés depuis les années 1970 pour étudier ce type d’ouvrages. Certains sont en deux dimensions (Hewlett et Randolph 1988, Low et al. 1994, Van Eekelen et al. 2003, Jenck et al. 2005, Chen et al. 2008) permettant de visualiser les phénomènes dans le sol. D’autres modèles sont en deux dimensions à symétrie axiale (Dinh et al. 2009). La plupart des modèles à trois dimensions (Demerdash 1996, Heitz et al. 2008, Van Eekelen et al. 2011) ne permettent pas la visualisation du champ de déplacements et contiennent un nombre d’inclusions rigides réduit, ce qui ne permet pas l’élimination des effets de bord.

Le comportement des sols compressibles renforcés par inclusions rigides verticales est également étudié par des expérimentations sur modèles centrifugés 3D à échelle réduite. Ces essais permettent d’appliquer des niveaux de contraintes du même ordre que dans les cas réels, cependant ils présentent d’autres limitations telles que la difficulté de représenter tous les matériaux explicitement (Baudoin et al. 2008).

Un modèle réduit tridimensionnel sous gravité normale a ainsi été développé au laboratoire 3SR de Grenoble, sur la base d’un modèle préliminaire initialement développé au Cnam de Paris et simulant le renforcement des sols par inclusions rigides

verticales, sous chargement monotone et également cyclique, mettant éventuellement en œuvre un renforcement supplémentaire par géosynthétique. Ce modèle comporte 16 à 20 inclusions rigides verticales. Une nappe de géosynthétique peut être placée en surface du sol renforcé. Des sollicitations sont appliquées en surface du massif par une membrane sous pression. Une instrumentation permet d’analyser le comportement du système. L’objectif de ce modèle n’est pas de simuler quantitativement le comportement d’un ouvrage réel (les règles de similitude n’étant pas toutes strictement respectées), mais vise à mieux comprendre les mécanismes qui se développent au sein du massif et plus précisément dans le matelas granulaire de transfert de charge, d’analyser l’effet du renforcement par géosynthétique et de mieux appréhender les mécanismes d’interaction entre ces diverses parties de l’ouvrage.

2 MODELE REDUIT

2.1 Dispositif expérimental

Le modèle réduit a été développé pour étudier les mécanismes qui se produisent plus spécifiquement dans la plate-forme de transfert de charge et dans le géosynthétique sous des charges monotones mais également cycliques. Il se compose d’une cuve de section carrée de dimension interne 1m x 1m, constituée de châssis en acier. La hauteur de la cuve peut être ajustée en ajoutant ou en supprimant des cadres, tandis qu’un espace vide en fond de cuve de 20 cm de hauteur est alloué à l'instrumentation (Fig. 1 et 2). Le modèle comporte seize inclusions en aluminium de 35 mm de diamètre et de hauteur 60 cm avec un espacement entre elles de 20 cm comme le montrent les figures 3 et 4. Le taux de recouvrement (proportion de la surface couverte par les inclusions) est ainsi de 2,4%.

928


Quatre demi-inclusions peuvent être utilisées près de la fenêtre transparente visant à visualiser les champs de déplacement du sol.

Figure 1. Coupe schématique du modèle réduit (coupe A-A de la figure 4)

Figure 2. Cadres empilables composant le modèle réduit

Figure 3. Photographie de l’intérieur de la cuve contenant les inclusions

Figure 4. Vue en plan schématique du modèle réduit.

2.2 Règles de similitude Les modèles physiques à échelle réduite présentent de nombreux avantages, comme la mise en évidence des mécanismes, l’obtention de résultats expérimentaux pour caler des modèles numériques, la conduite possible des essais jusqu’à la rupture. Une des difficultés majeures est cependant de satisfaire aux règles de similitude, afin d’appliquer les résultats observés sur le modèle réduit au problème en grandeur réelle. Ces règles sont établies à partir des équations générales de la mécanique, de l’équation de conservation de la masse et des lois de comportement des matériaux.

Dans un modèle avec une réduction d’échelle de n (ici n = 10) sous gravité normale, l’établissement des règles de similitude conduit au tableau 1.

Tableau 1. Facteur d’échelle pour un modèle sous gravité terrestre. Grandeur Notation Unité SI Facteur d’échelle Longueur L* m 1/n Pesanteur g* m/s2 1 Contrainte σ* Pa 1/n2

Déformation ε* --- 1 Masse volumique ρ* kg/m3 1 Angle de frottement Φ* --- 1 Module Young E* Pa 1/n

Dans notre cas, toutes les règles de similitude ne sont pas strictement respectées (E* et σ*). En revanche, l’utilisation de matériaux présentant des angles de frottement équivalents à ceux des matériaux réels, notamment pour le matelas de transfert de charge et le remblai, où les mécanismes sont principalement des mécanismes en cisaillement, apparaît ici justifiée et indispensable.

Le facteur d’échelle sur les contraintes est σ* = 1/10, ce qui indique que les contraintes appliquées dans le modèle devraient correspondre à des contraintes 10 fois plus élevées dans le prototype. Ceci n’est pas strictement respecté en tout point du modèle, mais l’application de la surcharge en surface est dimensionnée afin d’obtenir un niveau de chargement réaliste, voire surdimensionné afin de mettre clairement en évidence le fonctionnement de l’ouvrage.

2.3 Matériaux mis en œuvre

Les matériaux utilisés dans les essais sont le sol compressible, la plate-forme granulaire de transfert de charge, le sol du remblai et des géosynthétiques de renforcement. En ce qui concerne le sol compressible le paramètre principal est la compressibilité. Il a été simulé par un mélange de billes de polystyrène, de sable de Fontainebleau et une teneur en eau de 10% pour éviter la ségrégation. Ce matériau a des caractéristiques de compressibilité d’un sol compressible tout en étant relativement facile à mettre en œuvre. Dans notre étude, une couche de 5cm de gravier 2/4 mm (ø =55° au pic, ø = 37° au palier et c = 0kPa) est utilisée comme plate-forme de transfert de charge (PTC). Concernant le matériau du remblai, il est constitué de sable gris. Deux types de géosynthétiques (GSY) de renforcement bidirectionnel ont été utilisés : l’un (S1) de raideur à 5% d’élongation de 170 et 220 kN/m dans chacune des deux directions ; l’autre (S2), de raideur à 5% d’élongation de 316 et 400 kN/m dans chacune des directions. Les résultats d’essais avec géosynthétiques ne seront cependant pas détaillés dans ce document.

2.4 Instrumentation

L’analyse complexe des interactions sol-structure qui se développent dans ce modèle nécessite une instrumentation idoine. Pour ce modèle préliminaire, les paramètres suivants ont été mesurés dans la maille centrale :

• Tassement du sol compressible en trois points : D1, D2 et D3 (3 capteurs de déplacement potentiométriques 50 mm),

• Force en tête de deux inclusions : Ft1 et Ft2 et force en pied de deux autres inclusions (4 capteurs de force 10 kN)

• Contrainte à 5 cm au-dessus d’une inclusion, au niveau de l’interface entre la PTC et le remblai : Pptc (capteur de pression 500 kPa, de diamètre 30 mm)

• Pression appliquée en surface : Pm (capteur de pression dans la membrane 0-200 kPa).

A partir des capteurs de déplacement, des dispositifs de mesure du tassement en surface du sol compressible ont été confectionnés, par un système de tiges traversant le massif de sol compressible.

Les capteurs de force et de pression permettent d’estimer les reports de charge qui s’opèrent dans le matelas vers les inclusions et qui sont susceptibles d’évoluer au cours d’un

929


chargement cyclique en surface. Les capteurs de déplacement mesurent le tassement en surface du sol compressible (voir plus loin) et renseignent sur le niveau de déformation global dans le massif renforcé ou non par inclusions rigides, et permettent d’évaluer les performances de réduction de tassement.

En fonction des premiers résultats, l’instrumentation sera adaptée et étoffée.

2.5 Protocole expérimental

Les inclusions rigides sont installées dans la cuve, en contrôlant leur verticalité et espacement. Le mélange de sable, polystyrène et eau est versé dans la cuve, puis arasé au niveau de la surface des inclusions. Les tests avec géosynthétiques sont effectués en fixant celui-ci directement au-dessus du massif de sol compressible renforcé par inclusions. La PTC de 5 cm est placée en deux couches, légèrement compactées et le capteur Pptc est placé grâce à un gabarit. Le remblai de sable gris est mis en place par couches successives de 5 cm. La hauteur finale de la PTC + remblai est de 40 cm, mais elle pourra varier lors d’études ultérieures grâce à la modularité géométrique du dispositif développé. La membrane d’application de la charge en surface est disposée sur le massif, puis le dispositif est fermé par un couvercle, afin de mettre en pression la membrane et d’appliquer une charge verticale et homogène en surface.

2.6 Application du chargement

Le chargement appliqué sur le sol compressible (en base du matelas) est composé de deux phases successives : monotone puis cyclique.

Le chargement monotone se compose du poids propre du massif et de la première mise en pression de la membrane jusqu’à Pm = 20 kPa.

Le chargement cyclique (quasi-statique) est appliqué sous la forme présentée dans la figure 5 : une centaine de cycles d’amplitude 10 kPa est appliqué pendant une quinzaine d’heures, soit avec une période de 10 min environ.

L’application de ce chargement est pilotée par ordinateur, ce qui permet ainsi de la faire varier au besoin en termes de valeur de pression, d’amplitude et de fréquence.

Figure 5. Application du chargement cyclique

3 CAMPAGNE D’ESSAIS PRÉLIMINAIRE

3.1 Essais réalisés et objectifs

Des essais ont été réalisés dans le cadre d’une campagne préliminaire de faisabilité et de validation du dispositif expérimental développé, sous chargement cyclique, avec ou sans géosynthétique, avec ou sans PTC (voir Tableau 2).

Les principaux objectifs de cette campagne de faisabilité ne sont cependant pas d’exploiter les études paramétriques mais plutôt : - d’analyser le comportement global de l’ouvrage (par une analyse fine des mesures données par les capteurs), - d’analyser la répétabilité des essais (en effectuant des essais à configuration identique), - de détecter ainsi les améliorations à apporter au protocole expérimental et à l’instrumentation, - d’aider à définir les configurations d’essais à effectuer lors de la campagne suivante. T ableau 2. Configurations d’essais effectués.

Configuration IR GSY PTC Remblai

A Oui --- 5cm 35 cm

B Oui --- --- 40 cm

C Oui S1 5cm 35 cm

D Oui S2 5 cm 35 cm

E Non --- 5cm 35 cm

3.2 Quelques résultats d’essais préliminaires

Pour illustrer l’étude du comportement du modèle et de la répétabilité des essais, la figure 6 indique l’évolution des deux capteurs qui instrumentent la base des inclusions pour deux essais correspondant à la configuration A. Cette figure montre que les valeurs des deux capteurs Fp1 et Fp2 sont proches pour un même essai et que la répétabilité est bonne jusqu’à une valeur de contrainte moyenne en base de la PTC de 17 kPa, soit une pression appliquée en surface (Pm) d’environ 10 kPa. Au-delà, les valeurs de capteurs Fp1 et Fp2 divergent lors d’un même essai. Le protocole d’application de la pression en surface sera alors amélioré et, de plus, lors de la campagne d’essai suivante, la valeur de Pm sera limitée à une valeur plus faible afin de s’approcher d’un niveau de chargement plus réaliste.

La figure 7 indique l’évolution d’un des efforts mesurés dans le modèle (en tête d’une inclusion). Seuls les premiers cycles sont présentés sur cette figure. L’évolution sous chargement cyclique montre une légère réduction des valeurs maximum et minimum de la force lors des premiers cycles puis les valeurs deviennent constantes d’un cycle à l’autre. Lors de certains essais, une chute plus marquée de la valeur des efforts sur les inclusions a été observée lors des tout premiers cycles.

La figure 8 décrit l’évolution du tassement en surface du massif de sol compressible lors du chargement monotone puis lors des premiers cycles de chargement cyclique pour le même essai que précédemment. Le tassement augmente lors des premiers cycles puis tend à se stabiliser.

Ces résultats préliminaires montrent donc que la campagne suivante peut se focaliser sur l’observation des mécanismes lors des 10 à 20 premiers cycles et que l’application d’une centaine de cycles n’est pas nécessaire.

930


Figure 6. Analyse de la répétabilité sous chargement monotone

(configuration A)

Figure 7. Evolution de la force en tête d’inclusion (configuration B)

Figure 8. Tassement en surface du sol compressible (configuration B).

4 CONCLUSION

Un nouveau dispositif expérimental tridimensionnel a été développé en laboratoire afin d'étudier différents types de problèmes géotechniques dont, pour l’étude présentée ici, les mécanismes de transfert de charge dans les massifs de fondation renforcés par inclusions rigides, éventuellement renforcés par nappe géosynthétique, soumis à des charges monotones et cycliques en surface.

Des résultats préliminaires prometteurs ont été obtenus. L’analyse du comportement sous chargement cyclique montre que les efforts sur les inclusions diminuent et que les tassements augmentent au cours des 10 premiers cycles pour se stabiliser ensuite.

Cependant, l’objectif principal de cette étude est la validation du dispositif expérimental et l’obtention de recommandations pour la poursuite des essais. Le modèle nécessite effectivement quelques améliorations et compléments sur l’instrumentation et sur le protocole expérimental, dont les limitations ont été mises en évidence lors de cette campagne.

L’originalité de ce modèle réside d’une part dans sa modularité en termes de géométrie, de matériaux utilisés, d’application de chargement (étude sur la durée de vie de

l’ouvrage). Le modèle permettra donc facilement l’étude d’autres types de problèmes géotechniques. Le modèle conçu permet également d’obtenir un champ de déplacement dans le massif par la prise de photographies au niveau de la fenêtre de visualisation et l’application de méthodes de corrélation d’images. Bien que ce modèle réduit ne satisfait pas strictement aux conditions de similitudes et ne permet donc pas d’extrapoler les résultats à un ouvrage réel de manière quantitative, il a néanmoins pour objectif de contribuer à une meilleure compréhension des mécanismes de transfert de charge dans les massifs, notamment sous chargement cyclique. Il servira de base à la validation de modélisations numériques en milieu continu ou par éléments discrets.

5 REMERCIEMENTS

Les auteurs remercient la Direction Technique et de la Recherche de la Fédération Nationale des Travaux Publics (France) pour l’aide financière apportée à cette étude, Cléber Da Silva Pinto pour l’aide à la réalisation des essais, et son encadrant Pr. Márcio Almeida de l’Université de Rio de Janeiro.

6 REFERENCES

Antoine, P.-C. 2010. Etude de dalles sur sols renforcés au moyen d’inclusions rigides ou non. Thèse de l’Université Libre de Bruxelles.

Baudoin G., Thorel L., Raul G., Garnier J. 2008. Centrifuge modeling of 3D load transfer in reinforced soft soil. Foundations: Proceedings of the Second BGA International Conference on Foundations

Briançon L. et Simon B. 2011. Performance of Pile-Supported Embankment over Soft Soil: Full-scale experiment. Journal of Geotechnical and Geoenvironnemental Engineering, 138 (4), 551 – 561.

Chen Y., Cao W., Chen R. 2008. An experimental investigation of soil arching within basal reinforced and unreinforced piled embankments. Geotextiles and Geomembranes, 26 (2), 164-174.

Chevalier B., Briançon L., Villard P., Combe G. 2010. Prediction of Load Transfers in Granular Layers Used in Rigid Inclusions Technique - Experimental and Discrete Element Method Analysis.GeoFlorida 2010.

Demerdash M.A. 1996. An experimental study of piled embankments incorporating geosynthetic basal reinforcement. Thèse de doctorat. University of Newcastle-upon-Tyne.

Dinh A.Q., Canou J., Dupla J.-C. 2009. Essais de chargement à pression contrôlée sur le modèle physique 1g –Étude paramétrique Rapport n°3.09.3.12 du PN ASIRI.

Heitz C., Lüking J., Kempfert H.-G. 2008. Geosynthetic reinforced and pile supported embankments under static and cyclic Loading. Proceedings of the 4th European Geosynthetics Conference EuroGeo4, paper n°215, 8p. Edinburgh, UK.

Hewlett W. J. and Randolph M. F. 1998. Analysis of piled embankment. Ground Engineering, 21(3), 12-18. IREX. 2012. Recommandations pour la conception, le

dimensionnement, l'exécution et le contrôle de l'amélioration des sols de fondation par inclusions rigides. Presses des Ponts. 384p.

Jenck O., Dias D., Kastner R. 2005. Soft ground improvement by vertical rigid piles – Two dimensional physical modeling and comparison with current design methods. Soils & Foundations. 45 (6), 15-30.

Low B. K., Tang S. K., Choa V. (1994). Arching in piled embankments. Journal of Geotechnical and Geoenvironmental Engineering, vol. 120, n° 11, pp. 1917-1938.

Van Eekelen D. J. M., Bezuijen A., Oung O. 2003. Arching in piled embankments; experiment and design calculation. Int. Conf. on Foundations: Innovations, observations, design and practice, 2-5 September 2003, 889-894. Dundee, Scotland.

Van Eekelen S.J.M., Bezuijen A., Lodder H.J., Van Tol A.F. 2011. Model experiments on piled embankments. Part I. Geotextiles and Geomembranes, doi:10.1016/j.geotexmem.2011.11.002

931

Full-scale field validation of innovative dike monitoring systems

Validation de systèmes de surveillance innovants pour digues à grande échelle

Koelewijn A.R., Vries (de) G. Deltares

Lottum van H. IJkdijk Foundation & Deltares

ABSTRACT: Three large scale field tests on dikes have been carried out at the IJkdijk test site in the Netherlands. Two tests involvedpiping, micro-instability of the sand core and erosion from overtopping. Both dikes failed on micro-instability. The third test involvedslope stability with a deep sliding plane. All tests were done to validate monitoring systems and dike safety information systems.Several systems performed well.

RÉSUMÉ : Trois essais à grande échelle sur digues sont exécutés sur le site de l’Ijkdijk aux Pays-Bas. Deux essais étaient concernés par un phénomène de renard, de déstabilisation par fluidisation de sable du cœur de la digue et d’érosion par surverse. Ces digues s’éffondraient par fluidisation de sable. Le troisième essai impliquait le mode de rupture de pente. Tous les essais ont été effectuéspour valider les systèmes de surveillance des digues. Plusieurs systèmes ont donné de bons résultats.

KEYWORDS: dike, embankment, full-scale test, slope stability, piping, microinstability, monitoring, information systems.

1 INTRODUCTION TO THE IJKDIJK TESTS

1.1 The IJkdijk research program

The IJkdijk (Dutch for ‘calibration dike’) is a Dutch research program with the two-fold aim to test any kind of sensors for the monitoring of levees under field conditions and to increase the knowledge on dike failure mechanisms.

Since 2007, several purpose-built dikes have been brought to failure at the IJkdijk test site at Booneschans, in the North-East of the Netherlands. Past experiments include a large stability test (Zwanenburg et al. 2012) and four field tests on backward seepage erosion or piping (van Beek et al. 2011). The tests presented in this article include these and other failure modes. For the near future, a test on static liquefaction is planned.

Meanwhile, the outcome of these tests has been implemented in practice by instrumenting several regular dikes, i.e. embankments with the function to protect the hinterland against flooding. By the end of 2012, this advanced surveillance by sensor equipment had been placed in ten different dikes in the Netherlands, United Kingdom, Germany and China.

1.2 All-in-One Sensor Validation Test

The main purpose of the All-in-One Sensor Validation Test (AIO-SVT) was to test the predictive power of full-service dike sensor systems, i.e. sensor in and on dikes combined with data processing and an information system providing a timely, reliable warning in case failure may occur. The application of such systems into practice will be a major improvement to the current state-of-the-art of dike management. In addition, contributing sensor systems were also tested and validated on their own. Another reason to carry out this test, in accordance with the two-fold aim of the IJkdijk,is to learn more on dike failure mechanisms, including failure prevention methods.

The AIO-SVT involved three dikes, which were all brought to failure. First, the geotechnical design of each dike is described, followed by the instrumentation. Next, the results are described, first regarding the failures of the dikes, then for the monitoring systems and finally for the information systems. Finally, conclusions are drawn.

2 DESIGN OF THE EXPERIMENTS

The experiments were designed in such a way that each dike could fail to different failure modes. The duration of each experiment was planned to be at least several days, with a maximum of one week, to allow the participating companies to collect a reasonable amount of data under varying conditions.

2.1 West and East dikes

The West and East dikes, named after their respective locations on the test site, were in many ways comparable. Both test dikes were 3.5m high, 15m long and 15m wide at their base, see Figure 1. The lower part of each dike was made of a 0.7m well-compacted clay layer, with a 1.7m high less-compacted small clay dike at the upstream side on top, a sand core behind this small clay dike and a cover of organic clay. This composition is found in many smaller dikes around the country. The base consisted of a uniform sand with a thickness of 3m with an impermeable foil below, to separate this test layer from the subsoil. Under the West dike, the sand has a d50 of 0.296mm and a uniformity coefficient U=d60/d10 of 1.69. Under the East dike, the d50 is 0.180mm and U=1.73. The upstream reservoir is enclosed by a 3.7m high dike. The size of the reservoir is about 2000m3.

By design, failure could occur from piping through the base, micro-instability of the sand core and overtopping of the crest

Figure 1. Cross-section of West and East dikes.

932


Figure 2. Cross-section of South dike at start of test, showing settled geometry and indicating positions of reference monitoring.

and subsequent erosion of the downstream slope. The earlier tests on piping had a similar configuration, but with a more sound clay dike (van Beek et al 2011). Those tests failed to piping at reservoir levels ranging from 1.75m to 2.3m. In order to make piping less likely this time, in each test dike a piping prevention measure has been placed. In the West dike, piping is controlled by a controllable drainage tube at 3.7m from the downstream toe, while in the East dike a coarse sand filter has been placed as a rectangular box 0.5m wide, 0.5m deep around 3.5m from the downstream toe. The grain size of this filter varies from 1 to 2 mm, the grain size distribution is such that excessive loss of particles through this filter is prevented (Burenkova 1993).

2.2 South dike

The South dike was built on a 4.5m thick composition of soft peat and clay. After construction, it was 4m high, 50m long at crest level, with a crest width of 3m and side slopes of 1:1.5 (V:H). The core was made of sand, with a 0.5m thick clay layer. Figure 2 shows a cross-section of the dike at the start of the test, i.e. after consolidation resulting in a settlement of 0.99m.

The designed failure modes of this dike were slope stability with a deep sliding plane through the subsoil with a minimum deformation of 20cm and rupture of the clay cover by high pore pressures inside the sand core as a result of saturating this core with water.

3 INSTRUMENTATION

For the instrumentation a clear distinction is made between the reference monitoring and the instruments of the participating companies. The reference monitoring was required (and sufficient) to closely monitor the course of the tests, while the other instruments were validated and the measurements were used to make updated predictions of the failures.

A total of nine companies participated with their instruments – some in all tests, others in only one or two. Each of these companies were invited to use their own measurements to give an initial prediction of the failure mode and the conditions at which failure would occur, and to update this prediction at least every 24 hours.

Three companies providing dike safety information systems participated in all three tests. These companies had access to the data of the monitoring systems being validated through a central data base. The data of the reference monitoring was not disclosed during the tests.

3.1 West dike

The reference monitoring was primarily carried out with pore pressure meters: two to record the water levels in the upstream and downstream reservoirs, four lines of 17 meters each at the interface between the lower sand and the well-compacted clay layer at 0.9m, 2.5m, 4.3m and 11.2m from the downstream toe and a 3 by 3 grid of pore pressure meters at the bottom of the sand core: right behind the small clay dike and at 1.8m and 6.0m downstream, respectively. In addition, visual inspections

were carried out at regular intervals, an HD camera facing the downstream slope taking one frame every five seconds was used, rainfall data was recorded and the upstream and downstream discharges were measured.

The following instrumentation was installed by the seven companies participating in this test:

- glass fibre optics woven into geotextile, measuring temperature and strain approximately every metre in eight lines parallel to the toe of the dike, five at the sand/clay interface and three in the downstream slope;

- a Fast Ground Based Synthetic Aperture Radar system, measuring a two-dimensional displacement field of the downstream slope every five seconds;

- two vertical tubes, installed at the upstream crest line, measuring temperature and strain profiles over depth employing glass fibre optics;

- a thermic infrared camera facing the downstream slope, with a resolution of 640x480 pixels and an accuracy of 0.05 K;

- a ground penetrating radar system at 100 MHz, operated by moving it across the crest of the dike;

- two controllable drainage tubes with measurements of pore pressure, temperature and discharge, located close to the sand/clay interface at 3.7m from the downstream toe (lower tube) and right behind the small clay dike at the bottom of the sand core (upper tube);

- six pore pressure meters at the sand/clay interface, three at 0.5m from the downstream toe and three at 2.2m.

3.2 East dike

The reference monitoring at the East dike was almost identical to the West dike, but with four lines of 16 instead of 17 pore pressure meters at the sand/clay interface.

The six companies participating in this test installed the following:

- glass fibre optics woven into geotextile, measuring temperature and strain approximately every metre in eight lines parallel to the toe of the dike, five at the sand/clay interface and three in the downstream slope;

- two vertical tubes, installed at the upstream crest line, measuring temperature and strain profiles over depth employing glass fibre optics;

- an electric resistivity system employing two rows of 14 electrodes on the downstream slope;


- a ground penetrating radar system at 100 MHz, operated by moving it across the crest of the dike;

- ten pore pressure meters at the sand/clay interface, five at 0.7m from the downstream toe and five at 2.2m.

3.3 South dike

The reference monitoring at the South dike consisted of 34 pore pressure meters and six automatic inclinometers. Twentysix pore pressure meters were installed in two cross-sections each 13m from the centre line, as indicated in Figure 2, six pore pressure meters were installed in six water tanks on top of the crest and the remaining two were installed in the basin on the non-failing side of the dike and in the ditch which was excavated during the test to reduce the overall stability. The inclinometers were distributed along the centre line and both instrumented cross-sections.

The seven companies participating in this test installed the following:

- glass fibre optics woven into geotextile, measuring temperature and strain approximately every metre in three parallel lines along the whole length of the dike, on ground level and on two higher levels;

- a system of six extremely accurate inclination instruments, each mounted on top of a 5.6m steel rod placed on the slope of the dike (three on the side of the failure, three on the other side);

933


- a Fast Ground Based Synthetic Aperture Radar system, measuring a two-dimensional displacement field of the slope at the side of the failure every five seconds;

- a total of four tubes measuring temperature and strain profiles over depth employing glass fibre optics: two vertical tubes 5.5m long halfway the slope at the side of the failure, one vertical tube 3.5m long at the toe at the same side in the centre line and one horizontal tube along the whole toe of the dike;


- one controllable drainage tubes with measurements of pore pressure, temperature and discharge, located inside the sand core, close to the toe at the side of the failure;

- eight instruments measuring pore pressure, temperature and local inclination distributed over two cross-sections 10m away from the centre line, in each cross-section one instrument in the sand core close to the toe and three instruments distributed over depth in the soft soil deposits under the toe.

4 RESULTS OF THE EXPERIMENTS

4.1 West dike

The test on the West dike started on August 21st at 4:30 pm. Filling the reservoir about 1m already caused serious cracks in the upper part of the dike. Also, leakage through the small clay dike occurred. Compaction of this clay was not sufficient. Once the situation stabilized, the upstream level was increased again. At a head drop of 1.56m the first wells appeared and sand producing wells (piping) appeared at a head drop of 1.79m.

At 66.7 hours after the start of the test (t=66.7hrs), at a head drop of 2.02m, the lower drainage tube was partly opened as piping had already been detected at the third line of pore pressure meters, i.e. upstream of this tube. This had a clear effect on the pore pressures, as shown in Figure 3, and the piping process stopped.

Meanwhile, the sand core became saturated, as measured by the upper pore pressure meters. At t=63.6 hrs, the upper drainage tube was opened and the pore pressures in the core were controlled. At t=94.0 hrs, both tubes were closed and the pore pressures rised sharply. From t=97.6 hrs, sliding of the downstream slope started to occur.

At t=110.1 hrs, considerable sliding of the downstream slope had occurred. Piping had resumed too, but the continued sliding from micro-instability of the sand core caused so much settlement of the crest that at t=111.9 hrs (August 26th at 8:24 am) failure occurred.

4.2 East dike

The test on the East dike started on August 21st at 3:20 pm and ran parallel to the test on the West dike. In many ways, both tests were similar, although the controllable drainage tubes were missing at the East dike. Wells occurred at the downstream slope at a head drop of 1.60m and piping started at a head drop of 2.02m. However, piping was detected only in the two lines of pore pressure meters downstream of the coarse sand filter, upstream no piping could be traced. Apparently, this measure worked.

As the last stages of the hydraulic load were delayed in comparison with the West dike, severe settlements from micro-instability of the sand core occurred later. Here at t=138.9 hrs (August 27th at 10:18 am) failure from micro-instability occurred. Figure 5 shows an overview of both failures.

4.3 South dike

The test on the South dike started on September 3rd at 12:12 pm,

Figure 3. Pore pressures at sand/clay interface West

Figure 4. West dike close before failure.

Figure 5. West and East dikes after failure.

by infiltration of water into the sand core. The next day, a small excavation was made in front of the dike. This had a limited effect on the dike, as shown in Figure 6 by the horizontal displacements at the toe of the dike. The next day, a final excavation was made and on the basis of slope stability calculations it was decided to continue by hydraulic loading only. In order to acquire a lot of measurement data, several days were taken to raise the phreatic surface in the sand core and to fill the water tanks on top. Finally, failure occurred on September 8th, at 2:27 pm, after 122.26 hours, see Figure 7.

Table 1 shows the results of slope stability calculations at characteristic moments applying the models of Bishop (1955) and Van (2001). The latter is a geometrically more flexible variant to Bishop’s model. The results correspond well to the deformation behaviour shown in Figure 6: close to the critical value of 1, the deformations quickly increase. These results may even draw some suspicion, but it should be borne in mind that quite advanced soil investigations had been carried out prior to the test (Zwanenburg et al. 2011, Koelewijn and Bennett 2012) and detailed actual measurements of pore pressures were available. Moreover, the model by Bishop has already long ago been described as surprisingly accurate for conditions close to failure (Spencer 1967). Table 2 gives the measured values of the horizontal deformations during the last phase of the test for all

934


inclinometers except one at the East side, which failed. The pre-set deformation criterion for a successful test was exceeded at the moment the maximum pore pressures were recorded.

Table 1. Safety factors calculated for the South dike.

Situation, date, time Van Bishop

Dike completed, June 26, 5:00 pmStart of test, Sept. 3, 12:12 pm Before last excavation, Sept. 5, 9:00 am After last excavation, Sept. 5, 5:00 pm Start of last infiltration, Sept. 8,1:53 pm Maximum pore pressures, Sept. 8, 2:13 pm Visible failure, Sept. 8, 2:27 pm

1.46 1.74 1.24 1.05 1.01 0.92 0.94

1.50 1.82 1.38 1.08 1.05 0.95 0.98

Table 2. Horizontal deformations measured by inclinometers around failure, in mm.

Time East in toe

Middle - crest

Middle in berm

West in berm

West in toe

1:53 pm 2:13 pm 2:27 pm 2:30 pm

115145180225

145190430

1450

160200470

1650

140175310900

135155320830

Figure 6. Horizontal displacements at toe of dike until close to failure.

Figure 7. South dike during failure: fracturing of slope of ditch.

5 PERFORMANCE OF THE MONITORING SYSTEMS

All monitoring systems were judged by their accuracy, range, density of measurements, measurement frequency, redundancy, robustness, time to install and adjust, processing time, interpretation and quality of prediction. Note that several of these factors are not only influenced by the instrumentation, but also by the strategy adopted by the company. It should also be noted that successful application of any technique depends on the actual conditions and environment.

An extensive evaluation of the results by the above criteria indicated a good to excellent performance in these tests of the controllable drainage tubes, the thermic infrared camera system for piping and micro-instability (although faster processing of the measurements seems, in general, a point of improvement),

the tubes measuring strain and temperature profiles (design could be improved) and the ground based SAR (robustness to field conditions could be improved). The other systems performed as expected or worse.

6 PERFORMANCE OF THE INFORMATION SYSTEMS

The information systems were judged by their ability to combine data of different sources, the application of various techniques and methods to arrive at meaningful information, the clarity of statements and the quality of prediction.

Two companies performed well, one employing advanced data driven modelling and anomaly detection to improve finite element calculations, the other one focused more on an engineer’s approach employing both modern technology and visual observations to update their predictions during the test.

The third company restricted its efforts mainly to producing all kinds of graphical presentations of the measured data, but hardly combining data of different sources.

7 CONCLUSIONS

Each of the three test dikes failed according to one of the designed failure modes. Instrumentation of nine companies was tested, indicating a novel technique to measure strain and temperature, a thermic infrared camera system to detect piping and micro-instability and fast ground based SAR as promising new monitoring techniques, as well as a controllable drainage tube capable of preventing failure. Employing monitoring data led to an improvement of the prediction of failure, especially if different types of monitoring were used. Real-time advanced modelling further improves the knowledge on the actual and expected condition of dikes.

8 ACKNOWLEDGEMENTS

Acknowledgements are made to Staatsbosbeheer for providing the test site at Booneschans, the Dutch Ministry of Economic Affairs, Agriculture and Innovation for the financial support and all participating companies for their efforts.

9 REFERENCES

Beek, V.M. van, Knoeff, H. and Sellmeijer, H. 2011. Observations on the process of backward erosion piping in small-, medium- and full-scale experiments, European Journal of Environmental and Civil Engineering 15(8), 1115-1137.

Bishop, A.W. 1955. The use of the slip circle in the stability analysis of slopes. Géotechnique 5 (1), 7-17.

Burenkova, V.V. 1993. Assessment of suffusion in non-cohesive and graded soils, Filters in geotechnical and hydraulic engineering,Brauns, Heibaum, Schuler (editors), Balkema, Rotterdam, 357-360.

Groot, M. de, Mastbergen, D., Bezuijen, A. and Stoutjesdijk, T. 2011. Micro-instability at dike inner slopes, A feeling for soil and water. A tribute to prof. Frans Barends, Van, M., Haan, E. den and Deen, J. van (editors), Deltares Select Series 07/2011, Delft, 65-74.

Koelewijn, A.R. and Bennett, V.G. 2012. Levee failure prediction competition 2012, ijkdijk.rpi.edu.

Spencer, E. 1967. A method of analysis of the stability of embankments assuming parallel inter-slice forces, Géotechnique 17(1), 11-26.

Van, M.A. 2001. New approach for uplift induced slope failure, Proc. XVth Int. Conf. Soil Mech. Geot. Eng., Istanbul, 2285-2288.

Zwanenburg, C., Haan, E.J. den, Kruse, G.A.M. and Koelewijn, A.R. 2012. Failure of a trial embankment on peat in Booneschans, the Netherlands. Géotechnique 62 (6), 479-490.

935

Physical modeling of the vibration mitigation by an isolating screen

Modélisation physique de l'atténuation des vibrations par un écran isolant

Masoumi H., Vanhonacker P. D2S international, Leuven, Belgium

ABSTRACT: The vibrations generated by railway traffic in urban area can be mitigated using the isolating screens. Both experimental and numerical simulations have been used by authors to realize the vibration transmission through the ground and thesoil-barrier interaction. Since a full-scale test is usually expensive and has some difficulties and limitations in terms of the soilconditions and the cost of screen construction, a physical modeling of the problem in small-scale has been proposed. In frame of an European project, a test bench consisting of a soil container and an isolating screen has been fabricated. The container is filled with a very fine sand using the pluviation technique to guarantee the uniformity of the soil conditions and the repeatability of the test. Asmall foundation excited by a shaker at different frequency ranges is used as the vibration source. The soil responses are measured by accelerometers placed on the soil surface at different distances from the source. The isolating efficiency of a concrete screen has been examined. Results of experimental measurements show a reasonable agreement with those obtained by the numerical modeling. RÉSUMÉ : Les vibrations générées par le trafic ferroviaire dans les zones urbaines peuvent être atténuées par un écran antivibratoire.Les simulations expérimentales ou numériques ont été utilisées par les auteurs pour réaliser la transmission des vibrations par le solainsi que l'interaction sol-écran. Tandis qu'un essai à grande échelle est généralement cher et difficile à réaliser en termes deconditions du sol et de coût de construction, une modélisation physique du problème en échelle réduite a été proposée. Dans la cadred'un projet européen, un banc d'essai constitué d'un conteneur de sol et un écran isolant a été fabriqué. Le conteneur est rempli par un sable très fin en utilisant la technique de pluviation afin de garantir l'uniformité des conditions du sol et la répétabilité de l'essai. Unepetite fondation excitée par un excitateur à différentes gammes de fréquences a été utilisée comme source de vibrations. Les réponses du sol sont mesurées par des accéléromètres placés sur la surface du sol à différentes distances de la source. L'efficacité d'isolationd’un écran en béton a été examinée. Les résultats des mesures expérimentales montrent un accord raisonnable avec ceux obtenus par la modélisation numérique.

KEYWORDS: Small-scale test, pluviation, soil-structure interaction, vibration mitigation, isolating screen.

1 INTRODUCTION

To assess the efficiency of isolating screens, besides several numerical computations presented and discussed in the literature (Adam and von Estorff 2005 , François et al. 2010), a few researchers have been focused on experimental tests (Celebi et al. 2010). Since a full-scale test is usually expensive and has some difficulties and limitations in terms of the soil conditions and isolating screen construction, small-scale tests with their flexibility for selecting different soil conditions and screen properties are more relevant. A major difficulty facing the physical modeling of vibration problems in the soil is the repeatability of the test and the replication of the in-situ stress field. Other difficulties for realizing the boundary conditions in the infinity where there are no reflections, may be resolved by selecting an appropriate scale factor or a relevant size for the soil container.

The similarity of the conditions between the model (small-scale) and the prototype (full-scale) is guaranteed by the scaling factors. The scaling factor is defined to extrapolate the relation between the results of the small-scale testing to those of the prototype. These relations represent the effects of the geometric and the stress scale. Three different scale factors between the small-scale model and the prototype can be defined as follows (Altaee and Fellenius 1994), where the subscripts “m” and “p” denote to the model and the prototype, respectively:

(1) the geometric scale ratio N = Lp/Lm, that represents a linear relation between the corresponding dimensions in the full-scale prototype and the small-scale model,

(2) the effective stress scale ratio n = σ′p/ σ′m , that represents the ratio of the effective stress at a certain depth in the prototype to that at the corresponding depth in the model,

(3) the effective stress gradient ratio I = ′p/ ′m , that is the rate of change of stress with depth to that of the prototype. In a conventional physical testing, and for the normal gravity

condition (1g model), the product of the stress-gradient ratio (I) and the geometric scale ratio (N) is equal to unity when n = 1.

However, in a dry soil, the effective stress is equal to , and the scaling factor n is related to the geometrical scaling factor N

such that n = N( ), where and are the unit weights of the soil in the prototype and the model.

In a wave propagation problem (as a dynamic problem), an

additional scaling factor should also be considered for the time

or the frequency to guarantee the similarity of the stress wave

transmissibility in the model and the prototype. In a low strain

dynamic problem (a linear problem) where the influence of the

soil stress condition in the soil behavior can be neglected, the

dimensionless frequency ratio ( ) must be identical in both

small and full scale test, where is the excitation frequency,

and is the wave velocity. Therefore, it can be written that

( )p =( )m (1)

This results in the frequency scaling factor , and for identical wave velocity in model and in the prototype, the frequency scaling factor is equal to the inverse of the geometrical factor N. In table 1, the prototype to model ratio’s

936


for different physical units are presented where identical soil

properties (E, , ) in both model and prototype are

assumed. E, , and are the Young’s modulus, the density and the Poisson’s ratio, respectively. Table 1. Scaling factor for different physical units. Physical unit Prototype/Model

Length W, H, R N

Frequency 1/N

Time N

Velocity 1

Acceleration 1/N

Wavelength N

Dimensionless length W/λ, H/λ, R/λ 1

Since an unbounded half space soil medium is replaced by a container with limited dimensions, radiation conditions at the boundaries cannot be satisfied perfectly. It is well known that body and surface waves lose the most of their energy after traveling some cycles of motion or wavelengths through the soil (after 3 to 4 wavelength) because of the geometric and material damping. So appropriate container dimensions and the excitation frequency must be selected to reduce these effects. In principle, screen dimensions (width and length) are normalized with respect to the shear wavelength to be comparable at different frequencies.

2 TEST BENCH

The test bench consists of the following parts: 1) The container: a demountable box with a floor, and side walls. The walls are made by deformed galvanized-steel plates with 0.75 mm thickness. Interior of the container, side walls and the floor covered by wooden plates to provide a proper smooth surface. 2) The isolating screen: a concrete slab that can be covered by a thin layer of resilient material. The isolating screen is completely embedded in the soil medium. 3) The soil: a sieved, dried fine sand.

2.1 The soil treatment

The container is filled with Mol silica sand with an average grain size (D50) of 0.26 mm. The sand is properly sieved, washed and then dried. The soil treatment should be (1) repeatable, (2) operator-independent, and should results in (3) a tight tolerance in soil conditions (uniformity of the soil density). Investigation in different soil deposition methods (Miura and Toki 1982, Vaid et al. 1999) have shown that the pluviation method is less operator-dependent and more repeatable than the other methods such as the moist tamping, dry tapping (the sand being poured in layers) and pouring using a hand rotated flask. The density of pluviated specimen depends on (1) the fall height, (2) the depositional intensity, and (3) the uniformity of the sand raining. To provide a uniform density, it has been shown that the pluviation device should be raised continuously with a constant low fall height and a constant drop energy. Since an universal device does not exist for the soil deposition by the pluviation, a pluviation device compatible with the container dimensions has been designed and fabricated. The pluviation device consists of three main parts: 1) A tank or reservoir in the upper level to deliver the sand through a nuzzle, 2) A shutter that can be in open/close position to control the deliverance of sand. Shutter consists of a fixed perforated plate and a sliding plate. Opening the sliding plate let the sand to pass through the holes in the perforated plate.

3) A diffuser consisting of a guide box with two grids (sieves). The second grid is posed at the lower part of device and its holes have different direction to polarize the drop. The sand delivers from the top by its gravity through the opened-shutter. The minimum fall height can be modified by changing the position of the diffuser respect to the soil surface.

Figure 1. The sand deposition by the pluviation technique.

As the rate of the sand flow increases, because of the air turbulence, a non-flat surface of the sand is generated. This effect generally happens when the sand is raining with a high flow rate and higher fall height. The air turbulence and non-uniformity can be controlled by reducing the fall height as well as by decreasing the depositional intensity. The sand deposition has been performed using the pluviation device that moves on two rails over the container with a speed of approximately 0.18 to 0.2 m/s. The height of the sand drop is varied from 15 to 20 cm, figure 1.

2.2 Investigation on sand properties

The sand density has been measured conventionally during pluviation by posing small cylindrical receptacles in different depth. To measure in-situ density, a total of 16 receptacles were installed at different depths: 40 cm, 20 cm and at the surface of the sand. The receptacles were distributed along the container width at distances of 40 cm and 60 cm from the container sidewalls. Results show an average density of 1640 kg/m3 near to the surface, 1685 kg/m3 at 20 cm in depth and 1700 kg/m3 at 40 cm in depth. In addition, upon completion of the soil pluviation, the uniformity of the soil stiffness (at the top layer) is examined by the impedance test. The test configuration consists of a small steel foundation, two accelerometers installed on the foundation, and a hammer, figure 2. The foundation response due to several hammer impacts is measured. A set of points on the sand surface has been selected for the impedance test. Figure 3 shows the mobility function of the foundation measured due to the impact hammer test. Results show a resonance frequency range from 120 to 130 Hz at different measurement points. Since the foundation is rigid, the dynamic foundation-soil system can be modeled with a dynamic system with a single degree of freedom with the foundation mass and the soil stiffness.

937


Figure 2. Impedance test using a hammer impact on the foundation. So, the equation of motion of the system can be written as:

(1)

, where and denote to the real and imaginary part of the vertical soil stiffness, and is the mass of the foundation.

Figure 3. Mobility functions of the foundation.

At low dimensionless frequencies < 0.25, the soil stiffness is approximately equal to the static vertical impedance of a rigid foundation, underlying on a homogeneous half space, (Sieffert and Cevaer 1992):

(2)

, where is the shear modulus of the soil and is the foundation ratio. Using a curve-fitting technique based on the least-square method, each parameters of the equation of motion (1) can be identified. Therefore, the shear modulus of the upper layer of

the sand can be determined. For = 5 cm, ν= 0.33, the shear modulus in the center area of the container is almost uniformly distributed with an average value of 12.5 MPa. Near the sidewalls, however, the non-uniform distribution is observed. The sand properties (the density and the shear modulus) measured by the density test and the impedance test will be used in the numerical modeling.

3 MEASUREMENT SETUP

The isolating screen is installed at the middle of the container. The screen is a concrete plate of 2.0 m x 0.4 m x 0.04 m. The measurement configuration consists of a small foundation posed on the soil surface where the dynamic force is applied and 10 accelerometers placed at the measurement points. The small foundation is excited at the frequency band of interest and the free field vibrations are measured symmetrically on both sides of the foundation, figure 4. This configuration enables us to

simultaneously measure the non-isolated responses (on the side without the screen) and the isolated responses (on the side where the screen is installed).

Figure 4. Overview of the measurement setup (the section view).

A shaker device is used for the excitation generation. The type, the amplitude and the frequency content of the excitation can be controlled by means of a wave generator software that feeds into a power amplifier, figure 5.

Figure 5. Shaker device and acquisition system.

A random vibration from 100 to 900 Hz is used. To obtain a reasonable coherency, the excitations were applied for a period of at least 3 minutes. Based on four frequency ranges, four separate measurements were performed. Figure 6 shows the configuration of the measurements for the concrete barrier test. The efficiency of the isolating barrier is determined by introducing the insertion loss factor.

Figure 6. Measurement configuration for the concrete barrier test.

The insertion loss is defined using the peak particle velocity (PPV) obtained at each measurement points.

(9)

where the peak particle velocity (PPV) is defined as the maximum value of the impulse response function (IRF) at each measurement points. Induced vibration due to railways traffic is mostly dominated in a frequency range from 10 to 60 Hz. According to the frequency

938


range of interest and the dimension of the container, a geometrical scaling factor of 15 would be suitable.

A 2.5-dimensional coupled FE-BE method is used for modeling of the problem. In 2.5-dimensional modeling, a longitudinally invariant geometry of the structure (the barrier) is assumed. In this methodology, 2.5D FEM is used to model the structure (the screen) and the soil impedance as well as the free field vibrations are computed by means of 2.5D BEM. This methodology has been already examined for different applications such as railway tracks, roads, tunnels, dams, trenches, and pipelines by François et al. (2010).

Table 3 Frequency ranges in the full scale and the small-scale test. Figure 8. Average insertion loss at different excitation frequencies. Small-scale test N=15 Full-scale test

100-300 [Hz] 6.7 – 20 [Hz]

300-500 [Hz] 20 - 33.3 [Hz]

500-700 [Hz] 33.3 - 46.7 [Hz]

700-900 [Hz] 46.7 – 60 [Hz]

4 CONCLUSIONS

A test bench has been fabricated for the examination of the isolating screen efficiency. Results of the small-scale test show a reasonable agreement with those obtained by the numerical modeling. This confirms the accuracy of the numerical prediction for further investigation.

The barrier has a width of 0.04*15 = 0.60 m and a depth of 0.4*15 = 6 m. Table 3 shows the frequency ranges that have been applied for the measurements. The same soil properties as obtained in the test bench are considered. A soft layer over a homogeneous half space is considered. The soil characteristics are presented in table 4. The soil has a material damping of 5%.

In addition, results show that the selected concrete screen with a

depth of 6 m ( ) is not efficient enough to mitigate the vibrations at frequencies lower than 30 Hz. At higher

frequencies where is greater than one, however, higher efficiency has been obtained.

5 ACKNOWLEDGEMENTS

The results presented in this paper have been obtained within the frame of EUROSTAR SOILVIBES project "Railways vibration mitigation in transmission path".

Table 4 Soil properties in numerical modeling.

Layer depth Young’s modulus Density Shear wave velocity

3 [m] 33.5 [MPa] 1660 kg/m³ 85 [m/s]

65 [MPa] 1690 kg/m³ 120 [m/s]

This project is funded by IWT Vlaanderen, the Institute of the Promotion of Innovation by Science and Technology in Flanders. Their financial support is gratefully acknowledged.

Figure 7 shows the variation of the insertion loss versus the distance. Results of the experimental test bench (dark line) are compared with those of the numerical modeling (gray line).

6 REFERENCES

The frequency range as well as the distance from the source is presented in the real scale.

Adam M. and von Estorff O. 2005, Reduction of train-induced vibrations by using open and filled trenches. Computers and Structures, 83:11–24.

Altaee A. and Fellenius B.H. 1994, Physical modeling in sand. Canadian geotechnical journal 31, 420-431.

Celebi E., Firat S., Beyhan G., Cankaya I., Vural I., and Osman K.. 2009, Field experiments on wave propagation and vibration isolation by using wave barriers. Soil Dynamics and Earthquake Engineering, 29:824–833.

François S., Schevenels M., Galvin P., Lombaert G., and Degrande G.. 2010, A 2.5D coupled FE-BE methodology for the dynamic interaction between longitudinally invariant structures and a layered half space. Computer methods in applied mechanics and engineering, 199(23-24):1536 – 1548.

Garnier J., Gaudin C., Springman S.M., Culligan P.J., .Goodings D, Konig D., Kutter B., Phillips R., Randolph M.F., and Thorel. L. 2007, Catalogue of scaling laws and similitude questions in geotechnical centrifuge modelling. International Journal of Physical Modelling in geotechnics, 7(3):1–24.

Miura S. and Toki S. 1982, A sample preparation method and its effect on static and cyclic deformation-strength properties of sand. Soils and Found., 22(1):61–77.

Sieffert J.G. and Cevaer F. 1991, Handbook of impedance functions, surface foundations, Ouest editions.

Vaid Y. P., Sivathayalan S., and Stedman D. 1999, Influence of specimen-reconstituting method on the undrained response of sand. Geotechnical Testing Journal, 22(3):187–195.

Figure 7. Insertion loss versus distance from the source.

An average insertion loss can also be calculated for each barrier over all distances. A reasonable agreement between the experimental and numerical simulation is observed, figure 8.

939

The Drained Strength of Soft Clays with Partially Penetrating Sand Columns at Different Area Replacement Ratios

La résistance drainée des argiles molles avec des colonnes de sable pénétrant partiellement à différents taux de remplacement

Najjar S., Sadek S., Bou Lattouf H. Department of Civil and Environmental Engineering, American University of Beirut

ABSTRACT: Granular columnar inclusions are generally used to improve the mechanical properties of soft clays. The objective ofthis paper is to investigate the long term behavior of clay/sand column composites as represented by the fully drained loadingcondition, for cases where the soft clay is reinforced by floating or partially penetrating sand columns. For this purpose, consolidateddrained triaxial tests (CD) were performed on back-pressure saturated normally consolidated Kaolin specimens that were consolidatedand tested at confining pressures of 100 kPa, 150 kPa, and 200 kPa. The sand columns were penetrated to 75% of the depth of the claysample to represent a partially penetrating condition, while the main parameter that was varied in the study was the area replacementratio which was varied from 7.9% to 17.8% to 31.2%. Results indicated that the positive effects of sand columns on strength areminimal for small area replacement ratios and increase gradually as the area replacement ratio increases. The average percentimprovements observed for area ratios of 17.8% and 31.2% were 20% and 32%, respectively. These results indicate that partially penetrating columns may provide effective strengthening for soft clays, provided that a suitable area replacement ratio is adopted indesign.

RÉSUMÉ: L’inclusion de colonnes granulaires est généralement adoptée pour améliorer les propriétés mécaniques des argiles molles.L'objectif de cet article est d'étudier le comportement à long terme des composites d'argile / sable, représentée par la condition dechargement complètement drainée, pour les cas où l'argile molle est renforcée par des colonnes de sable partiellement pénétrantes. A cet effet, des essais triaxiaux drainés consolidés (CD) ont été réalisées sur des spécimens de Kaolin saturés et normalement consolidés. Des échantillons consolidés à des pressions de confinement de 100 kPa, 150 kPa et 200 kPa ont été testés. Les colonnes de sable ont été établis jusqu’à à 75% de la profondeur de l'échantillon d'argile pour représenter un état partiellement pénétrant, tandisque le principal paramètre qui a été modifié dans l'étude était le taux de remplacement qui a varié de 7,9% à 17,8% à 31,2% de lasurface de section du spécimen. Les résultats indiquent que les effets positifs de colonnes de sable sur la résistance sont minimes pour les petits taux de remplacement et augmentent progressivement avec l'augmentation du taux de remplacement. Les améliorations observées en moyenne pour des rapports de surface de 17,8% et 31,2% étaient de 20% et 32%, respectivement. Ces résultats indiquentque les colonnes partiellement pénétrantes pourraient être utilisées pour renforcer les argiles molles, à condition que le rapport deremplaçant approprié soit choisi dans le design.

KEYWORDS: soft clay, sand columns, stone columns, consoldiated drained triaxial tests, soil improvement, floating columns

1 INTRODUCTION

Granular columnar inclusions in the form of sand drains/columns or vibrated stone columns are commonly used to improve the mechanical properties of soft clays. Historically, experimental research studies have been designed to investigate the behavior of sand/stone column-reinforced clay systems in the laboratory using 1-g tests that are conducted in one dimensional loading chambers (Hughes and Withers 1974, Muir Wood et al. 2000, Malarvizhi & Ilamparuthi 2004, McKelvey et al. 2004, Ayadat and Hanna 2005, Ambily & Gandhi 2007, Gniel & Bouazza 2009, Murugeson & Rajagopal 2010, and Fattah et al. 2011).

The limitations of 1-g model tests were recognized by many researchers who resorted to testing soft clay specimens that were reinforced with sand/stone columns under triaxial conditions where the stress state, the drainage conditions, and the loading rate could be controlled. Examples of such studies include the work reported in Juran and Guermazi (1988), Sivakumar et al. (2004), Black et al. (2006), Black et al. (2007), Andreou et al. (2008), Najjar et al. (2010), Black et al. (2011), and Sivakumar et al. (2011).

For cases involving sites with deep deposits of soft clay, the use of sand/stone columns that fully penetrate the soft clay layer is prohibitive and may not be practically achievable. As a result,

the use of partially penetrating columns is common as a practical soil improvement scheme.

Current design methods for stone columns do not reflect the effect of the degree of column penetration in the soft clay on the response of clay/stone column system. As a result, there is a need for investigating the behavior of clays with partially penetrating columns using an experimental framework in which the stress state and the drainage conditions could be controlled.

The objective of this paper is to investigate the load response of soft clay that is reinforced with partially penetrating sand columns in a triaxial framework. The parameters that were varied in the experimental program are the area replacement ratio which was varied from 7.9% to 17.8% to 31.2% and the effective confining pressure which was varied from 100 to 150 to 200 kPa. All tests were conducted using columns that penetrated the soft clay to a depth that is equal to 75% of the height of the clay sample. Since the sand columns are expected to act as drains that will facilitate radial drainage, fully drained tests were conducted to represent the long term behavior of the clay/stone column system and to provide an upper bound of the response for practical loading conditions in the field where the clay surrounding the columns is expected to be partially drained.

940


2 EXPERIMENTAL PROGRAM

In total, 12 isotropically consolidated drained (CD) triaxial tests were performed on consolidated kaolin specimens having a diameter of 7.1 cm and a length of 14.2 cm. Tests were conducted on control specimens and specimens that were reinforced with single sand columns having diameters of 2cm, 3cm, and 4cm with a column penetration ratio Hc/Hs of 0.75. The 2-cm, 3-cm, and 4cm diameter columns represent area replacement ratios Ac/As of 7.9%, 17.8%, and 31.5% respectively. All sand columns were placed in pre-drilled holes in the center of the clay specimens. All specimens were saturated using a back pressure of 310 kPa and isotropically consolidated under effective confining pressures of 100, 150, or 200 kPa. In all tested specimens, the measured “B” value was greater than 0.96 indicating an adequate degree of saturation. Samples were then sheared in drained conditions at a strain rate of 0.25% per hour (~0.06mm/min). All tests were terminated at a maximum axial strain of about 12%.

2.1. Material Properties

The clay used in the testing program is a kaolin clay with a liquid limit of 55.7%, a plasticity index of 22.4%, and a specific gravity of 2.53. Consolidation and strength properties for the clay are presented in Najjar et al. (2010). Ottawa sand which classifies as poorly graded sand (SP) according to the Unified Soil Classification System was used to construct the sand columns. For sand specimens prepared at a dry density of 16.2 kN/m3 (relative density of 44%), Najjar et al. (2010) reported an effective peak friction angle of 33o based on consolidated undrained triaxial tests with pore pressure measurement. In this study, isotropically consolidated drained triaxial tests were conducted on sand specimens with a height of 14.2 cm and a diameter of 7.1 cm at confining pressures of 100, 150, and 200 kPa to determine the friction angle of the sand. The resulting effective friction angle was found to be equal to 35o. The difference between the measured effective friction angles from the CU+U and CD tests could be attributed to the respective mean effective stresses at failure which were an order of magnitude greater for the undrained tests.

2.2. Sample Preparation

Kaolin clay powder was mixed with water at a water content of 100% (i.e. 1.8 times its liquid limit) to form a slurry. The slurry was then poured into custom-fabricated consolidometers in preparation for one-dimensional consolidation. Dead weights were used to consolidate the specimens from slurry to a vertical effective stress of 100 kPa. The water content at the end of the consolidation stage was relatively uniform (~53%) throughout the depth of the sample. The average bulk density for all the clay specimens prepared was about 16 kN/m3. A detailed description of the sample preparation and testing procedure is presented in Najjar et al. (2010).

The sand columns were formed from Ottawa sand at a dry density of about 16.2 kN/m3. These sand columns were prepared by pouring 3 layers of dry Ottawa sand in cylindrical pre-cut and stitched geosynthetic fabrics. The fabrics were initially inserted in a glass tube having the same inner diameter as the sand column, and the sand layers were densified by vibration. Water was then added to the sand column to reach a water content of about 20%. The saturated sand column was then frozen for 24 hours (Fig. 1a). The geosynthetic fabric was cut and detached from the sand column. The frozen sand column was then inserted into a hole drilled at the center of the clay specimen (Fig. 1b) and allowed to thaw. The reinforced clay specimen (Fig. 1c) was then transferred to the triaxial cell and saturated using a back pressure of 310 kPa.

Figure 1. Installation process of sand columns.

3 TEST RESULTS AND ANALYSIS

The automated triaxial test setup “TruePath” by Geotac was used to conduct CD tests on control and reinforced clay specimens saturated at a back pressure of 310 kPa. The samples were then isotropically consolidated under confining pressures of 100, 150, or 200 kPa and sheared drained at a strain rate of 0.25% per hour, while measuring volume change through drain lines connected to the porous stones at the top and bottom of the sample. The measured volume change reflects a global change in the composite sample and do not provide information on local changes in the water content in the sand column and the surrounding clay. Throughout the tests, the total confining pressure was kept constant as the vertical stress was increased in compression.

3.1. Mode of Failure

The mode of failure was characterized by bulging of the clay specimen. The bulging was slight and relatively uniform along the height in samples reinforced with the smallest area replacement ratio of 7.9% (see Fig. 2a). As the area replacement ratio increased, the bulging was significant and concentrated in the lower half of the clay specimen, indicating stress and strain concentration in the unreinforced portion of the specimen. For the largest area replacement ratio of 31.2%, clearly defined shear planes formed in the lower half of the sample as indicated in Fig. 2c.

To investigate the mode of failure of the sand columns, the same test specimens were split along their vertical axes to expose the columns and the surrounding clay (Figs. 2a-2c). The figures indicate that relatively uniform bulging of the sand columns occurred with depth, with the specimens at the higher area replacement ratios showing signs of punching of the sand columns into the unreinforced clay.

3.2. Stress-Strain Response

The variation of the deviatoric stress and volumetric strain with axial strain is presented in Figs. 3, 4, and 5 for tests with replacement ratios of 7.9, 17.8, and 31.2%, respectively. The stress-strain curves exhibited consistent increases in deviatoric stresses with strains as the samples were sheared towards critical state conditions. In this paper, failure is defined at an axial strain of 12%, which is the maximum strain measured.

(a) Ac/As= 7.9% (b) Ac/As= 17.8% (c) Ac/As= 31.2% Figure 2. Internal and external modes of failure.

941


Figure 3. Variation of deviatoric stress and volumetric strain with axial strain (Ac/As=7.9%).

Results presented in Figs. 3 to 5 indicate that for the smallest area replacement ratio used in this study (Ac/As=7.9%), no improvement was evident in the stress-strain response compared to the control clay specimen. In fact, slight reductions in the load-carrying capacity were measured at all levels of strain. For the higher area replacement ratios, clear and consistent improvements in the stress-strain response were observed for all effective confining pressures. With regards to the volumetric strains, results indicate that the measured volumetric strains were contractive for all the specimens tested. However, the volumetric strain at failure was found to decrease for specimens that were reinforced with 3-cm and 4-cm sand columns (higher area replacement ratios), compared to control clay specimens and specimens reinforced with 2-cm columns. This reduction in volumetric strains at failure for the reinforced clay specimens is expected and is due to the dilative nature of the sand columns, particularly at higher area replacement ratios.

3.3. Effect of Sand Columns on Deviatoric Stress at Failure

The percent improvement in the deviatoric stress at failure for the series of tests involving area replacement ratios of 7.9%, 17.8%, and 31.2% was calculated and presented in Fig. 6 as a function of the effective confining pressure. Results in Fig. 6 indicate that the use of 2-cm diameter sand columns (area ratio of 7.9%) did not result in increases in the deviatoric stress at failure. For the higher area replacement ratios of 17.8% and 31.2%, improvements ranging from 17% to 25% and from 28% to 38% were observed in the deviatoric stress at failure, respectively.

The calculated percent improvement in the deviatoric stress at failure was also plotted in Fig. 7 as a function of the area replacement ratio. Interestingly, the results in Fig. 7 indicate that for the smallest effective confining pressure of 100 kPa, the percent improvement increased at the same rate as the area replacement ratio was increased from 7.9% to 17.9% to 31.2%. For the tests conducted at the higher confining pressures of 150 kPa and 200 kPa, the rate of improvement in the deviatoric stress at failure decreased as the area replacement ratio was increased from 17.9% to 31.2%. This decrease in the percent improvement could be attributed to the mode of failure observed for the samples reinforced at an area replacement ratio


-5

-4

-3

-2

-1

00 2 4 6 8 10 12

Vol

umet

ric S

train

(%)

050

100150200250300350

0 2 4 6 8 10 12

Dev

iato

ric s

tress

(kP

a)

Axial Strain (%)

controlArea Replacement = 7.9%Area Replacement = 17.8%Area Replacement = 31.2%

-5

-4

-3

-2

-1

00 2 4 6 8 10 1

Vol

umet

ric S

train

(%)

2

050

100150200250300350

0 2 4 6 8 10 1

Dev

iato

ric s

tress

(kP

a)

Axial Strain (%)2

ControlArea Replacement = 7.9%Area Replacement = 17.8%Area Replacement = 31.2%

of 31.2% and tested at confining pressures of 150 kPa and 200 kPa. For these cases, clear shear planes formed at the bottom of the sample (see Fig. 2c) indicating a possible premature failure in the lower-half of the sample due to elevated stresses in the sand columns that are bearing on the unreinforced clay.

3.4. Effect of Sand Columns on Shear Strength Envelope

Figure 8 shows the effective Mohr-Coulomb envelopes corresponding to the different area replacement ratios used in this study. As expected the Mohr-Coulomb failure envelope for the specimens reinforced with the smallest area replacement ratio of 7.9% was almost identical to that of the control clay, with an effective apparent cohesion c’ = 0 kPa and an effective friction angle of about 21o.


-5

-4

-3

-2

-1

00 2 4 6 8 10 1

Vol

umet

ric S

train

(%)

2

050

100150200250300350

0 2 4 6 8 10 12

Dev

iato

ric s

tress

(kP

a)

Axial Strain (%)

controlArea Replac. = 7.9%Area Replac.= 17.8%Area Replac. = 31.2%

942


-10

0

10

20

30

40

50

100 150 200

Per

cent

Impr

ovem

ent i

n D

evia

toric

stre

ss (k

Pa)

Effective Confining Pressure (kPa)

Area Ratio = 7.9%Area Ratio = 17.8%Area Ratio = 31.2%

0

50

100

150

0 50 100 150 200 250 300

She

ar S

tress

(kP

a)

Ef fective Normal Stress (kPa)

Ac/As= 31.2% (c'=18 kPa, '=21o)

Ac/As=7.9%(c'=0 kPa, '=20.7o)

Controlc'=0 kPa, '=21o)

Ac/As= 17.8(c'=0 kPa, '=23o)

Figure 6. Dependency of the percent improvement in deviatoric stress at failure on the effective confining pressure

Figure 8. Failure envelops for control and reinforced clay specimens.

5. ACKNOWLEDGEMENTS

Figure 7. Dependency of the percent improvement in deviatoric stress at failure on the area replacement ratio.

-10

0

10

20

30

40

50

5 10 15 20 25 30 35

Per

cent

Impr

ovem

ent i

n D

evia

toric

stre

ss (k

Pa)

Area Replacement Ratio (%)

100 kPa150 kPa200 kPa

The authors would like to acknowledge the support of the University Research Board (URB) at the American University of Beirut for funding this research program.

6. REFERENCES

Alamgir M., Miura N., Poorooshasb H.B. and Madhav M.R. 1996. Deformation analysis of soft ground reinforced by columnar inclusions. Computers and Geotechnics 18(4), 267-290.

Ambily A.P. and Gandhi S.R. 2007. Behavior of stone columns based on experimental and FEM analysis. Journal of Geotechnical and Geoenvironmental Engineering 133(4), 405-415.

For specimens reinforced with an area replacement ratio of 17.8%, the friction angle ' was found to increase to 23o

(compared to 21o for the control clay) with the effective cohesion intercept c’ remaining at zero. On the other hand, samples with an area replacement ratio of 31.2% showed no improvements in the friction angle compared to the control specimens ('=21o), but were associated with a non-zero c’value of 18 kPa. The non-zero c’ could be related to the relative reduction in the percent improvement in the deviatoric stress for samples tested at the higher confining pressure of 150 kPa and 200 kPa as indicated in Figs. 6 and 7.

Andreou P., Frikha W., Frank R., Canou J., Papadopoulos V. and Dupla J.C. 2008. Experimental study on sand and gravel columns in clay. Ground Improvement 161(4), 189-198.

Ayadat, T. and Hanna, A. M. (2005). “Encapsulated stone columns as a soil improvement technique for collapsible soil.” Ground Improvement, 9(4), 137-147.

Bauman V. and Bauer G.E.A. 1974. The performance of foundations on various soils stabilized by the vibro-compaction method. Canadian Geotechnical Journal 11(4), 509-530.

Black J., Sivakumar V., Madhav M.R., and McCabe B. 2006. An improved experimental set-up to study the performance of granular columns. Geotechnical Testing Journal, 29(3), 193-199.

Black J.V., Sivakumar V. and McKinley J.D. 2007. Performance of clay samples reinforced with vertical granular columns. Canadian Geotechnical Journal 44, 89-95.

Black J.V., Sivakumar V. and Bell A. 2011. The settlement performance of stone column foundations. Geotechnique 61(11), 909-922. 4. CONCLUSIONS

Based on the results of 12 consolidated drained triaxial tests the following conclusions can be drawn on the effect of partially penetrating sand columns on the drained response of soft clay:

Fattah M.Y., Shlash K.T. and Al-Waily M.J.M. 2011. Stress concentration ratio of model stone columns in soft clays. Geotechnical Testing Journal 34(1), 1-11.

1. The mode of failure of the test specimens was governed by bulging that was relatively uniform for specimens reinforced at a small area replacement ratio of 7.9% and concentrated in the lower half of the specimens for the higher area replacmenent ratios of 17.8% and 31.2%. For an area ratio of 31.2%, specimens tested at confining pressures of 150 kPa and 200 kPa exhibited clear shear planes in the lower half of the specimens indicating elevated stress concentrations in the unreinforced clay.

Gniel J. and Bouazza A. 2009. Improvement of soft soils using geogrid encased stone columns. Geotex. and Geomem. 27(3), 167–175.

Hughes J.M.O. and Withers N.J. 1974. Reinforcing of soft cohesive soils with stone columns. Ground Engineering 7(3), 42-49.

Juran I. and Guermazi A. 1988. Settlement response of soft soils reinforced by compacted sand columns. Journal of Geotechnical Engineering 114(8), 930–943.

Malarvizhi S.N. and Ilamparuthi K. 2004. Load versus settlement of claybed stabilized with stone and reinforced stone columns. Proceedings of Geo-Asia-2004, Seoul, Korea, 322-329.

2. The specimens tested with the lower area replacement ratio of 7.9% did not show any improvement in the load carrying capacity. For the higher area replacement ratios of 17.8% and 31.2%, average improvements of 20% and 32% were observed in the deviatoric stresses at failure, respectively. For the higher confining pressures of 150 kPa and 200 kPa, the rate of improvement in the deviatoric stress at failure was found to decrease as the area replacement ratio was increased from 17.9% to 31.2%. This could be due to the premature formation of shear planes in the lower half of the specimens

McKelvey D., Sivakumar V., Bell A. and Graham J. 2004. Modeling vibrated stone columns in soft clay. Proceedings of the Institute of Civil Engineers Geotechnical Engineering 157(3), 137-149.

Muir Wood D., Hu W. and Nash D.F.T. 2000. Group effects in stone column foundations: Model tests. Geotechnique 50(6), 689-698.

Murugesan S. and Rajagopal K. 2010. Studies on the behavior of single and group of geosynthetic encased stone columns. Journal of Geotechnical and Geoenvironmntal Engineering 136(1), 129-139.

Najjar S.S., Sadek S. and Maakaroun T. 2010. Effect of sand columns on the undrained load response of soft clays. Journal of Geotechnical and Geoenvironmental Engineering 136(9), 1263-1277.

3. An analysis of the Mohr-Coulomb envelopes indicated that for an area replacement ratio of 17.8%, ' increased from 21o (control clay) to 23o, while for an area replacement ratio of 31.2%, c’ increased from 0 (control clay) to 18 kPa with ' remaining constant at 21o.

Sivakumar V., McKelvey D., Graham J. and Hughus D. 2004. “Triaxial tests on model sand columns in clay. Canadian Geotechnical Journal 41, 299-312.

Sivakumar V., Jeludine D.K.N.M., Bell A., Glyn D.T. and Mackinnon P. 2011. The pressure distribution along stone columns in soft clay under consolidation and foundation loading. Geotechnique 61(7), 613-620.

943

Physical modeling of arch action in undercut slopes with actual engineering practice to Mae Moh open-pit mine of Thailand

Modélisation physique de l'effet de voûte dans les pentes en déblai en suivant la pratique de l'ingénieur pour la mine à ciel ouvert à Mae Moh en Thaïlande

Pipatpongsa T. Global Scientific Information and Computing Center, Tokyo Institute of Technology

Khosravi M.H., Takemura J. Department of Civil and Environmental Engineering, Tokyo Institute of Technology

ABSTRACT: In the field of mining engineering, a stable arch formed across a pit is beneficial to the design of an undercut slope; therefore, prediction of the maximum stable undercut width under which the slope does not collapse is needed. The relation between a stable width and an inclined angle has been obtained experimentally to confirm the developed theoretical relations. A series of simple experiments using a block of compacted moist sand confined by parallel rigid walls has been conducted by varying the thickness, width and length. The actual engineering application was immediately tested at the Mae Moh open-pit mine in Thailand. The factor of safety in fully saturated condition with hydro-static water pressure on bedding shear was evaluated. It is concluded that this novel procedure in mining is practically realizable and results in reductions in massive excavation, transportation and dumping of unstable rock mass, as well as saving an amount of time and expense.

RÉSUMÉ : En génie minier, la formation d'une voûte stable au-dessus de l'excavation fournit un avantage pour la conception d'une pente en déblai; par conséquent, la prédiction de la largeur du déblai maximale, qui ne provoque pas d'effondrement, est nécessaire.La relation entre une largeur stable et un angle d'inclinaison a été obtenue expérimentalement pour confirmer les relations théoriques.Une série d'expériences simples utilisant un bloc de sable humide compacté confiné par des mur rigides parallèles a été menée envariant l'épaisseur, la largeur et la longueur. La pratique de l'ingénieur est immédiatement appliquée pour la mine à ciel ouvert à MaeMoh en Thaïlande. Le coefficient de sécurité dans un état complètement saturé avec la pression hydrostatique sur la zone decisaillement a été évalué. Il est conclu que cette nouvelle procédure de l'exploitation minière est réalisable, entraînant une réduction duvolume d'excavation, du transport et du déversement de la masse rocheuse instable, ainsi qu'une économie de temps et d'argent.

KEYWORDS: arching effect, undercut slope, excavation, physical model, open-pit mining. 1 INTRODUCTION

Evaluating the stability of slopes is one of the most important activities in geotechnical engineering. The existence of a stable scarp in some slope failures along oblique faults can be evidence of an arching effect in those slopes. Pipatpongsa et al. (2009) reported the existence of some stable scarps in huge slope failures in the Mae Moh coal mine. Exposed scarps of a slope failure remains stable if the material has a sufficient strength to resist the load transferred to the stable adjoining parts. This phenomenon of load transfer from the yielding part of the material to the adjacent stationary parts is known as the arching effect (Janssen 1895 and Terzaghi 1936).

300 m

Unstable area

A

A’

Figure 1. Unstable Area 4.1 (as of September 2010)

The relation between a stable width and an inclined angle has been obtained experimentally to confirm the developed theoretical relations (Khosravi 2012) for (a) a strip arch with slip failure in laterally supported sand blocks, (b) a segmented arch with arch-shaped failure in mild undercut slopes and (c) a circular arch with buckling failure in steep undercut slopes. For the design purpose of undercut mining, this particular relation provides the maximum span of the undercut where load could laterally be transferred to vertical planes of a neighboring rock mass. A case study of an undercut slope at the Mae Moh open-pit mine in Thailand is presented.

2 SITE DESCRIPTION

The actual engineering application of the developed theory was immediately tested at the Mae Moh open-pit mine in Thailand. The Mae Moh open-pit lignite mine primarily supplies coal to generate electrical power in Thailand. The mine, under the operation of the Electricity Generating Authority of Thailand (EGAT) since 1952, is located approximately 630 km north of Bangkok in Lampang province. Currently, the annual production of the mine is about 16−17 million tons/year with a volume of excavated overburden of around 60−80 million m3/year. Its pit has a maximum width of about 4 km and a maximum length of about 9.5 km. Green clay in the bedding shear zone has caused problems of various scales. Slopes in the Mae Moh mine are prone to plane failure once they dip out of the slope face and strike parallel to the bedding shears.

944


According to the full core drilling, which consists of up to 900 holes have been drilled in the Mae Moh mine during the period 19872005. The thickness of the bedding shear zone ranges from 10 to 80 mm. About 40% of the bedding shear zone comprises continuous layers of clay seam. In this study, the targeted area is Area 4.1, shown in Fig.1. Its cross-section is shown in Fig.2.

Potential failure plane(Shearing zone, green clay)

Lignite layer

Unstable rock mass

Borehole NEI N29 (Depth: 50m)

A A’

Figure 2. Cross-section A-A’ of Area 4.1 (Courtesy of EGAT)

Block of moist Silica sand No. 6

Position of block before slippage

αf

Figure 3. Slippage of the sand block along the lateral supports (after Khosravi et al. 2012)

Side supports

Side supports

α=40o

T=0.05 m

Figure 4 Arch-shaped failure in mild undercut slopes at the maximum undercut span (after Khosravi et al. 2012)

Side supports

Side supports

α=60o

T=0.05 m

Figure 5. Buckling failure in steep undercut slopes at the maximum undercut span (after Khosravi et al. 2012)

3 PHYSICAL MODEL

Khosravi et al. (2009, 2010, 2011) have conducted a series of simple experiments using a block of compacted moist sand confined by parallel rigid walls by varying thickness, width and length. The inclined angle of the bedding plane was gradually

increased until the block started to slip (see Fig.3). Also, some laboratory-scale undercut slope physical model tests were conducted under both 1G and centrifugal acceleration fields. The existence of passive arching effects in the slope models can be confirmed by means of earth pressure recordings and image processing techniques. In the undercut slopes, some parts of the load are transferred from the yielding portion of the slopes to the stiffer sides. The level of load transfer depends on the stiffness and strength of the lateral supports. Two types of slope failures can be expected: an arch-shaped failure (see Fig.4) in the central part of the slope for the strong sides, and side buckling (see Fig.5) leading to total failure of the slopes for the weak sides. In addition, the performance of a counterweight balance, which is considered a technique to stabilize undercut slopes with weak sides, was demonstrated through a series of physical models and confirmed that a wider undercut span in front of the slope can be realized (Khosravi et al. 2012).

4 THEORETICAL BACKGROUND

In chemical engineering, a stable arch formed across the orifice of a hopper causes difficulty in discharging of cohesive material; therefore, determination of the minimum diameter which destabilizes the arch action is required. On the other hand, in mining engineering, a stable arch formed across a pit is beneficial to the design of an undercut slope; therefore, prediction of the maximum undercut width which does not cause it to collapse is needed. Jenike’s (1961) model for arch formation has laid the foundation for understanding the behavior of a static system of cohesive materials confined by hopper walls (Walker 1966 and Walters 1973). This study extends a basic idealization of a stationary system used by Jenike (1961) to the stability of a laterally confined rigid block inclining on a stiff bedding plane. The following similar assumptions were adopted in the present study with an additional consideration of interface resistance: (a) the resistance supporting the arch is characterized by unconfined compressive strength, and (b) the load breaking the arch is due to its own weight and to the force exerted by the material above the arch. The mechanism involved and its implication on instability can be explained in that if the load induced by weight of the arch is greater than the unconfined compressive strength and the interface resistance, the arch will collapse and therefore the widest possible span or the failure width of block Bf of a stable arch can be predicted.

The authors (Khosravi 2012) have recently developed equations to describe the instability phenomena of undercut slopes based on Jenike’s (1961) theory of cohesive arching in hoppers, as shown in Eq.(1) which can be alternatively expressed by Eq.(2) in terms of the inclined angle at failure f for a given span of undercut B.

sin tan cosc

fi i

kBc T

(1)

1sin cosi cf i

c kT B

i

(2)

where α: inclined angle, T: thickness of block, i: interface friction angle, ci: interface adhesion, c: unconfined compressive strength, : bulk unit weight, : friction angle of material, k: arching coefficients:

k=0 no arching k1=cos strip arch with soil slip k2=1 segmented arch with stable scarp k3=4/π circular arch with slope buckling

The arching effect is the ability of soil to transfer load

945


laterally to a more rigid lateral/basal base by trajectories of the major principal stress. Failure happened along the shear plane generated by relative displacement. Because the undercut of a steep slope generates more stress relief, the shear zone is bigger with a wider shape at the top. While the shear zone of a mild slope is smaller with a wider shape at the bottom. This difference causes many failures of arches. Subsequent stacks of arches can form in a mild slope until reaching the collapse of a whole slope, while failure will happen aggressively for a steep slope due to slope buckling without many local failures.

Characterization of each type of failure is different by means of arching coefficients, k, based on theoretical mechanics and validated by the results of the physical model test. Three values of the coefficient are provided for (1) a strip arch with soil slip, (2) a segmented arch with stable scarp and (3) a circular arch with slope buckling. For a slope with no arching, the arching coefficient is merely zero.

5 APPLICATIONS TO SITE CONDITIONS

Since bedding shear zone in the clay seam layer is considerably thin, excessive pore water pressure can be dissipated in a short time. The drained shear strength obtained from a constant-volume direct shear test with measurement of vertical stress change is considered applicable to the site condition (Ohta et al. 2010). Wangsa et al. (2012) and Pipatpongsa et al. (2011) examined the mechanical properties of G1 green clay which is associated with a bedding shear zone in Area 4.1. The residual friction angles with zero cohesion-intercept obtained from multi-stage reversal constant volume direct shear box test are ranged from 12 to 17. Therefore, the minimum value of 12 was considered as a critical case. Moreover, consideration of hydro-static pressure is required in engineering practice. Four cases are considered below.

A) Failure width of passive arching slope in dry condition

sin tan cos

c

fi

kB (3)

B) Failure width of passive arching slope in dry condition with hydro-static pressure on bedding shear plane

sin 1 tan cosc

fw i

kB

(4)

C) Failure width of passive arching slope in fully saturated condition with no hydro-static pressure on bedding shear plane

1 sin tan cosc

fw i

kB

(5)

D) Failure width of passive arching slope in fully saturated condition with hydro-static pressure on bedding shear plane

1 sin 1 tan cosc

fw w i

kB

(6)

As the last condition is the most critical case, Eq.(6) is

employed to determine the failure width in the implementation at the site. Based on various laboratory and field experiments, the material parameters (EGAT 1985, 1990, Khosravi et al. 2011 and Wangsa et al. 2012) are selected for the analysis as summarized in Table 1. The contribution of the arching effect can be evaluated by a factor of safety. The safety factor for a

two-dimensional slope (planar condition) is simply calculated by Eq.(7) and Eq.(8) for dry and submerged conditions, respectively. Based on Eq.(3) and Eq.(6), the factor of safety for three-dimensional slopes (arching effect condition) can be calculated by Eq.(9) and Eq.(10) for dry and submerged conditions, respectively.

Table 1. Geometry and material parameters of the green clay seam and hale required for calculating safety factor of the undercut slope. s

Inclined slope, 18o

Residual interface friction angle, i 12o

Bulk unit weight of shale, 19.12 kN/m3

Unit weight of water, w 9.81 kN/m3

Residual UCS of shale, cr 0.33 MPa

Peak UCS of shale, cp 4.50 MPa

Designed UCS of shale, cd 1 MPa

UCS = unconfined compressive strength Table 2. Calculated safety factors against a width 130 m for dry and submerged conditions under two and three dimensions using residual, peak and designed values of unconfined compressive strength of shale

ith an arching coefficient assigned to k2=1. w 2D (planar

slope) 3D (arching

effect) Safety factor dry

submerged

dry submerged

Residual rock strength

1.2 0.4

Peak rock strength

16.9

4.9

Designed rock strength

0.6 0.2

3.8 1.1

2 ,tantan

iD drySF ,

2 ,

1 tan1 tan

w

D submergedw

SF

i (7), (8)

3 , sin tan cos

c

D dryi

kSFB

(9)

3 , 1 sin 1 tan cos

c

D submergedw w i

kSFB

(10)

Using the material parameters shown in Table 1, the safety

factor determined from Eqs.(7)(10) are shown in Table 2. In the calculation, the arching coefficient assigned to k2=1 for mild slopes with supporting ground for the maximum exposed width 130 m. Safety factors based on a planar condition for both dry and submerged conditions are less than one which might conclude that the slope cannot be undercut. However, an arching effect allows a higher factor of safety; therefore, if the shale above the clay seam has not been weathered into weak soft rock, mining at Area 4.1 with the span of 130 m is possible. The undercut span at Area 4.1 is varied as a function of the unconfined compressive strength of the shale; thus, if the unconfined compressive strength of shale on the slope could be maintained at 1 MPa at the least, the undercut span of 130 m at the clay seam level can be reached safely in a short term.

The width of Area 4.1 in the Mae Moh lignite mine is about 300 m and the length about 250 m along the pit wall. The total depth of 33 m in this area for lignite mining was planned by digging 3 benches with a height of 11 m each. According to EGAT’s mining plan, Area 4.1 is divided into 2 stages of excavation, namely stage 1 for 180 m and stage 2 for 120 m

946


measuring at the G1 clay seam level which is located at a depth of 1112 m. Stage 1 is subdivided into 1A and 1B. Despite a crack along the existing fault zone with a length of about 50 m, observed in May 2012 at the left side corner of the pit, the process of excavation and mining for stage 1A to a width of 120 m was successfully achieved over December 2011 to July 2012. At the end of stage 1A, the total movement towards the pit at the clay seam level measured by an inclinometer was 24 mm and the slope had already been mobilized along the bedding shear plane. For stage 1B, an increment width of 60 m was excavated during July to October 2012 across a rainy season, while the slope on the right side of the excavated pit was being dumped with overburden “claystone” from the pit up to 100,000 m3 covering a length of about 60 m to provide a counterweight. Supported dump was on an incline plane, limestone rock bunds were constructed underneath dumped material for reinforcement. Due to excessive movement more than 60 mm in September 2012, the inclinometer tube was deformed and not functioning. Fig.6 shows the beginning of stage 1B observed in October 2012, when the total excavated width reached 190 m.

8 REFERENCES

EGAT 1985. Thailand-Australia Lignite Mines Development Project, Geotechnical Report, Mae Moh Mine, Thailand.

EGAT 1990. Interim Review of Geotechnical Strength Data. Mae Moh Mine, Thailand.

Janssen H.A. 1895. Versuche über getreidedruck in silozellen. Zeitschr d Vereines deutscher Ingenieure 39, 1045-1049.

Jenike A.W. 1961. Gravity flow of bulk solid. Bulletin of the Utah Engineering Experiment Station 108.

Khosravi M.H. 2012. Arching effect in geomaterials with applications to retaining walls and undercut slopes, Department of International Development Engineering, Graduate School of Science and Engineering, Tokyo Institute of Technology, PhD Dissertation.

Khosravi M.H., Ishii Y., Takemura J. and Pipatpongsa, T. 2010. Centrifuge model test on compacted sand slopes undercut by In-flight excavator. Proceedings of Geo-Kanto, 136-139.

Khosravi M.H., Pipatpongsa T., Leelasukseree C. and Wattanachai P. 2009. Failure mechanisms in arched excavation of sloped earth using model test. Proceedings of Geo-Kanto, 241-246.

Khosravi M.H., Pipatpongsa T., Takahashi A. and Takemura J. 2011. Arch action over an excavated pit on a stable scarp investigated by physical model tests. Soils and Foundations 51(4), 723-735.

6 CONCLUSIONS The slope stability problem in the Mae Moh lignite mine in

Lampang province in Thailand has been briefly reported. The presence of a weak shearing zone in the clay seam between the layer the shale caused trouble in the northeast pit. Area 4.1 in the northeast pit is one of the potential failure slopes; part of the lignite and rock mass had been left in front of this slope as a counterweight to prevent a huge landslide. Mining in the unstable slope was considered expensive. The newly developed moving-pit mining method based on the physical model and theoretical developments was introduced as an applicable method for mining in Area 4.1. In order to apply this method, two stages of excavation were planned. The process of excavation and in-pit dumping must be done in sequence. At the clay seam level, the total excavated width of 190 m with an exposed width of 130 m and an area of dumped claystone of 60 m with limestone rock bunds underneath was found stable. It is concluded that this novel procedure for mining is practically realizable and results in reductions in massive excavation, transportation and dumping of unstable rock mass, as well as saving an amount of time and expense.

Khosravi M.H., Pipatpongsa T., Takemura J., Mavong N. and Doncommul P. 2011. Investigation on shear strength of shale at the Mae Moh open-pit mine, Proceedings of the 4th Thailand-Japan International Academic Conference, 51-52.

Khosravi M.H., Tang L., Pipatpongsa T., Takemura J. and Doncommul P. 2012. Performance of counterweight balance on stability of undercut slope evaluated by physical modeling. International Journal of Geotechnical Engineering 6(2), 193-205.

Ohta H., Pipatpongsa T., Heng S., Yokota S., Takemoto M. 2010. Significance of saturated clays seams for the stability of rainfall-induced landslides. Bulletin of Engineering Geology and the Environment 69(1), 71-87.

Pipatpongsa T., Khosravi M.H., Doncommul P. and Mavong N. 2009. Excavation problems in Mae Moh lignite open-pit mine of Thailand. Proceedings of Geo-Kanto, 459-464.

Pipatpongsa T., Heng S., Likitlersuang S., Mungpayabal N. and Ohta H. 2011. Investigation of mechanical properties of clay seam in bedding shears of the Mae Moh open-pit mine of Thailand, Proceedings of the International Conference on Advances in Geotechnical Engineering, 209-214.

Terzaghi K. 1936. Stress distribution in dry and saturated sand above a yielding trap-door. Proceedings of First International Conference on Soil Mechanics and Foundation Engineering, 307-311.

Walker D.M. 1966. An approximate theory for pressures and arching in hoppers. Chemical Engineering Science 21 (11), 975-997.

Walters J.K. 1973. A theoretical analysis of stresses in axially-symmetric hoppers and bunkers. Chemical Engineering Science 28(3), 779-789.

7 ACKNOWLEDGEMENT This research work was funded mainly by the Electricity

Generating Authority of Thailand (EGAT) under the research project grant “Stabilization of Alternative Excavations and Strengthening of Supporting Materials for Pit Wall in Area 4.1”. In addition, the financial support provided by JSPS KAKENHI Grant Numbers 23760441 and JSPS Asian CORE Program is truly appreciated. The authors would like to express their gratitude to all staffs of Mae Moh Geotechnical Department of EGAT and particularly acknowledge Mr. Prajuab Doncommul, EGAT and Dr. Cheowchan Leelasukseree, Chiang Mai University for their collaborations and contributions.

Wangsa R., Wongsiriworakul J., Mungpayabal N., Pipatpongsa T. and Wattanachai P. 2012. Residual shear strength of Mae Moh clay seam measured by multiple reversal direct shear box test under constant load, Proceedings of the 17th National Convention on Civil Engineering, GTE012, 1-10.

Elev. +170

Elev. +181

Elev. +192Elev. +203

11 m

11 m

11 m

Exposed width 130 mDumped claystone

width 60 m

Figure 6. The total excavated width of 190 m, consisting of an exposed width of 130 m and an area of a dumped claystone of 60 m measuring at the level of clay seam (as of October 2012)

947

Critical State Modelling of Soil-Structure Interface for Advanced Design

Modélisation à l'état critique d'interface sol-structure pour la conception avancée

Sarma D., Sarma M.D LM-IGS, M-ISSMGE, Independent Consultant (Southern Africa), Gaborone, Botswana

ABSTRACT: Information on the influence of impregnation of cementitious slurry at the soil-structure interface of bored cast in-situfoundation is inadequate in available literatures. Moreover, influence of such impregnation on negating the detrimental effects ofsmear zone, formed by construction tools, surrounding the borehole is also unknown. In classical foundation engineering, influencesof smear and impregnation are neither considered as dependent functions in determining contributory or negative shaft resistances, nor in shaft and base resistance interaction. This ignorance contributes empiricism in bearing capacity evaluation recognising it as oneof the possible causes of variation of field performance with respect to prediction. Solution to these problems has been exploredthrough field and simulated laboratory studies of smear and impregnation, developing new device and technique. Further, an approachto interface modelling of soil-structure is presented considering impregnation.

RÉSUMÉ : L'information disponible dans la littérature sur l'influence de l'imprégnation de coulis de ciment à l'interface sol-structure de fondations coulées en place est inadéquate. En outre, l'influence de l'imprégnation sur la négation des effets néfastes de la zone de souillure, dus aux outils de construction, entourant le trou de forage est également inconnue. Dans les travaux de fondation classiques,les influences des souillures et l'imprégnation ne sont jamais considérées comme des paramètres liés dans la détermination du frottement négatif ou positif, ni dans la résistance de pointe. Cette ignorance contribue à l'empirisme en cours dans la détermination de la capacité portante et peut être l'une des causes possibles de variation des performances sur le terrain par rapport à la prédiction. Une solution à ces problèmes a été explorée in situ et en laboratoire en développant un nouveau appareillage et une nouvelle technique. Enoutre, une approche de modélisation des interfaces de sols structure est présentée en prenant en compte l'imprégnation.

KEYWORDS: Impregnation, smear zone, soil-structure interface, effective diameter, shaft resistance.

1 INTRODUCTION

Barring the situations where a permanent casing is left in the borehole, in all other bored cast in-situ deep foundations, fresh concrete comes directly in contact with the ground. During the process of concreting, cementitious slurry from the body of the unset concrete of the cast in-situ deep foundation starts impregnating and upon setting, strengthens the surrounding soil within the impregnation depth. Physical evidence of surrounding cement-impregnated soil becoming a part of the foundation shaft, stated by many authors, was reconfirmed (Sarma, 1992). In such a case it is apparent that the adhered soil shall behave as an integral part of the foundation, increasing its effective diameter. Other researchers (Berezantzev, 1965; Sowers, 1979) also found evidences of such a phenomenon, investigation on which is, however, limited. Such impregnation may increase negative shaft resistance causing serious treat to both friction and end bearing foundations, positive contribution of which, may enhance shaft resistance for all types of cast in-situ foundations. Consideration of such phenomenon in design can contribute immensely in economic aspects. Therefore, extent and effect of such impregnation into the surrounding soil have been studied in this paper for wide variety of soils. These are aimed at ascertaining the soil-structure interface strength mobilisation for bored cast in-situ foundations considering smear zone, which is formed by construction equipments. Necessary equipment for field simulation was designed and developed (Sarma, 2000) for impregnation study and eventually for simulation modelling. Investigations through microscopic and staining techniques were carried out for insights into the effects of smear and impregnation with associated effect of shrinkage. Ignorance of this complex soil-structure interface interaction contributed empiricism in bearing capacity evaluation witnessing uncertainty in field performance.

2 FORMATION OF SMEAR ZONE

The process of installation of cast in-situ deep foundation disturbs the sub-soil formation surrounding the borehole with respect to its virgin state. The mechanical process involved in the boring changes physical characteristics weakening the structure of soil to reduce its shear strength within the zone of influence of the boring tools. During the progress of boring, soil in the upper portion of borehole squeeze inward due to the loss of lateral support and their shearing during withdrawal and reinsertion of boring-tool further disturbs the structure of clays or in case of sand reducing its density despite stabilisation under drilling mud. This process repeatedly continues during each insertion and withdrawal of the boring tool. Further, the surging effect that occurs during repeated withdrawal of boring equipment further loosens the soil. The process of boring also changes the particle orientation at the interface. Thus a smear or distortion zone is formed circumscribing the borehole. The extent of smear zone is thus function of the sensitivity or relative density of the surrounding soil, magnitude of vibration and disturbances caused by the boring equipments, properties of the soil responsible for propagation of velocity wave for the vibration and disturbances caused during boring etc.

3 DEPTH OF SMEAR ZONE AND INTENSITY OF SMEAR

Meyerhoff G.G (1976) reported that depth of smear surrounding bored pile might exist up to 1 inch (25 mm). Besides that, there is not much information in the literatures about the depth of smear zone associated with cast in-situ deep foundations. This is perhaps due to the fact that it is rather a problem primarily associated with workmanship for installation of such foundation along with a host of secondary factors and thus any effort for rational analysis of the problem bristles with difficulties. The depth of smear or distortion zone depends on the type of boring equipments, method of boring and more importantly on the nature and status of the soil. Effort has been made to fulfil parts

948


of the objectives of the research (Sarma, 2000) to evaluate the depth of smear zone, intensity of smear, and effect of impregnation under laboratory test conditions.

4 IMPREGNATION AND ITS EFFECT ON INTERFACE STRENGTH

Fresh concrete forces the softened/loosened soil partially outward from the borehole owing to its larger unit weight and pressure head developed during placement and concreting operations. This creates further disturbances to the soils. Specific gravity of the fresh concrete being higher than that of the in-situ soil and due to positive pressure developed particularly in tremie action cementitious slurry, which is one of the constituents of the fresh concrete, always exists at higher than the ground water pressure unless an artesian flow of higher order reverse the situation. Therefore, in general, a tendency for impregnation of cementitious slurry from the body of the freshly cast deep foundation towards the less stressed zone surrounding the borehole always exists until a state of equilibrium of slurry pressure in the pores of surrounding soil is attained. Such impregnation of cementitious slurry alters the physico-chemical characteristics of soil within the impregnated zone and upon setting strengthens smear/distortion zone within the impregnation depth.

5 MOBILISATION OF SHAFT RESISTANCE

The philosophy of soil-structure interface strength based on the effect of distortion or smear zone and impregnation of cementitious slurry together, can be visualised in the following way:

Figure 1 (a). Idealised profiles at soil-structure interface, cemented rength

egnated strength of soil is lower an its undisturbed strength

, cemented

rength less than undistur

tial remoulded strength dep

y be idealised within the ext

te due to the com

to stabilise the borehole,

ffective dia

st more than undisturbed strength (After Sarma, 1992)

It was asserted that the impregnated cement slurry improves the shear strength of part of the distortion zone surrounding cast in-situ deep foundation (Sarma, 1992). Also during curing, concrete absorbs moisture from the surrounding soil and thus soil gets consolidated to give better strength. This gain in strength diminishes with increasing distance from the foundation. On the other hand, intensity of remoulding / loosening effect that causes shear strength to decrease in the surrounding soil reduces away from the foundation. Hence the rupture surface for mobilisation of shaft resistance does not exactly lie at the interface of the concrete of the foundation and soil. It is possible that the weakest surface may exists away from the body of foundation as a result of the two opposing effects mentioned above. Therefore soil up to the weakest surface may adhere to the foundation surface behaving as its integral part. In such a case slip or local yield that occur when the shear stress reaches the adhesive (or yield) strength may not occur at soil-structure interface rather between adhered and surrounding soil. Two different possibilities have been presented. Fig. 1(a) portrays potential rupture surface in case

cement impregnated strength of soil is higher than its undisturbed strength and Fig. 1(b) portrays potential rupture surface in case cement imprth

Figure 1(b). Idealised profiles at soil-structure interface

st bed strength (After Sarma, 1992)

In both the figures soil-structure interface shear strength is presented in the ordinates. Therefore, any point towards left of the ordinates will represent the body of foundation. The profile of the shear strength of soil due to the effect of smear is shown with minimum remoulding strength at the soil-structure interface where extent of smear is maximum and strength increases away from the interface. It will be undisturbed shear strength at the end of smear zone due to the diminishing affect of smear. Any point at the profile of shear strength of soil within smear zone will represent par

ending upon the severity of smear. On the other hand cemented soil strength, i.e., soil shear

strength due to the effect of impregnation of cementitious material, will be maximum at the soil-structure interface. The cemented soil strength has a diminishing trend away from the body of the foundation up to the remoulded shear strength of soil at the end of impregnation. This ma

ent of average impregnation depth. The combined effect of cemented soil strength, which has

diminishing tendency, and remoulded shear strength which has increasing tendency from the body, may act as the mobilised shear strength of soil surrounding the shaft. It is clear from the figure that potential rupture surface exists at the point where mobilised shear strength attains a minimum value. The soil up to the potential rupture surface will act as a part of the shaft that affects increased diameter. With the fact that the potential rupture surface exists away from the shaft, the average shear strength of the soil within the impregnated zone may be either lower or higher than that of the undisturbed sta

bined effect of smear and cemented strength. While formulating the philosophy of this radical concept of

soil-structure interface strength it is considered that the depth of impregnation is less than the depth of smear. The depth of impregnation being more than the depth of smear is possible only in case of granular soil. This is ruled-out as bentonite slurry, which is generally usedimpregnates prior to cement slurry.

6 FIELD EVIDENCES OF IMPREGNATION

Piles pulled out of soil frequently appear with a skin of soil sometime several mm thick adhering tightly to the surface of the pile thus becoming a part of the pile itself (Bowles, 1988). Field evidence of such a phenomenon was noticed and reconfirmed by this author too during excavation for construction of deep pile caps. Due to soil adhering to the pile surface the e

meter, at which shaft resistance mobilises, increases. Field investigation reveals that soil becoming a part of pile,

with a thick skin of adhered soil, is prominent in case of cast in-

949


situ deep foundation installed by conventional / hydraulic assisted augers or bailers where no drilling mud is used. Such installations where Direct Mud Circulation (DMC) technique is used, a thick soil becoming a part of the foundation is not generally observed during excavation for caps. This has been confirmed during bored pile foundation for the hostel building of Indian Institute of Technology, Guwahati, and in several other cases. This is due to the fact that the impregnation of bentonite that takes places during the process of boring, fills the voids adjacent to the bored surface. Moreover, the thixotropic property of the bentonite particles left no room for cement slurry to impregnate further, thus negating the chances for a thick skin of soil becoming a part of the structure. This confirms the possibility of mobilisation of shaft resistance closer to the surface of the structure or at the surface itself depending upon the nature of soil. In such a case the adhesion that occurs between two different materials may govern the process of mobilisation of shaft resistance rather than cohesion between adhered to surrounding soil.

LOPMENT OF DEVICE FOR IMPREGNATION

a hol

icpum used to compress air inside the first compartment.

side the rectangular com

tedas 1

without anyisturbance to the impregnated and smeared face.

aw without causing any undue damage to the required lane.

RVATION FOR SMEAR ZONE

se two possibilities might occur separately or simultaneously.

7 DEVESTUDY

Notwithstanding the effect of impregnation and soil becoming part of bored cast in-situ deep foundation, available literatures do not give much information on the magnitude of impregnation and its dependent factors. Furthermore, no technique for measuring such impregnation could be known from present literatures. Therefore, a new method was developed in which the pressure exhibited by cementitious slurry during placement of fresh concrete in borehole was simulated in laboratory for allowing impregnation through the soil sample collected from borehole. Such laboratory simulation involved development of concept, fabrication of device, and performing trial tests. Final version of the impregnation test equipment, incorporating minormodification upon trial tests, was used for impregnation study. The equipment comprised of a closed rectangular concreting compartment of size 150x200x200 mm fabricated from thick steel plates with detachable top lid fixed by high tensile nuts and bolts. At the top lid a non-return-air valve and a pressure gauge were fixed to pump compressed air in and to monitor air pressure inside the compartment respectively. In one of the side plates of the compartment a hole of 75 mm diameter was made and a threaded socket was welded along the circumference of the hole so that a sampler could be threaded into the socket. Cylindrical hollow samplers of 75 mm diameter, 150 mm long, and 2 mm thick having threads at both the ends were used for sampling and for fixing with the socket for impregnation test. At the other end of the sampler, a threaded cap was fixed with

e in it, plugged by jute wick, to allow water to come out. All the joints have been made airtight. The non-detachable

joints are sealed by resinous epoxy, threaded joints that require frequent removals are sealed by jute fibre soaked in zinc solution and non-threaded joints are sealed by rubber gasket kept in highly a compressed state. A remote control pneumat

p was

8 IMPREGNATION TEST PROCEDURE

Samples for impregnation tests were collected from shallow depth, generally, two numbers at the same depth from each auger-borehole by horizontal sampling. Shallow sampling depth was preferred in order to collect samples experiencing maximum disturbance from repeated insertions and withdrawals of boring tool. Collected samples were kept for twenty-four hours inside the sampler to regain its natural state to the possible extent. A pumice stone wrapped by filter paper was placed by trimming soil in the driven end of the sampler. After covering the end by cap, the sampler was inserted into the socket with smeared face of the sample towards the rectangular

compartment. Nominal mix of concrete of ratio 1:2:4 (1 part cement: 2 parts sand: 4 parts 20 mm nominal size aggregate) with 10% extra cement, having slump of 120 mm was poured in the rectangular compartment and properly placed. Without any delay the top lead was fixed and air was pumped by the pneumatic pump till air pressure in

partment reaches the required limit. Bowles (1982) indicated that the critical maximum pressure

of concrete would occur at a depth within 10 to 20 times the diameter of pile. So for 400 mm diameter pile, the extent of height of fresh concrete was worked out within 4 and 8m andcorresponding extents of impregnation pressures were adop

and 2 Kg/cm2 for fresh concrete of specific gravity 2.5. The pressure had to be maintained at required limit as

sometime it was found necessary to compensate little pressure drops as a result of impregnation. Although initial setting time of cement is 30 minutes, pressure in the compartment was maintained for 90 minutes to simulate field condition to the possible extent. The pressure in the compartment was released and the sampler was then removed from the compartment. Sampler cap, pumice stone and pieces of filter paper were also removed for extrusion of the sample from sampler. For extrusion, pressure was applied in the face at the outer side where pumice stone and G.I. cap was fixed such that no disturbance occurs to the impregnated and smeared face of the sample. The samples were stored in dry placed

9 PREPARATION OF BLOCK SAMPLES FOR OBSERVATIONS

For visual observation it is necessary to prepare block samples with at least one plane face. Of the whole sample, since the face through which impregnation occurs was important, it was necessary to cut along the direction of impregnation. Prior to cutting, the samples were saturated with toluene-epoxy solution under vacuum desiccators. With the process of de-airing, the sample absorbed toluene-epoxy solution and upon curing gets strengthened. After 15 days of strengthening the samples were found to be suitable for cutting. The cutting was done by a thin metal sp

10 MICROSCOPIC OBSEAND IMPREGNATION

Microscopic investigation was carried out under a high resolution polarised microscope. The prepared sample for investigation was placed under microscope and observed through lens ‘X100’ and subsequently with ‘X10’. In case of study of samples under high magnifying lens, the particle re-orientation or particle crushing due to smear effect was found very difficult to identify from other randomly oriented soil particles that naturally exist in the soil. Furthermore, it was found to be very difficult to distinguish between the particles of impregnated cement from that of other similar whitish materials scattered in the soil matrix. Under low magnifying lens (X10), however, some irregular cylindrical veins of deep brown tinted materials and whitish materials were seen at random. Later those deep brown tinted material was identified as the epoxy resin used for stabilisation of the study samples and whitish material was identified as impregnated cement particles respectively. The identification was confirmed upon comparing the samples of hardened epoxy and hardened cement under the same microscope. While in case of epoxy resin, cracks formed due to desiccation of surface during preparation of samples was identified as the chief reason for impregnation, two possibilities were identified for impregnation of cementitious materials in the form of cylindrical veins. The

950


In the first possibility, the flow of cementitious slurry might take place through the shear channels formed during the process of boring. In such an event there had the possibility that the smear zone would comprises series of lumps at random surrounded by shear channels without altering soil properties within the lumps. In the second possibility, flow of cement slurry had not taken place through the smeared face of the borehole wall uniformly and depending upon variations in permeability cement slurry impregnated in the shape of cylindrical veins. Less permeability might be either due to undisturbed state of soil or due to a lesser effect of smear. As undisturbed state of soil at the face of the borehole was not in conformity with practical experience, then, the lesser effect of smear would seem to govern. However, it could not be ascertained that impregnation of cement particles took place through the smeared face to the extent of formation of cylindrical veins, since it was found very difficult to distinguish between the particles of impregnated cement from that of other similar whitish materials scattered in the soil matrix.

The findings on the smear zone and impregnation could not therefore be properly interpreted from overall perspective due to limitation of the study under microscope. However, the only distinct photomicrograph showing soil-structure interface is presented in Fig. 2 (a). From the other photomicrographs, mapping of a distinguishable demarcation line between natural soil and its impregnated counterpart was not found convenient and therefore alternative technique had to be explored.

11 STAINING TECHNIQUE

Hutchison (1974) formulated a procedure to identify carbonated ingredients in sedimentary rocks. The chemical treatment carried out on rock sample turns the carbonated ingredient into pink colour. Initially, this ‘staining technique’ was adopted to identify the extent of impregnation in natural soil as cementitious material is predominantly enriched with carbonate particles. The staining technique was started by washing the exposed face of the sample with 1.5% HCl solution for 10 to 15 seconds and then immersing in the reagent for 10 to 15 seconds and drying under sun rays. The basic chemicals used for staining tests were 1.5% HCl solution, alizerin red-s and potassium ferocyanide. For treatment, two solutions were prepared from the above basic chemicals, one alizerin red-s solution (ARS) and other potassium ferocyanide solution (PFS). Both the solutions were prepared by dissolving solvent in 100cc of 1.5% HCl acid. For ARS, quantity of alizerin red-s was 0.2 gm while for PFS, quantity of potassium ferocyanide was 2.0 gm. The reagent was prepared by mixing ARS and PFS in ratio 3:2 for 30 to 45 seconds. This process, however, gave slight tint of colour from which distinguishable demarcation of natural soil with impregnated counterpart was not possible. Eventually after many trials an appropriate staining technique, suitable for the impregnated soil samples, was developed modifying over the staining technique suggested by Hutchison (1974). Prominent demarcation was noticed after the treatments, resulting in two distinguishable colours of the sample viz., pink and green. In order to identify material of impregnation, hardened epoxy and cement pellet were treated by the same technique. Hardened epoxy did not give traceable change in colour while cement pellet turned pink. Thus cement-impregnated portions of the samples were ascertained. The reason behind natural soil turned green was explored and found that alkaline material turned pink while acidic material turned green upon reaction with the reagent. pH test conducted in Government Agricultural Laboratory confirmed pH values of soil ranges from 5.00 to 6.70, which confirmed acidic nature.

12 MEASUREMENT OF IMPREGNATION

To facilitate the measurement of impregnation close-up photographs of the samples were taken with a magnification of

about two times rendering easy measurement of impregnation. A typical photograph with a pasted scale showing dimension of 5 mm is presented in Fig 2 (b). The scale of each photograph was derived by dividing measured length of the pasted scale in mm by 5. Impregnation mapping of all samples were then drawn from the photographs. A typical mapping is presented in Fig. 3 (a).

(a)

(b)

5 mm

Figure 2. (a) Photomicrograph of soil-structure interface, (b) Coloured profile of cement impregnated soil at the soil-structure interface (After Sarma, 2000)

The mean depth of impregnation (Iav) was determined from the relationship as follows:

SL

Iiz

av where, Aiz is the area enveloped by impregnation zone, Lizlength of impregnation zone, and ‘S’ the scale. Similarly, the maximum depths of impregnation were marked and measured multiplying by scale factor.

Aiz (1)

13 FACTORS AFFECTING IMPREGNATION AND FINDINGS

From the test results of samples collected from nine different locations, the average impregnations of different types of samples were found varying from 0.57 to 4.56 mm and maximum impregnations were varying from 1.42 to 7.6 mm. The peak values of average and maximum impregnation depths corresponding to 1 kg/cm2 slurry pressure were 2.84 mm and 6.3 mm respectively. Similarly, peak values of average and maximum impregnation depths corresponding to 2 Kg/cm2

slurry pressure were 4.56 mm and 7.6 mm respectively.

951


(a)

(b) Figure 3. (a) Typical impregnation mapping, (b) Response of impregnation versus D50×e (After Sarma, 2000)

As a general trend it was observed that both average and maximum impregnation depths increase with increase in slurry pressure. Individual plots between impregnation and other parameters like, over consolidation ratio, liquid limit, plasticity index, activity, coefficient of permeability, void ratio, revealed no definite trends. Therefore, plots were tried with composite parameters. For plots with maximum impregnation versusD50×e (where D50 is the size of mesh through which 50% of soil passes and ‘e’ is the ), no mean line could be drawn as points were ay be due to the fact that

t it being a single function of void ratio and ore size. However, points of average impregnation versus

1 Kg/cm2 and 2 Kg/cm2

(2)

itigating the effects of shrinkage. -soil becoming part of

sive collapse with

irtual o shrinkage and its effect on the shaft y been completed. Formation of smear

in-situ void ratio scattered. This m

maximum impregnation depth depends on the depth of local fissures and as the depth of fissures vary considerably from sample to sample, the depth of maximum impregnation too would vary withoupD50×e for slurry pressure ofrespectively were seen in a trend more or less around a straight line drawn through the mean points for each pressure (Fig. 3 b). Regression analysis was done for average impregnation versus D50×e and following equations of straight lines were developed for slurry pressure of 1 Kg/cm2 and 2 kg/cm2 respectively:

05.11023.0 3 VI av 12.11054.0 3 VI av (3)

where, ‘V’ is D50×e and ‘Iav’ is average impregnation in mm. The above equations show that average impregnation is a direct function of ‘D50×e’, which is a measure of both pore size and overall void ratio.

14 EFFECT OF SHRINKAGE

Shrinkage of concrete, which is a phenomenon associated with curing, is an important factor as it introduces an element of uncertainty in the location of the potential rupture surface that results in uncertainty in the prediction of shaft resistance (Sarma, 1992). It is therefore of considerable importance in mobilisation of shaft resistance.

With the setting action of concrete, the process of shrinkage also continues. The magnitude of shrinkage, however, depends on various factors among which effects of size of aggregate, elastic properties of aggregate, concrete used, contamination of concrete by clay particles are important.

Among the influencing factors as stated above, the most important influence is exerted by aggregates. The size and grade of aggregate do not influence the magnitude of shrinkage directly but large aggregate permits use of linear mix and hence results in lower shrinkage (Nevile, 1981).

The elastic properties of the aggregate determine the degree of restraint offered. Presence of clay particles in concrete lowers its restraining effect increasing shrinkage. Even if, aggregates used in concrete are free from clay particles, during the process of tremie concreting it may carry clay particles from borehole wall and prone to higher shrinkage.

Based on the shrinkage strain of 3×10-4, recommended by Indian Standard (I.S. 456-1978), reduction of diameter is expected varying from 0.09 to 0.36 mm for diameter of pile 300 to 1200 mm. This reduction of diameter shall give rise to virtual gap around the shaft leading to virtual loss of contact with borehole wall.

There may be mixed opinion whether such gap has any practical significance or not. Generally it is expected that the soil of the borehole trends to fill-up such gap by collapsing

nder active pressure muHowever, from the evidence of cementedsuch structures, it is possible that progresgradual shrinkage occurs outside the zone of impregnation as concrete brings-in cement-impregnated surrounding soil during its shrinkage. Eventually, shrinkage may have effect outside impregnation zone and strain softening the potential failure surface further. Therefore, separate effect of shrinkage has not been considered while presenting the alternative concept of shaft resistance mobilisation based on impregnation. Nevertheless, effect of shrinkage may be prominent in case of bored cast in-situ deep foundation installed by DMC technique where, impregnation depth and soil becoming part of the pile may not be significant.

15 CONTINUING DEVELOPMENTS

The model of average impregnation is presented for the range of maximum concrete pressure for pile of 400 mm diameter. As the information on critical depth of maximum concrete pressure is available, the model can be extended for higher diameter shaft of deep foundation. With these findings determination of effective diameter of bored cast in-situ deep foundation is possible for variety types of soil more importantly information on the location of potential rupture surface for mobilisation of shaft resistance. Among other two indeterminate factors, nam age and smear zone, development of a vely, shrinkcollapse model due tresistance had alreadzone is a problem primarily associated with types of equipments used and installation workmanship. Developments over the conventional construction equipments and method of installation had already been done, new equipments fabricated, and prototype piles constructed for performance evaluation. The objectives of these new equipments were to keep the depth of smear minimum and within the impregnation depth. With these new patented equipments (viz., valve auger, scraper unit, and

952


vibratory tremie) full scale piles constructed for performance tests which indicated promising results in comparison to the piles installed by conventional technique.

16 CONCLUSIONS

The extent of impregnation and smear zone surrounding borehole is a function of the soil properties and installation technique. Impregnation of cementitious slurry alters the physico-chemical characteristics of soil within the impregnated zone and upon setting strengthens smear zone within the impregnated depth. Cement-impregnated soil becoming a part of the cast in-situ deep foundation is almost certain. The resulting larger effective diameter is causing potential threat from negative shaft resistance. However, its positive contribution enhances shaft resistance vis-à-vis design economy. Due to the opposing nature of smear and cemented-soil characteristics, soil-structure interface strength depends on the properties of partially remoulded soil at the potential rupture surface, location of which is dictated by the depth of impregnation. Using newly developed simulation device soil samples from various auger-boreholes were impregnated at 1 and 2 Kg/cm2 slurry pressure. Microscopic observations on impregnated samples were not found promising for determination of the extents of smear and impregnation. Modified staining technique enabled mapping of impregnation profile for determination of maximum and average impregnation depths. Although both impregnation depths increase with increase in slurry pressure, their plots with common soil parameters revealed no definite trends. In the plots of impregnation depth with composite parameter D50×e, which is a measure of both pore size and overall void ratio, no mean line could be drawn for maximum impregnation, however,

impregnation rendered model of s location of rupture surface for

onstruction of prototypes, performance of which were found to be promising..

17 ACKNOWLEDGEMENT

In order to supplement a research (Sarma, 2000), Mr. A. Deb undertook detailed investigation on impregnation (Deb, 1995) implementing the scheme and using the device developed by the Author1. His support is gratefully acknowledged. The experimental investigations referred to in this paper were part of Author1’s PhD work (Sarma, 2000), which were conducted under the supervision of Prof. P. K. Bora, PhD (Birmingham), Head of Civil Engineering (Retired), Assam Engineering College, India. His encouragement is gratefully acknowledged.

regression analysis for average straight line. The model indicateauger bored soil-structure interface that can be used for determination of shaft resistance. Further due to linear shrinkage the progressive collapse (strain-softening) may occur at the rupture surface or at the interface when impregnation is not prominent. This may be detrimental particularly for higher diameter shafts. The maximum unit shaft resistance is possible when the depth of smear is less and remains within impregnation depth. In order to minimise the extent of smear zone, new equipments were fabricated, patented, and used forc

18 REFERENCES

Berezentzav, V.G. 1965, Design of Deep Foundation. 6th ICSMFE, Vol. 2.

Bowles, J.E. 1982. Foundation Analysis and Design. Third Edition. McGraw-Hill.

Bowles, J.E. 1988. Foundation Analysis and Design. Fourth Edition. McGraw-Hill.

Dev, A. 1995. A Study of Adhesion Factor of Bored and Cast In-place Piles in Cohesive Soils. M.E. thesis submitted to Gauhati University.

Hutchison, G.S. 1974. Laboratory Handbook of Petrographic Techniques. First Edition, Wiley Inter-science Publication. New York.

I.S: 456-1978, Indian Standard Code of Practice for Plain and Reinforced Concrete. Third Revision, March 1987, Bureau of Indian Standard, New Delhi.

Meyerhoff, G.G. 1976. Bearing Capacity and Settlement of Pile Foundations. Journal of Geotechnical Engineering Division, ASCE. Vol. 102, GT3.

Nevile, A.M. 1981. Properties of Concrete. Third Edition, ELBS. Sarma, D. 1992. Bored and Cast In-situ MRP - A New Approach in Pile

Foundation, M.E. thesis submitted to Gauhati University. Sarma, D. 2000. Bored Cast In-situ Pile with CSR - A New Approach in

Pile Foundation,. PhD Thesis, University of Gauhati, Assam, Republic of India.

Sowers, G.F. 1979. Introductory Soil Mechanics & Foundations: Geotechnical Engineering, Fourth Edition. Collier Macmillan International.

953


1

A Study on the Influence of Ground Water Level on Foundation Settlement in Cohesionless Soil

Étude de l’influence de la variation du niveau d’eau sur le tassement des fondations superficielles reposant sur sol granulaire

M. A. ShahriarJames Cook University,Townsville, Australia. email: [email protected]. Sivakugan James Cook University,Townsville, Australia. email: [email protected]. M. Das California State University, Sacramento, USA email: [email protected]. Urquhart James Cook University,Townsville, Australia. email: [email protected]. Tapiolas James Cook University,Townsville, Australia. email: [email protected]

ABSTRACT: Settlement calculation is an important part in the design of shallow foundations resting on granular soils. Rise of groundwater level is believed to increase the settlement significantly and had been a topic of research for many years. Terzaghi (1943)suggested that the submergence of soil mass reduces the soil stiffness to half, which in turn doubles the settlement. Since then, variousresearchers proposed correction factors to account for the additional settlement due to water table fluctuation. However, a comprehensive settlement testing and its numerical modeling to account for the influence of ground water level has not been reportedin the literature. The objective of this paper is to quantify the effect of water table rise on settlement through laboratory testing over wide range of footing shape, soil density, water table depth and stress level. The tests were carried out within a settlement tank. Thefootings under working load were subjected to water table rise, and the additional settlements were measured. The experimental setup was modelled in FLAC and the results were compared with the laboratory tests. The results obtained will be valuable in verifyingTerzaghi’s intuitive reasoning and explaining the observed additional settlement of footings found in the literature.

RÉSUMÉ : Le calcul du tassement est un élément important dans la conception des fondations superficielles reposant sur les solsgranulaires. L’augmentation du niveau d’eau souterrain est supposée augmenter de façon significative le tassement et avait été unsujet de recherche pendant de nombreuses années. Terzaghi (1943) a suggéré que la submersion du dépôt de sol réduit la capacité duce dernier de moitié, ce qui à son tour double le tassement. Depuis lors, plusieurs chercheurs ont proposés des facteurs de correction pour tenir compte du tassement additionnel en raison de la fluctuation du niveau d'eau dans le sol. Toutefois, on ne reporte pas d’étudeexpérimentale et/ou numérique dans la littérature pour tenir compte de l'influence du niveau de la nappe phréatique sur le tassement des fondations superficielles. L'objectif de cette étude est de quantifier l'effet de la variation du niveau d’eau sur le tassement par lebiais d’essais au laboratoire sur une large gamme de forme de semelle, de densité du sol, de niveau de charge et de profondeur de lanappe phréatique. Les essais ont été réalisés dans un réservoir de tassement. Les semelles sous chargement ont été soumises à unevariation du niveau d’eau et des tassements supplémentaires ont été enregistrés. Le montage expérimental a ensuite été modélisé àl’aide du logiciel FLAC et les résultats ont été comparés avec ceux obtenus au laboratoire. Les résultats obtenus seront utiles pourvérifier le résonnement intuitif de Terzaghi et pour expliquer le tassement supplémentaire des semelles rapporté dans la littérature.

KEYWORDS: correction factor, granular soil, settlement, shallow foundation, water table.

1 INTRODUCTION

Shallow foundations such as pad, strip or raft footings are often preferred by geotechnical engineers when the soil conditions are suitable. Bearing capacity and settlement are the major considerations in designing shallow foundations on granular soils. The designers try to ensure sufficient safety factor against bearing capacity failure and to limit the settlement within a

tolerable value. More than 40 settlement prediction methods for footings on cohesionless soils are available in the literature (e.g. Terzaghi and Peck 1967, Schmertmann et al. 1978, Burland and Burbidge 1985, Mayne and Poulos 1999). These methods recognized that the major influencing factors for shallow foundation settlements are the applied pressure, soil stiffness and depth, width and shape of foundation.

A Study on the Influence of Ground Water Level on Foundation Settlement in Cohesionless Soil

Étude de l’influence de la variation du niveau d’eau sur le tassement des fondations superficielles reposant sur sol granulaire

Shahriar M.A., Sivakugan N., Urquhart A., Tapiolas M. James Cook University, Townsville, Australia.

Das B.M. California State University, Sacramento, USA

954



2

Seasonal fluctuations such as floods or heavy rainfalls can raise the water table up to or beyond the footing level and produce additional settlements of shallow foundations. The soil loses its stiffness when submerged, and settles more. Substantial additional settlement may occur when the groundwater level changes, which can exceed the tolerable limit for settlement and threaten the integrity of structure. Very few works have been found in the literature investigating the influence of fluctuating water level on shallow foundation settlements. Some researchers suggested using a water table correction factor, which can be used as a multiplier on the settlements predicted for footings resting on dry sands, to get the settlements in submerged condition. Limited laboratory model tests have been conducted in the past, which did not cover the effect of foundation shape or varying stress level on additional settlement induced by water table rise. In this paper, the authors have described a comprehensive laboratory test program carried out to quantify the additional settlement due to rise in water table with varying footing shape, soil density, water table depth and stress level. This was followed by modeling the experimental set up in geotechnical modeling software FLAC, and the results were compared with the experimental data.

2 WATER TABLE RISE AND CORRECTION FACTOR

Terzaghi (1943) made an intuitive suggestion that when dry sand becomes saturated, the soil stiffness (Young’s modulus) reduces by approximately 50%. He noted that, the effective vertical stress on soil under the water table reduces roughly to half; which reduces the effective confining stress by 50%. This leads to loss of stiffness of saturated soil to half of that in the dry condition. As a result, settlement in soil below the water table gets doubled.

When the water table rises to some depth below the footing, a correction factor for the new location of water table is used in the design of shallow foundations. The settlement under dry conditions is multiplied by this factor, to give the settlement expected due to the water table rise. The correction factor Cw is greater than or equal to 1 and increases with rise in water table. It is defined as:

Cw = dry sand in settlementlevel footing the below table water withsettlement

)1(

Various researchers (Terzaghi and Peck 1948, Teng 1962,Alpan 1964, Bazaraa 1967, Peck 1974, Bowles 1977) proposed correction factors to quantify the additional settlement due to the water table rise below the footing. The depth below the footing where the water table fluctuation will not have any effect is not unanimously agreed upon. The depth of embedment of the footing also affects the influence of water table on settlement, as the surcharge due to embedment increases the settlement in raised groundwater level. Throughout this paper, the correction factor for water table, foundation width, depth of water table below the foundation and the depth of embedment are denoted by Cw, B, Dw and Df, respectively, as illustrated in Figure 1.

Shahriar et al. (2012) made a critical review of the current state-of-the-art for predicting shallow foundation settlement due to rise in water table in granular soil. Theoretical studies by Vargas (1961), Brinch Hansen (1966) and Bazaraa (1967) suggested a maximum correction factor of 1.7, when the water table rises to the base of the foundation. Limited field investigations suggest that submergence of granular soil doubles the settlement when compared to dry condition, agreeing with Terzaghi’s proposition. Numerical modeling conducted by Shahriar et al. (2012) shows that the settlement gets doubled in submerged sand if linear elastic model is used, but the use of

hyperbolic non-linear elastic soil model gives higher additional settlements at high stress levels.

Ground Surface

Df Footing Level

B Dw Water Table

Figure 1. Schematic diagram of a shallow foundation.

Very little laboratory studies have been conducted so far and contradictory results have been found. Agarwal and Rana (1987) conducted tests on square footings of three different sizes. Their results support Terzaghi’s proposition that the settlement gets doubled when the sand gets submerged. Murtaza et al. (1995) also used three different sized square footings and conducted the tests with loose, medium dense and dense sands. The results showed 8 to 12 times more settlement in submerged condition. Morgan et al. (2010) carried out settlement tests with a square footing in two different types of soils and found that the increase in settlement in submerged sand can be 5.3 times the dry sand. However, these experimental programs were small in scale and none of these considered the effect of varying footing shape and stress level.

3 LABORATORY MODEL STUDY

A Perspex rectangular tank 800 mm x 800 mm in plan and 600 mm high was built to carry out the settlement test. Various footing shapes were used. A circular footing of 100 mm diameter and square and rectangular footings with B/L =1.0, 0.75, 0.50, 0.25 were used where the width, B was fixed to 100 mm in each case. A locally available granular soil was used. In a model footing having smaller dimensions, the settlement might get affected by change in soil stiffness in a partially saturated area. From laboratory testing, it was observed that the capillary rise is higher in well graded soil. Hence, the finer particles were sieved out from the test soil to get a uniformly graded soil with soil grains large enough to significantly reduce the capillary height. The rate of capillary rise of the sieved soil was then tested using soil filled Perspex tubes protruding from water. At five minutes, the capillary height observed were 40 mm and 53 mm in loose and dense sands respectively. Five minutes was the maximum time to get the water level static during the settlement tests, so the capillary rise is expected to be limited within the range of 40-53 mm. In fact, the height of capillary rise was limited to 50 mm for most of the time during the tests. This height is reasonable when compared to the footing width (100 mm). In case of granular soil, the elastic modulus of the soil is a key parameter in predicting foundation settlement, and Vanapalli and Mohamed (2007) showed that the elastic modulus of unsaturated soil can be significantly influenced by matric suction. However, by limiting the capillary rise within a shorter range, the unsaturated zones in the model tests were kept quite small and hence, their effect on the overall settlement was negligible. The soil properties of sieved out sand are: effective size D10=0.67 mm, co-efficient of uniformity, Cu=1.64, co-efficient of curvature Cc=0.89, specific gravity, Gs=2.61, maximum and minimum dry densities =1.53 t/m3 and 1.382 t/m3 respectively. Two different relative densities (37.6% and 77.4%) of the sand were used. Since the model tests represent the larger footings with higher densities in the field, maximum relative density was limited to 77.4%.

955



3

The tank was filled with sand in multiple lifts. The height of each lift was equal to the foundation width. The mass of soil for each layer was determined from the required density. Soil was poured through a funnel moving around the tank and to achieve a uniform density, a specific height of fall was maintained. A wooden float was used to compact and level the soil top after every lift. The density achieved by compaction was checked by putting square cans at various levels and reasonable accuracy was observed. Water was supplied through rubber tubing attached to a nozzle located at the bottom of the tank. Water table was raised at a lift height of 100 mm (foundation width, B)from the bottom of the tanks up to a depth of B below the footing level. Then the rise was reduced to B/5 until the water table reached footing level. The height of water table rise was monitored by a glass tube attached to the soil tank. The load was applied with a hydraulic jack. Settlement for each water table lift was obtained by averaging the two dial gauge readings placed on top of the footings. Figure 2 shows a close view of the experimental setup used in the tests.

Rectangular Footing (B/L=0.25) Rectangular Footing (B/L=0.5) Square Footing (B/L=1.0) Rectangular Footing (B/L=0.75) Circular Footing

(a)

Square Footing (B/L=1.0) Rectangular Footing (B/L=0.5) Circular Footing

Figure 2. Experimental Setup with model footing, dial gauges and loading arrangement.

Initially, pressure-settlement curves were obtained for each case by applying vertical pressure in increments and measuring corresponding settlements in dry condition. Then double tangent method was used to determine bearing capacity of the footings. This means the ultimate bearing capacity was taken as the intersection of the two tangents drawn from the two linear segments of the load-settlement plot. The working load was taken as one-third of the bearing capacity, keeping the factor of safety at 3. In the next step, the footings were subjected to working loads and the water level was raised gradually from bottom of the tank up to the footing level.

4 INTERPRETATION OF EXPERIMENTAL RESULTS

From the additional settlements measured at various water table depths, the water table correction factor diagrams were obtained. Figure 3 shows the correction factor diagrams for various footing shapes in loose and dense condition. The figure shows that the additional settlement due to water table rise is higher in loose sands, with Cw ranging from 4.9 to 7.6 times the settlement in dry condition. Footings on dense sand experienced less additional settlements than in loose sands, with Cw ranging from 2.9 to 4.4. The results indicate significantly higher additional settlement due to rise in water table than what was suggested by Terzaghi (1943).

It is evident from the curves in Figure 3 that the increment in correction factor is not linear with water table rise, instead, settlement increases at a faster rate when the water table gets closer to the footing. The stress level immediately below the footing is very high, which causes significant additional settlements.

Figure 4 shows the load-settlement curves for square footing resting on dense sand in dry (solid line) and submerged

Figure 3. Water table correction factor diagrams for model footings on a) loose sand, b) dense sand.

(b)

condition (dotted line). It shows that the additional settlement in submerged sand rises from 2.92 to 3.25 times as the applied pressure rises from 40 kPa to 75 kPa. This reflects the effect of stress level on additional settlement due to submergence. The bearing capacity of soil gets reduced while submerged, which induces high additional settlements at higher stresses.

Figure 4. Applied pressure-settlement curves for 100 mm square footing in dry and submerged condition.

5 NUMERICAL MODELING OF EXPERIMENTAL SETUP

The authors modeled the experimental setup in FLAC 6.0 (Itasca, 2008), a finite difference code used in geotechnical modeling. A hyperbolic non-linear elastic model was used in the simulation. The model relies on the nonlinear stress-strain relationship suggested by Kondner and Zelaska (1963):

max31

31

)(1

)(

iE(2)

956



4

where: (σ1-σ3)max = asymptotic value of stress difference axial strain Ei = initial tangent modulus i.e., the slope of curve

While modeling, the initial Young’s modulus was assumed to be 5 MPa for dry sand considering the lower soil stiffness in small scale footings. Following Terzaghi’s (1943) suggestion that the Young’s modulus reduces by 50% in submerged sand, the initial Young’s modulus in this sand was taken as half of that of the dry sand. The asymptotic stress difference relates closely to the ultimate strength of the soil mass and was taken as the bearing capacities of footings on dry and submerged sands obtained from pressure-settlement curves derived from the model tests. The test on circular footing placed on dense sand was modeled in this paper. The rise of water table depth was simulated using appropriate parameters and correction factors at various water table depths were observed.

Figure 5. Water table correction factor diagram for 100 mm diameter circular footing obtained from experimental results and numerical modeling.

Figure 5 shows the comparison of water table correction factor diagrams obtained from numerical modeling (dotted line) and experimental results (solid line). The diagrams were similar in shape, both being curved rather than linear as previously proposed by some researchers. Also, both the curves indicate that the effect of water table depth is negligible at a greater depth, whereas settlement increases rapidly as the water table gets closer to the footing base. The assumed soil parameters may contribute to the differences in correction factors obtained from numerical modeling and laboratory testing.

6 SUMMARY AND CONCLUSIONS

Laboratory model tests were carried out to investigate the effect of various factors on increase in shallow foundation settlement when subjected to fluctuation in ground water level. Additional settlements at various water table depths were observed and water table correction factor diagram for each case was obtained. The results show significant increase in settlement as the soil immediately below the footing level gets saturated. The results clearly indicated that the increment is higher in soils having lower density; however, the increment is significant even in dense soils. The effect of footing shapes on additional settlement in saturated sand was not evident from the results. Comparison of applied pressure-settlement curves in dry and submerged sands suggest that the additional settlement due to submergence increases with the stress level. Modeling a circular footing in FLAC and its comparison with test data confirms that the correction factor diagram is not linear, and the correction factor increases at a faster rate in the vicinity of the footing. The results obtained will help to understand how the fluctuating

water level affects the shallow foundation settlements on granular soils and will allow designers to apply appropriate correction factors for water level rise. There is a scope for further investigations to identify the effect of other important factors (e.g. depth of embedment, footing width, and soil gradation) in settlement behaviour of shallow footings with changing groundwater level. More laboratory testing with different initial densities might be useful to develop water table correction factor charts for varying relative densities and shear strength parameters. Also, advanced soil models can be used to study the effect of rising water table on shallow foundation settlement on cohesionless soils.

7 REFERENCES

Agarwal K.G. and Rana M.K. 1987. Effect of ground water on settlement of footing in sand. Proceedings, Ninth European Conference on Soil Mechanics and Foundation Engineering,Dublin, 2, 751-754.

Alpan I. 1964. Estimating the settlement of foundations on sand. Civil Engineering and Public Works Review, 59(700), 1415-1418.

Bazaraa A.R. 1967. Use of the standard penetration test for estimating settlements of shallow foundations on sand. Ph.D. dissertation, Department of Civil Engineering, University of Illinois, Champaign-Urbana.

Bowles J.E. 1977. Foundation Analysis and Design, 2nd Ed. McGraw-Hill, New York.

Brinch Hansen J. 1966. Improved settlement calculation for sand. The Danish Geotechnical Institute Bulletin No. 20, 15-19.

Burland J.B. and Burbidge M.C. 1985. Settlement of foundations on sand and gravel. Institution of Civil Engineers, 78(1), 1325-1381.

Itasca 2008. FLAC 6.0: User’s guide. Itasca consulting group, Minneapolis, USA.

Kondner R.L.A. and Zelasko J.S. 1963. A Hyperbolic Stress-Strain Formulation of Sands. Proceedings of Second PanAmerical Conference on Soil Mechanics and Foundation Engineering,Brazil, 289.

Mayne P.W. and Poulos H.G. 1999. Approximate displacement influence factors for elastic shallow foundations. Journal of Geotechnical and Geoenvironmental Engineering, ASCE,125(6), 453-460.

Morgan A.B., Sanjay K.S., and Sivakugan N. 2010. An experimental study on the additional settlement of footings resting on granular soils by water table rise. Soils and Foundations, 50 (2), 319-324.

Murtaza G., Athar M., and Khan S.M. 1995. Influence of submergence on settlement of footing on sand. Journal of the Institution of Engineers (India), 76 (5), 51-54.

Peck R.B., Hanson W.E., and Thornburn T.H. 1974. Foundation Engineering, 2nd Ed. John Wiley & Sons, New York.

Schmertmann J.H., Hartman J.P., and Brown P.R. 1978. Improved strain influence factor diagrams. J. Geotech. Eng. Div., ASCE, 104(8), 1131-1135.

Shahriar M., Sivakugan N., and Das B.M. 2012. Settlements of shallow foundations in granular soils due to rise of water table – A critical review. International Journal of Geotechnical Engineering, J Ross Publishing, 6(4), 515-524.

Teng W.C. 1962. Foundation Design. Prentice-Hall Inc., New Jersey. Terzaghi K. 1943. Theoretical Soil Mechanics. John Wiley & Sons,

New York. Terzaghi K. and Peck R.B. 1948. Soil Mechanics in Engineering

Practice, 1st Ed. John Wiley & Sons, New York. Terzaghi K. and Peck R. B. 1967. Soil Mechanics in Engineering

Practice, 2nd Ed. John Wiley & Sons, New York. Vanapalli S.K. and Mohamed F.M.O. 2007. Bearing capacity of model

footings in unsaturated soils. Experimental unsaturated soil mechanics. Springer-Verlag. Berlin Heidelberg, Germany, 483-493.

Vargas M. 1961. Foundations of tall buildings on sands in Sao Paulo, Brazil. Proceedings of Fifth International Conference on Soil Mechanics and Foundation Engineering, Paris, 1, 841-843.

957

Water injection aided pile jacking centrifuge experiments in sand

Essais en centrifugeuse d’installation de pieux vérinés dans le sable avec injection d’eau

Shepley P., Bolton M.D. University of Cambridge

ABSTRACT: Jacked piles have several advantages over conventional piling techniques; namely the low noise and vibration outputfrom the process. However they are often difficult to install in hard ground conditions. A supplementary water injection technique can be used to reduce installation loads. Water is injected from the pile toe at high pressure and flow rates into the ground, achievinglarge reductions in installation loads. In this research, water injection aided pile jacking was modelled on a geotechnical centrifuge. A dense, fine sand was used and multiple pile installations were completed in order to investigate the effect water injection has on installation loads. To complete the modelling, a high water pressure supply for use on the centrifuge was developed and systems tomaintain centrifuge balance were implemented. This paper also identifies and validates a method for calculating pressures of interest based on the limited measurement locations available.

RÉSUMÉ : L’installation des pieux vérinés présente plusieurs avantages comparativement aux techniques de pieux conventionnelles, en particulier cette méthode est peu bruyante et génère de faibles vibrations. Cette technique est par contre difficile dans les solsdenses et compacts. L’injection d’eau supplémentaire permet de réduire les charges d’installation. L'eau est injectée au niveau de la base des pieux dans le sol sous forte pression et à haut débit, ce qui permet une réduction importante des charges d’installation. Danscette étude, l’installation des pieux vérinés assistée par injection d’eau a été modélisé dans une centrifugeuse géotechnique. Un sabledense et fin a été utilisé et de nombreuses installations de pieux ont été réalisées pour évaluer l'effet de l'injection d'eau sur les chargesd'installation. Pour compléter la modélisation, une nouvelle alimentation d'eau sous haute pression pour utilisation en centrifuge a étédéveloppée et un système de maintien de l'équilibre de la centrifuge a été implémenté. Cette recherche identifie et valide égalementune méthode de calcul des pressions d’intérêt basée sur des localisations de mesure limitées sur le terrain.

KEYWORDS: jacked pile, centrifuge, sand, water jetting, water injection

1 INTRODUCTION

Jacked, or silent, piling is an increasingly important method for pile installation. Construction projects in urban or sensitive areas desire a low-impact means of installing pile walls or piled foundations. This is due to their low disruptive nature – producing little noise and few ground vibrations (White et al. 2000). In addition, they often require fewer enabling works due to the smaller machinery (Goh et al. 2004).

However, jacked piles are restricted by the maximum deliverable installation force. Often this is limited by the available kentledge for counterweight systems. In the case of the jacked piling system produced by Giken Seisakusho Ltd., a Japanese piling contractor, reaction force is provided by the previously installed piles in the pile wall. Three or four piles are used in tension to provide the installation load for the subsequent pile. In this case the load limit is set by the capability of the machine, not necessarily the available reaction force. In all cases, if the piling load approaches the load limit of the machine, the installation rate may fall to an uneconomical level or even pile refusal.

To reduce or prevent these situations, a supplementary installation technique can be used. The aim of any such technique is to maintain the advantages of the installation, with low noise and vibration levels, but also reduce the installation load so that piles can be jacked into hard ground. Many techniques exist to achieve this, such as surging, pre-augering and gyropiling. However, the use of a supplementary water

injection during the pile installation to reduce the installation forces is of particular interest for this study.

Modelling of the water injection technique has been completed using the Turner Beam Centrifuge at University of Cambridge. A high pressure water supply was developed for use on the centrifuge in order to replicate the high water pressures and flow rates experienced in the field. This paper will outline the current use of the technique, in addition to the centrifuge modelling completed.

2 DEVELOPMENT OF WATER INJECTION

Water jetting has been in common use for decades, mainly for offshore pile installation (Tsinker 1988). The offshore setting provides a large water source and no nearby structures that may be affected or damaged. Typical flow rates for this early technique exceeded 1500 litres per minute in all soil types. In addition, water jetting was found to be disruptive to the soil fabric around the installed pile. The ground was liquefied so that the pile could be installed under self weight. This resulted in global particle rearrangement where large particles sank to the bottom of the pile installation.

The technique has since been improved to allow its more widespread use. Required flow rates were reduced to below 1000 litres per minute following a review from Tomlinson and Woodward (2008).

If the water jetting technique is used in conjunction with another pile installation method – pile jacking with supplementary water injection, then the flow rates can be

958


reduced further. There is no longer a reliance solely on the water flow to install the pile. Instead the pile is jacked and the water injection is used to reduce the required installation loads. Flow rates for this method reduce to less than 300 litres per minute, and depend on the size and type of pile being installed.

The aim of water injection is to aid pile installation with minimal impact to the surrounding ground. Water injection should only be required during periods of high pile installation loads. During these phases, high water injection rates would be required to reduce the installation loads. Once the installation loads are sufficiently reduced, the flow rate can be reduced unless pile loads begin to increase again.

Despite the variety of full scale testing completed, there is still uncertainty over the water injection technique. The main unknown is the governing mechanism. Some options have been suggested, most recently the scour system outlined by Schneider et al. (2008), however further research is required to investigate the technique further.

3 CENTRIFUGE MODELLING

Initially, the aim of the centrifuge testing was to find an effect on the pile installation load when using the water injection system.

3.1 Model construction

A body of fine sand was prepared to a relative density of 80 % in a centrifuge container, 850 mm in diameter, to a depth of 320 mm. This was saturated from the base with de-aired water.

The sand was prepared so that it possessed a low permeability by mixing fine Fraction E silica sand with a commercially available builders sand. To ensure continuity between tests, the sand was repeatedly sampled and the particle size distribution (PSD) was found for different batches using the Single Particle Optical Sizing (SPOS) technique. Figure 1 shows the particle size distribution of the mixed sand compared with the Fraction E and builders sand components.

Figure 1. PSD comparison of the mixed sand for testing with standard sand types, Fraction E and a builders sand.

3.2 Model pile

A bespoke instrumented model pile was constructed for the testing program. A stainless steel tube of 12 mm outside diameter was used, with a water delivery pipe running through the centre. Stainless steel was chosen due to its strength, hardness and resistance to corrosion – preventing buckling during testing or surface abrasion over multiple installations. This ensured consistency over all the installations. A photograph of the pile is shown in Figure 2.

Strain gauges were used to monitor the axial load at the pile toe and the pile head. Two full Wheatstone bridges were used at each location.

The water delivery pipe was a 2.5 mm internal diameter plastic pipe. This terminated at a detachable nozzle at the pile toe which could be easily changed between tests. Different nozzles were used throughout the test program. Nozzles using only a central orifice will be assessed in this paper. These were modelled on small orifice plates, with a nozzle diameter of 0.5, 1.0, 2.5 and 3.0 mm.

Figure 2. Photograph of the model pile as used, with nozzle attached at the toe and visible strain gauges at the pile head.

3.3 Water injection system

In order to model the water injection technique, a new system was required to provide high pressure water to the pile at a relatively high flow rate. Previous centrifuge testing of water jetting used low flow rates and pressures, due to the chosen pumping system.

Typical pumping systems for use one board a centrifuge package are based on a syringe pump. Such systems are commonly used for modelling excavations, where fluid is drained from a region to simulate ground volume loss, or for simulating pile jetting, such as the jetted spudcan experiments of Gaudin et al. (2011). Syringe pumps are limited by the actuator used to drive the piston. The actuator provides a high degree of control over the flow, but also restricts its use to low flow rate and low pressures. In addition, syringe pumps typically have a small volume capacity, meaning it is difficult to maintain high flow rates for a long period of time.

To avoid this issue during testing, the new system developed derived water pressure from the radial acceleration down the centrifuge arm. Water was provided to the slip rings at typical mains pressure (around 200 kPa) and then fed to the package through a pipe running down the beam. Moving through the gravitational field gives an increase in pressure according to:

2225.0 ringslippackageringsslippackage rrPP (1)

where P is the pressure at the package and slip rings

measured in Pascal, ω is the angular velocity of the centrifuge in rad/s and r is the radius from the centre of the beam of the package and slip rings in metres.

This procedure developed peak pressures at the model of 1.2 MPa and sustainable flow rates of up to 3.5 litres per minute. Water pressure and flow rate were monitored at the centrifuge model, a short distance from the pile toe. This location was chosen for the simplicity of mounting a pressure transducer and a turbine flow meter in the water delivery system. In addition, a solenoid valve was used to allow or terminate flow to the pile.

Pressure at the pile toe could be calculated following the centrifuge test using pipe flow theory as laid out by Goforth et al. (1991). Loss factors can be confirmed by comparing calculated values with data taken during a flow test – where the pile toe is suspended above the sand surface and water is passed through the system. The calculations can then be extended to allow for different toe positions in the acceleration field and the toe pressure at all pile depths can be found.

Flow rate control was achieved using a manually operated flow tap before the slip rings. This controlled the water flow

959


delivered to the centrifuge. Any changes were made by hand during the centrifuge flight. The on board instrumentation is monitored to ensure that a consistent and appropriate flow was being delivered to the pile. The position of this control tap governed the peak flow rate and was unchanged throughout a single installation.

3.4 Maintaining balance

The centrifuge at Cambridge is balanced using a fixed mass counterweight. This is cumbersome to change during a test week and cannot be changed mid-flight. Therefore the mass of the experimental package had to remain constant throughout the centrifuge test, despite adding water to the package at very high flow rates.

A passive standpipe system was designed in order to drain excess water out of the experimental package into the centrifuge chamber. The standpipe was positioned within the sand body near the edge of the container – remote from any pile locations. A set of holes at the base of the standpipe linked the water level in the standpipe to the water table in the sand body.

Holes at the top of the standpipe allowed water to drain out of the package through a set of drainage pipes. If the water level exceeded the design water level at any point, water would exit the package by draining through these top holes.

To monitor the success of the standpipe, pore pressure transducers were used. A series of these were positioned in the sand body to monitor the pore pressures around an advancing pile installation. Additionally, these transducers provided knowledge of the water table position in the model. A further transducer was placed at the base of the standpipe to check that the drainage system was functioning.

3.5 Testing program

All centrifuge tests to be presented in this paper were completed at an acceleration of 60g. According to length scaling, this modelled a 720 mm diameter, close-ended tubular pile installed to a depth of 11.4 m. For the purpose of future discussion, all future units will be at the model scale.

A soil stabilisation loop was completed before the first installation in order to prevent excessive change of the sand body between the first and subsequent flights. Following this, multiple pile installations were completed in a single flight using the centre's 2D actuator (Haigh et al. 2010). Piles in a single flight were spaced at 140 mm (12Dp), but final pile spacing was close to 70 mm (6Dp). A typical pile layout is shown in Figure 3.

The nozzle at the pile toe was changed between flights to investigate the importance of the nozzle layout. The nozzles restricted the peak achievable flow rate, in addition to attracting further pressure losses at the pile toe.

4 RESULTS

The discussion of results will be split into sections to discuss the success of the water injection system and pile installation information.

4.1 Water injection system

The novel water injection system proved to be successful. The feeder pressure from the mains water supply provided a relatively steady pressure of 200 kPa during testing. The flow rate to the beam was controlled using the manual control tap; a variety of flow rates were possible using this simple control.

Multiple flow rates were essential in order to calibrate the loss factors in the pipe between the measurement point and the pile toe. Increased confidence in the calculation could be achieved if more unique flow rates were tested. Figure 4 shows a plot of the data points used to find the loss factors for four different nozzle sizes.

On comparing these flow test results, the effect of changing the nozzle becomes immediately apparent. As predicted, the smallest nozzle attracts the largest pressure losses; denoted by the steeper gradient lines of best fit in the figure. This is a similar result as monitoring the pressure loss from small orifice plates blocking flow through a pipe and highlights how the loss factors are dominated by the nozzle used.

With the larger nozzle sizes, larger flow rates were achievable with smaller losses. There is little to no difference between the 2.5 mm and the 3.0 mm diameter nozzles due to their similar size to the feeder pipe. The 2.5 mm nozzle acts as a continuation of the feeder pipe, and the 3.0mm nozzle effectively reduces the sharpness of the pipe exit; both have little effect on the pressure loss.

4.2 Maintaining balance

The standpipe system maintained the balance of the centrifuge. As shown in Fig 5, the pressure of the standpipe remains constant throughout the flight plotted. The two dotted lines for the standpipe PPT represent brief periods where the instrument failed during the test.

Figure 3. Typical pile layout in a single test week. At least four flights are completed at 140 mm pile spacing in each flight.

Figure 4. Energy loss per unit volume of water passing between the pressure line and the termination nozzle. All lines of best fit shown have a correlation R2 value greater than 0.94. The smallest diameter nozzle attracted the largest loss, as expected. The 2.5 mm and 3.0 mm nozzles attracted the same loss due to their relative size to the feeder pipe.

960


Figure 5. Illustration of a water injection event and subsequent model drainage. Water is added to the model at around 3 litres per minute for nearly 200 seconds (highlighted by the shaded region). This causes the sand body to fill with water, represented by the increase in the normalised pressure. Meanwhile, the stand pipe water level remains constant and the pressure unchanged. (A dotted line represents a brief period where the instrument failed.)

During an injection event, a small difference is evident in the

system. This is indicated by the increase in the sand body water pressure during the injection phase. A small pressure difference between the sand body and the standpipe arises, driving water flow into the standpipe. Any additional water in the standpipe drains away through the drainage holes at the top of the standpipe.

The base pressure of the standpipe remains constant throughout the process, showing that the standpipe maintains a constant water height, as designed. With time, the pressure difference between the standpipe and the sand body reduces. This will slow the rate of drainage through the standpipe, until the pressures become equal and no excess water is present in the model after a time greater than 1100 seconds.

4.3 Installation load results

Multiple centrifuge tests were completed on identical sand bodies. The sand bodies were tested for their continuity via a control installation, without water injection. This was effectively a penetration test and gave a reference to compare the water injection aided installations to. There was good agreement between the different control installations over the multiple sand bodies used.

Figure 6 shows some installation data. The pile head loads for three installations are shown – a control installation and two water injection installations, one using a 1.0 mm diameter nozzle and the other a 3.0 mm diameter nozzle. In addition, the model flow rate is plotted, to show the link between the delivered flow rate and reduction in load when compared to the control installation.

Initially, load is generated in all installations as all piles are installed to a depth of 2Dp without the aid of water injection. At this depth, the water supply to the pile is activated and the load reduces to zero. The flow rate is allowed to stabilise at this level as the pile installation continues.

The difference between the two nozzles is apparent, with the smaller nozzle allowing a smaller peak flow rate to be pushed through the pile. Despite a significant flow rate of 1.3 litres per minute, there is little reduction in the pile load. The maximum load reduction is experienced at the shallower depths, where the load can be reduced to zero. Load reduction then diminishes with depth as the pressure at the pile toe becomes closer to the hydrostatic pressure in the sand body.

Figure 6. Comparison of effects of nozzles. Diameters 1.0 and 3.0 mm nozzles are compared with the no injection installation. The flow rate delivered to the pile is displayed for all installations in the right plot.

5 CONCLUSION

A water injection system has been successfully developed and tested on the centrifuge at University of Cambridge. Peak pressures of 1.2MPa and peak flow rates of 3.5 litres per minute were delivered to the model pile. The system has been tested to find the energy loss in the pipe line, with the aim of calculating the toe pressure during an installation.

To maintain centrifuge balance during high flow rate events, a standpipe system was developed and its performance closely monitored. This proved successful for the duration of testing.

In addition, the system has been used to complete multiple installations of water injection aided jacked piles. Different nozzles were tested during the experimental program to investigate their different effects. Whilst the effect of injection reduces with depth, it was discovered that the larger nozzles were the most effective at reducing installation loads.

6 ACKNOWLEDGEMENTS

The authors would like to thank Giken Seisakusho Ltd. for their continued support throughout the duration of the research.

7 REFERENCES

Gaudin C., Bienen B. and Cassidy M.J., 2011. Investigation of the potential of bottom water jetting to ease spudcan extraction in soft clay. Geotechnique, 61(12) 1043-1054.

Goh T., Shiomi T., Yamamoto M., Ikeda T. and Motoyama M. 2004. A solution for road construction. In 6th Malaysian road conference, Kuala Lumpur.

Goforth G.F., Townsend F.C. and Bloomquist D. 1991. Saturated and unsaturated fluid flow in a centrifuge. In Centrifuge in soil mechanics, 497-502, Ko and McLean.

Haigh S.K., Houghton N.E., Lam S.Y., Li Z. and Wallbridge P.J. 2010. Development of a 2D servo-actuator for novel centrifuge modelling. In 7th international conference on physical modelling in geotechnics, 239-244, Zurich.

Schneider J.A., Lehane B.M. and Gaudin C. 2008. Centrifuge examination of pile jetting in sand. In 2nd IPA workshop, 17-24 New Orleans.

Tomlinson M. and Woodward J. 2008. Pile design and construction. Taylor & Francis, London

Tsinker G.P. 1988. Pile jetting. Journal of geotechnical engineering 114(3), 326-334.

White D.J., Sidhu H.K., Finlay, T.C.R., Bolton M.D. and Nagayama T. 2000. The influence of plugging on driveability. In 8th international conference of the deep foundations institute, 299-310. New York.

961

Shear Behaviour of Rock Joints under CNS Boundary Conditions

Comportement en cisaillement de joints rocheux en condition de raideur normale constante

Shrivastava A.K. Delhi Technological University, Delhi, India

Rao K.S. Indian Institute of Technology, Delhi, India

ABSTRACT: The shear behaviour of rock joints depends up on many factors, the correct evaluation of this is possible only if these parameters are properly considered during experimental investigation, constitutive modelling and numerical modelling. Which is important for safe and economical design of underground openings in jointed rocks, stability analyis of rock slopes, risk assessment ofunderground waste disposal, design of foundation on rock and design of rock socketed piles. These concerns invite accurate quantification of shear strength of unfilled and infilled joints, proper understanding of the basic mechanics of discontinuity and theprinciples involved in their shear deformation. This can be done through in-situ or laboratory large scale testing on natural rock or laboratory testing on physical model. In the present paper the detail account of test results of direct shear tests performed on large sizemodeled unfilled and infilled rock joints under different boundary conditions is systematically presented. It is observed that the constant normal stiffness (CNS) conditions better simulate the field conditions of the loading and shear strength predicted under CNScondition is more than the constant normal load (CNL) conditions for both unfiled and infilled joints.

RÉSUMÉ : Le comportement en cisaillement de joints rocheux dépend de nombreux facteurs, dont l'identification n'est possible quepar approche expérimentale numérique ou rhéologique. Cela est important pour la conception sécuritaire et économique de cavités souterraines dans les roches fracturées, l'analyse de stabilité des talus rocheux, l'évaluation du risque d'élimination des déchetssouterraine, la conception de fondation au rocher et la conception pieux. Une quantification précise de la résistance au cisaillement des joints remplis ou non, ainsi qu'une bonne compréhension des mécanismes de base de la discontinuité et des principes appliqués àleur déformation en cisaillement sont nécessaires. Ceci est possible grâce à des essais en vraie grandeur in situ ou en laboratoire sur modèle physique. Dans le présent document, le détail des résultats des essais de cisaillement direct effectués en vraie grandeur avecou non un remplissage des joints sous différentes conditions aux limites sont systématiquement présentées. On a observé que les conditions de la rigidité normale constante (CNS) simulent mieux les conditions sur le terrain et que la résistance au cisaillementprédite sous condition de CNS est plus grande que pour les conditions de charge normale constante (CNL) pour les deux types de joint.

KEYWORDS: Shear Behaviour, Rock Joints, CNL, CNS, Direct Shear, Infill, Unfill, Dilation, Shear Strength, Deformation.

1 INTRODUCTION

Rock joints are mechanical discontinuities having geological origin. These discontinuities are present in the form of joints, faults, bedding planes or other recurrent planar fractures in the rock mass. In general, strength and deformability properties of these discontinuities are quite different from those of intact rock, and in many cases, the discontinuities completely dominate the shear and deformation behaviour of the in situ rock mass in a given stress conditions.

The presence of infill or gouge material in the joints further reduces the shear strength. The sources of infill material include products of weathering or overburden washed into open joint, water conducting in discontinuities, precipitation of minerals from the ground water, by-products of weathering alterations along joint walls, crushing of parent rock surfaces due to tectonic and shears displacements, and thin seams deposited during formation. In general, infill materials may consist of partially loose to completely loose cohesionless soil or fine grained clay. Normally fine-grained clays are more frequently found as fillers and are more troublesome in terms of structural stability. Thickness of the infill material varies from micrometers to several meters and it plays an important role in shear behaviour. In tectonically crushed zones, the infill thickness may exceed several meters.

These rock joints unfilled or infilled are the weakest plane which tries to slide or shear one over the other due to construction of foundations of a structure and tunnels or

highways and railways on the rock slopes. Hence, for safe and economical analysis of all the above cases it is important to understand the strength and deformation behaviour of the rock joints under direct shearing conditions. Shrivastava and Rao (2009) discussed in details the influence of factors like (a) boundary condition (b) shear rate (c) joint roughness (d) size of joint i.e. scale effect (e) joint condition i.e. unfilled joint/in filled joint on the direct shear strength of rock joints.

There are two boundary conditions i.e constant normal load (CNL) and constant normal stiffness (CNS) boundary conditions under which the shear behaviour of rock joints can be studied. The planar rock joints can be investigated in the laboratory by using a conventional direct shear apparatus where the normal load is kept constant (CNL) during the shearing process. This particular mode of shearing is suitable for situations where the surrounding rock freely allows the joint to shear without restricting the dilation or there is no dilation during the shearing process, thereby keeping normal stress constant during shearing process. Shear testing under a constant normal load (CNL) boundary condition is only beneficial for cases such as non-rereinforced rock slopes or planar rock joints, but natural rock joints are seldom planar.

However, for non- planar discontinuities, shearing results in dilation as one asperity overrides another, and if the surrounding rock mass is unable to deform sufficiently, then an inevitable increase in the normal stress occurs during shearing. At any time t if the normal stress Pn (t) then increase in normal stress

962


on the shear plane at any time t +Δt is equal to Pn (t) + kn Δ δv (t+ Δ t), where kn is the stiffness of the surrounding rock mass and Δ δv (t+ Δ t) is the dilation restricted in the given interval of time. Therefore, shearing of rough joints under such circumstances no longer takes place under constant normal load (CNL), but rather under variable normal load where stiffness of the surrounding rock mass plays an important role in the shear behaviour. This particular mode of shearing is called as shearing under constant normal stiffness (CNS) boundary conditions. For analysis and design of tunnels, foundations and rock slopes, shear tests results under CNL condition are not appropriate. A more representative behaviour of joints would be achieved if the shear tests were carried out under boundary conditions of constant normal stiffness (CNS).

In past decades numerous shear models have been proposed based on experimental, analytical and numerical study to find out the shear behaviour of rock joint. These models available in the literature fail to appropriately determine shear behavior of rock due to limitations of boundary condition i.e. CNL boundary condition is used for modeling like (Patton 1966, Barton 1973 and 1976, Haberfield and Johnston 1994 and Yang and Chiang 2000).

But very few studies are available under CNS condition i.e. constant normal stiffness conditions. To study the shear behaviour under CNS conditions, the conventional direct shear test apparatus working under CNL boundary condition is modified by different researchers like, (Obert et al. 1976, Ooi and Carter 1987, Johnston et al. 1987, Indraratna 1998, Gu et al. 2003 and Kim et al. 2006) to be used for CNS boundary conditions.

Despite frequent natural occurrence of infill material, filled discontinuities have been studied much less, perhaps because of the difficulties arising from sampling, testing or due to increased number of variable parameters for constitutive and numerical modelling. Due to limited research, it is a common practice to assume the shear strength of an infilled joint equal to the infill material alone, regardless of its thickness. Kanji 1974 reported that the shear strength of the infilled joint is lower than that of the infill material. Hence this assumption will lead to unsafe designs. These uncertainties in estimation have motivated the present work.

2 PHYSICAL MODELLING OF ROCK JOINTS

It is difficult to interpret the results of direct shear test on natural rock because of difficulty in repeatability of the sample. To overcome this problem a model material is searched which can easily be handled and reproducibility of the sample can be ensured. To achieve this different brands of plaster of Paris and dental plasters at different moisture content and curing period in isolation or combinations have been tried. Finally, plaster of Paris is selected because of its universal availability and its mould ability into any shape when mixed with water to produce the desired joints and also long term strength is independent of time once the chemical hydration is completed. To characterize model material a series of physical and mechanical tests on a number of specimens prepared by mixing the prescribed quantity of water with plaster of Paris powder were carried out. The prescribed percentage of water is decided so as to achieve proper workability of the paste and required strength to simulate the soft rock. Different water cement (POP) ratios were tried in order to obtain desired strength and workability. The ratio which is finally selected is 0.60.

The physical and engineering properties of the model material were determined in the laboratory as per the suggested methods of ISRM 1977 and 1979. The average uniaxial compressive strength and tangent modulus at 50% of peak axial stress of model material at 0.60 water cement (POP) ratio and after 14 days of air curing is 11.75 MPa and 2281 MPa respectively. Thus, the material can be classified as ‘EL’ based

on Deere and Miller 1966 classification chart, indicating that the material has very low strength (E) and low modulus ratio (L). The cured plaster of Paris samples showed a consistent uniaxial compressive strength (σc) in the range of 10.58 to 13.22 MPa and a Young’s modulus of 1856 to 2631 MPa. These ranges of strength and modulus values are suitable for physically and mechanically simulating the behaviour of jointed rocks like siltstone, sandstone, friable limestone, clay shale and mudstone.

2.1 Preparation of unfilled rock joint samples

The asperity plate of 150-150 angle designed and fabricated by Rao and Shrivastava 2009 has been used to produce desired asperity in the sample as shown in Fig. 1(a). The plaster of Paris with 60% of the moisture is mixed in the mixing tank for 2 minutes and then the material is poured in the casting mould which is placed on the vibrating table. Vibrations are given to the sample for a period of 1 minute and then the sample is removed from the mould after 45 minutes and kept for air curing for 14 days before testing.

2.2 Preparation of infilled rock joint samples

The infill material is selected to simulate the field conditions. In the present work combination of fine sand and mica dust both passing through 425micron sieve and plaster of Paris is selected. The selected composition is plaster of Paris 40%, fine sand 50% and mica dust 10% mixed together with water 45% by weight of total mass of the material. The uniaxial compressive strength of the 7 days air cured infill material is 3.47 MPa and direct shear tests carried on the infill material gave friction angle and cohesion, 28.80 and 0 respectively.

The infill joint with required thickness as shown in Fig. 1 (b) is created on the sample with the help of infill mould developed by Shrivastava et al. 2011.

Figure 1. Photograph of simulated rock joints (a) unfilled (b) infilled.

The samples are placed on the mould and tighten at suitable point so that the required thickness of the infill material is created. The infill material is spread over the lower sample and the asperity plate is put over the infill material and the asperity plate is compressed from the top with the help of C- clamps so that the uniform pressure is applied on the sample and the same

963


asperity is created on the infill material. The upper mould is now placed over the lower mould with the help of the guide rod and movable screw the correct placement and thickness of the infill material is insured. The whole assembly is now compressed from the top with the help of C- clamp, after 30 minutes the sample is removed from the mould and kept for air curing for 7 days before testing.

3 SHEAR BEHAVIOUR

ferent initial normal stress (P ) ranging from 0.1

d is having the facility to collect data and plot online gra

actuator to provide the programmed for

ck joints are set to be 8 kN/mm y condition.

view of large scale direct shear machine (Rao and Shr stava 2009).

ate of shearing is maintained as 0.5mm/min during each test.

To study the effect of CNS boundary condition and infill material on the shear behaviour of the rock joints the extensive tests were planned and conducted under different boundary conditions on the equipment designed and developed by the authors as shown in Fig. 2, on 150-150 asperity unfilled and infilled joint at dif i

to 2.04 MPa. In this equipment Normal and shear load is applied through

an electro hydraulic servo actuator unit which works on closed loop principle. The displacements are measured by LVDT’s mounted on the specimen. The data acquisition system has 16 channels, 2 channels for load cell, 6 channels for LVDTs and remaining 8 channels are free for additional input. The data acquisition system converts the mechanical and electrical signals in to the digital data. The output of signal is connected to CPU via cord. The load and deformation values are stored at desired intervals as note pad data. The direct shear software develope

phs. In this apparatus CNL and CNS boundary conditions are

reproduced by an electro hydraulic servo-valve which under the control of an electronic controller controls the application of hydraulic power to a linear

ce to the test specimen. The thickness of the infill material (t) and height of asperity

(a) is maintained at 5mm for the present case. The normal stiffness (kn) of surrounding rofor CNS boundar

Figure 2 Close up iva The effect of shearing rate for different asperity joint under

different boundary conditions have been studied by Rao et al. 2009 and they found that the effect of increasing shear rate for shearing rate > 0.5mm/min is to increase the peak shear stress for the same initial normal stress and for shearing rate ≤ 0.5 mm/min, the effect of shearing rate is not much on the peak shear stress. Hence, for the present case r

The shear behaviour of 150 -150 asperity unfilled and infill joint under CNL (kn=0 kN/mm) and CNS (kn=8 kN/mm) boundary condition is plotted as shown in Fig.3 and Fig.4 respectively. The stress – displacement behaviour is characterized by a well defined peak. It is clear from the test result that CNL boundary condition always under predicts the shear strength of the joint as compared to CNS boundary condition for the same initial normal stress. This is due to increase in normal stress at the shearing surface during the shearing because of restriction in dilation imposed by simulated surrounding rock stiffness.

Shrivastava and Rao 2011 reported the variation of nomal stress with shear displacement under CNL and CNS boundary condition for similar type of synthetic rock joints. The normal stress on the shear plane remains constant during testing for CNL conditions. However, for CNS conditions normal stress increases as asperity slides on over the other. Variation of normal stress under CNS conditions exactly follows the shape of the asperity, but angle of inclination is different.

The shear strength of the infill joint is less than that of unfill joint for both CNS and CNL condition, when tested under the same Pi. But for CNS boundary conditions % decrease of shear strength of infill joint is lower at higher Pi. It may be due to failure of infill material under increased compression.

The shear stress and displacement behaviour curve of modelled rock joint can be divided into three zones. In the zone I predominantly sliding of the sample take place without shearing of the asperity. The limit of the zone-I depends upon the shear strength of the material and shear stress increases at higher rate with small shear displacement in this zone. In zone-II, shearing of the asperity is more predominant than the sliding. The limit of the zone-II is up to maximum shear stress, in this zone rate of increase in shear stress decreases with shear displacement. Zone-III is the last zone where all the asperity is sheared off. Due to deposition of the crushed material on the joints, shear stress decreases or increases slightly with shear displacement depending upon CNL or CNS conditions.

Probable strength envelope is found by joining the peak shear stress of different stress path and plotted as shown in Fig.5.

Shear Displacement (mm)0 5 10 15 20 25

She

ar S

tress

(MP

a)

0.0

0.5

1.0

1.5

2.0

2.5

3.0

CNL, Pi=0.10 CNL, Pi=0.31 CNL, Pi=0.51 CNL, Pi=1.02 CNL, Pi=2.04 VNL, Pi=0.05 VNL, Pi=0.10 VNL, Pi=0.31 VNL, Pi=0.51 VNL, Pi=1.02 VNL, Pi=2.04

MPa

CNL, kn=0 kN/mmVNL, kn=8 kN/mm

Figure 3. Shear behaviour of 150-150 unfilled joint under CNL and CNS boundary condition.

The shear test result on 150 - 150 asperity modelled rock joint reflects that the strength envelope for both CNL and CNS boundary condition is curvilinear and curvature is same up to low Pi i.e Pi <0.09 σc of the sample and after that the curvature of the strength envelope is change. The change in the slope of the strength envelop indicates that the complete shearing of the asperity at that normal stress and sliding of samples takes place after that normal stress. At low normal stress, shear strength significantly increased because of the enhanced shearing resistance offered by the angular asperities. However, at higher

964


normal stress increased degradation of the asperities is associated with decrease in increase of the shear strength and at very high normal stress the complete shearing of the asperity takes place and there is no effect of boundary conditions on shear strength.

II. The % increase in shear strength of unfilled joint under CNS conditions as compared to CNL conditions is as high as 221 for Pi=0.10 MPa.

Shear Displacement (mm)0 5 10 15 20

She

ar S

tress

(MP

a)

0.0

0.5

1.0

1.5

2.0

CNL, Pi=0.10CNL, Pi=1.02CNL, Pi=2.04 CNS, Pi=0.10CNS, Pi=1.02CNS, Pi=2.04

MPa

CNL, kn=0 kN/mCNS, kn=8 kN/m

mm

i

III. The effect of the infill material in the joint is to reduce the shear strength and a maximum reduction of 35% is observed for 5mm infill thickness under CNS conditions at P =0.10 MPa.

IV. The effect of boundary conditions on the shear strength of non planar unfilled/ infilled rock joints decreases with increase in Pi, the effect is almost nil for Pi≥0.18σc.

5 REFRENCES

Shrivastava A.K. and Rao K.S. 2009. Shear behaviour of jointed rock: a state of art. IGC-Guntur, 245-249.

Patton F.D. 1966. Multiple modes of shear failure in rock and related materials. PhD Thesis, University of Illinois, Urbana.

Barton N 1973. Review of a new shear strength criterion for rock joints. Engineering Geology 7, 287–332.

Barton N. 1976. The shear strength of rock and rock joints. Int. J. Rock Mech. Min. Sci. and Geomech. Abst. 13, 255-279.

Yang Z.Y. and Chiang D.Y. 2000. An experimental study on the progressive shear behaviour of rock joints with tooth-shaped asperities. Int. J. Rock Mech. Min. Sci. 37, 1247–1259. Figure 4. Shear behaviour of 150-150 infilled joint under CNL and CNS

boundary condition. Obert L., Brady B.T. and Schmechel F.W. 1976. The effect of normal stiffness on the shear resistance of rock. Rock Mech. 8, 57-72.

Ooi L.H. and Carter P.J. 1987. A constant normal stiffness direct shear device for static and cyclic loading. Geotechnical Testing Journal 10, 3-12.

Initial Normal Stress (MPa) 0.0 0.5 1.0 1.5 2.0 2.5

Pea

k S

hear

Stre

ss (M

Pa)

0.0

0.5

1.0

1.5

2.0

2.5

3.0

CNL unfillCNS unfillCNL infillCNS infill

Johnston I.W., Lam T.S.K. and Williams A.F.1987. Constant normal stiffness direct shear testing for socketed pile design in weak rock. Geotechnique 37, 83-89.

Indraratna B., Haque A. and Aziz N. 1998. Laboratory modelling of shear behaviour of soft joints under constant normal stiffness condition. J. Geotechnical and Geological Engineering 16, 17-44.

Gu X. F., Seidel J. P. and Haberfield C. M. 2003. Direct shear test of sandstone- concrete joints. Int. J. of Geomechanics 3, 21-33.

Kim D.Y., Chun B.S. and Yang J.S. 2006. Development of a direct shear apparatus with rock joints and its verification tests. Geotechnical Testing Journal 29, 1-9.

Kanji M.A. 1974. Unconventional laboratory tests for the determination of the shear strength of soil-rock contacts, Proc. 3rd

Congr. Int. Soc. Rock Mech., Denver 2, 241-247. ISRM 1977. Suggested method for determining water content,

porosity, density, absorption and related properties and swelling and slake-durability index properties.

Figure 5. Strength envelope of 150-150 unfilled and infilled joint under CNL and CNS boundary condition. ISRM 1979. Suggested method for determining the uniaxial

compressive strength and deformability of rock materials. The increase in shear strength for unfilled joints for CNS conditions varies from 221% to 6% of the CNL conditions when P increases from 0.10 MPa to 2.04 MPa.

Deere D.U. and Miller R.P. 1966. Engineering classification and index properties of rock, Technical Report No. AFNL-TR-65-116, Albuquerque, N.M : Air Force Weapons Laboratory.

iThe curvilinear strength envelope for infilled joint is

observed as presented in Fig.5 at all range of Pi. But peak shear stress of the infill joint is always less than that of unfilled joint. Maximum reduction in shear strength of the infill joint as compared to unfill joint for CNS condition is observed to be 35% at Pi=0.01 MPa and % reduction in shear strength decreases with increase in the Pi. At high Pi, the close look on the sheared sample reflected breakage of the infill material, which has resulted into more participation of the joints, hence less reduction in shear strength.

Shrivastava A.K. 2012, Physical and Numerical Modelling of Shear Behaviour of Jointed Rocks Under CNL and CNS Boundary Conditions. Ph.D. Thesis, IIT Delhi.

Shrivastava A.K., Rao K.S. and Rathod, G.W. 2011. Shear behaviour of infill joint under CNS boundary condition. IGC - Cochi, 981-984.

Rao, K.S., Shrivastava, A.K. and Singh Jattinder, 2009. Universal large scale direct shear testing machine for rock. INDOROCK- New Delhi, 157-168.

Shrivastava A.K. and Rao K.S. 2011. Shear behaviour of non planar rock joints. 14th ARC on Soil Mechanics and Geotechnical Engineering, Hong Kong, China, 1-6.

4 CONCLUSIONS

The experimental studies on physically modeled unfilled and infilled rock joints have been conducted to understand the effect of boundary conditions and infill thickness on shear behavior. The conclusions made from the test results are summarized below:

I. CNL boundary condition is not suitable for non planar rock joints and it under predict the shear strength, which makes the design uneconomical.

965


1

Experimental study on compaction grouting method for liquefiable soil using centrifuge test and X-ray tomography

Etude expérimentale sur la CPG pour le sol liquéfiable par centrifugation et tomographie à rayons X

Takano D., Morikawa Y. Geotechnical Engineering Division, Port and Airport Research Institute, Japan

Nishimura S. Faculty of Engineering, Hokkaido University, Japan

Takehana K. Engineering Department, Geodesign Co., Ltd., Japan

ABSTRACT: Compaction grouting, an in-situ static compaction technique by means of grout injection, has been increasingly adopted for improving the liquefaction resistance of loose sandy ground in recent years. The surrounding ground’s stress changes and densification induced by grout injection are considered to be an important cause for the stabilization effects. The present study investigates characteristic of ground deformation by simulating compaction grouting processes in micro focus X-ray tomography. 3D Volumetric Digital Image Correlation (V-DIC) techniques were applied to the tomographic images. V-DIC analysis of in-situ-acquired tomographic images provides a characterization of porosity, displacement and strain field of model ground. Additionally,simulation of ground injection and shaking table test was carried out in a geotechnical centrifuge to evaluate stress changes and liquefaction resistance of improved sandy soil.

RÉSUMÉ : Coulis compactage, une technique in-situ compactage statique à l'aide d'injection de coulis, a été de plus en plus adoptéepour améliorer la résistance à la liquéfaction du sable loos terrain ces dernières années. Changements de contrainte du sol environnantet de densification induite par injection de mortier sont considérées comme une cause importante pour les effets de stabilisation. Laprésente étude examine caractéristique de la déformation du sol en simulant les processus de compactage coulis au point tomographieà rayons X micro. 3D volumétrique Digital Image Correlation (V-DIC) techniques ont été appliquées aux images tomographiques. V-DIC analyse in-situ acquis par images tomographiques fournit une caractérisation de la porosité, de déplacement et champ dedéformation du sol modèle. En outre, la simulation de l'injection de sol et de test table à secousses a été réalisée dans unecentrifugeuse géotechnique pour évaluer les changements de résistance au stress et la liquéfaction de sol sablonneux améliorée.

KEYWORDS: compaction grouting, ground improvement, liquefaction, sand, full field measurement.

1 INTRODUCTION

In-situ static compaction by means of grout injection (CPG) is widely used as a countermeasure against liquefaction in loose sandy ground (e.g. Boulanger and Hayden 1995, Miller and Roycroft 2004). An increase in the liquefaction resistance of sand caused by compaction grouting is presumed to derive from three possible mechanisms. They are, (i) increase in the lateral confining stress, (ii) densification and (iii) reinforcement by hydrated and hardened grout piles. Especially, the surrounding ground’s stress changes and densification induced by grout injection are considered to be an important cause for the stabilization effects. However, systematic studies of ground condition changes due to compaction grouting have been limited in number.

The author’s research group has been studied about ground behavior due to grout injection especially focusing on stress change due to grout injection. The present study evaluate characteristic of ground deformation by simulating compaction grouting processes in 1g and centrifuge model test. In 1g test, the deformation process of the model ground by grout injection was visualized using X-ray computed tomography (XRCT). Moreover, 3D Volumetric Digital Image Correlation (3D V-DIC) techniques were applied to the x-ray tomographic images in order to discuss deformation of the ground quantitatively. 3D V-DIC analysis of in-situ acquired tomographic images provides a characterization of porosity, displacement and strain field of model ground. Additionally, simulation of ground injection and shaking table test was carried out in a geotechnical centrifuge to evaluate stress changes and liquefaction resistance of improved sandy soil.

2 SUMMARY OF TESTING

A microfocus X-ray tomographic scanner and a beam-type geotechnical centrifuge owned by Port and Airport Research Institute Japan were used in this study. Figure 1 illustrates a typical arrangement of grout piles with regular spacing in a triangular pattern. In order to simulate completion of this improvement arrangement in a physical modeling in laboratory, a cylindrical soil container with diameter of 60 mm (see Fig. 2) and hexagonal cylinder with diameter of 100 and 160 mm (see Fig. 3) are used for the XRCT test and the centrifuge test, respectively. The difference of diameter of soil container in centrifuge test presents the different pile spacing or improvement ratio and rigid wall of soil container simulates improved area by pre-injected grout piles. The ground was

Figure 1. Illustration of compaction grouting (bottom-up pattern)

４

Improved zoneby one grout pile

Lifting rod

Pile diameter, d

Pile spacing, x

１

２

３

５

６

７

Injection

966



2

Figure 2. Schematic view of XRCT test setup Figure 3. Schimatic view of test setup in centrifuge (CPGs30).

Table 1. properties of Soma Silica Sand #5 Table 2. List of test cases in centrifuge

prepared by air-pluviation targeting the relative density of 50 % and Soma Silica Sand #5 is used as ground material. The properties of Soma Silica Sand #5 are shown in Table 1. The injected grout prepared mixing the Soma Silica Sand #5, the Kawasaki Clay, Portland cement and water at the ration of 40 %:60 %:12 %:50 % by weight.

60106

130254

100342

442

Injection rod

Scanned area

Unit: mm

As the procedure of XRCT test, the soil container was mounted on turn table of the XRCT scanner after preparing model ground and then grout pile was injected in a bottom-up sequence through the injection rod. All injection and CT scanning processes are carried out in-situ condition (in-situ meaning x-ray scanning at the same time as injection) and full volume of cross sectional images are recorded every two step of grout injection. The injected volume was controlled to make the grout pile diameter of 20 mm. The details of the injection system can be referred in Nishimura et al. 2012. In 1g model test using miniature model ground as XRCT test in this study, it was concerned that the uplifting deformation caused by grout injection will dominate because of lower effective confining pressure therefore densification effect will not be evaluated correctly. In order to avoid this effect, 10 kPa of overburden pressure was applied at tip of injection rod by filling up soil container above the model ground by stainless cubes with 2 mm of diameter. Reconstructed volume images were analyzed using 3D V-DIC in order to evaluate the deformation process during grout injection.

In centrifuge tests, the model ground was prepared as same procedure as CT test. All the test procedure, saturation of model ground, injection of grout pile and shaking table test were carried out under a centrifugal acceleration of 30 G. Viscous degassed fluid (30 mm2/sec of dynamic viscosity) was used as pore water. The cross section of model ground presents shearing zone by one grout pile thus small diameter of container presents small pile spacing or large improvement ratio. Table 2 shows a list of test cases in centrifuge, in which case CPGd18sA and CPG30sB models 1.8 m and 3.0 m of pile spacing (in prototype scale) equivalent to 13.7 % and 4.9 % of improvement ratio and case CPG30nA models ground without grout injection, respectively. As a shaking condition, sinusoidal wave with 2 Hz of frequency (in prototype scale) and 50 waves are applied increasing the amplitude of vibration. During the shaking, horizontal earth pressure, response acceleration of ground and pore water pressure was measured.

3 RESULTS AND DISCUSSIONS

3.1 X-ray tomography

Figure 4 presents density distribution of the model ground at each scanning steps as a result of the x-ray tomography. After grout injection, the increase of ground density can be observed at the side of the grout piles with increase of 25 % of relative density thus densification effect may occur mainly at lateral side o grout piles. The distribution of vertical and horizontal displacement as results of DIC analysis is shown in Figure 5. The upper line of Fig. 5 presents incremental vertical displacement between initial-2nd injection (Step A), 2nd-4th injection (Step B) and 4th-6th injection (Step C) and the lower line presents incremental horizontal displacement, respectively. Horizontal displacement is localized in the area around the grout pile through all the steps. On the other hand, vertical displacement beneath the grout pile can be observed at Step A injecting deeper position and the downward displacement is decrease with injecting position shallower. These features indicate that the mechanism of ground deformation caused by grout injection can be considered as a cavity expansion for deep injection and cone uplift for shallow injection. The map of maximum shear strain and volumetric strain calculated from DIC results are shown in Figure 6. The general tendency observed from Fig. 6 is that shear strain is localized around the

model ground

horizontal crossection

LVDT

model ground

vertical cross-section

grout pile

DT1

A11

A12

P1

P2

E1

E2

puressure cell

pore waterpressure cell

accelerationgage

0.69 (23)

3.5 (116)

11.1

(370

)7.

8(2

60)

1.8

(60)

3.3

110)

(4.

5(1

50)

3.3

(110

)

3.5 (116)

1.7 (58)

2.4 (80)

digassed water

3.0

(100

)grout pile

3.5 (116)

plane viwe of a improved site

1.5(

50)

units：[m] in prototype scale ([mm] in model scale）

Specific gravity,Gs

2.65Median particlesize, D50

0.35 mm

Maxumum voidratio, emax

1.115Coefficient ofuniformity, Uc

1.5

Minumum voidratio, emin

0.71

Shear resistanceangle, φ' (fromDr=50 %)

36.1°

Shear resistanceangle, φ' (fromDr=90 %)

40.4°

Soma Silica Sand #5

CaseImprovement

ratio (%)

Pilespacing

(m)

Pilediameter

(m)

Relativedensity

(%)CPGd18s 13.7 1.8 0.7 55CPGd30n - - - 41.3CPGd30s 4.9 3 0.7 70.4

967



3

Figure 4. Density distributions as the results of XRCT scan Figure 5. Vertical (upper line) and horizontal (lower line) incremental displacement. grout pile and volumetric strain in this area shows contraction. Moreover, another localized area is also developed from the edge of grout pile to ground surface and it is indicating that upper ground of localized area is uplifted due to grout injection. Dark color in the map of volumetric strain presents volume contraction and this zone gradually decrease the injection depth becomes shallower. The average calculated volumetric strain is 5 % of contraction, 23.5 % of increase of relative density, in Step A, which agrees against the density map of tomography. Figure 7 shows the relationship between volumetric strain and maximum shear strain at the same pixel in Fig. 6. One of the features common to all the steps is that constrain volumetric strain mainly appears at small shear strain and expansion at

Depth

( mm

)

0-10-20-30 10 20 30260250

130140150160170180190200210220230240

r (mm) r (mm)0-10-20-30 10 20 30

r (mm)0-10-20-30 10 20 30

Step A Step CStep B

0 0

incremental maximum shear strain

Depth

( Figure 6. Incremental maximum and volumetric strain field

.04

mm

)

0-10-20-30 10 20 30260250

130140150160170180190200210220230240

r (mm) r (mm)0-10-20-30 10 20 30

r (mm)0-10-20-30 10 20 30

Step A Step CStep B

-0.04 0.04

incremental volumetric strain

contraction expantion

Step A Step CStep B

0-10-20-30 10 20 30260250

130140150160170180190200210220230240

0-10-20-30 10 20 30260250

130140150160170180190200210220230240

Step A Step CStep B-10 10

displacement (mm)

r (mm)

depth

(mm

)

0-10-20-30 10 20 30r (mm)

0-10-20-30 10 20 30r (mm)

r (mm) r (mm)0-10-20-30 10 20 30 0-10-20-30 10 20 30

r (mm)

depth

(mm

)

Figure 7. Relationship between volumetric strain and maximum shear strain.

-0.08

-0.06-0.04

-0.020

0.02

0.040.06

0.08

0 0.05 0.1 0.15 0.2 0.25 0.3

-0.08

-0.06-0.04-0.02

00.02

0.040.06

0.08

0 0.05 0.1 0.15 0.2 0.25 0.3

-0.08

-0.06-0.04

-0.020

0.02

0.040.06

0.08

0 0.05 0.1 0.15 0.2 0.25 0.3

Δεv

Δγmax(a)　Step A

(b)　Step B

(c)　Step C

expansion

contraction

expansion

contraction

expansion

contraction

Δεv

Δεv

Δγmax

Δγmax

968



4

20 400 10060 80

4

2

0

6

8

Time (sec)

large shear strain. Especially in Step C, volumetric strain tends to expand with smaller shear strain compare to injection at deeper depth (Step A and Step B). This derive from failure mode at shallower depth tend to be uplift mode due to low confining pressure, thus it can be said that a certain grout pile diameter or injection volume is not largely contribute to densification especially at shallower depth.

3.2 Centrifuge test

Figure 8 presents the changes in horizontal stress recorded during shaking table test on centrifuge, expressed as K = (horizontal effective stress, σ’h) / (initial vertical effective stress, σ’v0). The results were fileted to remove the effect of variation due to shaking. The K value before shaking starts (0 sec ~ 3 sec) presents the value after grout injection. The influence of size of shearing area by one grout pile or improvement ratio is clear, with larger improvement ratio resulting in higher K value after grout injection. The residual K value after shaking test in Cases CPG18s and CPG30s keeps high K value even 250 m/s2 of acceleration was applied to the ground. This feature indicates that the improved ground by CPG possibly to keep its improvement effect in terms of residual K value. Figure 9 shows time histories of ratio of excess pore water pressure, Ru = (excess pore water pressure ∆u) / (initial vertical effective stressσ’v0). Liquefaction was observed in Case CPG30n, without grout injection, from beginning of the dynamic loading. In contrast, the remarkable increase of Ru cannot be observed in the cases with grout injection even in the case CPG30s with lower improvement ratio. This is indicating that the increase of K value due to grout injection provides the increase of liquefaction resistance even in lower improvement ratio. 4 CONCLUSIONS

Ground behavior due to compaction grouting was studied by x-ray tomography and centrifuge tests, with particular interest in the density change and the increase of liquefaction resistance.

Density change and ground response due to grout injection were not only visualized by x-ray tomography but also discussed quantitatively based on results of image analysis, Volumetric Digital Image Correlation (V-DIC). The mechanism of ground deformation caused by grout injection can be considered as a cavity expansion for deep injection and cone uplift for shallow injection. The densification of ground was mainly observed in the area around the grout pile. However, it can be also said that a certain grout pile diameter or injection volume is not largely contribute to densification especially at shallower depth because of lower confining pressure. The influence of the improvement ratio appeared in the residual K values after grout injection which was observed in centrifuge tests. As the results of shaking table test, the increase of K value due to grout injection provides the increase of liquefaction resistance even in lower improvement ratio of about 5 %. 5 REFERENCES

Boulanger R.W. and Hayden R.F. 1995. Aspects of compaction grouting of liquefiable soil. Journal of Geotechnical Engineering, ASCE, 121 (12), 844-855.

Miller E.A. and Roycroft G.A. 2004. Compaction grouting test program for liquefaction control, Journal of Geotechnical and Geoenvironmental Engineering, ASCE, 130 (4), 355-361.

Nishimura S., Takehana K., Morikawa Y. and Takahashi H. 2012. Experimental study of stress changes due to compaction grouting. Soils and Foundations, 51 (6), 1037-1049.

GL -3.3 mGL -7.8 m

shaking

Kvalue　

(σh’

/σv0

’)

20 400 10060 80

Time (sec)

GL -3.3 mGL -7.8 m

shaking

Kvalue　

(σh’

/σv0

’)

4

2

0

6

8

20 400 10060 80

Time (sec)

shaking GL -3.3 mGL -7.8 m

Kvalue　

(σh’

/σv0

’)

4

2

0

6

8

(a) CPG18s (c) CPG30n (b) CPG30s

Figure 8. Time history of K values

20 400 10060 80

0.6

0.4

0.0

0.8

0.2

1.0

Time (sec)

Ru

GL -3.3 mGL -7.8 m

shaking

20 400 10060 80

0.6

0.4

0.0

0.8

0.2

1.0

20 400 10060 80

0.6

0.4

0.0

0.8

0.2

1.0GL -3.3 mshaking GL -3.3 mshaking

Time (sec)R

u

GL -7.8 m

Time(sec)

Ru

GL -7.8 m

(a) CPG18s (c) CPG30n (b) CPG30s

Figure 9. Time history of ratio of Ru

969

A model study of strains under footings supported by floating and end-bearing granular columns

Une étude sur modèle réduit des contraintes sous semelles isolées reposant sur des colonnes granulaires flottantes et encastrées

Tekin M. GEOMED, Geotechnical Consultancy, Investigation Supervision & Trd. Co.Inc. Ankara, TURKEY

Ergun M.U. Civil Engineering Department, Middle East Technical University, Ankara, TURKEY

ABSTRACT: A model study was performed in order to examine settlements in the presence of floating and end-bearing granular columns and without columns (untreated) under D=100 mm and D=200 mm circular loading plates (footings). It is aimed to findeffective length in floating type granular columns that provides significant settlement improvement under footings. In addition to total footing settlements recorded by displacement transducers, subsurface displacements both along the column and in the untreated soilbelow column were measured by miniature borehole settlement gages for different column lengths and loading plate diameters. Settlements and strain distributions obtained from displacement measurements with depth show the role of column length insettlement reduction.

RÉSUMÉ : Une étude sur modèle réduit a été réalisée pour examiner les tassements du sol en présence ou non de colonnes granulaires flottantes et encastrées, sous des plaques de chargement circulaires de diamètre D=100 mm et D=200 mm (semellesisolées). Le but est de trouver la longueur optimale des colonnes granulaires flottantes permettant d’obtenir une amélioration significative des tassements sous semelles. Pour différentes longueurs des colonnes et différents diamètres de fondations, lestassements des semelles ont été enregistrés à l’aide des transducteurs. En plus de ces enregistrements, les déplacements en profondeur, le long des colonnes et en dessous de celles-ci ont été mesurés par des gauges miniatures. Les tassements et les distributions descontraintes déduits des enregistrements de déplacements en profondeur ont montré le rôle de la longueur de la colonne dans la réduction des tassements.

KEYWORDS: model test, granular column, floating stone column, settlement, subsurface displacement

1 INTRODUCTION

Granular columns as semi-rigid inclusions are used to reduce settlements under foundation loads in compressible soils. The column length is a significant design decision in settlement reduction in deep compressible soils.

An experimental research was performed to find out an effective length in floating type columns to get benefit of them as settlement reducers.

2 LABORATORY TESTS

2.1 Material Properties

The model tests were carried out in loading tanks designed as large oedometers (d=410 mm, h=390 mm). The clay ‘foundation’ soil loaded in model test was obtained by consolidating commercially available remoulded kaolinite type clay from paste (moisture content w=42 %) inside the loading tank under a pressure of v=50 kPa. Kaolin clay has a plasticity index of PI=22 % (liquid limit, LL=51% and plastic limit, PL=29 %). Plasticity of kaolin eliminates swelling and shrinkage problems.

Foundation soil has an average undrained shear strength of cu=25 kPa. Coefficient of compressibility, mv values for the pressure intervals in main tests are given in Table 1.

Granular columns with a relative density of 80 % were formed by compacting sand material. Grain size distribution of the sand has been arranged based on Dc/D (column diameter/typical particle size) ratio in stone column applications. The grain size distribution and physical properties of sand are shown in Table 2 and Table 3 respectively.

T able 1. Laboratory consolidation test results of clay foundation soil

Test No 50-75 kPa

mv (m2/kN) 75-100 kPa mv (m2/kN)

Test No:1 Test No:2 Test No:3

0.000409 0.000442

0.000324 0.000345 0.000363

T able 2. Grain size distribution of sand used in granular column

Sieve D(mm) Percent finer than D(%)

50 70 100 200 325

0.315 0.210 0.149 0.074 0.035

100 70 38

6 0

T able 3. Physical properties of sand with relative density of 80%

USCS emin emax Gs

SP 0.961 0.581 2.683 42

2.2 Experiment Set-Up and Test Procedure

Following the completion of consolidation inside loading tank, clay foundation height is fixed to H=290 mm by removing some soil at the top. To form granular columns, soil was drilled by 20 mm auger to the desired depth with the aid of steel template placed on soil and then the bored hole was filled with sand and

970


rammed in a controlled manner to have a relative density of 80% throughout the column length. In each test, two granular columns were instrumented for subsurface settlements, the central column and one of the closest columns to the centre. Drilling of instrumented columns was done throughout the foundation soil (290 mm). A miniature magnetic switch apparatus has been developed based on the measurement principle of magnetic extensometer to get subsurface displacements (Tekin 2005). The apparatus consists of an antenna rod with a base, ring type magnets, magnetic switch probe and a worm gear device designed to move the probe inside the antenna rod slowly. After drilling, long antenna rod (d=6 mm) was inserted into this hole and its base was placed at the bottom of the tank. The gap between the bored hole and outer surface of antenna rod was filled with sand or clay till the desired level of settlement measurement. Then ring type magnet (d=13 mm) was passed through the rod. At least eight magnets were placed along 290 mm depth. A hollow ramming device was used to compact sand at the same relative density of other columns and for compression of clay. When all columns were prepared, the surface was flattened; loading plate and loading frame were mounted. Two displacement transducers were located on the loading plate to record plate settlement (total footing settlement). Initial (reference) magnet locations were determined by moving the probe slowly inside antenna rods.

Footing load of 75 kPa was applied on clay soil improved by granular columns. At the end of consolidation (test), final readings were taken throughout the antenna rods.

In the test series with 200 mm plate, at the end of 75 kPa loading, test was repeated for 100 kPa loading with settlement measurements. Figure 1 shows experimental set-up.

Tests without granular column installation (untreated soils) but with the same loading levels and instrumentation were done for each loading plate diameter. These reference tests were the basis of comparisons to find out the contribution of granular columns to settlement reduction for different column lengths.

Figure 1. Experiment Set-Up

3 TEST RESULTS

Scope of this research involved not only the efficiency of floating granular column length but also effect of column number (area replacement ratio) and comparison of end-bearing columns for footings (B=100 mm, B=200 mm) and 1-D unit cell (B=410 mm) concepts on settlement improvement. Column length, loading plate diameter, area replacement ratio and loading levels were the variables, but clay thickness

(H=290mm) and column diameter (20 mm) were kept constant in all tests. However, this paper includes only the behavior of floating granular columns loaded under B=100 mm and B=200 mm plates, therefore only related test series and their findings were given in this paper. The details of the test series selected are summarized in Table 4. Area replacement ratio is 0.22 in all tests. End bearing series in Series III are also included.

T able 4. Summary of test results

Series No

LoadingPlate

DiameterB(mm)

Applied Pressure

(kPa) L (mm) L/B

I 100 75 Untreated (no columns)

III 100 75

60 100 140 210

290 (eb)

0.6 1.0 1.4 2.1 2.9

IV 200 75,100,125 Untreated (no columns)

V 200 75,100

100 140 200 275

0.5 0.7 1.0

1.375

In series I and III, the ratio of clay height (H=290 mm) to

plate diameter (B=100 mm) is H/B=2.9. Due to the limitation in tank height, H/B ratio is 1.45 (H=290 mm/ B=200 mm) in Series IV and V. There may be some boundary effects in Series IV and V due to lower H/B ratio. However they do not influence the general trend of displacements and strain behavior encountered both along the columns and the clay layer underlying columns. Similar results have been found for floating columns both in B=100 mm and B=200 mm tests.

As seen in Figure 2 and Figure 3, cumulative displacements occurring within the clay layer underneath column decrease significantly in the columns longer than L/B=1.0.

Air jack

Frame

Loading Plate end of column (B=410 mm)

L/B=0.6 LVDT

L/B=1.0

Loading TankL/B=1.4

L/B=2.1

L/B=2.9 (eb)

Figure 2. Normalized settlement versus normalized depth graphs of Series I and III (B = 100 mm, 75 kPa)

971


Figure 3. Normalized settlement versus normalized depth graphs of Series IV and V (B = 200 mm, 75 kPa)

The strains at every cm depth interval are determined from

the settlement-depth graphs obtained with LVDT and miniature magnetic switch apparatus measurements and then plotted as strain vs. normalized depth graphs as illustrated in Figure 4 and Figure 5.

In floating granular columns, the efficiency of column length of L/B=1.0 to 1.2 is observed in strain vs. normalized depth graphs shown in Figure 4 and Figure 5. It is obvious that they reduce the strains considerably throughout their length when compared to those in the untreated case. They also manage to decrease the strains in underlying clay layer significantly.

Figure 4. Strain versus normalized depth graphs of Series I and III (B = 100 mm, 75 kPa)

However, a trend of increasing strains is observed in clay

layer starting from the end of column under the columns shorter than L/B=1.0. Strains make a peak and then decline (L/B=0.6 with B=100 mm plate and L/B=0.5 and L/B=0.7 with B=200 mm plate). In the test Series III (B=100 mm) with L/B=0.6 these peak strains are even higher than those in the untreated case for the same depth. All the findings show that the columns shorter than L/B=1.0 are not adequate to reduce the strains in clay layers underlying floating type granular columns.

end of column end of column L/B=0.5

L/B=0.5

L/B=0.7

L/B=1.0

L/B=0.7

L/B=1.0

L/B=1.375 L/B=1.375

Figure 5. Strain versus normalized depth graphs of Series IV and V (B = 200 mm, 75 kPa)

Uppermost 0.2B to 0.3B depth of columns are strained in end bearing columns, then strain value decreases and keep constant till the end of column. The columns with L/B=1.375 in test series V can also be considered as end-bearing columns.

Ratio of the strain values in the columns to those in the

untreated clay at the same depth are presented in Figure 6 and Figure 7. Effectiveness of the ratio L/B=1.0 to 1.2 is again clear. Strain ratios under short columns with L/B<1.0 length are nearly the same as those in the untreated condition, in other words L/B<1.0 column length is insufficient for settlement reduction.

end of column end of column

L/B=0.6

L/B=1.0

L/B=1.4

L/B=2.1

L/B=2.9 (eb)

L/B=0.6

L/B=1.0

L/B=1.4

L/B=2.1

L/B=2.9 (eb)

Figu o versus normalized depth graphs of Series I and III (B = a)

re 6. Strain rati 100 mm, 75 kP

In columns with L/B>1.0 length, columns contribute to strain

reduction in the clay soil underneath till 2.0 B depth. The increase in strain ratio after 2.0 B implies that cumulative settlement decreases considerably below 2.0 B in the untreated cases so that any improvement effort has no contribution to settlement reduction below this depth. This also explains similarities in the behavior of L/B=2.1 floating and end bearing columns.

972


Figure 7. Strain ratio versus normalized depth graphs of Series IV and V (B = 200 mm, 75 kPa)

Figure 8 and Figure 9 are the graphs of settlement reduction

factor β with depth, in other words the cumulative settlement ratio of improved to untreated soil at any depth. Similar behavior to variation of strain ratio above is observed.

Figure 8. Settlement reduction factor versus normalized depth graphs of Series I and III (B = 100 mm, 75 kPa).

Figure 9. Settlement reduction factor versus normalized depth graphs of Series IV and V (B = 200 mm, 75 kPa)

Settlement reduction factors along the columns are almost the same as those measured in total settlements (cumulative

settlement) for L/B > 1.0 in all floating type granular columns (Figure 10). Settlement reduction factors in the underlying clay soil are also similar for lengths of L/B >1.0 but it is much higher (0.7) for L/B=0.6 which means poor improvement.

end of column

L/B=0.5

L/B=0.7

L/B=1.0

L/B=1.375

Figure 10. Settlement reduction factor graphs of Series I and III (B = 100 mm, 75 kPa)

4 CONCLUSIONS

Strains measured along granular columns and below in clay under loaded circular plates indicate the following:

a) Strains are small below granular columns whose length is roughly equal to side dimension of the plate (L/B=1.0-1.2). This explains why floating columns are effective in ground improvement.

b) Strains below short columns in clay with L/B<1.0 show an increase and a peak indicating probable high stress transfer through columns. Improvement by short granular columns is limited.

L/B=

end of column

0.6

c) Longer columns with L/B>2.0 show similar behavior to end bearing columns (i.e. very limited displacement) L/B=1.0

Settlement reduction factors decrease with increasing column lengths. These factors along the columns are similar to the factors below the columns for lengths L/B>1.0.

L/B=1.4

L/B=2.1 5 REFERENCE

Tekin M 2005. Model study on settlement behavior of granular columns under compression loading. Ph.D. thesis submitted to the graduate school of natural and applied sciences of Middle East Technical University Ankara TURKEY, 223 pages.

end of column

L/B=0.5

L/B=0.7

L/B=1.0

L/B=1.375

973

Modélisation physique du blocage d’un écoulement d’eau dans un sol par injection d’un produit de colmatage

Physical modelling of blocking phenomenon, by injection of a clogging product, of water flow through soils

Truong Q.Q., Dupla J.-C., Canou J. Université Paris-Est, ENPC, Laboratoire Navier, Marne-la-Vallée, France

Chevalier C. Université Paris-Est, IFSTTAR, Marne-la-Vallée, France

Chopin M. Geomechanical Consult. Intern., Lausanne, Suisse

Fry J.J. EDF-CIH, Savoie Technolac, Le Bourget-du-Lac, France

RESUME : On présente la problématique du blocage d’un écoulement d’eau dans un sol à partir d’un produit de colmatage. Celle-ci est abordée à partir d’une modélisation physique unidimensionnelle du phénomène, basée sur des essais d’injection localisée, en conduite d’écoulement. Les essais sont réalisés à partir d’un dispositif prototype développé dans le cadre du projet national ERINOHconsacré à l’étude des phénomènes d’érosion interne dans les sols. La problématique concerne la possibilité de blocage d’un écoulement d’eau parasite sous une digue ou un barrage, pouvant évoluer vers la rupture de l’ouvrage par érosion interne et renard hydraulique. Après une présentation du dispositif d’essai et du protocole expérimental associé, on présente un résultat typique obtenu sur un gravier de référence à partir de l’injection d’un produit composé d’un mélange binaire de bentonite et d’un filler spécifique, caractérisé par une viscosité et un seuil d’écoulement. On décrit en particulier le mécanisme de blocage complet de l’écoulement et on donne finalement des éléments relatifs aux conditions de blocage en termes des paramètres pression d’eau et pression d’injection.

ABSTRACT : This communication is related to the issue of permanent water flow blocking process by injection of specific clogging products. The latter is tackled through a physical modelling of the blocking phenomenon, based on localized injection tests in a flowpipe. The tests have been carried out on a prototype experiment developed within the framework of the ERINOH French national project, devoted to research on the internal erosion phenomena in soils. The problematic here concerns the possibility of blocking a parasite flow of water under a dike or a dam, which may lead toward the failure of the structure by internal erosion and piping. After presenting the experimental setup and testing procedure, a typical test is presented and analysed, obtained for injection of a cloggingproduct composed of a blend of bentonite and filler, characterized by a given viscosity and yield stress. In particular, the blocking mechanism observed is described and elements are finally given on the blocking conditions in terms of the parameters water pressureand injection pressure.

MOTS-CLÉS : blocage d’écoulement d’eau, érosion interne, coulis d’injection, renard hydraulique

KEYWORDS : water flow blocking, internal erosion, injection grout, piping 1 INTRODUCTION

Le développement d’écoulements d’eau parasites à travers ou sous les digues et les barrages constitue un problème très important associé aux ouvrages hydrauliques, car ceux-ci peuvent entraîner, par érosion interne et développement de renards hydrauliques, des ruptures parfois catastrophiques de ces ouvrages (voir, en particulier, Fry et al. 1997, Monnet 1998, Lautrin 2002, Bendahmane 2005, Fell et al. 2005). Dans le cas où l’écoulement a déjà pris une ampleur significative en terme de débit, il est donc intéressant d’essayer de développer des méthodes permettant de bloquer rapidement, en temps réel, l’écoulement parasite. Les méthodes envisagées ici sont basées sur l’injection de produits de colmatage capable de résister à l’entraînement par l’écoulement en vue de bloquer définitivement ce dernier. L’objectif est donc de trouver des produits d’injection spécifiques associés à des protocoles d’injection appropriés en vue d’arriver à bloquer des écoulements caractérisés par des débits relativement importants dans des sols relativement perméables. Dans ce contexte, un travail de recherche a été développé dans le cadre du Projet national français ERINOH, avec pour objectif l’étude, dans une configuration unidimensionnelle

(conduite), des mécanismes de blocage d’un écoulement permanent par injection, localement d’un produit de colmatage. Après une présentation du dispositif d’essai prototype, développé de manière spécifique pour cette recherche, et de la procédure expérimentale associée, on présente et on analyse les résultats d’un essai typique réalisé sur un gravier relativement perméable injecté par un mélange de bentonite et d’un « filler » spécifique, en mettant l’accent sur les mécanismes de blocage observés. Finalement, sur la base d’une série d’essais réalisés, on donne quelques éléments sur les conditions de blocage de l’écoulement, en termes de relations entre les paramètres principaux caractérisant le problème. 2 DISPOSITIF EXPERIMENTAL ET PROTOCOLE D’ESSAI

2.1 Dispositif expérimental

Le principe de l’expérimentation est de générer, dans un premier temps, un écoulement d’eau permanent dans une colonne de sol contenue dans une conduite. Dans un deuxième temps, on va chercher à bloquer cet écoulement par l’injection locale d’un produit de colmatage, en un certain point le long de la conduite.

974


L’objectif est alors d’analyser le processus d’injection et d’essayer de comprendre les mécanismes de blocage éventuel de l’écoulement d’eau. Les principaux paramètres caractérisant un essai sont la nature du sol étudié, les caractéristiques de l’écoulement (débit, pressions, gradient hydraulique) et les paramètres d’injection (nature du produit, pression d’injection). Un schéma fonctionnel du dispositif expérimental est présenté sur la figure 1 où sont indiqués les principaux éléments du dispositif ainsi que l’instrumentation mise en œuvre. La conduite a une longueur utile de 3 m et un diamètre intérieur de 18 cm. Elle est composée d’un assemblage de six modules en plexiglas de 50 cm de long. Chaque extrémité est équipée d’un module conique permettant d’uniformiser l’écoulement. Le module n°2 en partant de l’amont est équipé d’un connecteur spécifique permettant de réaliser l’injection. La conduite est instrumentée avec six capteurs de pression (numérotés de 1 à 6), dont la localisation est indiquée sur la figure 1. Deux débitmètres (petite et grande capacités) sont installés en amont de la conduite. Un turbidimètre est installé en aval de la conduite et permet d’identifier la sortie éventuelle de particules fines de la conduite. L’écoulement d’eau claire, en circuit fermé, est généré par un surpresseur qui permet d’imposer une pression constante régulée à l’entrée de la conduite quel que soit le débit d’eau circulant dans le système. Un réservoir d’eau claire est intercalé dans le circuit et permet de disposer d’une réserve d’eau suffisante pour assurer l’écoulement. L’écoulement peut être dévié vers un autre réservoir dans le cas où une quantité importante d’eau chargée en particules sort de la conduite.

2.2 Protocole d’essai

La réalisation d’un essai est relativement lourde et implique la réalisation de plusieurs opérations successives. La colonne de sol est tout d’abord reconstituée dans la conduite, par compactage de couches successives en empilant les uns sur les autres, les six modules, en position verticale, sur un support pivotant spécifique. Des grilles sont positionnées à chaque extrémité pour retenir le matériau mis en place dans la conduite. Ce support permet ensuite de basculer la conduite à l’horizontale puis de la reprendre au pont roulant pour la positionner sur son support d’essai. Les modules coniques d’extrémité sont ensuite fixés, l’ensemble étant finalement raccordé au reste du circuit. L’écoulement d’eau permanent est ensuite progressivement établi, en appliquant des paliers croissants de pression en vue de bien saturer le sol contenu dans la conduite.

Figure 1. Schéma général du dispositif expérimental

Figure 2. Vue générale du dispositif d’essai, depuis l’amont, montrant la

conduite assemblée sur son support, équipée du tube d’injection, le support pivotant, le surpresseur et le réservoir d’injection

Le produit à injecter est préparé en parallèle puis est versé dans le réservoir d’injection. Le tuyau d’injection est saturé en coulis puis raccordé à la canne d’injection. Un robinet permet d’isoler le réservoir du tube d’injection. La pression d’injection souhaitée est ajustée dans le réservoir. L’injection démarre lorsque l’on ouvre le robinet. Pendant l’injection, l’observation visuelle à travers le plexiglas permet d’observer comment se propage le coulis. Pendant l’essai, on réalise l’acquisition des données sur tous les capteurs de mesure, ce qui permet d’obtenir la répartition des pressions le long de la conduite, l’évolution du débit d’eau pendant l’injection et l’évolution du débit d’injection, permettant ensuite de réaliser une analyse complète de l’essai. L’essai est poursuivi jusqu’au blocage éventuel de l’écoulement d’eau. En cas de blocage, on maintient la pression d’eau en amont pendant une certaine durée pour observer s’il n’y a pas de déblocage éventuel au cours du temps. 3 DESCRIPTION D’UN ESSAI TYPIQUE

3.1 Caractéristiques de l’essai

Le sol utilisé pour cet essai est un gravier propre 4-10 mm, de taille moyenne d50= 7 mm. Le matériau est mis en place dans un état de compacité moyen, correspondant à une masse volumique sèche de 1,51 t/m3 et une porosité n égale à 0,43. Le débit de l’écoulement d’eau permanent est de 0,49 l/s, correspondant à une pression imposée au niveau du surpresseur de 20 kPa. Le produit d’injection est un mélange de bentonite et d’un filler spécifique (PKA filler) en suspension dans l’eau. La préparation du coulis passe d’abord par la préparation de la suspension de bentonite que l’on doit laisser s’hydrater pendant 24 heures. Le filler est ensuite progressivement rajouté à la suspension, le mélange étant maintenu en agitation. On mesure ensuite les caractéristiques mécaniques du mélange avec un rhéomètre, en termes de contrainte seuil τ0 et de viscosité μp . Le produit injecté ici est caractérisé par un seuil de cisaillement τo = 46 Pa et une viscosité plastique ηp =41 mPa.s.

3.2 Phase de génération de l’écoulement

L’écoulement d’eau est progressivement établi en appliquant des paliers croissants de pression sur le surpresseur, par incréments de 50 kPa, jusqu’à une valeur maximale de 350 kPa, correspondant à un débit maximum de 3,23 l/s. Pour chaque

975


Figure 3. Distribution de la pression d’eau le long de la conduite pour les différents paliers de pression imposés en amont

palier de pression imposé, la distribution des pressions d’eau le long de la conduite est représentée sur la figure 3 en fonction de la distance par rapport à l’entrée de la conduite. Cette phase de l’essai permet de vérifier que l’ensemble du dispositif fonctionne correctement, qu’il n’y a pas de fuite et permet aussi de saturer le matériau en expulsant, pour les pressions les plus élevées, les bulles d’air restant piégées dans le matériau. La pression d’entrée est finalement redescendue jusqu’à la valeur nominale choisie pour l’essai et la phase d’injection peut démarrer.

3.3 Phase d’injection

Pour cet essai, la pression d’injection appliquée dans le réservoir d’injection est de 60 kPa. Dès que le robinet situé à la sortie du réservoir d’injection est ouvert, l’injection démarre. On continue de faire l’acquisition de toutes les mesures pendant cette phase, avec, en plus, la variation de masse du réservoir d’injection qui permet d’avoir accès au débit d’injection ainsi qu’à la masse totale de coulis injecté. La figure 4 présente l’évolution, en fonction du temps depuis le début de l’injection, des pressions d’eau mesurées le long de la conduite ainsi que l’évolution du débit d’eau dans la conduite. En ce qui concerne le débit d’eau, on constate une décroissance rapide de celui-ci jusqu’à obtention du blocage complet de l’écoulement au bout de 20 secondes d’injection. L’évolution des pressions d’eau mesurées sur les six capteurs le long de la conduite est en bon accord avec le phénomène de blocage progressif observé. La pression mesurée sur le capteur n°1 augmente régulièrement pendant le colmatage et se stabilise au niveau de la pression imposée en amont par la pompe (20 kPa). La pression mesurée sur le capteur 2 continue à augmenter légèrement au-delà de cette valeur car ce capteur, plus proche du point d’injection, « ressent » la pression d’injection. En aval du point d’injection, les pressions mesurées sur les quatre capteurs décroissent comme normalement attendu. Les trois derniers capteurs (n°4, 5 et 6) se stabilisent clairement autour de la pression atmosphérique, imposée par les conditions aux limites aval (sortie de conduite). Le capteur n° 3 reste significativement au dessus de cette valeur car il « sent » encore la pression d’injection régnant au niveau de la canne d’injection.

3.4 Interprétation du phénomène de blocage

La figure 5 montre, à la fois, l’évolution du débit d’eau et l’évolution du débit d’injection pendant la phase d’injection. Comme précédemment décrit, le débit d’eau décroît

Figure 4. Evolution du débit d’eau et des pressions d’eau le long de la conduite pendant l’injection

régulièrement et rapidement jusqu’au blocage complet de l’écoulement après 20 secondes d’injection. Le débit du coulis est maximum dès le début de l’injection puis décroît lui aussi rapidement. Lorsque le blocage de l’écoulement est atteint, le débit d’injection est encore de l’ordre de 0,1 l/s. Celui-ci continue ensuite à décroître jusqu’à atteindre une valeur résiduelle très faible pour un temps de 100s au bout duquel l’injection est stoppée. La figure 6 montre une photo du bouchon de blocage formé par le produit injecté après la fin de l’injection. Pour cet essai, le bouchon a une longueur de 74 cm environ, dont 47 cm situés en aval de la canne d’injection et 27 cm en amont de ce même point. Pendant l’essai, on peut visualiser la formation du bouchon, qui commence à se former dans la partie aval puis progresse ensuite vers l’amont, lorsque la partie aval est suffisamment consistante. Dans cet essai, on a injecté environ 8030 cm3 de coulis. A partir de la porosité du gravier (n=0,43), on peut estimer le volume de l’espace poreux correspondant à ces 74 cm, à savoir 8100 cm3. Cette valeur est très proche de la quantité de coulis injecté mesurée, ce qui montre que le coulis a bien rempli l’essentiel de la porosité du gravier et a formé un bouchon localisé autour de la canne d’injection.

Figure 5 – Evolution du débit d’écoulement d’eau et d’injection pendant

la phase d’injection

976


Il est intéressant de remarquer que, pour obtenir le blocage de l’écoulement, il faut avoir un différentiel de pression d’au moins 30 kPa entre la pression d’injection (mesurée au niveau du réservoir d’injection) et la pression d’eau amont (mesurée au niveau du surpresseur). De même, plus le différentiel de pression est important et plus la longueur du bouchon de blocage est élevée. On arrive à une longueur maximale de 140 cm pour un différentiel de 50 kPa (essai EB9). Il est aussi intéressant de remarquer que, pour un différentiel de pression donné, plus le niveau des pressions est élevé et plus, là aussi, la longueur du bouchon est élevée (comparaison des essais EB4 et EB6).

Figure 6 – Visualisation du bouchon de blocage

4 SYNTHÉSE DU PROGRAMME D’ESSAIS RÉALISÉ

Un certain nombre d’essais ont été réalisés sur le dispositif en vue d’étudier l’influence de différents paramètres d’essai sur les résultats obtenus. En fonction des paramètres d’essai, on observe trois types de comportements : essais avec blocage définitif de l’écoulement, essais avec blocage temporaire et essais sans blocage. Le blocage définitif est défini lorsque, en maintenant la pression d’eau en amont, sur le surpresseur, pendant deux heures, on n’observe pas de déblocage de l’écoulement. Le blocage temporaire correspond au cas pour lequel, après un blocage initial complet de l’écoulement, on observe avant deux heures d’attente, un déblocage progressif de l’écoulement. Le non blocage correspond au cas où on n’arrive pas à bloquer l’écoulement pendant la durée de l’injection.

Par ailleurs, la répétabilité des essais, évaluée à partir des essais EB2 et EB3, est satisfaisante (longueurs de bouchon de 50 cm et 56 cm respectivement). 5 CONCLUSION

On a présenté dans cette communication des essais d’injection de coulis spécifique destinés à bloquer un écoulement d’eau permanent en conduite. Cette recherche a été réalisée dans le cadre du Projet national français ERINOH et a, en particulier, nécessité le développement d’un dispositif expérimental assez lourd qui a pu être entièrement validé.

En particulier, neuf essais ont été réalisés sur le gravier 4/10 mm et le coulis dont les caractéristiques ont été décrites ci-dessus (essais EB1 à EB9), pour différentes combinaisons de pression d’écoulement et de pression d’injection. Les principaux résultats obtenus sont présentés dans le tableau 1 en termes des résultats des essais (blocage définitif, blocage temporaire ou non blocage) et de la longueur des bouchons de blocage obtenus.

Sur la base des essais réalisés, on peut déjà dire qu’il est possible de bloquer un écoulement d’eau à partir d’une suspension bentonite/filler assez concentrée, pourvu que la pression d’injection soit suffisamment élevée par rapport à la pression d’eau régnant dans l’écoulement au niveau du point d’injection. Trois résultats ont pu être obtenus, en fonction des paramètres d’essai, à savoir le blocage définitif de l’écoulement, le blocage temporaire et le non blocage. A partir des essais réalisés dans une configuration unidimensionnelle simplifiée, on peut donc déjà donner certains éléments sur les conditions nécessaires qui devront être respectées, en conditions réelles in situ, pour assurer le succès d’une telle opération.

Tableau 1. Récapitulatif des résultats obtenus pour un ensemble d’essais éalisés sur le gravier 4/10 mm et le coulis bentonite/filler

6 REMERCIEMENTS

Les auteurs remercient le Projet national ERINOH pour son soutien dans le cadre du développement de ces recherches.

r

Nomd’essai

Pressionimposée par la

pompe pp(kPa)

Pressiond’injection

pi (kPa) Observations

Longueurde blocage

(cm)

EB1 20 40 Non-blocage -

EB2 20 50 Blocage définitif 50








7 RÉFÉRENCES

Bendahmane F. 2005. Influence des interactions mécaniques eau-sol sur l’érosion interne. Ph.D. thesis (french), university of Nantes, 160p. Fell R., MacGregor P., Stapledon D., Bel G. 2005 GeotechnicalEngineering of Dams. Taylor & Francis Edition, 912p. Foster M., Fell R., Spannagle M. 2000. The statistics of embankment dam failures and accidents. Canadian Geot. J., vol 37 : 1000-1024. Fry J. J., Degoutte G., Goubet A. 1997. L’érosion interne : typologie, détection et réparation. Barrages & Réservoirs, n°6, 126p. Johan L. 2009. Remedial Injection Grouting of Embankment Dams With Non-Hardening Grouts. Thesis of KTH Architecture and the Built environment, Stockholm, Sweden, 170p. Lautrin D. 2002. Vieillisement et réhabilitation des petits barrages en terre. Cemagref editions, 239p. Monnet A. 1998. Boulance, érosion interne, renard. Les instabilités sous écoulement. R.F.G., n° 82, p 3-10. Raul Flores B., Ramirez-Reynaga M., Macari E.J. 2011. InternalErosion and Rehabilitation of an Earth-Rock Dam. J. of Geot. and Geoenv. Eng., vol. 137, no2, pp. 150-160.

On peut voir à partir du tableau 1 que, pour les essais réalisés dans ces conditions, on n’a pas observé de blocage temporaire (uniquement blocage définitif ou pas de blocage).

977

Hydraulic conductivity and small-strain stiffness of a cement-bentonite sample exposed to sulphates

Conductivité hydraulique et module de cisaillement initial d'un échantillon de ciment-bentonite exposé aux sulfates

Verástegui-Flores R.D., Di Emidio G. Laboratory of Geotechnics, Ghent University, Belgium

Bezuijen A. Laboratory of Geotechnics, Ghent University, Belgium & Deltares, The Netherlands

ABSTRACT: Cement-bentonite slurries have often been used for geoenvironmental applications to isolate contaminated areas. A potential issue with all cement-treated materials is their durability, especially when applied in chemically aggressive environments. Inthis paper, the small-strain shear modulus (G0) and the hydraulic conductivity (k) of a cement-bentonite sample in contact with water and an aggressive sodium sulphate solution were investigated. Bender elements were installed in a flexible-wall hydraulic conductivity cell, to simultaneously monitor both G0 and k. As expected, permeation with clean water had no significant effect on the cement hydration, e.g. G0 continued to increase and k decreased gradually with time. However, after prolonged permeation with sulphates, a decrease of G0 and a gradual increase of k were recorded. These observations suggest that contact with sulphates produces degradation of the cemented structure that results in loss of strength and development of a network of interconnected fissures withinthe sample that increases the hydraulic conductivity.

RÉSUMÉ : Des barrières de ciment-bentonite ont été souvent utilisées pour des applications géo-environnementales afin d'isoler les zones contaminées. Un problème potentiel avec tous les matériaux cimentés est leur durabilité, en particulier lorsqu'ils sont mis enplace dans des environnements chimiquement agressifs. Dans cet article, le module de cisaillement initial (G0) et la conductivitéhydraulique (k) d'un échantillon de ciment-bentonite en contact avec de l'eau et une solution agressive de sulfate de sodium ont étéétudiés. Des bender elements ont été installés dans une cellule de conductivité hydraulique avec une paroi flexible, afin de surveillersimultanément G0 et k. Comme prévu, l'infiltration à l'eau claire n'a eu aucun effet significatif sur l'hydratation du ciment, parexemple, G0 a continué d'augmenter et k a diminué progressivement avec le temps. Cependant, après l'infiltration avec les sulfates, soit une diminution de G0 et une augmentation progressive de k ont été enregistrées. Ces observations suggèrent que le contact avec les sulfates produit la dégradation de la structure cimentée qui entraîne une perte de la résistance au cisaillement et la formation d'unréseau de fissures interconnectées dans l'échantillon qui augmente la conductivité hydraulique.

KEYWORDS: clay, cement, sulphate attack, hydraulic conductivity, small-strain shear modulus.

1 INTRODUCTION

Low permeability vertical barriers (cut-off walls) are often used to control groundwater flow and to isolate polluted soil. They are constructed by excavating a vertical trench. During excavation, the trench is filled with a slurry to prevent collapse. When the slurry is a mix of cement, bentonite clay and water, the barrier is denominated cement-bentonite (CB) cut-off wall (Jefferis 1981). The design of a CB cut-off wall is based on the characterization of the hydraulic conductivity, the strength of the cement-clay mix and eventually chemical compatibility with local groundwater.

Traditionally, the mechanical properties and hydraulic properties of the cement-clay mix are studied separately on different specimens (e.g. Opdyke and Evans 2005). Mechanical properties are usually evaluated by unconfined compression testing; however, the amount of data obtained is often limited to a few curing times and is usually subjected to scatter. On the other hand, hydraulic properties are evaluated by hydraulic conductivity tests; however, it is difficult to relate the hydraulic conductivity data alone to variations of strength, stiffness or microstructure of the cement-clay mix.

In this paper, an advanced testing method was used to simultaneously monitor both mechanical and hydraulic properties of a single sample. To that aim, a flexible-wall hydraulic conductivity cell was combined with a non-destructive technique to monitor the hardening of cement-clay samples. This technique uses bender elements (Dyvik and Madshus 1985) to measure the small-strain shear modulus, G0.

Such stiffness modulus is typically associated with small shear-strain levels (lower than 10−3 %). In general, G0 is governed by a number of factors such as stress history, stress level, void ratio, soil fabric, and the stiffness of the porous medium skeleton (Santamarina et al. 2001). Then, an increase of G0 can be expected with increasing interparticle cementation due to cement ageing. Conversely, a decrease of G0 can be expected when interparticle cementation is disrupted due to either mechanical or chemical degradation.

Experimental work was carried out on bentonite clay mixed with blast furnace slag cement. Monitoring of the small-strain shear modulus of cement-treated clay proved to provide valuable additional information to study the degradation of these materials.

2 MATERIALS AND SAMPLE PREPARATION

The samples studied in this research consist of a mixture of clay, cement and water. A sodium-activated bentonite clay, blast furnace slag cement of the type CEM III/B (ENV 197-1) with a nominal strength of 42.5 MPa and purified water with an electrical conductivity EC ≤ 2 μS/cm and a pH of about 7.6 were used. During hydraulic conductivity testing, the CB sample was initially permeated with purified water for 1 month to allow for further hydration of the cement products. After that period, the sample was permeated with a 25 g/L solution of Na2SO4. Such high Na2SO4 concentration was chosen here to accelerate the degradation process; however, it may be too high to represent common sulphate exposure levels in the field.

978


Some properties of the clay and the chemical composition of the clay and cement used here are summarized in Table 1 and Table 2, respectively.

T able 1. Physical properties of the bentonite clay

Parameter Value

Liquid limit 541.9

Plastic limit 67.0

Plasticity index 474.9

Swell index (ml/2g) 34

Cation exchange capacity (meq/100g) 73

T able 2. Chemical composition of the blast furnace slag cement & clay

Main component Cement Bentonite

SiO2 (%) 29.3 53.7

Al2O3 (%) 8.8 23.4

Fe2O3 (%) 1.2 5.9

MgO (%) 6.7 2.4

CaO (%) 47.1 1.9

Na2O (%) 0.2 2.2

The prepared samples consist of 80% water, 16% cement and 4% bentonite (by weight). This composition is in agreement with other studies in the literature (e.g. Ryan and Day 1986, Jefferis 1992, Opdyke and Evans 1995). First, a slurry of bentonite and water was mixed with a high-speed shear mixer for 5 minutes. The slurry was poured in a closed container and allowed to hydrate for about 24 hours. Subsequently, a slurry cement and water (accounted for in the final composition) was prepared to obtain a water cement ratio of about 0.5. Finally, the cement slurry and the 24-hour hydrated bentonite slurry were mixed in a dough mixer for about 10 minutes.

Then the fresh CB slurry was poured in stainless-steel moulds to prepare cylindrical specimens. The moulds were lightly vibrated to ensure that any trapped air bubbles were removed. The bottom and top ends of the mould were sealed with plastic foil to prevent moisture loss. Then, the samples were allowed to cure for 7 days in a conditioned room at 18°C. After that period, when the samples showed enough strength to be handled, their ends were flattened with a spatula and they were carefully extruded out of the moulds. Samples with a diameter of 100 mm and a height of 60 mm were used for hydraulic conductivity tests. 3 METHODS

The hydraulic conductivity and small-strain shear modulus of a CB sample were studied in a flexible-wall hydraulic conductivity cell provided with bender elements. In parallel, the hardening of the CB mix in contact with water was monitored through bender element testing in a bench top setup to provide a reference of G0 increase under normal curing conditions.

3.1 Bench top bender element test

The small-strain shear modulus of the CB mix was evaluated (starting from a fresh state) by bender element testing (Shirley and Hampton 1978, Dyvik and Madshus 1985) in a bench top setup (Fig. 1). The bender elements used here are of the type T220-A4–203Y (Piezo Systems, Inc.). The effective bender element length penetrating in the sample was about 4.5 mm.

The bench top bender element setup consists of two translucent polymethyl methacrylate (PMMA) plates that hold a U-shaped rubber mould with an open space for housing a cemented sample. The bender element transmitter and receiver are fixed to the PMMA plates, one in front of the other and vertically aligned. All parts are held together by four sets of screws and nuts resting on rubber disks to avoid wave propagation through the apparatus itself. More details on this setup are given by Verastegui-Flores et al. (2010).

Testing was started immediately after a fresh CB mix was prepared. The mix was poured into the rubber mould and it was allowed to cure in a conditioned room at constant temperature (18°C). In order to avoid desiccation, the sample was kept all the time under a thin layer of purified water. Bender element measurements were performed on a regular basis from the first day of curing up to 2 months approximately. In bender element testing, G0 is determined out of the propagation velocity (Vs) of shear waves generated and detected by the transmitter and receiver bender elements installed on opposite sides of a sample. G0 is estimated as:

G0 = Vs2 (1)

where is the density of the sample. Vs is evaluated as follows:

Vs = L / ts (2) where L is the tip-to-tip distance between the transmitter and receiver bender elements, and ts is the travel time of the shear waves from the transmitter to the receiver. ts is evaluated out of the signal recordings. In this research, ts was evaluated by means of two methods. The first one consists of visually identifying the first direct arrival from the output signal (e.g. Dyvik and Madshus 1985, Jovičić et al. 1996; Viggiani and Atkinson 1995). Clearly, the success of this method depends on the quality of the recorded signal. The second method used here was the cross-correlation method, first introduced by Viggiani and Atkinson (1995). The cross-correlation analysis measures the level of correspondence or interrelationship between two signals of similar nature and it produces the time shift between them, which is equivalent to the travel time of the shear wave. Although some authors argue that the cross-correlation method may not be suitable for bender element interpretation (e.g. Arulnathan et al. 1998), it produced very consistent results. Both methods produced a similar outcome.

Figure 1. Bench top bender element setup

979


3.2 Flexible-wall permeability cell with bender elements

A flexible-wall permeability cell was provided with bender elements (one in the base pedestal and the other on the top cap) to enable the simultaneous monitoring of G0 and k of a cement bentonite sample. Moreover, height changes during permeation could also be monitored through a cathetometer placed next to the cell (Fig. 2).

The parameter k was evaluated out of a falling-head test performed in a conditioned room at constant temperature (18°C) and at an isotropic effective stress of 30 kPa. The sample was first permeated with deionized water for about 1 month (1.7 pore volumes of flow). Next, the deionized water was replaced with the 25 g/L solution of Na2SO4 and the test was continued for a period of about 250 days (about 10 pore volumes of flow).

Figure 2. Flexible-wall hydraulic conductivity cell provided with bender elements and a cathetometer 4 RESULTS

4.1 Small-strain shear modulus

Figure 3 summarizes all G0 measurements carried out in the benchtop bender element setup and in the modified flexible-wall hydraulic conductivity cell. As expected, the results of the benchtop bender element setup, where the CB sample was cured in pure water, showed a gradual increase of G0 in time due to normal cement hydration. Verastegui-Flores et al. (2010) showed that the G0 increasing trend of clay treated with blast furnace cement could be fairly-well characterized through a logarithmic function. Clearly, all measurements up to a sample age of 90 days were in excellent agreement with such rule.

0

50

100

150

200

250

300

350

400

450

0 50 100 150 200 250

Smal

l-str

ain

shea

r m

odul

us (M

Pa)

Age (days)

Benchtop setup measurement

During permeation with water

During permeation with sulphates

Prediction of G0 cured in pure water

Figure 3. Impact of sulphate attack on the small-strain shear modulus of a cement-bentonite sample

Similary, G0 measurements in the flexible-wall permeability cell are in agreement with measurements out of the benchtop bender element setup during the first phase of the tests, when the sample was permeated with pure water for about one month. However, when the permeation with the 25 g/L Na2SO4 solution started, the normal cement hydration process was clearly disrupted (Fig. 3). G0 measurements up to 250 days of permeation, when compared to the expected G0 trend in contact with pure water, show that the stiffness of the CB sample was significantly reduced due to contact with sulphates. Such reduction is the result of interparticle cementation degradation and it could also indicate severe fissuring affecting the original structure of the CB sample. Clearly, a decrease in G0 suggests a decrease of strength as well as both parameters are strongly linked to interparticle cementation.

Deterioration of the cement hydration products by sulphates is a well-known durability problem in cement mortars exposed to high concentrations of sulphate ions. The most common manifestations of sulphate attack in concrete are expansion, caused by formation of ettringite and gypsum within the matrix of a specimen, and loss of strength. A similar phenomenon was observed in CB samples.

4.2 Hydraulic conductivity

Figure 4 summarizes all k measurements as well as sample height changes during permeation with water (for one month) followed by permeation with a 25 g/L Na2SO4 solution for a total period of about 250 days.

1,E-10

1,E-09

1,E-08

1,E-07

1,E-06

0 50 100 150 200 250

Hyd

raul

ic c

ondu

ctiv

ity

(m/s

)

Age (days)

Water Sulphates

Prediction of permeability to water

(a)

60

61

62

63

64

65

0 50 100 150 200 250

Sam

ple

heig

ht (m

m)

Age (days)

(b) Figure 4. Impact of sulphate attack on the (a) hydraulic conductivity and (b) the height of a cement-bentonite sample during permeation with water and a 25 g/L Na2SO4 solution in a flexible wall cell.

980


As expected during the first phase of permeation with pure water, a gradual decrease of k with increasing time was observed. This feature of cement-bentonite samples has been reported before in the literature (e.g. Fratalocchi et al. 1998, ICE 1999) and an equation describing such trend has also been proposed:

k(t) = k28d (t / 28)-n (3)

where k28d is the hydraulic conductivity at an age of 28 days, t is the age of the sample in days and n is a constant. Based on existing data and Eq. 3, it was possible to predict the permeability to water of the CB sample vs. time (Fig. 4a). In the second phase of the test, the CB sample was permeated with a Na2SO4 solution. As a result of deterioration of the sample an immediate increase of k was expected. However, in the early phase of sulphate attack (age between 30 to 70 days) the permeability to sulphates seems to be lower than the expected permeability to water. This feature may have been caused by formation of gypsum in the pores within the sample (Santhanam et al. 2003) which may result in a gradual clogging of the pores. Gypsum primarily deposits in the fissures and in voids, because these provide the best sites for nucleation. After a while, when the formation of ettringite becomes significant, the affected areas tend to expand, then, fissures start to appear which will gradually lead to an increased hydraulic conductivity. In fact, figure 4 shows that expansion of the sample and increase of hydraulic conductivity start at approximately the same time.

The hydraulic conductivity of a cement-clay mix is not a simple function of the porosity, but depends also on the size, distribution, shape, tortuosity and continuity of the pores that change during the cement hydration and sulphate attack.

5 CONCLUSIONS

Traditionally, the mechanical and hydraulic behaviour of cement bentonite samples are studied separately on different specimens. In this research, a flexible-wall hydraulic conductivity cell was provided with bender elements to measure the hydraulic conductivity as well as the small-strain shear modulus.

Monitoring of G0 was shown to provide valuable quantitative information to study the deterioration effects of sulphate attack on a cement-clay mix. As expected, during permeation with deionized water an increase of G0 and a decrease of permeability were observed due to normal cement hydration. On the other hand, after sustained contact with sulphates a significant decrease of G0 and increase of k were measured. These observations suggest that the sulphate attack produces severe degradation of interparticle cementation and perhaps also severe fissuring affecting the macrostructure of the CB sample which in turn may lead to a loss of strength. 6 ACKNOWLEDGEMENTS

The authors would like acknowledge Jonathan Mawet for his assistance during this research.

7 REFERENCES

Arulnathan R., Boulanger R.W. and Riemer M.F. 1998. Analysis of bender element tests. Geotechnical Testing Journal 21(2), 120-131.

Dyvik R. and Madshus C. 1985. Lab measurements of Gmax using bender elements. Proc. ASCE Annual Convention: Advances in the art of testing soils under cyclic conditions, Detroit, pp. 186–196.

European Standard EN 197-1. 2000. Cement - Part 1: Composition, specifications and conformity criteria for common cements.

Fratalocchi E. and Pasqualini E. 1998. Permeability over time of cement bentonite slurry walls. Proceedings of the 3rd International Congress on Environmental Geotechnics, ICEG. Lisbon, pp. 509-514.

Institution of Civil Engineers. 1999. Specification for the Construction of Slurry Trench Cut-off Walls. Thomas Telford, London, UK.

Jefferis S. A. 1981. Bentonite cement slurries for hydraulic cut offs. Proc. 10th ICSMFE, Vol. 1, 425-440.

Jefferis S. A. 1992. Contaminant-grout interaction. Proc. of the Specialty Conference Grouting, Soil Improvement and Geosynthetics, New Orleans, pp. 1393 – 1402.

Jovičić V., Coop M.R. and Simic M. 1996. Objective criteria for determining Gmax from bender element tests. Géotechnique 46(2), 357-362.

Opdyke S.M. and Evans J.C. 2005. Slag-cement-bentonite slurry walls. Journal of Geotechnical and Geoenvironmental Engineering 131(6), 673–681.

Ryan C. and Day S. 1986. Performance evaluation of cement-bentonite slurry wall mix design. Proc. HMCRI Conference, Washington.

Santamarina J., Klein K., and Fam, A. 2001. Soils and Waves: Particulate Materials Behaviour, Characterization and Process Monitoring. Wiley, Ltd., Chinchester, England.

Santhanam M., Cohen M.D., Olek J. 2003. Mechanism of sulfate attack: a fresh look Part 2. Proposed mechanisms. Cement and Concrete Research 33, 341 – 346.

Shirley D.J. and Hampton L.D. 1978. Shear-wave measurements in laboratory sediments, J. Acoust. Soc. Am. 63, 607–613.

Verástegui Flores R.D., Di Emidio G. and Van Impe W. 2010. Small-strain shear modulus and strength increase of cement-treated clay. Geotechnical Testing Journal 33(1), 62–71.

Viggiani G. and Atkinson J.H. 1995. Interpretation of bender element tests. Géotechnique 45(1), 149 – 154.

981

Centrifuge modelling of bored piles in sands

Modélisation en centrifugeuse de pieux forés dans le sable

Williamson M.G. University of Cambridge and Ove Arup & Partners

Elshafie M.Z.E.B., Mair R.J. University of Cambridge

ABSTRACT: As part of a series of experiments to investigate the effects of tunnelling on bored piles carried out at the Cambridge University Geotechnical Centrifuge, a new and novel model pile design is presented. The pile is semi-circular in cross section allowing the sub surface displacements around the pile toe to be monitored using particle image velocimetry (PIV). The new piledesign, along with the loading mechanism, ensured the load was transmitted predominantly though the pile centroid, which reducedthe bending effects which have previously caused significant errors in these types of problem. The pile load cells at the head and thebase were also placed along the pile centroid to minimise the effects of bending on the load measurement. The paper presents theresults of a centrifuge pile loading test, illustrating the excellent response of the load cells within their working range and the highquality PIV data which was obtained through this novel modelling approach.

RÉSUMÉ : Dans le cadre d’une série d’expériences réalisée dans centrifugeuse géotechnique de l’Université de Cambridge pourexaminer les effets tunnels sur les pieux forés, un nouveau modèle de conception de pieu est présenté. Le pieu est semi-circulaire en coupe transversale permettant aux déplacements sous la surface autour de la pointe du pieu d’être surveillés par imagerie de lavélocimétrie des particules (PIV). La nouvelle conception du pieu et le mécanisme de chargement assurent que la charge soittransmise à dominante par le centroïde du pieu, ce qui a réduit les effets de flexion qui avaient précédemment causé des erreurssignificatives dans ce type de problème. Les cellules de charge fixées en tête et en pointe des pieux étaient aussi placées le long ducentroïde des pieux pour minimiser les effets de flexion sur la mesure de la charge. L’article présente les résultats d’un test dechargement de pieux en centrifugeuse, montrant une excellente réponse des cellules de charge dans la zone de travail et la grandequalité des données PIV obtenues par cette nouvelle approche de modélisation.

KEYWORDS: pile settlement centrifuge bored tunnelling base PIV modelling

1 INTRODUCTION

The ability to predict the effect of tunnelling on piles is increasingly important no only from a safety, but also economic point of view. To provide a better understanding of the mechanisms influencing the effects of tunnelling on bored piles in sands centrifuge modelling has been carried out at the Cambridge University Geotechnical Centrifuge (Schofield 1980).

A significant amount of research work at Cambridge and worldwide in recent years has been carried out using particle image velocimetry (PIV) developed at Cambridge (White et al. 2003) as a tool to understand subsurface mechanisms for soil-structure interaction problems. The ability of this method allows both 2D plane strain and 3D plane of symmetry problems to be modelled. The latter of these methods was used in this research using plane strain (2D) tunnelling movements and applying these to a non plane strain (3D) pile loading setup.

As part of this research work a novel pile and pile loading system were designed at Cambridge University to model bored pile behaviour in sands and it is the pile design aspect of the research which is described in detail within this paper.

1.1 Background

Accurate modelling of plane of symmetry pile behaviour in the centrifuge has been attempted by previous researchers (Marshall 2009 and Lu 2010) however a major drawback has been the difficulty in providing an accurate measurement of the axial pile load.

The cross section of a plane of symmetry pile is semi-circular; this creates a variation in the flexural stiffness of the pile in its two principal bending modes. This results in the piles having a propensity to bend towards or away from the plane of symmetry (its minor axis) rather than parallel to the plane of symmetry (its major axis), see Figure 1.

This bending can lead to significant errors when strain gauged load cells (Marshall 2009) are used to measure the axial load of the piles such that their accuracy cannot be relied upon.

Loading these piles has also proven to be difficult owing to their position within a centrifuge package (extremely close to the plane of symmetry). This has previously prevented researchers from loading these piles within their centroid accurately and hence led to greater bending problems. Loading along the centroid of the piles at the pile head and placement of the strain gauged load cells along the pile centroid would allow a significant proportion of this error to be mitigated through bending compensation within the strain gauge bridges.

Placement of strain gauges on the centroid of a plane of symmetry pile has been attempted previously (Lu 2010) however as the strain gauged bridges were only ‘quarter’ Wheatstone Bridges these still suffered from a lack of true bending and temperature compensation. However the results were shown to be a significant improvement on the attempts to position strain gauged load cells away from the pile centroid.

982


2 PILE SPECIFICATION

To produce a pile and pile loading system capable of high quality bored pile behavioural simulation a wholly new design was required of both the pile and the pile loading system.

2.1 Pile Cross Section

Researchers in the past have used a single wall to represent a series of piles to simplify the problem to 2D plane strain (White and Bolton 2004), while others have used square piles to simulate the behaviour of circular piles.

Such solutions are not suitable for the problem type given the inherent 3D nature of the problem and the significant edge effects when using a plane of symmetry respectively. As such a semi-circular design was deemed most appropriate to provide the correct 3D stress and strain field around the pile.

Figure 1. Schematic cross section of plane of symmetry pile.

2.2 Axial Load Measurement

Shaft friction piles in the centrifuge have previously been shown to produce inaccurate mechanisms for piles in sands. The effect of shear band dilatancy on lateral stresses conditions means that neither the load nor the mobilisation strain along the shaft can be replicated concurrently (Lehane et al 2005).

To remove this error and to investigate a worst case scenario of an end-bearing bored pile in sand, the shaft was sleeved against friction.

Following on from this it was therefore only necessary to measure the load at the pile head and base to ensure that the sleeving was working.

In an attempt to reduce the errors associated with pile bending on axial load measurement load cells situated along the pile centroid were considered a suitable option.

2.3 Loading System

The pile loading system was designed to ensure that loading remained through the centroid of the cross section and simulated a bored pile. (This system is described in detail in Williamson 2013).

2.4 Final Specification

The final specification was therefore set based on the requirements described:

Pile to be semi-circular in cross section Axial load measurement at head/base Smooth face of shaft Pile must remain in contact with symmetry plane Loading through pile centroid Load measurement along pile centroid Loading to simulate bored pile behaviour

3 PILE DESIGN

The final pile design is shown in Figure 2.

3.1 Pile Body

The pile body was machined from 2014-T6 aluminium, with aluminium strain gauged load cells attached at the head and the base. The pile was 15 mm in diameter and had an overall length of 355 mm including the base and load cells.

3.2 Load Cells

The load cells were connected as full Wheatstone Bridges with 4 No 350 Ω strain gauges used, 2 active and 2 inactive.

These load cells are situated precisely on the centroid of the pile and connected securely between the pile shaft and base.

Figure 2. Final pile design.

3.3 Calibration

The pile load cells were calibrated before and after being affixed into the piles. This was to investigate the effect of bending during calibration resulting in the considerable free length of the pile which has previously found to be an issue (Marshall 2009).

A new calibration setup was designed to better replicate the effective lengths in the centrifuge (~75 mm), which were small in comparison with the overall free length of the pile (355 mm).

A comparison in the calibration factors between a load cell calibrated individually and the same load cell within the pile is shown in Figure 3. Clearly the agreement between the calibration factors was good, and hence the system was then taken forward to be used in the centrifuge where the effects of high acceleration loads could be tested on the pile.

983


4 CENTRIFUGE SETUP

4.1 Centrifuge Package and Instrumentation

The centrifuge setup consisted of a strongbox, with a loading frame/system and PIV cameras/lights (see Figure 4). The strongbox had a Perspex face against which the piles rested, a steel U-frame and an aluminium back, which provided very high stiffness boundaries around the soil.

The PIV markers were calibrated against known target positions. The pile displacements measured through PIV, were also monitored with linear variable differential transformers (LVDT’s).

To verify the values provided by the head and base load cells, Novatech F259 Miniature Diaphragm 1 kN Loadcells were attached to the loading system so that the change in head load could be measured independently.

Figure 3. Pile calibration – before affixing to the pile body and after affixing to the pile body.

4.2 Sand

The piles were placed firmly against the Perspex face and the sand rained down over them. Fraction E Leighton Buzzard Sand was used in the experiments at a relative density of 76% ±2%.

4.3 Experimental Procedure

Piles A and B (see Figure 4) were used to provide data on the effects of tunnelling on bored piles, whereas Pile C was used to provide details of the pile capacity and load settlement response.

The centrifuge was accelerated slowly to the desired 75g, with photographs being taken from each of the PIV cameras at various stages. The pile loads were monitored during the acceleration phase after which loads were added to the pile heads using a specially designed loading system, which applies only the dead load of a pile cap for Piles A and B.

Pile C had a different loading system which also did not apply any load until the desired 75g had been reached. This pile also had a pile cap but with a loading pin capable of applying a head beyond that of the cap dead weight.

All the piles were sleeved to the base (for details see Williamson 2013) so that only the base component was a factor in the pile loading.

5 RESULTS

5.1 Load Cell Response

The response of the load cells to the changing head load for Pile C is shown in Figure 5.

Clearly the comparison between the commercially available Novatech head load cell and the response of the new strain gauged load cells is very good compared with the ideal 1 in 1 slope. No greater than 8% variation between the different types of load cell is found.

The variation between the head and base load cell response is relatively constant at around 80 N between 420 N and 1000N. It was found that the pile toe began to move away from the face slightly at 1000 N and hence the divergence is likely to be attributed to some high level bending to the base load cell.

Piles A and B operate at working loads of between 100 and 120 N depending upon the experiment. Within this range the variation between the change in head and base loads is small, but the incremental loading is shown to be linear. This would perhaps indicate slightly higher amounts of bending on the head load cell. The calibration factors for the head load cells though linear were slightly less consistent throughout the tests than the base load cells. It is thought that their shorter length and the slight inconsistency in the contact point between the pile cap and the load cell could have affected the calibration factor, though it was shown to vary by no more than 10 % throughout the test series.

Figure 4. Centrifuge setup – illustrating the piles, loading system and PIV setup.

984


6 CONCLUSIONS

The paper has presented and discussed the details of developing and testing a novel model pile loaded in the centrifuge using a novel pile loading setup. The new pile, which adopted a semi-circular cross-section and was sleeved along its shaft to minimise friction, enabled a test load from the loading setup to be predominantly transmitted axially along the pile with minimal bending effects, which have been traditionally difficult to deal with. Within the working load range, the variation between the change in head and base loads was found to be very small. The vertical displacement PIV data at the pile toe shows the importance of the modelling details in obtaining high quality data.

7 ACKNOWLEDGEMENTS

Figure 5. Pile C load cell response – comparison of novel load cells with commercially available Novatech head load cells.

The authors would like to thank the EPSRC and Ove Arup and Partners for their support and funding for this project as well as the technical staff at Cambridge University Engineering Department for the construction of the items described in this paper.

8 REFERENCES

5.2 Load Settlement Response

The PIV response of the pile toe and the surrounding soil is shown in Figure 6 for the load relating to 10% of the pile diameter or 520 N (2.8 MPa). The data around the pile toe is of exceptional quality, and it is possible to see the rigid cone beneath the pile toe with the emanating strain bulb centred beneath the pile toe.

Schofield A. 1980. Cambridge geotechnical centrifuge operations. Géotechnique 30 (3), 227-268.

White D.J., Take W.A. and Bolton M.D. 2003. Soil deformation measurement using particle image velocimetry (PIV) and photogrammetry. Géotechnique 53 (7), 619-631.

Marshall A.M. 2009. Tunnelling in sands and its effects on pipes and piles. PhD Thesis. University of Cambridge.

Lu W. 2010. Axisymmetric centrifuge modelling of deep penetration in sand. PhD Thesis. University of Nottingham.

White D.J. and Bolton M.D. Displacement and strain paths during plane-strain model pile installation in sand. Géotechnique 54 (6), 375-397.

Lehane, B.M., Gaudin, C. and Schneider, J.A. 2005. Scale effects on tension capacity for rough piles buried in dense sand. Géotechnique 55 (10), 709-719.

Williamson M.G. 2013. The effects of tunnelling on bored piles. PhD Thesis. University of Cambridge.

Figure 6. Vertical displacement contours (mm) at a pile displacement of 10% DPile.

985

Stability and performance of ground improvement using geocell mattresses underextreme weather

La stabilité et les performances de l'amélioration du sol en utilisant des matelas géocellules dansdes conditions météorologiques extrêmes

Xu Y., Wang J.P.Department of Civil and Environmental Engineering, Hong Kong University of Science and Technology, Kowloon, HongKong

ABSTRACT: Owing to the rapid change in our living environment, extreme weather and its ensuing effect should become a focus offuture engineering designs. Therefore, the underlying scope of this study is to evaluate the performance of a relatively new method inground improvement under a harsh condition, simulating its response to intense rainfall as a result of extreme weather. The modeltests show that the ground with the geocell reinforcement is of higher bearing capacity, compared to a natural ground withoutreinforcement. In addition, the tension cracks around the footing, which was observed in the natural ground, would not develop in thereinforced soil. The cause to the improved bearing capacity in the soil is in line with the finding through element testing suggestingthat the geocell with soil infilled in the pocket as an integrated material equivalently becomes a cohesive material. On the other hand,the deformed geocell under the footing would develop a high passive earth pressure, larger than the water pressure possibly resultingin the tension cracks around the footing in a natural ground without the geocell reinforcement.

RÉSUMÉ: En raison de l'évolution rapide de notre cadre de vie, les conditions météorologiques extrêmes et leurs effets devraientdevenir un foyer d'études techniques futures. Par conséquent, le champ d'application de base de cette étude est d'évaluer laperformance d'une méthode relativement récente dans l'amélioration du sol dans des conditions difficiles, en simulant sa réponse à desprécipitations intenses dues à des conditions météorologiques extrêmes. Les essais sur modèle indiquent que le sol renforcé à unecapacité portante meilleure qu'un sol naturel non renforcé. En outre, les fissures de traction autour de la fondation ne se développentplus dans le massif renforcé par géocellules. La cause de l'amélioration de la capacité portante du sol est à relier à l'observationexpérimentale que l'élément géocellule remblayé avec de la terre devient équivalent à un matériau cohésif. D'autre part, la géocelluledéformée sous la semelle engendrerait une pression des terres passive élevée, plus grande que la pression de l'eau qui règenrait dansles fissures de traction développées autour de la semelle dans un sol naturel.

KEYWORDS: Geocell, ground improvement, extreme weather

1 INTRODUCTION

Geocell is a relatively new form of geosynthetics mainly usedfor geotechnical engineering. Different from the commonlyused geogrid, geocell is considered a three-dimensional soilreinforcement material (Wang, 2007, Tafreshi and Dawson2010). Its applications to engineering include soil retainingsystems (Wesseloo et al., 2008, Ling et al. 2009, Leshchinsky etal. 2009,), ground improvements (Krishnaswamy et al. 2000,Dash et al. 2001, Leshchinsky and Ling 2012), and erosioncontrol (Wu and Austin 1992). The results of element testing(Rajagopal et al. 1999) suggest that the geocell with soil infilledin the cell pocket as an integrated material equivalentlybecomes a cohesive material with its friction angle remainingmore or less the same as the soil infill. Utilizing such atechnique as for strengthening the soil, the dynamicperformance of the soil retaining system with geocell wallfacing was found satisfactory under the intense shaking of theKobe earthquake, with full-scale shake table tests (Wang 2007,Ling et al. 2009, Leshchinsky et al. 2009). In addition toengineering aspects, geocell applications could reduce someconstruction expense partly due to easy and rapid installation ofit, which can be constructed by low-skill crew without heavymachinery (Wang 2007, Ling et al. 2009).

Extreme weather and the ensuing effect, such as heavyrainfalls and floods, is the underlying cause to some recentcatastrophes. For example, in 2009 the Shiaolin landslide inSouth Taiwan, destroying a local village completely and

causing more than 400 casualties, was a result of an abnormalrainfall event brought by the Typhoon Morakot. The cumulativerainfall in three days reported at 1,700 mm is nearly equal to theannual rainfall of 1,800 mm around the region in the past fewdecades (Tsou et al. 2010).

However, owing to the continuing development of globalcivilization in need of keeping “exploiting” our living environment, global warming and extreme weather would notbe expected to calm until we have a sound and effectiveresponse, say, the advent of new technology and the change ofour living style and mind.

Therefore, this study aims to investigate the performance ofthe geocell-reinforced ground under a harsh environment (i.e.,intense rainfall condition), which could be anticipated during itsservice life owing to the changing climate and extreme weather.This study is mainly assisted with laboratory works of geocellapplications, and the details including experimental designs andsetups, results and discussions are summarized and given in thispaper.

2 EXPERIMENTAL DESIGNS, SETUPSAND MATERIALS

The experiment is to create a harsh condition for the ground tosimulate its response to intense rainfall events as a result ofextreme weather. Therefore, the water content in the soil modelis added to a level around 20%, at which a thin layer of watercan be observed on top of the model ground surface, simulating

986


a situation that the ground is nearly immersed by water becauseof intense rainfall. The model tests were carried out with asandbox fabricated in-house, which was deisgned at 1,670 mmin length, 550 mm in width, and 1,300 mm in height. Inaddition, the sandbox is perfectly sealed, preventing the waterfrom draining out during testing. Therefore, a desired conditionthat the ground is excessively and rapidly saturated by heavyrainfall could be best simulated. In addition, for a bettermodeling of the unreinforced natural ground, the model wasprepared at a loose to moderate condition with the unit weightaround 19.6 kN/m3, corresponding to a relative desntiy around40% at 20% water content.

The sandy soil used in the model tests is the so-called ChinaStandard ISO sand (Figure 1). The specific gravity of this soil is2.65, with the maximum and minimum dry densities equal to1.91 g/cm3 and 1.62 g/cm3, respectively. With our in-housetriaxial tests, the angle of internal friction is around 38 degreesat a relative density of 70%.

Figure 2 shows the grain size distribution of this sand. It isworth noting that although the soil is also categorized into a SP(i.e., poorly graded sand) type of soil according to the UnifiedSoil Classification System, its size is not particularly uniformwith some presence of small to large sand particles (see Figure1), compared to the “popular” sand, such as the Toyuora sand orNevada sand.

Figure 1. The China Standard ISO sand.

0.01 0.1 1 100

20

40

60

80

100

Cum

ulat

ive

perc

entf

iner

(%)

Grain size (mm)

D10=0.18mmD30=0.30mmD50

=0.83mmD

60=0.99mm

Figure 2. Grain size distribution of the China Standard ISOsand.

Figure 3 shows the appearance of geocell used in this study,which is manufactured by Beijing Orient Science & TechnologyDevelopment Co., Ltd. (BOSTD) in China. The pocket size ofthe geocell is rather comparable to most commercial geocells inaround 200 mm x 200 mm (Wang, 2007). Note that the geocellsample, courtesy of the manufacturer, is of no perforated holeson it. Also note that the 75-mm-high geocell sample isspecifically adopted in this study, for a better fit to thedimension of the experimental layouts.

Figure 4 shows the installation of geocell during modelpreparation. The instrumentation includes LVDTs to measurethe ground settlement and heave, and the particle imagevelocimetry (PIV) technique (White et al. 2003) to analyze thesoil displacement in the ground subject to strip loading.

Figure 5 shows the schematic diagram of the experimentlayout. With the footing’s width in 100 mm, the ratio betweenthe footing width to the geocell pocket size is around 0.5, closeto the optimum ratio suggested in a related study (Dash et al.2001b). Note that a layer of geotextile was installed between thegeocell mattress and the foundation for preventing the wash-away of sand infills from the pocket. A total of three tests, twowith geocell reinforcement and one without reinforcement, arereported in the following.

Figure 3. Top view of geocell mattress after expanding.

Figure 4. Installation of geocell in the soil model.

Loading strip

Load cellHydraulic cylinder

B=100mm

Geocell mattressGeotextile

Figure 5. The schematic diagram of the experiment layout.

Pocket size = 200 mm

Geocell height = 75 mm

Foundation constructed by theChina Standard ISO Sand, in 40%relative density and 20% watercontent

987


3 RESULTS AND DISCUSSIONS3.1 Bearing Capacity

Figure 6 shows the relationships between load and settlement onthe footing. At a given settlement, the ground with more layersof geocell is indeed of higher bearing capacity. Given 0.5 timesof the footing width (i.e., 0.5B) being the tentative designsettlement, the model tests show that the bearing capacities ofthe ground reinforced with one-layer and three-layer geocellmattresses are 1.4 and 2.2 times of the natural ground, withoutany reinforcement.

The increase of bearing capacity with the geocellreinforcement could be in line with the finding (Rajagopal et al.1999) that the geocell filled with soil as an integrated materialequivalently becomes a cohesive material, and in the meanwhilethe angle of internal friction remains more or less the same asthe sand infills. Therefore, in use of the Terzaghi’s bearingcapacity theory (see the textbook of Das 1999), the extramaterial cohesion adds the overall bearing capacity to theground. More importantly, unlike the shear strength contributedby the frictional behavior becoming nominal as the effectivenormal stress is reduced significantly with the increase of porepressure, the soil strength contributed by cohesion isindependent of external stress and water pressure, or it shouldcome to existence regardless of external stress condition.

60

50

40

30

20

10

00 25 50 75 100 125 150 175 200 225

Bearing load (kPa)

Foot

ing

settl

emen

t,s/

B(%

)

Without reinforcementWith one-layer geocell reinforcementWith three-layer geocell reinforcement

Figure 6. The relationships of load and settlement in the threemodel tests.

3.2 Tension cracks

Figures 7 and 8 show the model ground surface with andwithout the geocell reinforcement. For a “natural” ground without reinforcement, major tension cracks were observed veryclose to the footing. On the other hand, tension cracks in the tworeinforcement tests were found located much further away fromthe footing, developing within the soil inside of the two ends ofthe geocell mattress.

The development of cracks close to the footing in the non-reinforcement test should be caused by the excessive pore waterpressure excited in the soil due to external loading. As thelateral soil pressure less than water pressure, the tension crackshould start developing. However, as the soil is reinforced bygeocell, the deformed geocell under the footing tends to shrinkthe size of pockets, resulting in a large passive earth pressurethat is larger than the water pressure, and therefore, thedevelopment of tension cracks is not allowed around the footingwith the geocell reinforcement. This also possibly explains thatthe cracks would develop within the soil at the two ends of thegeocell mattress, because the level of deformation in geocell isrelatively small and the corresponding passive earth pressure is

not large enough to compensate the excited pore pressure in thewater-immersed ground.

Figure 7. The side view of the model ground surface: (a) withgeocell mattress; (b) without geocell mattress.

Figure 8. The top view of the model ground surface: (a) withgeocell mattress; (b) without geocell mattress.

3.3 Ground surface settlement and heave

Figure 9 shows the ground surface settlement at differentdistances from the footing captured with LVDTs. In the three-layer geocell model, the ground surface tends to settle in arelatively large area, owing to the geocell-soil mattress acting asan integrated system. Simply speaking, the geocell-soilcomposites far away from the footing were pulled down owingto the geocell’s structure, causing the ground settlement alsoobserved relatively far from the footing.

On the other hand, for the natural ground without the geocellreinforcement, the soil adjacent to the footing was pushedupwards because of soil failure occurring right under the footingthat would have formed a failure surface because of differentlevels of soil movement. It is worth noting that this mechanismand pattern in the ground deformation is well documented in abearing capacity test (Das 2007).

The ground deformation captured with LVDTs is on thesame page of the displacement field suggested by the PIVsystem, as shown in Figure 10. For the natural ground, the PIVdisplacement vector (Figure 10b) was pointing upwards near theground surface, but at the same locations, the downwarddisplacement vectors (Figures 10a) were observed as the groundwas reinforced by geocell. It is worth noting that thedisplacement fields of the reinforced ground are relativelyrandom compared to the natural ground, which should resultfrom the fact that the surface processed by PIV is neither acompletely reinforced soil nor a completely un-reinforced soil,as the boundary condition of the geocell structure shown inFigures 3 and 4.

4 CONCLUSIONS

This paper summarized the experimental work of using geocellin ground improvement under an intense rainfall condition,which recently recurs with an increasing rate owing to climatechange and extreme weather. The result shows that theinstallation of geocell can indeed effectively improve thebearing capacity of the loose-to-moderate ground subject tohigh water content as a result of intense rainfall. The increasedbearing capacity should possibly result from the “equivalent

Tension crack

Tension crack

Tension crackFooting

(a) (b)

(a) (b)

988


cohesion” as geocell and sand are integrated as a compositematerial. In addition, the deformed geocell inducing a largepassive earth pressure in the soil within cell pockets wouldprevent the development of tension cracks close to the footing,which was observed in the natural ground without the geocellreinforcement.

50

40

30

20

10

0

-10

0 100 200 300 400

Natural ground without reinforcement

Geocell-reinforced ground

Distance from footing, d/B (%)

Surf

ace

settl

emen

t,h/

B(%

)

Figure 9. Ground surface settlement profile (negative values implyingground heave).

Figure 10. Soil deformation suggested with the PIV technique:(a) three-layer geocell-reinforced ground; (b) natural groundwithout reinforcement.

5 REFERENCES

Das B. M., 1999. Principles of Foundation Engineering 4th ed. PWSPublishing.

Dash, S.K., Krishnaswamy, N.R. and Rajapopal, K. 2001. Bearingcapacity of strip footings supported on geocell-reinforced sand.Geotectiles and Geomembranes 19(4): 235-256.

Krishnaswamy N.R., Rajagopal K. and Madhavi Latha G. 2000. Modelstudies on geocell supported embankments constructed over a softclay foundation. Geotechnical Testing Journal, 23(1), 45-54.

Leshchinsky B. and Ling H.I. (2012). Effects of geocell confinement onstrength and deformation behavior of ballast. Journal ofGeotechnical and Geoenvironmental Engineering, ASCE.(accepted)

Leshchinsky D., Ling H.I., Wang J.P., Rosen A. and Mohri Y.2009. Equivalent seismic coefficient on geocell retentionsystems. Geotextiles and Geomembranes 27, 9-18

Ling, H.I., Leshchinsky, D., Wang, J-P., and Rosen, A. 2009. Seismicresponse of geocell retaining walls: Experimental studies." Journalof Geotechnical and Geoenvironmental Engineering, ASCE,135(4), 515-524.

Rajagopal K., Krishnaswamy N.R. and Madhavi Latha G. 1999.Behaviour of sand confined with single and multiple geocells.Geotectiles and Geomembranes 17, 171-184.

Tafreshi S.N.M. and Dawson A.R. 2010. Comparison of bearingcapacity of a strip footing on sand with geocell and with planarforms of geotextile reinforcement. Geotectiles and Geomembranes28(1), 72-84.

Tsou C.Y., Feng Z.Y. and Chigira M. 2010. Catastrophic landslideinduced by Typhoon Morakot, Shiaolin, Taiwan. Geomorphology127, 166-178.

Wang J.P. 2007. Full-scale shaking table tests of reinforced retainingwalls with geocell facing (Ph.D Thesis). Columbia University, NewYork City

Wesseloo J., Visser A.T. and Rust E. 2008. The stress-strain behavior ofmultiple cell geocell packs. Geotectiles and Geomembranes 27, 31-38.

White D.J., Take W.A. and Bolton M.D. 2003. Soil deformationmeasurement using particle image velocimetry (PIV) andphotogrammetry. Géotechnique 53(7), 619-631.

Wu K.J. and Austion D.N. 1992. Three-dimensional polyethylenegeocells for erosion control and channel linings. Geotectiles andGeomembranes 11(4-6), 611-620.

(a)

(b)

Documents

Technical Committee 104 Physical Modelling in Geotechnics · Technical Committee 104 Physical Modelling in Geotechnics ... Ils incluent notamment des expériences de laboratoire menées