a0 Brasilia, DF, Brazil

ndicalorproetichetdvasoilhesoil

of utmrced soependthe sodd toal charreforen devederstaneferen

quantify soilreinforcement bond under these circumstances. Inregion B, soil and reinforcement can deform laterally, so a planestrain test similar to the in-soil tensile test can be used in this case.Region C shows a situation where soil and reinforcement aresheared, so the direct shear test with the reinforcement inclined to

the literature for the direct shear test apparatus. In conventionaldirect shear devices, the bottom half of the box can be occupied bythe same soil as the top, a different soil or a rigid block. Usually themain differences among test arrangements are related to the waythe geosynthetic is xed and to the procedure adopted to apply thenormal stress to the soil sample top. This stress can be applied bya rigid and free top plate, a rigid top plate not allowed to rotate,a top plate xed to the top half of the cell or a exible pressurisedbag. In spite of some advantages (Jewell,1981), the use of a top plate

q Mercer Lecture 20072008.* Corresponding author. Tel.: 55 61 3273 7313x217; fax: 55 61 3273 4644.

Contents lists availab

Geotextiles and G

sev

Geotextiles and Geomembranes 27 (2009) 368390E-mail address: palmeira@unb.brSome of the test devices developed have become classical andFig. 1 (Palmeira and Milligan, 1989a; Mendes et al., 2007) to someextent helps to understand the reasons for that. This gure showssome possible failure or deformation mechanisms of a reinforcedsoil wall depending on the region and loading conditions consid-ered. In region A (Fig. 1) sliding of the soil mass on the reinforce-ment surface can occur, so direct shear tests can be employed to

adherence between these materials.

2. Direct shear tests

2.1. Inuence of boundary and test conditions

Different apparatus and boundary conditions can be found inSoilreinforcement interaction isdesign and performance of reinfointeraction can be very complex, dproperties of the reinforcement andused as reinforcing materials can aproblem because of their geometricstrain rate dependent behaviour. Theand theoretical analyses have beedecades in order to improve the unment interaction, with particular rsynthetic materials.0266-1144/$ see front matter 2009 Elsevier Ltd.doi:10.1016/j.geotexmem.2009.03.003ost importance for theil structures and thising on the nature andil. Some geosyntheticsthe complexity of theacteristics or time and, different types of testsloped during the lastding of soilreinforce-ce to the use of geo-

being pulled-out, so pull-out tests would be applicable. It should bepointed out that all these test types have limitations in simulatingthe actual conditions found in a reinforced soil structure.

Soilgeosynthetic interaction has been modelled theoreticallyand experimentally in many different ways for the last threedecades. Themain objective of this paper is to discuss experimentaland numerical modelling tools for the understanding and evalua-tion of such interaction, which is relevant to most of the applica-tions of geosynthetics. In the following sections, the characteristicsand limitations of some tests like those depicted in Fig. 1 are pre-sented and discussed, as well as their potential for the study ofother relevant aspects of soilgeosynthetic interaction, besides the1. Introduction the shear plane can be employed. In region D the reinforcement isSoilgeosynthetic interaction: Modelling

Ennio Marques Palmeira*

University of Brasilia, Department of Civil and Environmental Engineering, FT, 70910-90

a r t i c l e i n f o

Article history:Received 10 November 2008Received in revised form13 March 2009Accepted 14 March 2009Available online 23 April 2009

Keywords:GeosyntheticsSoilgeosynthetic interactionTestingAnalysis

a b s t r a c t

Interaction between soils areinforcement in geotechnenvironmental protection wdepending on the type andsome experimental, theorbetween soils and geosyntreinforcement. The main amethods for the study ofmethods are addressed. Ta better understanding of

journal homepage: www.elAll rights reserved.geosynthetics is of utmost importance in applications of these materials asengineering. That is also the case for some applications of geosynthetics inks. The mechanisms of soilgeosynthetic interaction can be very complex,perties of the geosynthetic and the soil. This paper presents and discussesal and numerical methods for the study and evaluation of interactionics, with particular reference to the applications of these materials in soilntages and limitations of some traditional experimental and theoreticalgeosynthetics interaction are presented and new applications of theseneed for improvements in experimental and theoretical techniques forgeosynthetic interaction is highlighted.

2009 Elsevier Ltd. All rights reserved.nd analysisq

le at ScienceDirect

eomembranes

ier .com/locate/geotexmem

xed to the box top half is limited to tests on dilative samples.Pressurised bags are more practical for large scale devices andguarantee a uniform normal stress distribution along the samplesurface. One should expect that different boundary conditions arelikely to produce some differences in test results.

Another aspect that must be considered is the inuence offriction along the internal faces of direct shear boxes with the tophalf xed, particularly for dilative materials, as shown in Fig. 2(a).Friction along the box frontal face restricts soil dilation, whichincreases the normal stress on the shear plane and consequentlythe shear strength measured. Assuming that friction along the boxfrontal wall is fully mobilised, a rather crude analysis of equilibriumof the top sample half yields to the following equation, whichestimates the increase on themeasured shear strength as a functionof the friction angle (dsw) between soil and box internal face and ofthe soil friction angle (4).

dimensions and surface characteristics of the geotextile. In tests ongeogrids one can have a test arrangement where a rigid base or soilis placed underneath the grid. In such cases, friction will bedistributed between the soil particles in between the transversemembers and the solid surface of the geogrid. Jewell et al. (1984)have derived expressions to evaluate bond between soils andgeogrids under such conditions. However, the shear stress distri-bution along the interface length will depend on how the geo-synthetic specimen is xed to the apparatus. Fox and Kim (2008)discussed and quantied the occurrence of progressive failure indirect shear tests on geomembraneGCL interfaces. A series of largedirect shear tests performed by these authors indicated thatprogressive shear failure occurred when the intended failuresurface did not have the lowest shear strength of all possible slidingsurfaces and when the geosynthetic specimens were clamped.Progressive failure reduced the measured peak shear strength andincreased the shear displacement at peak. The development ofprogressive failure has also been observed in large scale directshear tests (Palmeira, 1987b; Palmeira andMilligan, 1989b,c) and innumerical analyses of ramp tests, as commented later in this paper.

Regarding direct shear tests on geotextile specimens at the boxcentral plane, the results obtained may also be inuenced bygeotextile distortion. Fig. 3 shows direct shear test results with thetraditional arrangement of a nonwoven geotextile at the box mid-height, with the geotextile specimen glued to a rigid block in the

F

f (F)

QQ

60

(%

)

a

b

Dense granular soilFriction

Soil movementdue to dilation

Fixed tophalve

E.M. Palmeira / Geotextiles and Geomembranes 27 (2009) 368390 369Dsso

11 tan dswtan 4 1 (1)

where:Ds is the shear strength increase caused by side friction, so isthe shear strength of the soil for a perfectly smooth box internalsidewall, dsw is the friction angle between soil and box internal walland 4 is the soil friction angle.

The graph in Fig. 2(b) shows predictions by equation 1, where itcan be observed that signicant increases on the measured soilshear strength can occur for shear boxes with xed top half andrough internal walls, depending on the roughness of the box walland on the soil friction angle. The inuence of the box internal wallcan be veried by the results of large scale direct shear tests onsamples of dense Leighton Buzzard sand using an apparatus withthe top half xed (Fig. 2c, Palmeira, 1987a). The soil samples hada total volume of 1 m3 (1 m 1 m 1 m). In one of the tests, thefrontal face consisted of a steel plate with an anti-rust coating. Inthe other test, the wall had a smooth painted surface which waslubricated with double layers of plastic lms and oil. The results inFig. 2(c) show the marked inuence of the friction along the frontalface on the shear strength obtained, as suggested by the rathercrude analysis presented in Fig. 2(b), as well as on the shape of theshear stress versus shear displacement curve.

In direct shear tests with the reinforcement at box mid-heightone has to take into account the type of reinforcement that is beingtested for a proper analysis of the test results. In the case of a geo-textile, if the specimen is appropriately xed to the testing box,a rather uniform shear mechanism will develop along the soilgeotextile interface, and some interlocking between soil particlesand geotextile bres may take place, depending on the particle

B

C

D

AFig. 1. Interaction mechanisms in a geosynthetic reinforced soil wall (modied fromPalmeira and Milligan, 1989a).0 10 15 20 25 300

0.5

1

1.5

2

/

Shear displacement (mm)

Lubricated wall

rough wall

= 30 kPa

Q

1m

1m

Fixed top halve

5

0 10 15 200

10

20

30

40

50

Sh

ear stren

gth

in

crease

30

Soil-wall friction angle ()

37

4550Soil frictionangle 11

1tantan sw

5

c

Fig. 2. Inuence of side friction in direct shear tests on dilative soils. (a) Side friction

mobilization. (b) Estimate of shear strength increase due to side friction. (c) Inuenceof side friction in a large scale direct shear test.

box bottom half (Tupa, 1994). Tests were performed on the soil(ne sand) alone and on the sandgeotextile interface. An addi-tional test with the geotextile xed to two rigid blocks was alsocarried out subjecting the geotextile to simple shear conditions.The results obtained show that the initial shear displacements ofthe soilgeotextile interface were those caused by the distortion ofthe geotextile. After some level of geotextile distortion hasoccurred, relative movement between soil particles and geotextiletook place. This behaviour is relevant to numerical analysis ofreinforced soil structures, because numerical solutions requireinterface shear stiffness as one of the input parameters. Numericalanalyses conducted by Hatami and Bathurst (2005) have shownthat the soilreinforcement interface stiffness can have a signi-cant inuence on the lateral displacements of a reinforced wallface. So, due care must be taken when choosing the value of thesoilnonwoven geotextile interface shear stiffness, as it may bedependent on the geotextile shear stiffness. Under eld conditionsthis inuence will depend on the construction characteristics andon the level of impregnation of the geotextile by soil particles. Ifthe geotextile is heavily impregnated, the mobility of the bres

E.M. Palmeira / Geotextiles and Geom370will be restricted and distortion under simple shear conditions islikely to be small.

The interpretation of the results of direct shear tests on sampleswith the reinforcement intercepting the shear plane (region C inFig. 1) is quite complex. This can be veried by test results by Dyer(1985) using photo-elasticity, which showed the inuence of thepresence of the reinforcement on the state of stress in the soilsample, as presented in Fig. 4(a)(c). In these experiments, crushedglass was used in substitution to soil and photo-elasticity tech-niques were employed. The normal stress was applied to thesample top by a rigid plate. The photograph in Fig. 4(a) was takenduring an unreinforced test, whereas those in Fig. 4(b) and (c) weretaken during tests on samples reinforced by vertical and inclinedsteel grids, respectively. Bright regions in the photographs areregions of high compressive stresses, whereas dark regions areregions of low stress levels. The pattern of light fringes obtainedshows that the presence of the reinforcement changes signicantlythe state of stress in the sample in a very complex way. In addition,it can also be noticed some interaction with the top and bottomrigid boundaries of the shear box in the reinforced tests, whichmayinuence test results.

The inuence of the top boundary on direct shear test resultscan be identied in tests on unreinforced and reinforced sand

0

100

0 2 4 6 8

(kP

a)

Shear displacement (mm)

Sand only

Geotextile only(simple shear)

Geotextile + sandFig. 3. Inuence of nonwoven geotextile distortion on the results of direct shear tests.shown in Fig. 5(a) and (b). The shear box used in these tests haddimensions 250 mm 250 mm 150 mm and the reinforcementwas a steel grid. Top boundary conditions tested were a rigid topplaten, a top platen xed to the box top half and a pressurisedrubber bag (Palmeira, 1987b). In Fig. 5(a) the result for the unre-inforced sand shows that some differences of peak shear strengthcan occur and that the tests with a more movable top boundarycondition yield to a slightly less fragile behaviour of the sand afterpeak strength. The inuence of the type of top boundary is morenoticeable in the test with a vertical steel grid reinforcement in thesample, as shown in Fig. 5(b). This is likely to be in part a result ofthe interaction between the reinforcement and the rigid bound-aries, as indicated in the photo-elastic study (Fig. 4b), and of thedifferent levels of constraint imposed to soil deformation,depending on the top boundary condition considered. Similardifferences in peak interface shear strength results were obtainedby Hsieh and Hsieh (2003) in direct shear tests on sandgeo-membrane interfaces, with the geomembrane at the box middleplane and under different conditions of rigidity of the box topboundary.

Fig. 6(a)(c) shows results of large scale direct shear testscarried out on 1 m3 reinforced sand samples with and withoutinclined reinforcements (Palmeira and Milligan, 1989b). Totalpressure cells were positioned along the reinforcement plane(Fig. 6a), along the failure plane (Fig. 6b) and along the rigid baseof the shear box (Fig. 6c). It can be noted the non-uniform stressdistributions in these gures. Fig. 6(a) shows that the normalstress acting on the reinforcement depends on its type and stiff-ness, because that stress is a function of the shear load mobilised.Fig. 6(b) shows that the presence of the reinforcement causesa signicant increase on normal stresses on the shear plane,which will yield to increased strength of the reinforced sample incomparison to the unreinforced one. Fig. 6(c) shows the non-linear nature of the normal stress distribution along the box rigidbottom boundary caused by the presence of the reinforcement,which can also be identied in the photo-elastic results presentedin Fig. 4.

The marked effect of the presence of the reinforcement can alsobe visualised in the nite element results presented in Fig. 7 (Sieira,2003). The soil in this analysis was modelled as an elastic perfectlyplastic material with MohrCoulomb failure criterion. Interfaceelements between soil and reinforcement were used. Fig. 7 showsregions of maximum shear stresses in the soil mass at failure fora test under a normal stress of 100 kPa. Both the nite element andthe photo-elastic results shown in Fig. 4 are consistent and enhancethe complex nature of the state of stress in the reinforced soil massin a test with the reinforcement inclined to the sample centralplane.

The complex nature of the results presented above indicatesthat direct shear tests with inclined reinforcement may providequalitative data, but the extrapolation of its results to real rein-forced structures is still rather limited.

2.2. Summary on direct shear tests

A summary regarding direct shear tests is presented below:

It is a good test for determining interface shear strength. Boundary conditions may inuence test results, particularly forsmall shear boxes.

It is difcult to assess soilgeotextile interface shear stiffness,which is relevant for numerical simulations of reinforced soilstructures such as walls and steep slopes.

The interpretation of tests with inclined reinforcements is very

embranes 27 (2009) 368390difcult.

5). (

E.M. Palmeira / Geotextiles and Geomembranes 27 (2009) 368390 3713. Pull-out tests

3.1. Inuence of boundary conditions

Pull-out tests are relevant for the study of the anchoragestrength of reinforcements (Fig. 1), particularly in the case of geo-grids. In this type of test, one has also to consider the inuence ofboundary conditions, which is more relevant as the test is notstandardised throughout the world. Besides, the interpretation ofthe results of tests on geogrids is not easy due to the complex

Fig. 4. Results of photo-elastic studies in the direct shear test (Dyer, 198geometry of such materials and inuence of effects such as soildilation and reinforcement loadstraintime characteristics, forinstance.

Someworks in the literaturehave shownthat the conditionsof thefrontal face of the box can have a marked effect on the test results(Palmeira, 1987b; Palmeira and Milligan, 1989a; Johnston and Rom-stad, 1989; Farrag et al., 1993; Lopes and Ladeira, 1996; Raju, 1995;Moraci and Montanelli, 2000; Sugimoto et al., 2001, for instance).

1.5

1.0

0.5

00 5 10

= 30 kPa

free top platenfixed top platenpressurised bag

Shear displacement (mm)

a

Fig. 5. Inuence of the rigid top boundary on direct sFig. 8(a)(d) shows some typical boundary conditions of the pull-outbox found in the literature. In the traditional test arrangement the soilis in contactwith the rigid frontal face and it is important that frictionalong this interface be minimised (Fig. 8a). The reduction of frictionalong this interface is usually achieved by using layers of plastic lmsand oil or grease (Palmeira, 1987b; Abramento, 1993, for instance).Alternatives tominimisepossible inuencesof thebox frontal facearethe use of sleeves, as shown in Fig. 8(b) (Farrag et al., 1993; Wilson-Fahmy et al., 1994; Perkins and Cuelho, 1999, for instance) orarrangements where the reinforcement length which is effectively

a) Unreinforced; (b) Vertical reinforcement; (c) Inclined reinforcement.tested is distant from the front wall (Fig. 8c, Palmeira, 1987b). Thelatter is practical only for tests on geogrids with regular patterns ofapertures (square or rectangular), when it is easy to remove some ofthe frontal transverse members of the grid. Another possibility is tohave a exible or movable frontal face made of pressurised bags(Fig. 8d) or plates (Sugimoto et al., 2001), respectively. Whittle et al.(1991) described a pull-out test devicewhere a exible boundarywasused at the rear face of the soil sample.

1.5

1.0

0.5

00 5 10

pressurised bag

free top platen

15

reinforcement

Shear displacement (mm)

b

hear test results. (a) Unreinforced. (b) Reinforced.

Geom= 30 kPa

400

200

00 0.5

x/L

metal grid

woven geotextile

= 30 kPa nx

L

(kPa)

n

a

b

E.M. Palmeira / Geotextiles and372To perform large scale pull-out tests to investigate the inuenceof boundary conditions is a hard working and time consuming task.Therefore, numerical methods can be used as tools to improve theunderstanding of some factors that may inuence test results. Oneof the key questions is to what extent the size of the pull-outapparatus may affect the test results. Rather limited experimentaldata suggests that the size of the apparatus may have a signicanteffect on the results obtained (Dyer, 1985; Palmeira and Milligan,1989a; Farrag et al., 1993; Lopes and Ladeira,1996). This inuence isalso suggested by results from numerical analyses (Dias, 2003).Fig. 9 shows results of pull-out tests on a reinforcement buried insand obtained using the nite element method. In the simulations,the box height was varied between 0.3 m and 1 m. Soil wasmodelled as an elasticplastic material and the reinforcement asa linear elastic material placed at the sample mid-height. Thelengths of the box (2 m) and of the reinforcement (0.5 m) were keptconstant. A rigid and lubricated box frontal face with 6 interfacefriction angle was assumed in the simulations as well as along the

10 0.50

50

100

150

x/L

x/L

At peak - reinforced specimen

Predicted for unreinforced specimen

= 30

x

L

n

(kPa)

n

75

50

25

00 1.00.5

xL

reinforcement

yb

(kPa)

yb

c

Fig. 6. Internal total stresses in a large scale direct shear test (Palmeira and Milligan,1989b). (a) Normal stresses on the reinforcement. (b) Normal stresses on the samplecentral plane. (c) Normal stresses at the box bottom.other internal faces of the apparatus. It can be noted that the lowestbox height yielded a stiffer pull-out response and a highermaximum pull-out load. Little inuence of the box height wasobserved for heights greater than the reinforcement length. Similarconclusions were obtained by Farrag et al. (1993) and by Lopes andLadeira (1996) in pull-out test experiments where the thickness ofthe soil was varied. As the reinforcement length tested in pull-outtests is typically smaller than 1 m, the numerical study describedabove and experimental results suggest that under these condi-tions the soil sample height should be greater than 0.6 m.

The nite element method can also be used to assess theinuence of the presence of a sleeve at the box front wall in anexample of a test with a reinforcement with a tensile stiffness (J) of200 kN/m in a hypothetical box with the dimensions shown inFig. 10(a) (Dias, 2003). In this case, the soil was also modelled as anelasticplastic material and the reinforcement as a linear elasticmaterial. Fig. 10(b) shows the comparison between predictions oftest results with no sleeve and a lubricated frontal wall and withsleeves 15 cm and 30 cm long. The results in this gure show thatfor the conditions analysed the presence of the sleeve yieldedmaximum pull-out loads greater than that observed in the case ofa lubricated frontal face. The sizes of the sleeves used did notinuence signicantly the maximum pull-out load obtained. Incontrast, Farrag et al. (1993) found a maximum pull-out load ina pull-out test with a 20 cm long sleeve approximately 20% greaterthan that obtained in a test with a sleeve 30 cm long. Lopes andLadeira (1996) found a maximum pull-out load in a test without

Fig. 7. Maximum shear stress distribution in a reinforced sample in a direct shear test FEM results (Sieira, 2003).

embranes 27 (2009) 368390sleeve (and a smooth, but apparently not lubricated frontal face)10% greater than that obtained in a test with a 20 cm long sleeve.Thus, these contrasting results suggest that a more comprehensivestudy on how to minimise the inuence of the front box wall onpull-out test results is required.

Another possible arrangement without the use of a sleeveconsists of placing the reinforcement away from the wall, as showin Fig. 11 (Palmeira, 1987b). Geogrids (particularly those with welldened square or rectangular apertures) can be tested cuttingunnecessary transverse members and isolating the longitudinalmembers from the surrounding soil by pipes, for instance. In Fig. 11the result of the test on a 1 m3 soil (dense Leighton Buzzard sand)sample with a steel grid away from the front wall (rst transversemember 300 mm away from the box face) is compared to thetraditional test arrangement with a lubricated frontal face. It can beobserved that the difference between maximum pull-out loads issmall, but the arrangement with the grid away from the wallpresents greater and increasing pull-out loads for pull-outdisplacements above 7 mm. This is in part due to the fact that therst bearing member in this type of arrangement keeps on nding

soil

lubricated frontal

face

length tested

P

flexible front face

a Lubricated frontal face

nt a

Pp

Pp

Pp

con

E.M. Palmeira / Geotextiles and Geomembranes 27 (2009) 368390 373rather undisturbed soil ahead and the testing length of the gridremains the same, whereas in the traditional arrangement trans-verse members leave the soil sample and the length of the grideffectively tested decreases during the test.

Sugimoto et al. (2001) performed tests on polymeric gridsburied in sand using a large pull-out test apparatus witha movable frontal face, as shown in Fig. 12. This gure also pres-ents comparisons between results of tensile strain distributionsalong the geogrid length for tests with movable and xed frontalfaces under 49 kPa normal stress and for a pull-out load of15.4 kN/m. Different distributions of strains along the geogridlength can be seen, depending on the mobility of the frontal faceconsidered.

Numerical simulations also indicate that the rigidity of the topboundary inuences pull-out test results, particularly for lowheight samples. Fig. 13 shows numerical predictions of pull-outforces versus pull-out displacements in a conventional box

sleeve

b Sleeve

c Reinforcemewall

Pp

Fig. 8. Typical boundaryarrangement and two box heights (0.3 m and 1 m) obtained by Dias(2003). The soil was modelled as an elasticplastic material and thereinforcement as a linear elastic material. Interface elements wereused at the soilreinforcement interface. Rigid and exible topboundaries were simulated. The results show that the higher thesample the less the inuence of the top boundary rigidity on thetest results. These numerical results and experimental ones (Pal-meira and Milligan, 1989a; Farrag et al., 1993; Lopes and Ladeira,1996) clearly suggest that large pull-out boxes should be preferred.

Pu

ll-o

ut lo

ad

(kN

/m

)

0.5 m

Pp H

L = 2 m

= 25 kPa

6 friction angle atinterfaces

Fig. 9. Inuence of the pull-out box dimensField pull-out tests are alternatives for those in the laboratory(Bergado et al., 1992, 1993; Bakeer et al., 1998; Becker, 2006, forinstance). This type of test is usually difcult to perform, because ofless controlled operational conditions in the eld in comparison tothose in the laboratory. Clamps for the reinforcement and reactionfor the applied pull-out load have to be carefully designed. In testsin real reinforced structures one has also to consider the possibleinuence of the conditions of the wall frontal face (if any) on thetest results. Tests with the reinforcement buried in embankmentscan be performed, however, due care should be taken in theinterpretation of the results from these types of tests because of theusually low stress levels on the geosynthetic required for the test tobe practical. Besides, different and some times unrealistic failuremechanisms may occur, depending on the thickness of soil abovethe geosynthetic layer, uniformity of ll material properties andboundary conditions.

way from the

d Movable/flexible frontal face

ditions of pull-out tests.3.2. Modelling of soilgeosynthetic interaction in pull-out tests

Analytical and numerical methods can be employed for thestudy of soilreinforcement interaction (Palmeira, 1984; Abra-mento, 1993; Bergado and Chai, 1994; Sobhi andWu,1996; Madhavet al., 1998; Gurung and Iwao, 1999; Gurung, 2001; Perkins, 2001;Palmeira, 2004; Khedkar and Mandal, 2009, for instance). Abra-mento (1993) discussed the use of the shear-lag model for theevaluation of interaction between soil and reinforcement and his

Pull-out displacement (mm)

0

5

10

15

20

25

0 2 4 6 8 10 12

H = 0.3 mH = 0.6 mH = 1.0 m

ions numerical analyses (Dias, 2003).

elements (plane strain conditions). Fig. 14(a) and (b) shows theresults obtained in terms of pull-out loads versus pull-outdisplacements and Fig. 14(c) shows the predicted and measuredtensile strains distributions along the grid length. These resultsshow that, in spite of some simplifying hypotheses usually assumedin this type of analysis (value of interface shear stiffness, forinstance), it is possible to make predictions t reasonably well themeasurements, particularly for geotextiles. However, as is commonin this type of analysis, this may not necessarily be achieved withthe actual or expected values of some of the input parameters. Inthe case of geogrids, if the grid is assumed as an equivalent roughplanar reinforcement, predictions may deviate from measure-ments, depending on the grid geometrical characteristics and soiltype. Besides, when relevant, the loadstraintime relationship forthe inclusion may also affect the results, if this is only approxi-mately accounted for, which may in part explain some deviationsbetween results in Fig. 14(b).

= 25 kPa

Pp 1 m

1.2 m

sleeve

J = 200 kN/m

20

25

30b

/m

)

a

E.M. Palmeira / Geotextiles and Geomembranes 27 (2009) 368390374study approached the use of planar and continuous reinforcingmaterials. The Finite Element Method can also be used as a tool for

0

5

10

15

0 10 20 30 40 50 60Pull-out displacement (mm)

Pp

(kN

no sleeve, lubricatedsleeve 15 cm long

sleeve 30 cm long

Fig. 10. Inuence of the presence of a sleeve at the box frontal face numericalanalyses. (a) Pull-out test with frontal sleeve. (b) Inuence of the presence and lengthof the frontal sleeve.the back-analysis of the results of pull-out tests (Wilson-Fahmyet al., 1994; Yogarajah and Yeo, 1994; Perkins, 2001; Dias, 2003).Fig. 14(a)(c) shows some comparisons between measurementsand predictions by the nite elements method of pull-out tests ona geotextile and on a geogrid (Dias, 2003). The tests modelled werecarried out on 1 m3 dense sand samples. The woven geotextile wasmade of polyester with a tensile stiffness of 4000 kN/m. The geo-grid was made of high density polyethylene, with average tensilestiffness of 550 kN/m. The soil was modelled as an elasticplasticmaterial with MohrCoulomb failure criterion. The reinforcementswere assumed as linear elastic materials. The geometry of thegeogrid transverse members was considered in the grid modellingand its longitudinal members were modelled as equivalent planar

30

20

10

00

P

L

1

P

L

d

2

Pp

Pp

Pp

(kN/m)

Fig. 11. Results of tests with the reinforcement awaBecause the large pull-out box used in the tests described abovehad total pressure cells along its frontal face, it was also possible tocompare predicted and observed normal stresses along thatboundary. The rigid pull-out box internal frontal face was lubri-cated with double layers of plastic lms and grease, yielding to aninterface friction angle of 6. Fig.15 shows results of predictions andmeasurements for a test on an HDPE geogrid in terms of theincrements of normal stress (Dsx) along the box internal frontalface (y is the elevation along the box face and h is half of the boxheight) normalised by the average shear stress (sb) along thereinforcement length (Palmeira and Dias, 2007). Again, a reason-able comparison between predicted and observed normalisednormal stresses can be observed.

Finite Element Method analyses usually assume a geogrid asa continuous equivalent rough planar reinforcement. However, thepull-out response of this type of reinforcement is fundamentallydependent on shape and geometrical characteristics of the grid,with particular reference to the transverse bearing members.Fig. 16(a) and (b) show results of pull-out tests performed byTeixeira (2003) on a commercially available geogrid, made ofpolyester, with apertures 23 mm 23 mm, buried in dense gran-ular material. Firstly, tests were performed on the geogrid with itsoriginal geometrical characteristics. Then, the transverse memberswere removed and only the longitudinal members were tested(Fig. 16a). It can be observed in Fig. 16(b) that the transversemembers were responsible for a signicant fraction of themaximum pull-out load of the geogrid, but friction along longitu-dinal members was equally important for that matter. Longitudinalmembers alone show a strain softening behaviour after peak, while

10 20 30 40

GeogridL = 500 mmd = 300 mmDense sand

1

2

Pull-out displacement (mm)y from the box frontal face (Palmeira, 1987b).

the entire grid shows a rather constant maximum pull-out loadwith increasing pull-out displacements (Fig. 16b).

Other aspects that inuence the grid response under pull-outnot considered in conventional FEM analyses are the shape andbending stiffness of the transverse members. Fig. 17 showsa photograph of the bended shape of a grid transverse member atthe end of a large scale pull-out test (Santos, 2007). For transversemembers with low bending stiffness a progressive mobilisation ofbearing resistance will take place as the member bends under

However, the relative size of the soil particle with respect to thegrid transverse member thickness will inuence the bearing stress

1

2

3

4

Strain

(%

)

sand

geogrid

680 mm

600

mm

air pressure bagair pressure bag

air pressurebag

piston

Pp

Fig. 12. Results of tests obtained in a pull-out box wit

15

20

25a

(kN

/m

) Pp H

1 m

H = 0.3m

E.M. Palmeira / Geotextiles and Geom0

5

10

0 2 4 6 8 10 12 14Pull-out displacement (mm)

Pp

Pp H

1 m

Prigid plate = 25b

0

5

10

15

20

25

0 2 4 6 8 10 12 14Pull-out displacement (mm)

Pp

(kN

/m

)

H = 1 m

Pp H

1 m

Pp H

1 m

Prigid plate = 25

Fig. 13. Inuence of the box top boundary rigidity on the test results numericalanalyses (Dias, 2003). (a) H 0.3 m; (b) H 1 m.ratio. The data points in Fig. 18 are experimental results from pull-out tests on isolated transverse members and the scatter observedis mainly due to the effects of the relative sizes between gridtransverse members and soil particles (Palmeira and Milligan,increasing pull-out loads. Brown et al. (2007) also discuss theimportance of transverse members bending stiffness on theperformance of geogrid reinforced ballast. This inuence has still tobe properly studied and quantied.

3.3. Soilgeogrid interaction under pull-out condition

As shown in the previous section, interaction between soil andgeogrid is quite complex. If the soil is assumed as a continuum,geogrid pull-out can be represented by the simultaneous pull-outof a series of plate like elements buried in the soil. The assumptionof the soil particles xed in between transverse members while thegrid is being pulled is certainly unrealistic. Jewell et al. (1984)proposed upper and lower bound theoretical solutions for thebearing stress at the transverse member normalised by the verticalstress as a function of the soil friction angle. These upper and lowerbound solutions are shown in Fig. 18 and were derived from anal-yses of plates buried in a homogeneous continuum. Moraci andGioffre` (2006) found good agreement between predictions bytheoretical solutions such as those proposed by Jewell et al. andresults from large scale pull-out tests on grids with large spacingbetween transverse members and little scale effects.

Distance from front face (mm)

0 100 200 300 400 5000

0

0

0

0

Fixed face

Movable face

= 49 kPaPp = 15.4 kN/m

h a movable frontal face (Sugimoto et al., 2001).

embranes 27 (2009) 368390 3751989a). Thus, solutions based on continuum cannot be directlyextrapolated to a particulate medium where the grid transversemember and the soil particles have similar dimensions (note thelogarithmic scale of the graph in Fig. 18).

The inuence of the relative sizes of soil particle and gridtransverse members can be clearly visualised in Fig. 19, where thenormalised bearing stress is plotted against the transverse memberthickness normalised by the average soil particle diameter. Theresults presented in this gure are those from tests on isolatedsmooth transverse members with circular, square, hexagonal andrectangular cross-sections buried in dense sands or crushed glass.The result of a test on a member of a polymeric grid (hexagon likecross-section) is also shown in this gure. Fig. 19 shows thatbearing stress depends on the transverse member shape with atfrontal face (square or rectangular) members providing slightlygreater bearing strengths. The results also show that the normal-ised bearing strength starts to be independent on the soil particlesize only for ratios betweenmember thickness and average particlediameter above 12.

05

10

15

20

25a

0 10 20 30 40Pull-out displacement (mm)

Pp

(kN

/m

)

Measured Predicted

0

20

40

60

0 10 20 30 40 50Pull-out displacement (mm)

Pp

(kN

/m

)

Measured Predicted

0

2

4

6

8

10

12

0 0.1 0.2 0.3 0.4 0.5 0.6Geogrid length (m)

Ten

sile strain

(%

)

MeasuredPredicted

PP

b

c

Fig. 14. Predicted and observed pull-out test results. (a) Loaddisplacement curves fora woven geotextile. (b) Loaddisplacement curves for an HDPE geogrid. (c) Tensilestrain distribution along the geogrid length.

-1

-0.5

0

0.5

1

3 2 1

PredictionsMeasurements

x/b

y/h

Fig. 15. Predicted and observed normal stresses along the box frontal wall (Palmeiraand Dias, 2007).

E.M. Palmeira / Geotextiles and Geom376P

P

L= 600mm

1

~ 26 members

Geogrid without the transverse members

231.5 3mma

embranes 27 (2009) 368390Another particular aspect of the interaction between soil andgeogrid is the possibility of interference between grid transversemembers. The discrete interaction between grid transversemembers and the surrounding soil is clearly visible in the results ofphoto-elastic studies conducted by Dyer (1985), some of themreproduced in Fig. 20(a)(c). These results show that the loaddistribution between transverse members of a steel grid is uniformonly if the members are sufciently apart from one another(Fig. 20a). As the distance between them is reduced (Fig. 20b)signicant non-uniformities in the distribution of bearing loadsamong transverse members can occur. This non-uniform loaddistribution is a result of interference between grid transversemembers. For a longer grid the interaction with soil is morecomplex but the photo-elastic results still show that the loaddistribution among transverse members is not uniform, as depictedin Fig. 20(c) (Dyer, 1985). In this gure the contour of the darkregion behind each transverse member has been highlighted by

2

P

P

L= 600mm

40b

60

20

0100 20 30 40 50

Pu

ll-o

ut fo

rce (kN

/m

)

Pull-out displacement (mm)

2

1

Fig. 16. Inuence of the grid bearing members on the pull-out response of a geogrid(Teixeira, 2003). (a) Types of tests carried out. (b) Test results.

a continuous line. These dark regions are regions of low stress levels

denition of the degree of interference (DI) and results of degree ofinterference against the ratio between transverse member spacing(S) and member diameter (B) obtained from tests on steel gridswith round transverse members buried in dense sand. The resultssuggest that negligible or no interference between transversemembers will occur only for S/B ratios above 40. In this case, eachmember of the grid would behave as it was isolated. The gure alsoshows predictions of DI values for round and square transversemembers by the Finite Element Method (Dias, 2003). It is inter-esting to note that the same trend of reduction of DI with the

Fig. 17. Bended grid transverse member at the end of a pull-out test (Santos, 2007).

(Palmeira 1987)

Silty sand-member of a PET geogrid (Teixeira 2003-v = 25, 50 and 100 kPa)Sand B-member of a HDPE geogrid (Palmeira 2005)

Crushed glass-member of a HDPE geogrid (Milliganet al. 1990)

Fig. 19. Results of pull-out tests on isolated transverse members with different cross-sections (modied from Palmeira and Milligan, 1989a).

E.M. Palmeira / Geotextiles and Geomcaused by the movement of the transverse members and it can beseen that they vary in size along the grid length. The movement ofa rigid grid in the soil leaves loose soil regions at the rear of eachtransverse member. Ahead of the member the soil is subjected toa passive state of stress whereas behind it an active state of stress isreached, creating those loose soil regions. This state of stress isrepresented by the dark regions in the photo-elastic results shownin Fig. 20(c). The size of this region and its proximity to the nextmember will inuence the pull-out load distribution amongtransverse members. Besides, pull-out tests on isolated bearingmembers in dense sands with large ratios between memberthickness and sand particle diameter have shown that the exten-sion of the soil mass ahead of the member affected by the failuremechanism can be as much as 6 times the transverse memberthickness (Palmeira and Milligan, 1989a). In situations where thesoil particles and the transverse members have similar dimensions,contact force chains can cause loading spreading over a large regionand be quite long (Fig. 20a and McDowell et al., 2006), yielding tomuch more complex interference conditions.

For a grid with large apertures (small solid fraction per unitarea), the larger the spacing between transverse members the lessthe inuence of interference between them. Under these condi-tions, interference can be quantied by the Degree of Interference(Palmeira, 1987b; Palmeira and Milligan, 1989a), which comparesthe actual pull-out strength of a geogrid with the pull-out strengthif there was no interference between transverse members, or as ifthe spacing between these members was very large (assumingnegligible skin friction between soil and grid). Fig. 21 shows the1000

100

10

120 30 40 50 60

b

b

v

vUpper bound

Lower bound

Grid transversemember

Soil friction angle ()

Fig. 18. Bearing stresses on isolated grid transverse members (Palmeira and Milligan,1989a).B/D50

00

50

100

10 20 30

v = 25 kPa

F F

FB

D50

b/(

v tan

)

b = F/B

Sand B - member of a metal grid - round sectionSand B - member of a metal grid - square section

Sand C - member of a metal grid - round section

Sand A - member of a metal grid - round section

= soil friction angle

v v

embranes 27 (2009) 368390 377increase of S/B is predicted by the Finite Element Method. It shouldbe pointed out that the results presented in Fig. 21 were obtainedfor steel grids.

For extensible grids the evaluation of the degree of interferenceis even more complex due to the non-uniform distribution oftensile strains along the grid length. Milligan et al. (1990) foundmore noticeable non-uniform distributions of bearing loads ina polymer grid in a photo-elastic study similar to that carried out byDyer (1985). This non-uniformity was certainly enhanced by thelower tensile stiffness of the polymer grid and the higher stresslevel used in the tests in comparison with the high tensile stiffnessof metal grids. Nevertheless, nowadays increasingly stiffer poly-meric grids are being produced and the results of Fig. 21 may beextrapolated to some extent to these materials.

A consequence of the inuence of grid geometrical character-istics, interference between transverse members and the assump-tion of the grid as a planar continuous element is that the apparentsoilgeogrid friction coefcient becomes dependent on the length

Fig. 20. Interference between grid transverse members (Dyer, 1985). (a) Large spacingbetween transverse members. (b) Short spacing between bearing members. (c) Test ona longer grid.

E.M. Palmeira / Geotextiles and Geom378of the geogrid (Palmeira, 1987b; Bergado et al., 1993; Moraci andGioffre`, 2006), which complicates the estimate of soilgeogridbond for design purposes. The results by Moraci and Gioffre` (2006)presented in Fig. 22 show the variation of the apparent frictioncoefcient of a geogrid as a function of the grid length and of thenormal stress. It can be seen that the shorter reinforcement was theone presenting the largest apparent friction coefcient. It is alsoshown that the lower the stress level the higher the friction coef-cient, in part due to a greater dilative response of the dense sandunder low stresses. These results suggest that in a real reinforcedsoil wall, for instance, different apparent friction coefcient valuesshould be considered in shallow and deep anchorage lengths.

3.4. Distribution of bearing forces among grid transverse members

It would be interesting to evaluate the non-uniform distribution

S

BP

Po

DI = 1 - P

nPo

Rigid steel grids

Degree of Interference:

S/B

n = number of bearingmembers

1.0

0.5

00 20 40

38

139

4

77

DI

n = 2 n = 3 n > 3FEM, B = 10mm, n = 3, squaremembersFEM, B = 10mm, n = 3, roundmembers

Experimental:

Fig. 21. Degree of interference versus the ratio between transverse member spacingand member thickness.

embranes 27 (2009) 368390of bearing loads among grid transverse members. However, theresults presented so far have shown that the particulate nature ofthe interaction between granular soils and typical grids makesmodelling of soilgrid interaction very complex. An additionaldifculty is that skin friction between soil and grid (Fig. 16) can bea signicant part of the bond strength between these materials andthe variety of shapes and geometrical characteristics of availablegrids adds to the complexity of the problem. The use of niteelement analysis may provide a good overall understanding of theresponse of a geogrid if the grid is substituted by an equivalentrough sheet, but limitations in the simulations do exist, like thevalue of interface shear strength and stiffness to be used in theanalyses, for instance. Finite element analyses have shown thatsoil-inclusion shear stiffness varies with the stress level (Perkins,2001; Dias, 2003). The use of the discrete element method (Cundalland Strack,1979; McDowell et al., 2006) is certainly very promising,but simulations are very time consuming and modelling of realisticsituations requires more powerful computers than those currentlyavailable. To assess the distribution of loads among grid transversemembers a simple alternative approach would be to model thegeogrid as a discrete material consisting of a succession of bearingelements (Palmeira, 1984).

The model shown in Fig. 23 can be used for the investigation ofload distribution along grid transverse members and, in particular,for the back-analyses of results of pull-out tests on geogrids

curves. A good agreement is also observed for the tensile straindistribution along the grid length at maximum pull-out force, as

Geom(Palmeira, 2004). Themodel takes into account a friction law for theinteraction between soil and longitudinal grid members. It alsoconsiders the bearing forcemember displacement relationshipand, when relevant, the loadtime relationship for the geogrid canbe accounted for. Based on these assumptions, equilibrium equa-tions can be derived and an iterative procedure used to obtain thepull-out response of the grid and the load distribution among gridtransverse members (Palmeira, 2004). By trial and error differentbearing loaddisplacement relations can be used to assess non-uniform load distribution among transverse members of geogrids.In such trials the result obtained from a pull-out test on an isolatedmember can be used as reference. In the actual grid the loaddisplacement curve obtained for the isolated member would bevalid only for the rst transverse member, even so if it was distantenough from the pull-out box front wall or sleeve. The loaddisplacement curves for the other grid members can be establishedbased on arbitrary reduction functions applied to the curveobtained for the isolated member (Palmeira, 2004), depending onthe transverse member considered. The choice of the function isarbitrary and the best choice will be the one for which the

0.4

0.8

1.2

1.6

L = 0.4mL = 0.9mL = 1.15m

2.0

0 20 40 60 80 100Normal stress (kPa)

Ap

paren

t co

efficien

t o

f frictio

n,

Fig. 22. Grid apparent friction coefcient versus normal stress (modied from Moraciand Gioffre`, 2006).

E.M. Palmeira / Geotextiles andpredictions by the model t the best the loaddisplacement curveand the tensile strain distribution along the grid length observed inthe pull-out test on the actual grid.

The model described above is useful for the back-analysis of theresults of pull-out tests on geogrids, particularly those with simplergeometries. However, it is certainly not practical for the predictionof geogrid pull-out responses in general, because the evaluation ofthe loaddisplacement relationship for isolated transversemembers is not straightforward. Besides, some mechanism ofinterference between transverse members has to be assumed,unless the effects of interference can be neglected.

The model in Fig. 23 and the results of some large scale pull-outtests (Palmeira, 1987b) can be used in the back-analysis of the non-uniform distribution of loads in grid transverse members. Thetesting equipment consisted of a large pull-out box capable ofaccommodating 1 m3 soil samples. A pressurised bag provided thenormal stress at the top of the sample. The internal frontal face of thebox was lubricated with double layers of plastic lms and oil. Thefrontal face had total pressure cells to measure normal stressestransmitted to this boundaryduring the tests anddenseanduniformLeighton Buzzard sand was used in the experiments. Woven geo-textiles, metal and polymeric grids were tested. During the tests,tensile strains were measured along the length of polymeric grids.The input data for the theoretical model described before isshown in Fig. 24(a)(c). The dimensions and geometrical charac-teristics of the polymeric grid tested, made of high density poly-ethylene, are shown in Fig. 24(a). Fig. 24(b) presents the result ofa pull-out test on an individual isolated transverse member of thepolymeric geogrid. The member has an approximately trapezoidalcross-section shape. The loadstraintime relationship for thegeogrid is presented in Fig. 24(c), where the results from tensiletests under different strain rates are shown (McGown, 1982).

Based on those input parameters, Fig. 25(a)(c) presentscomparisons between the results of a large pull-out test ona polymer grid under 25 kPa normal stress and a test speed of0.5 mm/min and the best t prediction based on an arbitrarilydistribution of forces on each transverse member (Palmeira, 2004).Fig. 25(a) shows a reasonable agreement between predicted andobserved geogrid pull-out force versus pull-out displacement

i = 1 i = n

B

a

ls

Sdo

lo

1 2

longitudinal segment no.

bearing member

FB

lo

FB

T

Skin friction Bearing Load-strain-timebehaviour

FB

Grid transversemember

Pp

Pp

o

s

1 2 n

1

2

j

max

Fig. 23. Discrete approach for geogrid modelling.

embranes 27 (2009) 368390 379shown in Fig. 25(b). Fig. 25(c) presents the distribution of loadsamong the transverse members for the best t hypothesis anda non-uniform load distribution can be noticed. This non-uniformload distribution is caused by interference between transversemembers and by different mobilisations of tensile strains along thegrid length.

If the same exercise is repeated for tests under higher stresslevels, results such as the ones presented in Fig. 26(a) and (b) areobtained at maximum pull-out force. The non-uniformity of thebearing force distribution increases and the maximum bearingforces may not necessarily act on the rst transverse member buton subsequent members. These results show the level ofcomplexity of the mechanism of interaction between soil andgeogrid in a pull-out test box and are consistent with the resultsfrom photo-elastic studies.

3.5. Summary on pull-out tests

At this stage, a summary of the conclusions on pull-out testsshould point out that:

The pull-out test is useful for determining anchorage strengthof geogrids.

PGeom70mm

200mm

lo =150mm

Pp

0123456789

10

Po (k

N/m

)

b

a

E.M. Palmeira / Geotextiles and380 Boundary conditions may inuence test results. Regardingminimising the inuence of the boxwall frontal face,a sleeve or a lubricated internal face can be used, but numericaland experimental results suggest some differences on pull-outresponses and maximum pull-out loads depending on thesolution adopted. Some authors have reported close values ofmaximum pull-out loads in tests with well lubricated boxfrontal faces, sleeve or with the reinforcement away from thefrontal face. However, the use of a sleeve seems to providegreater condence on the less inuence of the box frontal wallon the test results. The appropriate length of the sleeve is likelyto be dependent on the roughness (which should be reducedanyway) of the box wall, height of the sample and length andtype of reinforcement. The results available suggest that for thetypical dimensions of large scale pull-out devices the sleevelength should not be smaller than 30 cm.

Large scale tests should be preferred, particularly becauseincreasing the scale of the test reduces the detrimentaleffects of the boundaries. Experimental and numericalresults suggest that for typical reinforcement lengths (

Geom10

15

20

25

30

35

ll-o

ut fo

rce (kN

/m

)

v = 25 kPaa

E.M. Palmeira / Geotextiles and1987; Siel et al., 1987; Wu and Arabian, 1988; Kokkalis and Papa-charisis, 1989; Ling et al., 1992; Gomes, 1993; Azambuja, 1994;Tupa, 1994; Boyle et al., 1996; Palmeira, 1996; Palmeira et al., 1996;Helwany and Shih, 1998; Mendes, 2006; Mendes et al., 2007).Connement increases interlocking and friction among geotextilebres, yielding to a stiffer response of the geotextile. If in signicantamounts, the soil particles that intrude the bre matrix restrainbre stretching, also contributing to the increase of geotextiletensile stiffness. In some of the testing apparatus reported in theliterature, the geotextile ends are stiffened by impregnation withepoxy resin. So, only the central part of the specimen sufferssignicant deformation. It is important to point out that in this type

0 0.01 0.02 0.03 0.04 0.05Pull-out displacement (m)

0

5

Pu

Grid PG - Test resultsVariable bearing force dependingon the member considered

1 2 3 4Longitudinal segment number

0

1

2

3

4

5

6b

Ten

sile strain

(%

)

Best fit assumptionTest results

1 2 3 4 5Bearing member no.

0

2

4

6

8

10

12

Bearin

g fo

rce (kN

/m

) Bearing force distribution for best fit

c

Fig. 25. Predicted and observed pull-out test results using the discrete approach(Palmeira, 2004). (a) Pull-out force versus pull-out displacement. (b) Comparisonbetween strain distributions along the grid length. (c) Distribution of forces amonggrid transverse members.v = 50 kPa

1 2 3 4 5Bearing member number

0

5

10

15a

Bearin

g fo

rce (kN

/m

)

Bearing force distribution for best fit

0

10

20

30

Bearin

g fo

rce (kN

/m

)

v = 75 kPab

embranes 27 (2009) 368390 381of test arrangement friction between geotextile and conning soilalso takes place and this inuences the magnitude of tensile loadand tensile stiffness measured.

Fig. 28 shows results of in-soil tensile tests in terms of secanttensile stiffness versus tensile strain for tests with different soiltypes and a nonwoven needle-punched geotextile using a testarrangement similar to that shown in Fig. 27 (Palmeira et al., 1996).Soils varied from silts to coarse sand. The result for a test witha lubricated rubber membrane conning the geotextile is alsopresented and this test provided a lower bound for the test results.In the test with the membrane there is only connement of thegeotextile and no impregnation by soil particles. Besides, thelubrication of the membrane reduces frictionwith the geotextile, sothe inuence of friction on the test results is minimised. Indepen-dent on the conning material used the tensile stiffness is consid-erably increased in comparison to the result of the test in isolation.

In-soil tensile tests with different boundary conditions can befound in the literature. Different boundaries or test conditions willproduce different test results. Palmeira et al. (1996) and Araujo andPalmeira (2008) carried out numerical analysis (FEM) and observedthat soil arching is likely to occur in test arrangements such as thatshown in Fig. 27 due to the presence of the stiffer ends of thegeotextile specimen and to the compressible central region ofthe geotextile layer that is effectively stretched. Fig. 29 shows thedistributions of normal stresses on the geotextile specimen for anin-soil tensile test on a nonwoven geotextile conned by densesand (Araujo and Palmeira, 2008). A very non-uniform stressdistribution can be observed and the non-uniformity intensity willdepend on the compressibility of the geotextile central part and ofthe stiffened ends of the specimen. This effect is also likely to

1 2 3 4 5Bearing member number

Bearing force distribution for best fit

Fig. 26. Non-uniform distributions of bearing loads among geogrid transversemembers (Palmeira, 2004). (a) Normal stress of 50 kPa. (b) Normal stress of 75 kPa.

inuence test results to an unknown extent. The values of tensilestiffness obtained under low strain levels (say, less than 1%) in in-soil tensile tests are also likely to be somehow inaccurate because ofthe inuence of non-uniform movements of clamps at the begin-ning of the test and due to accommodation or closure of small gaps

4.2. In-soil tensile tests to assess the inuence of geotextileimpregnation by soil particles and geotextile damage

The presence of soil particles inside the geotextile can increaseits stiffness (Tupa, 1994), as the presence of these particles furtherreduces the space available for bre stretching, as schematicallyshown in Fig. 31 (Mendes et al., 2007). This gure also shows resultsof tensile tests under connement on a needle-punched polyesternonwoven geotextile (400 g/m2) for different levels of impregna-tion of the geotextile by soil particles and a conning stress of100 kPa. Impregnation of the geotextile was achieved by spreadingthe soil on the geotextile surface and by vibrating the specimen tohelp particle intrusion. The level of impregnation (l) was quantiedas the ratio between the mass of soil particles per unit area and themass of geotextile bres per unit area. In these tests, connement ofthe geotextile was provided by wooden plates with the face incontact with the geotextile specimen lubricated by layers of plasticlms and grease. Fig. 31 shows that the greater the mass of soilinside the geotextile the greater its tensile stiffness. However,a stiffer response of the geotextile due to impregnationwill dependon the feasibility of soil particles intrusion in the bre matrix andon the amount of geotextile void space occupied by these particles.Mendes et al. (2007) have observed that the shape of the intrudedsoil particles also inuenced the tensile stiffness increase, with

Ten

sile lo

ad

(kN

/m

)

Tensile strain (%)

Nonwoven geotextile

12

8

4

00 20 40

In isolation

In-soilgeotextile

geotextilestiffened end

soil

Fig. 27. Effects of connement on geotextile stiffness (McGown et al., 1982).

E.M. Palmeira / Geotextiles and Geomembranes 27 (2009) 368390382between components of the apparatus.An improved version of the apparatus with respect to arching

effects is presented in Fig. 30(a) (Boyle et al., 1996; Palmeira, 1996),where the entire geotextile length is stretched. In this case, the sidewalls also work as clamps. In this arrangement, both soil archingand friction between soil and geotextile are avoided. Fig. 30(b)shows that the secant tensile stiffness obtained for this arrange-ment is still signicantly greater than that obtained with the geo-textile in isolation, but smaller than the value obtained whenfriction between soil and geotextile specimen is allowed (Mendeset al., 2007). Fig. 30(c) (Araujo and Palmeira, 2008) shows that thetensile load variation along the geotextile specimen length (tensilestrain of 1.1%, dense sand conning the geotextile and 150 kPanormal stress) for this type of test arrangement is quite constant, incontrast to what is observed in the traditional apparatus with thegeotextile ends stiffened, where the tensile load in the geotextilevaries due to friction between soil and geotextile.

300

350soil Asoil B

soil Dsoil E0 2 4 6 8 10 120

50

100

150

200

250

Tensile strain (%)

Secan

t ten

sile stiffn

ess (kN

/m

) soil C

In-isolation

lubricatedmembrane

= 50 kPa

Fig. 28. Inuence of connement on the loadstrain behaviour of a nonwoven geo-textile (Palmeira et al., 1996).round soil particles being less effective than angular ones for thatmatter.

The in-soil tensile test can also be useful to evaluate theinuence of mechanical damages on the tensile properties ofgeotextiles. Fig. 32(a) shows photographs of a horizontal cut anda Y shaped cut in geotextile specimens at the beginning of anin-air wide strip tensile test and for a tensile strain of 30%. It canbe seen that the original horizontal cut evolves to an oval shapeand the Y shaped cut evolves to a heart like shape. Fig. 32(b)shows results of in-soil tensile tests on needle-punchednonwoven geotextile specimens (200 g/m2) with horizontal cutsof varying lengths. The soil used to conne the geotextile spec-imen was a uniform coarse sand and the conning stress usedwas equal to 100 kPa. Fig. 32(b) shows that the results obtainedfor geotextile specimens with horizontal cuts with varying

0 50 100 150 200Distance along the specimen length (mm)

0

50

100

150

200

250

25 kPa on top50 kPa on top

100 kPa on top 150 kPa on top

n

No

rm

al stress,

n (kP

a)

Stiff endFig. 29. Non-uniform normal stress distributions on the geotextile specimen due tosoil arching (Araujo and Palmeira, 2008).

Geom300

b

(kN

/m

)

= 100 kPageotextilea

E.M. Palmeira / Geotextiles andlengths are not much different from the result obtained for theintact and virgin geotextile specimen for tensile strains above 1%.Thus, connement can reduce the detrimental effects of damageson the behaviour of nonwoven geotextiles, which may be rele-vant for applications of such materials as reinforcement, sepa-rators and, to some extent, lters.

4.3. Summary on in-soil tensile tests

Regarding in-soil tensile tests, in summary:

The test is indicated for nonwoven geotextiles and can beuseful for the investigation of the effects of damages on theirmechanical behaviour.

00

100

200

5 10 15Tensile strain (%)

Sec

an

t ten

sile stiffn

ess

with frictionin air, = 0

without friction

0 20 40 60 80 100Distance along the specimen length (mm)

0123456789

Ten

sile lo

ad

(kN

/m

)

With friction Frictionless

Tensile strain: 1.1%Confining material: dense sand

c

Fig. 30. Results of in-soil tensile tests with and without friction between soil andgeotextile (Mendes et al., 2007; Araujo and Palmeira, 2008). (a) Frictionless in-soiltensile test. (b) Comparison between test arrangements. (c) Variation of tensile loadalong the geotextile specimen length (Araujo and Palmeira, 2008). The boundaries can also affect the results and to isolate theeffect of friction between soil and geotextile (lower bound testresult) the original test arrangement proposed by McGownet al. (1982) with lubricated rubber membranes conning thegeotextile or the arrangement where both soil and geotextilecan deform laterally should be preferred.

5. Ramp test

5.1. Inuence of sample sizes and boundary conditions

Ramp tests, or inclined plane tests, are relevant for the study ofthe stability of cover systems of waste disposal areas or for erosioncontrol in slopes. In slopes of waste disposal areas, besides thecover soil, more than one type of geosynthetic can be present, suchas geomembranes, geonets and geotextiles, to full different roles,as shown in Fig. 33(a). If the cover soil stability is not properlyaddressed, failure can occur. Failures of cover soils on geosyntheticlayers in slopes of waste disposal areas, reservoirs and ponds havebeen reported in the literature (Dwyer et al., 2002; Gross et al.,2002; Blight, 2007; Stark et al., 2008;Wu et al., 2008a, for instance),and two examples are shown in Fig. 33(b). The repair of such eventsis time consuming and expensive because usually a large fraction ofthe lining system must be repaired or substituted.

The ramp test can be a useful tool to study and quantify theinteraction between soils and geosynthetics in problems such asthose shown in Fig. 33(a) and (b), being one of its advantages thepossibility of simulating the actual low stress levels on the inter-faces caused by the relatively thin cover soil layer. The use ofconventional direct shear apparatus for tests under low stress levelscan result inmisleading interface strength parameters (Girard et al.,1990; Giroud et al., 1990; Gourc et al., 1996). The ramp test basicallyconsists of increasing the inclination of the ramp until a soil blockslides on the geosynthetic layer xed to the ramp surface, as shownin Fig. 34. Similar test arrangements can be employed for tests oninterfaces between different geosynthetics, such as geotextiles andgeomembranes, for instance (Palmeira et al., 2002). Variationsbetween equipment and testing conditions can be found in theliterature (Girard et al., 1990; Giroud et al., 1990; Koutsourais et al.,1991; Girard et al., 1994; Gourc et al., 1996; Izgin and Wasti, 1998;Lalarakotoson et al., 1999; Lopes et al., 2001; Wasti and Ozduzgun,2001; Palmeira et al., 2002; Palmeira and Viana, 2003; Wu et al.,2008b). Briaon et al. (2002) describe a ramp test apparatus capableof performing tests under submerged conditions. The resultsobtained with such apparatus were validated by the authors incomparison with results from large eld experiments.

If one examines the mechanics of the test (Fig. 34), assuminglimit equilibrium conditions and a trapezoidal distribution ofnormal stresses along the soilgeosynthetic interface, the followingequations for maximum and minimum normal stresses can bederived (Palmeira et al., 2002)

smaxs

4 6xL

(2)

and

smins

6xL 2 (3)

with

x cosa tan1h=L"

1h2#1=2

(4)

embranes 27 (2009) 368390 383L 2 cos a L

10

20

30

40

50

60

70

80

Secan

t ten

sile stiffn

ess

(kN

/m

)

T

Fig. 31. Inuence of the impregnation of a nonwoven geotextile

0 1 2 3 4 5 6Tensile strain (%)

0

200

400

600

800b

Secan

t stiffn

ess (kN

/m

)

Virgind = 10mm

d = 25mm d = 50mm

T

T

d

Not to scale

200mm

100m

m

a

Fig. 32. Inuence of connement on the tensile stiffness of mechanically damagednonwoven geotextiles (Mendes et al., 2007). (a) In-air wide strip tests on mechanicallydamaged geotextiles. (b) Inuence of connement.

E.M. Palmeira / Geotextiles and Geom384and

s W cos aL

(5)

where smax and smin are the maximum and minimum normalstresses on the interface at the edges of the soil block (Fig. 34), s isthe average normal stress on the interface, x is the distancebetween the lower soil block edge and the point of application ofthe normal force on the interface, a is the inclination of the ramp tothe horizontal, h is the soil block height, L is the soil block lengthand W is the weight of the soil block.

Equations 25 suggest that the level of non-uniformity of thedistribution of normal stress along the soilgeosynthetic interfaceis a function of the dimensions of the soil block. The lower the ratiobetween soil sample height (h) and length (L), the more uniform isthe distribution of normal stresses on the interface. This favours theuse of long boxes with thin soil layers for testing.

The approach presented above can be considered as a simpleapproximation to the actual stress conditions in the ramp test. Tocheck the assumption of trapezoidal normal stress distributionalong the interface, numerical analyses of ramp tests were per-formed using the computer code PLAXIS (Brinkgreve and Vermeer,1998). The soil properties relevant to this analysis were: soil unitweight equal to 17 kN/m3, Youngs modulus of 20 MPa, Poissonratio equal to 0.3 and friction angle equal to 35. MohrCoulombfailure criterion was used for the soil and the interfaces. Threelengths (0.5 m, 2 m and 10 m) of the soil block and different rampinclinations were investigated.

The variations of normal stresses on the interface with its nor-malised length obtained by the numerical analysis, for a ramp

0 1 2 3 4 5Strain (%)

0

0

0

0

0

0

0

0

Virgin = 4 = 8

by soil particles on its tensile stiffness (Mendes et al., 2007).

embranes 27 (2009) 368390inclination of 25 , are presented in Fig. 35 (Palmeira et al., 2002).The difference between maximum and minimum normal stressvalues increases considerably as the length of the block is reduced.For the 0.5 m long box the maximum normal stress was as much as5 times the minimum normal stress. A non-uniform normal stressdistribution on the interface associated with the low stress level isa complicating factor for the analysis that should be avoided orminimised. The predictions of maximum and minimum normalstresses by equations 2 and 3 are also presented in Fig. 35, whichshows that the approach of a trapezoidal normal stress distributionon the interface was reasonably accurate for the conditionsanalysed.

Fig. 35 also shows the effect of using a box with inclined sides, assuggested by Gourc et al. (1996), so as to these sides to coincide tothe vertical direction at a ramp inclination of 25, for the 0.5 m longbox. The results obtained by the numerical analyses show thata uniform stress distribution is obtained with this type of box, asexpected, in contrast with the results obtainedwhen the traditional

E.M. Palmeira / Geotextiles and Geombox shape is used. These results indicate that for shorter boxes theconguration with inclined sides should be employed, althoughsample preparation may be a bit more complicated in this box. Theresults of the analytical and numerical analyses provide usefulinformation for the design of this type of equipment.

The numerical analysis also allows the investigation of theevolution of force mobilisation in a geosynthetic layer and interfaceshear strength as the ramp inclination increases. Fig. 36(a) and (b)shows results fromthenumerical simulationof a ramptest (2 m longand 0.23 m high soil sample) on a sandgeotextile interface (Pal-meira et al., 2002). For this analysis, the geotextile tensile stiffness

Fig. 33. Geosynthetic layers in slopes of waste disposal areas. (a) Geosynthetics layers inrespectively).

max

W

Wsin

Soil

Geosynthetic

Box

L

H

RampNx

min

Fig. 34. Equilibrium conditions in the ramp test (Palmeira et al., 2002).embranes 27 (2009) 368390 385was assumed equal to 70 kN/m and the interface friction anglebetween soil andgeotextilewas assumedequal to 31. Thegeotextilehad its top extremity anchored to the ramp. The friction anglebetween the geotextile and the ramp facewas equal to 26. The sandlayer had a Youngs modulus equal to 20 MPa, Poisson ratio of 0.3,friction angle of 35 and density equal to 17 kN/m3. MohrCoulombfailure criterionwas used for the soil and values of shear stiffness forthe soilgeotextile and geotextileramp interfaces were assumedequal to 3000 kN/m3 and 25,000 kN/m3, respectively.

Fig. 36(a) shows a progressive mobilisation of tensile force onthe geosynthetic as the ramp inclination increases. The entiregeosynthetic length is tensioned only for ramp inclinations veryclose to the failure value (31). Mobilised interface friction anglesversus normalised interface length are presented in Fig. 36(b) andthe progressive nature of the failure mechanism can also be clearlyidentied. A similar type of mechanism is likely to occur under eldconditions, particularly for long slopes.

5.2. Ramp tests to assess the behaviour of reinforced cover soils

Regarding eld conditions, a possible solution to reduce tensileforces in a geomembrane in a slope of a waste disposal area is theuse of reinforcement in the cover soil, as schematically shown inFig. 37(a). From the practical point of view, the easiest arrangementis to install the reinforcement directly on the geomembrane.However, this is not the most efcient position for the reinforce-ment. A large ramp test can be employed to investigate the

waste disposal areas. (b) Failures of cover soil (Dwyer et al., 2002; Gross et al., 2002,

The ramp test results also suggest that the tensile forcemobilisedin the geomembrane can be signicantly reduced because of the

0.8

0.4

1.2

1.6

0 10

ramp inclination (degrees)30

27.5

25

15

Ten

sile fo

rce (kN

/m

)

Normalised interface length

0.80.60.40.2

Normalised interface length

face

frictio

n

gree

s)

10

25

30

b35

0 0.2 0.4 0.6 0.8 1

2527.5

30

a

Fig. 36. Progressive mobilisations of tensile forces and interface strength in a ramp test(Palmeira et al., 2002). (a) Tensile force distribution along the geotextile length. (b)Mobilised interface strength along the geotextile length.

Geomembranes 27 (2009) 368390potential benet of such solution (Palmeira and Viana, 2003).Fig. 37(b) shows the test arrangement for a ramp test apparatuswith 2 m long boxes. The geomembrane below the soil and thegeogrid buried in the soil are anchored to a rigid frame and loadcells allow for the measurement of mobilised tensile loads duringthe test. The elevation of the geogrid can be varied to investigate itsinuence on the box displacements and on geosynthetics loads.

Fig. 38 shows results of large scale ramp tests for the followingcases: a cover soil on a geomembrane, cover soil reinforced bya geogrid installed at its mid-height and also a situation with thereinforced cover soil on a geomembrane protected by a geotextile(Palmeira and Viana, 2003). The use of a geotextile on the geo-

Ramp inclination = 25

0

2.5

5.0

7.5

0 0.2 0.4 0.6 0.8 1Normalised interface length (x/L)

No

rm

al stress (kP

a) inclined side walls

Trapezoidal distribution, L = 2mTrapezoidal distribution, L = 0.5mBox length L = 10m

Box length L = 2mBox length L = 0.5m

Fig. 35. Numerical and analytical simulations of the ramp test (Palmeira et al., 2002).

E.M. Palmeira / Geotextiles and386membrane is a common measure to minimise damages to thegeomembrane that might jeopardize its performance as a barrier.The geogrid was made of polyester with apertures20 mm 20 mm, thickness of 1.1 mm and a tensile stiffness of200 kN/m. The geotextile, when present, was a nonwoven, needle-punched, geotextile made of polypropylene and with a mass perunit area equal to 200 g/m2. The geomembrane had a smoothsurface and was made of high density polyethylene. The soil wasa uniform coarse sand. Fig. 38 shows the variation of boxdisplacement versus ramp inclination. The presence of the geogridreinforcement increased markedly the ramp inclination at failurefrom 26 to 34. The arrangement with the geogrid and the geo-textile increased the ramp inclination at failure a bit further andprovided a less deformable system, with signicant box displace-ments starting to occur only for ramp inclinations above 28. Thus,the presence of the reinforcement layer in the cover soil can allowhigher slope inclinations or, for the same inclination of an unrein-forced system, the presence of the reinforcement can provideadditional safety against cover soil sliding on the geomembrane.

Fig. 39 presents the variation of ramp inclination at failure (a)for reinforced and unreinforced cover soils as a function of theposition of the geogrid reinforcement (Palmeira and Viana, 2003).The most efcient ratio between reinforcement elevation (y) andcover soil height (H) obtained in the tests fell within the range0.250.5. The combination of geogrid in the soil and geotextile onthe geomembrane provided the greatest benecial effect to thesystem stability.Mo

bilised

in

ter

an

gle (d

e

15

20

22.5

ramp inclination (degrees)presence of the reinforcement. Fig. 40 shows the reduction on theforce in the geomembrane due to the presence of the reinforcementin the cover soil at different values of normalised elevation. In all

geogrid

geomembrane

geogrid

y

L = 2m

soil

geomembrane

clamp

a

b

load cell

H

upper box

lower box

displacementtransducer

Fig. 37. Reinforcement of cover soils of waste disposal areas (Palmeira and Viana,2003). (a) Cover soil reinforcement. (b) Large ramp test apparatus.

materials. Less friction between the geomembrane and the geogridwill certainly lead to a better performance of that arrangement.

The results presented above show the potential of the use of theramp test to investigate the behaviour of reinforced cover soils inwaste disposal areas.

5.3. Summary on ramp tests

A summary of the conclusions on the ramp test is presentedbelow:

Ramp tests are appropriate for tests under low stress levels,which are typical in cover soils of slopes of waste disposal areasor in protection works against slope erosion.

They are also useful for testing interface strength of multiplelayers of geosynthetics or reinforced cover soils.

They are easy to perform and long boxes or boxes with inclinedlateral faces should be preferred.

Long boxes with one of the geosynthetics ends anchoredsimulate more accurately the progressive failure nature of theinterfaces in the region close to the slope crest. Nevertheless,depending on the dimensions of the test, it should be stillconsidered as a rather rough model of the expected behaviourunder eld conditions.

geomembraneramp

soil

geomembrane

ramp

soil

yH

geotextilegeogrid

geomembrane

ramp

soil

yH

geogrid

1 2

3

4

5

nt (m

m)

y/H = 0.5 2

geomembrane geogridgeomembrane

E.M. Palmeira / Geotextiles and Geomembranes 27 (2009) 368390 3870 10 20 30 400

1

2

3

To

p b

ox d

isp

lac

em

e 1

3cases the ramp inclination was equal to the value at failure for theunreinforced case (a 25.6 in Fig. 40). The combination of geogridand geotextile yielded the greatest reduction of tensile force in thegeomembrane. The elevation of the geogrid also affected markedlythat force, with the geogrid directly on the geomembrane being theleast effective arrangement. However, it should not be ignored thatthe efciency of the arrangement of the geogrid on the geo-membrane also depends on the friction angle between both

Ramp inclination (degrees)

With geogridWith geogrid and geotextileGeomembrane

Fig. 38. Effect of the presence of reinforcement on cover soil stability (Palmeira andViana, 2003).

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7y/H

20

25

30

35

40

With geogridGeomembrane only

Geogrid and geotextile

()

Fig. 39. Inuence of the reinforcement elevation on the ramp inclination at failure(Palmeira and Viana, 2003).

ramp

soil

yH

geotextile

geogrid

ramp

soil

Geomembrane only Geomembrane and geogrid

= 25.6 = 25.6With geogrid With geogrid and geotextile

ramp

soil

yHgeomembrane

Geomembrane, geogrid and geotextile

= 25.6

0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8y/H

0

20

40

60

80

100

Geo

mem

bran

e fo

rce red

uctio

n (%

)

= 25.6 degreesFig. 40. Reduction of mobilised geomembrane tensile forces against geogrid elevation.

Geom6. Conclusions

This paper presented and discussed typical experimental andnumerical methods for the study of soilgeosynthetic interaction.Emphasis was given to direct shear tests, pull-out tests, in-soiltensile tests and ramp tests. The main conclusions obtained aresummarised below.

Direct shear tests are simple tests to perform, but boundaryconditions may inuence the results of tests on reinforced soilsamples, particularly for small shear boxes. The soilgeo-synthetic interface shear stiffness, which is relevant fornumerical analysis is difcult to evaluate, unless moresophisticated equipment and testing techniques are employed.The interpretation of tests with the reinforcement crossing theshear plane is difcult and extrapolations of results orconclusions obtained under such testing conditions to realstructures should be done with caution.

Pull-out tests results can be highly sensitive to boundaryconditions. Lubrication of the wall internal frontal face or theuse of a frontal sleeve is a practical measure that minimises theinuence of the box frontal face on the test results. The use ofsleeve should be preferred as it may provide greater condenceon a signicant reduction of the box frontal face inuence, buttheoretical and experimental data available suggests thatfurther investigation is required to a better understanding onthe inuence of the box frontal wall on pull-out test results.Preferably, sleeve and lubrication of the box frontal face shouldbe used. For the usual dimensions of soil samples in large pull-out boxes, the sleeve length should not be smaller than 0.3 m.Because other boundaries of the pull-out apparatus and soilsample dimensions may also inuence test results, pull-outtests should be performed using large apparatus with sampleheights greater than 0.6 m (soil thickness above and below thereinforcement greater than 0.3 m) for the usual reinforcementlengths tested (typically less than 1 m long). Higher samplesmay be required if longer reinforcement layers are tested.Accurate methods to predict the pull-out strength andresponse of geogrids for use in practice are yet to be developed,which enhances the importance of the pull-out test.

In-soil tensile tests are useful for the study of the mechanicalbehaviour of geotextiles under connement, particularly thatof nonwovens. Factors such as impregnation of the geotextileby soil particles can increase even further the in-soil tensilestiffness of these materials, but this particular mechanism isstill difcult to anticipate or predict under eld conditions. In-soil tensile tests with the connement of the geotexile imposedby lubricated rubber membranes can provide lower boundvalues of nonwoven geotextile tensile stiffness.

Ramp tests are easy to perform and useful for the study of thestability of lining systems. They can also be employed to assessthe effectiveness of cover soil reinforcement and the inuenceof water ow on the cover system stability conditions, forinstance. Large testing equipment or boxes with inclined lateralfaces should be preferred to minimise boundary effects andnon-uniform normal stress distributions on the soilgeo-synthetic interface.

Numerical analyses using the nite element method and thediscrete element method can be useful tools for the under-standing of the pull-out behaviour of geogrids and interactionmechanisms in other types of tests. They can also be employedto evaluate the limitations or the inuence of boundaryconditions on the test results. As numerical solutions arebecoming more andmore sophisticated and user friendly, their

E.M. Palmeira / Geotextiles and388contributions to a better understanding on soilgeosyntheticinteraction will certainly increase markedly in the comingyears. However, users of computer codes must be sufcientlyeducated on numerical analysis and aware of modelling limi-tations and their implications on the accuracy of predictions.

Despite how complex a testing device for the study of soilgeosynthetic interaction may be, it should be pointed out thatthe testing techniques available are, in most cases, still rathercrude approximations of the behaviour of the geosynthetic inthe eld.

The standardisation of testing devices and procedures for theevaluation of soilgeosynthetic interaction is of utmostimportance in practical terms as well as for a faster improve-ment on the understanding on soilgeosynthetic interaction.

Acknowledgements

The author is indebted to several persons and institutionsthroughout the years, which collaborated in different ways withsome of the research works presented in this paper. In particular,the author would like to express his gratitude to several past andcurrent research students, to the University of Brasilia, CNPq-National Council for Scientic and Technological Development,CAPES-Brazilian Ministry of Education, FAP/DF-Federal DistrictFoundation for the Support to Research, Finatec/UnB and geo-synthetics manufacturers and suppliers. The author also thanks theInternational Geosynthetics Society (IGS), International Society ofSoil Mechanics and Geotechnical Engineering (ISSMGE) and TensarCo. for supporting the Mercer Lecture.

References

Abramento, M., 1993. Analysis and Measurement of Stresses in Planar Soil Rein-forcements. Ph.D. thesis, Massachusetts Institute of Technology, USA, 288 pp.

Araujo, G.L.S., Palmeira, E.M., 2008. Numerical Analysis of the In-soil Tensile TestUnder Different Boundary Conditions. Internal Research Report, University ofBrasilia, Graduate Programme of Geotechnics, Brasilia, DF, Brazil (inPortuguese).

Azambuja, E., 1994. Investigations on Mechanical Damages on Nonwoven Geo-textiles. M.Sc. thesis, Federal University of Rio Grande do Sul, 142 pp. (inPortuguese).

Bakeer, R.M., Abdel-Rahman, A.H., Napolitano, P.J., 1998. Geotextile friction mobi-lisation during eld pull-out test. Geotextiles and Geomembranes 16 (2), 7385.

Becker, L.B., 2006. Geogrid Behaviour in Reinforced Soil Walls and in Pull-out Tests.Ph.D. thesis, Pont