European Drying Conference - EuroDrying'2011 Palma. Balearic Island, Spain, 26-28 October 2011

VISUALIZATION OF LIQUID IN DRYING PARTICLE AGGREGATES BY X-RAY MICROTOMOGRAPHY AND USE FOR PORE-SCALE MODELLING

Yujing Wang, Abdolreza Kharaghani, Thomas Metzger, Evangelos Tsotsas

Thermal Process Engineering, Otto von Guericke University, Magdeburg, Germany,

Tel.:+49-391-6711362, E-mail: thomas.metzger@ovgu.de

Abstract: This study presents a quantitative approach to investigate liquid phase evolution during drying at the pore level. Random packings of glass beads or glass bead aggregates are used as model porous materials. During drying experiments, 3D spatial distributions of solid, liquid and gas phases are obtained by X-ray microtomography (µ-CT). Main aspects of the image analysis procedure are presented, which is used to extract geometrical features such as the positions of the primary particles that constitute the solid phase and the distribution and saturation of liquid at the pore scale.

Keywords: X-ray microtomography, image processing, liquid distribution, pore network model, volume-of-fluid method

INTRODUCTION

For better understanding of transport phenomena in porous materials during drying, many mathematical models of drying have been developed at different length scales and for different porous structures. Continuous models based on volume-averaged transport equations are used to describe macroscopic drying behavior of porous media. However, locally dominated phenomena such as gradient-free liquid flow and fractal phase distributions, which depend strongly on the structure of the capillary porous medium, can only be described by micro-scale models. Published pore-scale models of drying include pore networks and volume-of-fluid method.

Pore network models represent the void space of porous materials by a network of pores that are connected by throats. Traditionally, these models were used in the field of multiphase flow. Prat (1993) presented a pore network drying model, which includes capillary and gravity effects and is based on invasion percolation. Since then, many extensions and variants of the model have been presented by several groups. Experimental results obtained by drying of transparent etched networks have shown good agreement with modeling results (Laurindo and Prat, 1996). The role of liquid corner films for enhanced mass transfer has been pointed out (Laurindo and Prat, 1998) and also modelled (Yiotis et al., 2003). Simulations have been extended to 3D networks (Le Bray and Prat, 1999; Yiotis et al., 2006). Recently, Kharaghani (2010) presented an irregular pore network model, which is obtained by Voronoi tessellation of a random packing of mono-sized solid spheres that has been generated by

gravitational deposition using discrete element method. For simplicity, the network is built from cylindrical pores and drying is simulated with an established network model (see Fig. 1a for an example). Film effects are not yet accounted for, but are expected to play an import role in the so-called funicular state, i.e. when liquid bridges between the primary particles are interconnected. Nevertheless, for dense particle aggregates, the network model should provide a good basis for studying local phase evolutions during drying.

(a) (b)

Fig. 1. Liquid distributions (in blue) as simulated by (a) pore network model for dense aggregate and (b) volume-of-fluid method for highly porous structure.

However, for highly porous materials such as gels, the open pore space is not suited for representation by a network of cylindrical pores. And for such materials, the drying process does not only induce liquid redistribution but also mechanical effects such as shrinkage and cracks. The particle phase may be displaced by capillary forces, and in a network description this would require tricky topological changes. In order to describe phase distributions

more precisely and handle solid motion in a more natural way, Kirsch et al. (2010) have developed a 3D micro-scale drying model based on volume-of-fluid method to simulate interactions between solid, liquid and gas. Again, gel structure is represented by an aggregate of spherical particles (see Fig. 1b), the mechanical response of which shall be described by discrete element method.

In this study, for validating these two simulation tools, appropriate model systems have been produced from glass beads by means of packing and sintering. X-ray micro tomography has then been used to investigate the drying process of these particle aggregates and provide pore-scale data.

MATERIALS AND METHODS

Drying experimental set-up

To investigate the drying process, systems of glass beads (SiLibeads from Sigmund Lindner GmbH, Germany) with mean diameters of 0.8 mm and 1 mm have been chosen as solid phase and distilled water as liquid phase. For high sphericity and uniform size, the glass beads were carefully selected one by one.

The drying process is carried out for (1) a packing of 0.8 mm glass beads and (2) a highly porous structure from 1 mm glass beads. The latter has been produced by first sintering the glass beads in a graphite mould (see Fig. 2) to form linear and circular aggregates and then creating a packing of these. Sintering was done in an oven, by heating up to 600°C for 6 minutes and slowly cooling down. Some of the aggregates have been broken into smaller pieces to create a wide pore size distribution in the packing. Both packings were generated in a cylindrical plastic container (∅ 8 mm, with bed height of approximately 8 mm).

(a) (b)

Fig. 2. (a) Linear and annular cut graphite mould and (b) sintered glass beads.

Before the experiment, each sample was completely saturated with distilled water. Then, it was mounted on the sample holder inside the X-ray system (Fig. 3) and dried without air condition control. Since drying was very slow, drying rate is expected to have no influence on phase distributions (viscous and thermal effects are negligible). At certain intervals, the sample was gently sealed by sticky tape to interrupt the drying process and to perform an X-ray scan. The sample stays in the X-ray equipment during the whole drying and scanning process; thereby particle displacement can be prevented at maximum.

Fig. 3. Drying set-up in the X-ray microtomograph

(adapted from ProCon X-ray).

X-ray microtomography and image processing

The principle of tomography is that an X-ray beam is attenuated as it passes though the material. This attenuation mainly depends on material density, but also on atomic number and beam energy. Projections of the sample are obtained from many angles (by rotating the sample) and used for reconstruction of 3D images, in which each voxel has a grey value corresponding to the local attenuation coefficient.

A laboratory X-ray microtomograph (CT-ALPHA-160 by ProCon X-Ray GmbH, Germany; financed by European Fund for Regional Development) was used. It is equipped with a microfocus X-ray source and a 2300×2300 detector. The X-ray cone beam was generated at 90 kV and 160 µA. By adjusting distances between source, sample and detector and by grouping detector pixels, effective voxel sizes of 10 to 20 µm have been set, the (cropped) images containing approx. 8003 to 4003 voxels. For a full scan, 800 projections were recorded within 2 hours.

Fig. 4. Image processing steps with (a) binarized

image of liquid and (b) segmented particles.

For geometric analysis and improved image quality, several image processing steps have been performed (by use of MAVI, Fraunhofer) as shown in Fig. 4. First, the original reconstructed grey level image was cropped to suitable size. Then, a median filter was applied for reducing noise. The grey-level histogram represents solid, liquid and gas as three peaks. By use of threshold values the grey level image can be binarized into black and white images, where the white part represents the selected (in Fig. 4a, liquid is

white). Considering the solid phase, particle segmentation, based on the pre-flood watershed algorithm, was applied in order to obtain centre coordinates and radii of the individual particles (see Fig. 4 b). These data will be fed into the simulation tools. The clean up routines were applied for two purposes: filling gas bubbles in the glass beads and removal of the thin artificial liquid layer (due to use of one threshold value for the gradual transition of voxel attenuation coefficient between solid and gas). After these image processing steps, the segmented images, particle center coordinates and radii of separated particles can be further analyzed by MATLAB for models assessment.

RESULTS

The microstructure geometry of solid phase and the evolution of liquid phase during drying are shown as YX sections for the dense packing and YZ sections for the highly porous aggregate in Fig. 5. Related 3D visualizations are given in Fig. 6. As can be seen, the three phases can be nicely separated. The major physical effect of capillary flow from large pores into small pores is easily recognized: large pores dry out first, while the small voids stay saturated with liquid.

(a)

(b)

(c)

(d)

Fig. 5. X-ray sections of liquid distributions in drying packings of glass beads (a, b) and of sintered glass

beads (c, d); glass: red, water: green, air: blue.

(a)

(b)

Fig. 6. 3D visualization of partially saturated particle structures: (a) dense and (b) highly porous.

CONCLUSIONS AND OUTLOOK

This study has shown that X-ray microtomography is suited to characterize pore-scale liquid distributions during drying of dense and highly porous model structures, which are generated from glass beads or sintered glass bead assemblies, respectively. For assessment of the irregular pore network model, experimental particle centre coordinates and radii will be used. Based on these data, a virtual pore network will be generated and the cylindrical pore will be related to sets in 3D voxel space. The filling states of these pores can then be found from the experiment data and directly compared to the simulation results. Additionally, the role of liquid films shall be investigated with high resolution scans. As regards the volume-of-fluid model, the liquid distribution in highly porous structures can be compared to simulations at the level of individual voxels. Future improvements of the experiments include the use of conditioned air for drying and shall reduce scanning times.

REFERENCES

Kharaghani, A. (2010), Irregular pore networks and mechanical effects during drying of porous me-dia, Otto-von-Guericke-Universität, Magdeburg, Germany, PhD thesis.

Kirsch, C., T. Metzger and E. Tsotsas (2010), Micro-scale model for drying of highly porous particle aggregates, Proceedings 17th International Drying Symposium, Magdeburg, Germany, October 3-6, 2010, Vol. A, pp. 407-414.

Laurindo, J.B., M. Prat (1996), Numerical and experimental network study of evaporation in capillary porous media. Phase distributions, Chem. Eng. Sci. Vol. 51, pp. 5171-5185.

Laurindo, J.B., M. Prat (1998), Numerical and experimental network study of evaporation in capillary porous media: Drying rates, Chem. Eng. Sci. Vol. 53, pp. 2257-2269.

Le Bray, Y., M. Prat (1999), Three dimensional pore network simulation of drying in capillary porous media, Int. J. Heat Mass Transfer, Vol. 42, pp. 4207-4224.

Prat, M. (1993), Percolation model of drying under isothermal conditions in porous media, Int. J. Multiphase Flow, Vol. 19, pp. 691-704.

Yiotis, A.G., A.G. Boudouvis, A.K. Stubos I.N. Tsimpanogiannis and Y.C. Yortsos (2003), Effect of liquid films on the isothermal drying of porous media, Phys. Rev. E, Vol. 68, pp. 037303-1/4.

Yiotis, A.G., I.N. Tsimpanogiannis, A.K. Stubos and Y.C. Yortsos (2006), Pore-network study of the characteristic periods in the drying of porous materials, J. Colloid Interface Science, Vol. 297, pp. 738-748.