Finite element/percolation theory modelling of the micromechanical behavior of clayey soils
© Pérez-Rea et al.; licensee Springer. 2015
Received: 8 September 2014
Accepted: 13 February 2015
Published: 25 March 2015
A hybrid model for soils, which combines percolation theory and finite element method is presented. The internal soil structure is modelled via the finite element method, and percolation networks are used for analyzing its mechanical behaviour. Through a microscopic characterization of elastic properties of soil grains, the model is generated. The effective percolation threshold obtained is lower than that of the network geometric percolation. The effective mechanical properties predicted are successfully compared to published experimental results.
In civil engineering practice, building foundations design involves the best-possible estimation of the effective mechanical properties, such as soil modulus of elasticity and Poisson ratio, for predicting the final building settlements. In recent studies, there has been an attempt to relate hydraulic conductivity and other properties of fine soils to changes in soil structure at three different levels, namely, micro-, mini-, and macro-structure. The first considers the small pores among single particles; the second, includes spacing among pores and clots; and the third incorporates macropores, such as cracks, fissures, root voids, and visible pores (Mitchel et al. 1973; Collis and McGown 1974; Reddi and Thangavadivelu 1996). Thus, at the microstructure level, flow occurs through macropores and medium-size features may be unimportant. Nevertheless, meso- and micro-pores flow involves intermolecular forces rising from a particular geometry or from electronic properties of the medium or fluid (Knutson andSelker 1994; Hsu and Masten 2001; McNamara and Luthy 2005; Guo and Chorover 2006). Flow, as other conductive properties of porous materials, becomes dependent of the nature of the components, geometric layout and internal conductivity. In this context, physisorption is a general phenomenon that occurs whenever an adsorptive liquid or gas is brought into contact with the surface of a solid (the adsorbent), involving forces of the same kind as those responsible for the condensation of vapors (Sing et al. 1985). In addition to the attractive dispersion forces and the short-range repulsive forces, specific molecular interactions (e.g. polarization, field-dipole, field gradient-quadrupole) usually occur, influencing the effective mechanical behavior of the materials. Because of the above, soil structure and its composition, at a microscopic level, are important properties to consider for developing models of mechanical behavior to attempt predicting real behavior. In this context, in the last thirty years a family of powerful theoretical methods have been developed; these models have allowed the interpretation of experimental results and the prediction of many properties of disordered micro- and macroscopic systems with regular and non-regular structures (Sahimi 1994). Among others, percolation theory has played an important role for understanding these systems and their properties. Specifically, problems related to porous media have received particular attention. For instance, bi-phase flow problems have been understood and at least, under some conditions, can be interpreted as a percolation phenomenon. Statistical Physics of disordered systems has as an aim to provide methods such that from macroscopic properties of such systems to derive laws at microscopic level. Alternatively, to deduce microscopic properties from macroscopic information observed via experimental techniques. The fast development of such theories is due to the fact that it seems that there is a relation between microscopic elements and its effects on the macroscopic properties. Most of the works reported in the literature deal with problems of hydraulic and electric conductivity. However, there are conductive properties that could be modeled as the elastic properties. In addition, literature reports on materials with repeated octahedral cells have shown that the finite element method (FEM) can be applied to predict effective elastic properties for porous materials such as metallic foams (Deshpande et al. 2001). Micromechanical models using properties such as grain size distribution, individual behavior or random structures of different kinds of composites have been realized using FEM with good results ((Berbenni et al. 2007; Prahl et al. 2007); Giannopoulos et al. In press). Perhaps, continuum mechanics cannot be applied to the modeling of materials behavior to macroscopic levels, but the single mechanical behavior of materials components could be modeled with this microscopic viewpoint. Thus, a combination of the percolation theory and the finite element method could lead to a useful prediction of the effective mechanical properties of fine soils.
2. Elastic properties of percolating systems
The first approximation to the elastic behaviour of random percolating nets was based on the problem analogous to the electrical conductivity in gels (De Gennes 1976). This, in turn, is based on Born’s model (Born and Huang 1954) for the microscopic elasticity of a cell. Considering a simple cubic lattice with central forces only, the shear modulus is zero and the bulk modulus goes to zero when an infinitesimally small fraction of bonds are missing (Feng and Sen 1984). In this case, the rigidity threshold is at p = 1. Feng and Seng studied numerically a 2D triangular lattice subjected to a hydrostatic pressure, finding that the elasticity exponent value is numerically greater than the conductivity exponent. However, it has been shown that the problem of percolation associated with this type of lattice elasticity is different from the one associated to regular site or bond percolation (Giannopoulos et al. In press) (Kantor and Webman 1984). Kantor and Webman (De Gennes 1976) proposed a model for the elasticity of a percolating network. In that model, the stiffness threshold was identical to the geometric one. The structure of the infinite percolating cluster has been analyzed via an elastic kinematic model. From that, the critical exponent τ was obtained. Results showed that the elastic behaviour of a percolating net belongs to a class universally different fromconductivity and that the critical exponent τ is much greater than the conductivity exponent t. The lattice model provides a correct description of the elastic behaviour of macroscopically inhomogeneous composed materials, built with regions locally stiff (solids) and regions locally very soft (voids). Close to p c , the material macroscopic stiffness is determined by the elasticity of long and tortuous channels of rigid material, which are in the main skeleton of the percolating cluster.
3. Finite element model (FEM) applied to networks representing soil
Although the problem was thought of as a 2D truss for the loading analysis, the deflections analysis in each element takes into account the structural sections of beams. This is the way for realizing analysis to structural frames of buildings. The cross section was considered squared of 1 micron per side. The beam elements have a length of 10 microns. Each line between nodes is a beam element and it is equivalent to a net loop. For the modelling of nets with different porosities, APDL, from ANSYSMR, was used.
Full bonds networks were modeled with only one material. This material, called henceforth “M1”, had the elastic properties of clay grains and its probability of being occupied is p = 1. The grain size distribution was omitted in these models; however, characterizations at microscopic levels were made to determine the mean grain size (TMG) for modeling regular nets. A second material, called henceforth “M2”, with identical mechanical properties of M1 was introduced. This material represents the soil voids, i.e. the pores.
Starting with an occupied probability p, net bonds were randomly selected and changed from M1 to M2. Thus, a network composed of bonds of two different materials, but with identical mechanical properties is modeled. M1 occupies a fraction p of bonds of the network whereas M2 occupies a fraction 1-p of the same.
To minimize the influence on the mechanical behavior of the network, the M2 stiffness represented by the elastic modulus was reduced in a 1×10−8 scale. The bonds of this material become extremely soft within a rigid structure whose resisting skeleton are all M1 bonds. The stiffness of M2 must be selected close to zero to model pores network, although structurally those bonds exist. The mechanical analysis is done via FEM. Nets were subjected to axial vertical compression, varying from zero up to the value needed to fail the structure due to excessive deformation. All the nodes had free displacements, such as it happens in lab tests, except the lower nodes that were constrained in the three directions.
Results were analyzed to find the mean value of vertical and horizontal strains, as well as the Poisson ratio. All load-strain processes were analyzed in order to obtain the modulus of elasticity of each of the net models, thus obtaining a mean modulus of elasticity for each value of porosity at each net.
4. Results and discussion
The value of p cr experimentally determined was 0.3943, lower than the geometric percolation threshold for a 2D squared net (p cr = 0.5) and much less than the 2D hexagonal net (p cr = 0.6527). Soil porosity is represented by (1-p) where p is ratio of the occupied network by soil solids. Soil critical porosity for a minimum stiffness E is given by 1-p cr . Displacements at the experimental percolation threshold related to geometric percolation of networks are due basically to dimensionality. Experimental percolation threshold was determined in soil samplings located in a 3D space, whereas model nets were considered in 2D. A non percolating aggregate in 2D could be the cross section of a percolating aggregate in 3D. This could explain why percolation threshold for 3D nets are much less than those similar in 2D, because a lesser amount of connected bonds are needed to form a percolating aggregate. Although working with 2D nets could seem very simplified, limitations in memory and processing speed of available hardware are an important factor to select the type of net for approximating soil structure. This is due to the number of degrees-of-freedom needed for the stress–strain analysis. In 3D, even reducing the number of degrees-of-freedom, networks shall be smaller. Even with these limitations, and taking into account the displacements of the percolation threshold in the predicting equations (3a, 4a), acceptable results are obtained. The degree of approximation based upon a statistical analysis of the results of the proposed model is 78%. This is considered a reliable level if one considers that traditional models of soils mechanical behavior have a level less than 50%.
Although percolation models appeared in the 1950’s, their use in soil mechanics has been limited mainly to the analysis of hydraulic conductivity. In porous media, transport phenomenon and electric conductivity have also been studied. Nevertheless, their application to the mechanical properties of soils has been scarce. In this work, percolation theory was resorted to the study of the soil stress–strain behavior. A model of regular nets was proposed to predict the effective modulus of elasticity via micro-structural properties such as layout, size, and shape of clay particles. Specifically, and as a result of this study, it can be concluded that squared and hexagonal nets can be used to model the soil stress–strain behavior of soils, compacted with natural humidity and different degrees of compactness. Nets were subjected to external loadings and it was assumed that each loop had a single elastic behavior. It was also considered that the loading transfer occurs through the contact points of grains; this approach neglects stresses of pores. Thus, a two-phase composite arises, namely, a solid phase due to clay grains and the porous phase with air inside the pores. The hybrid model (percolation-FEM) has been successfully used herein. From this study, and comparing results to reported results, the squared net seems to fit much better laboratory data for remolded soils. The prediction model was calibrated from soil elastic properties determined in the laboratory. The accuracy of the model for predicting the effective modulus of elasticity is about 72% for a squared net and 70% for a hexagonal one.
The percolation threshold for experimental data was less than any of the two models with regular nets. This means that stresses in natural soil are not transmitted only by the contact between grains, represented as nodes in the nets. It is conjectured that inter-particles sliding, associated to the viscous effect, traditionally neglected, is responsible for the difference. The shift of the percolation threshold is necessarily associated to the viscous effect which, in turn, is due to the different relaxation times for the two soil phases, the solid phase and the liquid phase. In the models, the liquid phase was not considered, thus the percolation threshold coincides with the percolation threshold for squared and hexagonal nets. In laboratory tests, since the liquid phase was not isolated, it was included in the percolation process. The soil takes most of the load, whereas the liquid phase takes longer in dissipating the load.
It is important to underline that the shift of the natural percolation threshold is strongly associated to dimensionality. Indeed this was determined via soil test probes in 3D and the percolation thresholds are smaller in this dimension than in the plane. The simplification of a 2D analysis is justified by the limitations of processing equipment and the information availability.
Even though soil porosity is much greater that the critical porosity in geometric nets, it keeps standing the loads without collapsing. Although the critical elasticity exponent τ is universal, which means it is independent from the net geometry and depends only on dimension, the scaling functions are not universal?? Or are not independent??. However, they may be set for predicting the value of the final effective modulus of elasticity, which is very important for the analysis and design of foundations on saturated and unsaturated soils. It is interesting to highlight that the soil natural porosity is very close to the percolation threshold. Physically, the critical porosity is the minimum necessary for reaching the soil strength.
The authors acknowledge the support of the CINVESTAV, campus Querétaro and of Jacqueline Cañetas, of the Instituto de Física, UNAM for the mineralogical and SEM characterization, respectively.
- Berbenni S, Favier V, Berveiller M (2007) Micro–macro modelling of the effects of the grain size distribution on the plastic flow stress of heterogeneous materials. Comput Mat Sci 39:96–105View ArticleGoogle Scholar
- Born M, Huang K (1954) Dynamical Theory of Cristal Lattices. Oxford Univ. Press, New York, Chap. 5Google Scholar
- Collis K, McGown A (1974) The form and function of microfabric features in a variety of natural soils. Geotechnique 24. No. 2:223-254Google Scholar
- De Gennes PG (1976) On a relation between percolation theory and the elasticity of gels. J Phys (Paris), Lett 37:L 1View ArticleGoogle Scholar
- Deshpande VS, Fleck NA, Ashby MF (2001) “Effective properties of the octet-truss lattice material”.-. Journal of the Mechanics and Physics of Solids 49:1747–1769View ArticleGoogle Scholar
- Feng S, Sen PN (1984) Percolation on elastic networks: new exponent and threshold. Phys Rev Lett 52:216–219View ArticleGoogle Scholar
- Giannopoulos GI, Karagiannis D, Anifantis NK Micromechanical modeling of mechanical behavior of Ti-6Al-4 V/TiB composites using FEM analysis. Comput. Mat. Sci. In press.Google Scholar
- Guo M, Chorover J (2006) Leachate migration from spent mushroom substrate through intact and repacked subsurface soil columns. Waste Manag 26:133–140View ArticleGoogle Scholar
- Hsu I, Masten SJ (2001) Modeling transport of gaseous ozone in unsaturated soils. J Environ Eng, ASCE 127:546–554Google Scholar
- Kantor Y, Webman I (1984) Elastic Properties of random systems. Phys Rev Lett 52:1891–1894View ArticleGoogle Scholar
- Keedwell MJ (1984) Rheology and soil mechanics”. Elsevier Applied Science Publishers, London, UKGoogle Scholar
- Knutson JH, Selker JS (1994) Unsaturated hydraulic conductivities of fiberglass wicks and designing capillary wick pore-water samplers. Soil Sci Soc Am J 58:721–729View ArticleGoogle Scholar
- Lee J (2002) Polyolefin-based nanocomposites. M.Sc (Eng) Dissertation. Department of Chemical Engineering. Queen’s University, BelfastGoogle Scholar
- McNamara SW, Luthy RG (2005) Sorbent wicking device for sampling hydrophobic organic compounds in unsaturated soil pore water. 1:Design and hydraulic characteristics. J Environ Eng, ASCE 131:11–20Google Scholar
- Mitchel JK, Greenberg JA, Witherspoon PA (1973) Chemico-osmotic effects in fine-grained soils. J Soil Mechanics and Foundation Division 99:307–322, ASCE, No. SM4Google Scholar
- Pérez-Rea ML (2005) Predicción de Propiedades Mecánicas y Reológicas de suelos usando teoría de percolación “. Ph.D. Dissertation. Instituto Tecnológico y de Estudios Superiores de Monterrey, CEM, MéxicoGoogle Scholar
- Pérez-Rea ML, Rojas F, Castaño VM (2003) Nitrogen adsorption method for determining the meso slit-like size distribution of expansive soils. Mat Res Innovat 7:341–352, Springler VerlagView ArticleGoogle Scholar
- Prahl U, Papaefthymiou S, Uthaisangsuk V, Bleck W, Sietsma J, van der Zwaag S (2007) Micromechanics-based modeling of properties and failure of multiphase steels. Comput Mat Sci 39:17–22View ArticleGoogle Scholar
- Przemieniecki JS (1968) Theory of Matrix Structural Analysis. McGraw-Hill, New YorkGoogle Scholar
- Reddi LN, Thangavadivelu S (1996) Representation of compacted clay minifabric using random networks. J Geotech Eng 122(11):906–913View ArticleGoogle Scholar
- Sahimi M (1994) Applications of Percolation Theory. Taylor & Francis, BostonGoogle Scholar
- Sahimi M (1995) Flow and Transport in Porous Media and Fractured Rock”. VCH, GermanyGoogle Scholar
- Sing KSW, Everett DH, Haul RAW, Moscou L, Pierotti RA, Rouquérol J, Siemieniewska T (1985) Reporting Physisorption data for gas/solid systems IUPAC. Pure & Appl Chem 57(4):603–619View ArticleGoogle Scholar
- Skempton WA (1961) Effective Stress in Soils, Concrete and Rocks”. Pore pressure and suction in soils. In: Conference on the British National Society. Butterworths, LondonGoogle Scholar
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.