A sound physical basis for constitutive models is indispensable for the correct prediction of soil engineering behavior. However, the complex behavior of unsaturated soil due to its multiphase nature and the interactions among phases, has delayed the emergence of this type of model. Most researchers simply run various tests and superficially describe the results without understanding the complex physical laws supporting the data. Several researchers (Houlsby, 1997; Wheeler et al., 2003; Sheng et al., 2004) tried and found some physical bases for discovering the complexities of this subject matter. Li’s insights and background led to an overall thermodynamics-based constitutive framework for unsaturated soil (Li, 2007, 1). This comprehensive elastoplastic modeling framework produced a ba...[ Read more ]

A sound physical basis for constitutive models is indispensable for the correct prediction of soil engineering behavior. However, the complex behavior of unsaturated soil due to its multiphase nature and the interactions among phases, has delayed the emergence of this type of model. Most researchers simply run various tests and superficially describe the results without understanding the complex physical laws supporting the data. Several researchers (Houlsby, 1997; Wheeler et al., 2003; Sheng et al., 2004) tried and found some physical bases for discovering the complexities of this subject matter. Li’s insights and background led to an overall thermodynamics-based constitutive framework for unsaturated soil (Li, 2007, 1). This comprehensive elastoplastic modeling framework produced a basic triaxial elastoplastic unsaturated soil model (Li, 2007, 2) which successfully reproduced the observations reported by Sivakumar & Wheeler (2000), and Wheeler & Sivakumar (2000), including wetting, consolidation, and shearing, on compacted speswhite kaolin.

This thesis finds that the Camclay model parameters λ & N are not constants for saturated statically compacted speswhite kaolin. The Camclay model parameters λ & N in the model (Li, 2007, 2) was treated individually because the model simulation was based on calibrated λ & N . For more than half a century, researchers have recognized the impact of compaction condition on clay structures. However, the impact of the wetting stage on clay structure was not commonly accepted. How to investigate and quantify both impacts remains an unresolved problem. Because clay structure influences the compression of the soil, two parameters λ₀ & N₀ , the counterparts of classical λ & N , right after compaction are introduced in this research. The state right after compaction characterized by λ₀ & N₀ is a referenced state, in which λ₀ characterizes the compressibility and N₀ gives a picture of density after compaction. For tests with wetting under constant mean net stress, λ₀ & N₀ and λ & N are correlated through plastic volumetric strain ε_v^p to obtain a physically and mathematically admissible entry of λ₀ & N₀ into the consistent model for both the expansive and collapsive response of an unsaturated soil (Li, 2008). Subsequently, λ₀ & N₀ are linearly correlated with A.C. (air content) based on a physical and mathematical appreciation of the compaction plane in terms of specific volume against A.C.. The intercept A and C of the linear correlations characterize the density change upon C.E. (compactive effort). Therefore, A and C are correlated with C.E. by logarithmic relations which are similar to e − ln p' characterizing the consolidation stage. Since compaction and consolidation are similar, both depict the density change upon loading.

To calibrate these correlations, extensive wetting and isotropic consolidation tests on statically compacted speswhite kaolin have been performed. The test conditions include 1) the different combinations of C.E. and A.C. (or ACWC); 2) the same combinations under different mean net stresses during wetting. The purpose of applying different mean net stresses during the wetting process, is to produce various plastic volumetric strains ε_v^p, such that the impact of plastic volume change on the soil structure can be examined. Reliability of apparatuses and experimental techniques were measured by the consistency between verification test results and literature data.

Test results unequivocally showed that different static compaction conditions produced different soil parameters λ & N . Different plastic volume changes on wetting produced different λ & N although their compaction conditions were the same. In the past (Gens & Alonso, 1992), swelling was believed to be the ‘privilege’ of the so-called expansive soil containing active clay minerals, such as montmorillonite. Today, the test results show that the kaolin clay can also swell as long as the mean net stress p̅ during wetting, is low and soil density immediately before wetting, is high enough. Swelling indicated a hysteresis, whereas the elastic rebound showed none when both the wetting and isotropic consolidation stages were plotted together in v− p̃ space and then the specific volume before and after wetting, was connected by a straight line. The intercept and slope of the saturated normal consolidation line of the compacted clay were not constants, but were closely related to the compaction conditions and wetting processes.

A unique set of compacted soil parameters was determined for statically compacted kaolin by taking the C.E. and A.C. ( A.C. = n(1 - ACWC x G_s / e)) as two variables. A prototype model in Li (2008) has revealed that both the swelling and collapse upon wetting can be interpreted by the thermodynamic framework (Li, 2007, 1). Therefore, the model (Li, 2007, 2) founded in this framework is improved to a consistent model (Li, 2008) through the bounding surface technique (Dafalias & Popov, 1975, 1976; Krieg, 1975) to cover both the expansive and collapsive response of an unsaturated soil. These tests were simulated with this new model by incorporating the unified model parameters. This bounding surface model exhibits both the swelling and collapse upon wetting and the smooth transition from overconsolidated to normal behavior. The significance of this model is that the mean net stress dependent mechanism for volume change on wetting was achieved, thus elevating the consistency and accuracy for unsaturated soil modeling.