Experimental and Numerical Evaluation of Mechanical Response of Unsaturated Soils and Their Application to Embankments and Footings

Abstract

Naturally, soils are rarely fully saturated, and climatic and environmental conditions frequently cause moisture content to change. Due to the varying climate changes and its effects on compacted engineering structures, studying unsaturated soils is fundamental in predicting the behavior of soil systems for various applications like compacted embankments and shallow footings upon rainfall. Rainfall infiltration results in the development of excess positive pore pressures in compacted embankments/soils due to the loss of suction. The buildup of excess positive pressures can significantly impact the stability of the embankment and potentially leading to reduced bearing capacity incase of footings. Unsaturated triaxial testing allows researchers and engineers to replicate these real-world conditions in the laboratory, providing a more accurate representation of the in situ behavior of soils. Moreover, the experimental investigation of the cyclic behaviour of unsaturated materials is more complicated and complex than that of saturated materials owing to the necessary machinery, expertise, and time. To better understand the complexity and sensitivity of unsaturated cyclic and monotonic parameters to various determinants like suction path, loading speed, deviatoric stress amplitude, physical specifications, and others, further research is required in the field of unsaturated materials. In this study, the mechanical behaviour of unsaturated soils viz, sand, and silty sand was investigated by performing monotonic loading tests under suction-controlled triaxial conditions. A series of consolidated drained (CD) tests were conducted under various constant matric suction values and net confining pressures. Matric suction values of 50kPa, 100kPa, and 200kPa and net confining pressures of 50kPa, 100kPa, 150kPa, 200kPa, and 250kPa were considered. Experimental investigation of both soils revealed that as the matric suction increased, shear strength gets increased. During triaxial shearing, the samples depicted peak shear strength followed by post-peak softening and eventually reached a critical state. Such a typical soil response has been numerically modelled based on the frameworks of Barcelona Basic model (BBM) and Morvan model in order to reproduce the experimentally observed behavior. A series of undrained cyclic triaxial tests were conducted inorder to investigate the influence of suction and net confining pressure on the dynamic properties of unsaturated silty sand under both isotropic and anisotropic consolidation conditions. It is observed that incase of anisotropic consolidation, shear modulus is higher at all confining pressures than isotropically consolidated samples due to the reorientation of particles during anisotropic consolidation. Unsaturated soils are characterized by loss of suction upon infiltration. In this study, Barcelona basic model (BBM)- an isotropic constitutive critical state model, was implemented in FLAC for simulating the effect of varying rainfall intensities on compacted embankment. It was observed that large vertical deformations occur and considerably high positive pore water pressures get developed, accounting for the collapse of an embankment. The results also show that rainfall intensity and duration affect slope stability. During the rainfall, collapse compression predominates due to the slippage of particles, resulting in the rearrangement of soil fabric towards a configuration dependent on the fabric of the initial stress state. Fabric changes lead to the anisotropy of mechanical behaviour in unsaturated soils. This study implements an anisotropic model (ABBM) based on the Barcelona Basic model (BBM) in FLAC to analyze the wetting behaviour of a typical compacted embankment upon infiltration. It was observed that vertical surface displacements were higher in case of ABBM due to the rotation of the yield surface as rainfall increased. Additionally, it was found that as the value of the anisotropic evolution parameter rises above a certain threshold point, anisotropy reduces, and the behaviour tends to be the same as that given by the isotropic Barcelona Basic model (BBM) due to the change in preconsolidation pressure. Further, this study proposed an approach for evaluating the performance of embankments subjected to rainfall by accounting for the change in fabric anisotropy with the degree of saturation. It is well known that soil heterogeneity is a substantial source of variation in geotechnical parameters, which gives the system a random nature. Deterministic slope stability analysis techniques usually fail to appropriately address the uncertainty involved, hence probabilistic methodologies must be used to account for the variability of the input soil characteristics. To assess the performance of embankments using different constitutive models considering the uncertainty of various hydraulic and mechanical parameters, a reliability analysis was conducted using the Monte-Carlo method with the Mohr-Coulomb model, BBM, and ABBM models. A series of realizations were performed using the soil modulus and friction angle as random variables. The results revealed that the failure probability of embankments simulated using ABBM is higher than those obtained using Mohr-Coulomb or BBM. Due to rainfall, footings are usually subjected to effects of water table fluctuations which results in the variation of ultimate bearing capacity in unsaturated soils. However, the conventional bearing capacity theories don’t account for the variation of suction as the water table position changes. In this study, an equation based on Terzaghi theory is proposed to evaluate the variation of bearing capacity due to suction variations due to rainfall. The proposed equation results in higher bearing capacity values compared to the conventional Terzaghi equation. Further, a comparison of the bearing capacity obtained using the proposed equation is made with those obtained from Barcelona Basic model (BBM), Mohr-Coulomb model and Terzaghi equation. It is observed that BBM results in lower values of bearing capacity compared to Mohr-Coulomb model as the former considers the loss of suction upon infiltration. Soil properties like porosity, permeability and soil water characteristic (SWCC) influence the bearing capacity in unsaturated soils. Thus, probabilistic analysis is carried out by using an efficient kriging surrogate model in order to analyze the impact of variation of the associated parameters on the bearing capacity. The surrogate model can predict the surface displacements of footing upon infiltration with enhanced accuracy and computational efficiency. Considering the uncertainty associated with rainfall loads, it is observed that as the return period of rainfall increases, failure probability increases. Soil water characteristic curve (SWCC) parameters are essential to study the complex relationship between matric suction and water content in unsaturated soils. It is an essential tool for evaluation of hydraulic and mechanical properties of unsaturated soils which is primarily regulated by soil texture, density, composition etc. Laboratory estimation of SWCC is quite laborious and time consuming, so alternate indirect methods are highly desirable. In this study, a comparison of prediction models based on three machine learning techniques, i.e., Random forest (RF), Extreme gradient boosting (XGBoost) and multi-expression programming (MEP) is conducted for SWCC estimation. The accuracy of the models is testified by several standard indicators which reveal that SWCC prediction using RF is the most robust, although the predictions from the other two models are also reliable. Further, for practical application, the straightforward expression obtained from MEP has been utilised to estimate the SWCC in case of a typical embankment. To accomplish this, a comparison of the developed MEP model is done with a traditional physio-empirical model i.e., continuous form of Arya and Paris model (ACAP) based on the experimental data of 25 soils, and it was found that MEP outperformed this particular model. Since pore pressure represents a measure of instability in soils, it is evaluated using both the models for the embankment under rainfall. A marginal increase in pore pressure is observed while deploying the MEP model to predict the SWCC of fine-grained soils. Embankment failures can be prevented by introducing geocomposites to act as drains. The effect of the geocomposite layer on the pore pressure distribution and surface displacements of an unsaturated embankment upon infiltration has been studied numerically using deterministic and probabilistic approaches. The inclusion of the geocomposite layer leads to an increase of suction below the interface and a decrease in suction above it by functioning both as a capillary barrier and a drainage layer, thereby reducing the surface displacements upon infiltration. The load in the form of rainfall and the resistance such as suction of the embankment material being variable leads to a variability in the displacements, so reliability analysis has been carried out using hydraulic permeability and soil water characteristic curve (SWCC) parameters as random variables. To assess the probability of failure (P_f), surrogate model based on augmented radial basis function has been used. Probabilistic analysis revealed that the embankment with geocomposite has less P_f compared to the one without geocomposite considering the rainfall infiltration. Moreover, sensitivity analysis predicted that SWCC parameters influence the P_f of geosynthetics inclusive embankment under infiltration to a larger extent.