EXPERIMENTAL AND NUMERICAL INVESTIGATIONS OF LIQUEFACTION BASED FAILURE OF STRATIFIED SOIL

Abstract

Seismic failures originated from multiple earthquakes in the recent past and have caused greater loss to humanity and society. Additionally, researchers have profusely investigated the induced landslides, site subsidence and associated liquefaction triggering. A part of the research community was also fascinated to analyse the effect of soil stratification in liquefaction failure. Although, many studies have ignored the presence of silt patches, silt interlayers, or less permeable soil mass in the natural soil deposits. This ignorance of soil stratification has caused several failures in dams, embankments or other geotechnically important structures due to the presence of water films or sliding surfaces. Therefore, further research is still needed to clearly understand the role of each participating and nonparticipating layer in the stratified soil deposits in affecting liquefaction susceptibility. In this thesis, attempts have been made to investigate the response of stratified specimens under both the pre and post-liquefied stages using extensive cyclic triaxial and shake table testing. The cyclic triaxial tests were performed at varying confining stresses, cyclic stress ratio, and densities. Detailed analysis and experimentation on cyclic triaxial testing consist of varying silt thickness (t), locations (z), and the number of silt layers (n) used in the soil stratification. The phenomenon of void redistribution and strain-localisation has been noticed using test results. Also, micro characterisation of failed specimens was performed using SEM (scanning electron microscopic) images. Based on the results obtained through the extensive test series, useful multi-linear models were developed using regression analysis, to predict the liquefaction susceptibility of the specimens. Additionally, to develop confidence in the model's validity, the comparison has been made using the check test and past research studies. The monotonic loading was applied to simulate the postliquefaction effects, where the significance of silt layer location was also visible. The stress paths have been utilised to correlate the induced failures using phase transformation and failure lines. The geotechnical engineers and researchers greatly utilise the dynamic soil properties to investigate the site response analysis of prominent projects. The presence of a silt layer or less permeable material modifies the pore pressure development which ultimately affects the degradation characteristics of the soil strata. Therefore, in this research study, the hysteresis loops originating from extensive cyclic triaxial tests were used to evaluate the dynamic soil properties. The comparison depicted the critical role of the characteristics of the silt layer used in soil stratification. It has been observed that the degradation characteristics of the specimens are affected by relative density, deviator stress, strain rate and mode of specimen preparation. The modulus reduction curves have been presented and validated with the established studies to ensure the efficacy of the results obtained. Many research studies focused on stress or strain-based theories for investigating the initiation of liquefaction and related failures. These investigations adopted the concept of equivalent shear stress and strain-based failure models to replicate earthquake-induced liquefaction. On the other hand, few research works depicted the role of energy-based formulations in understanding liquefaction triggering and suggested that the energy-based theories are more conservative than other conventional methods. These studies considered energy dissipated in the hysteresis loops and correlated with the developed pore pressure after performing cyclic triaxial tests. Most of the earlier work on the energy dissipation approach has neglected the presence of silt layer presence or soil stratification while investigating liquefaction-induced failures. A part of this thesis is based on the dissipated energy-based approach to closely observe the failure mechanism of liquefaction triggering inside the homogenous and stratified soil mass. The salient parameters of the silt layer, namely, thickness (t), location (z), number of silt layers (n), cyclic stress ratio (CSR) and density critically modified the dissipated energy of the soil specimens. Therefore, these influencing parameters have been considered to develop multi-linear regression models for evaluating normalised dissipated energy required to attain liquefaction (Wn,max). Model testing is significantly useful in finding the key reasons and associated failures of liquefaction triggering. These testing facilities provide a clear appearance of failure surfaces and deformations due to seismic instability. Previously, the model testing apparatus like shaking table, vertical tubes, and centrifuge facility have been utilised in many liquefaction studies of dams, embankments, and bridges having homogeneous and stratified soil profiles. The observation of water films and sand boils have been limited to a few key research studies due to the associated complications. This is because of the criticalities and unpredictabilities involved in assessing the weakest liquefaction zone in the stratified soil mass. The present study further attempts to investigate the behaviour of stratified soil systems using modified 1-g shaking table tests. The key parameters like thickness, density, and the number of soil layers were considered in the test series. Also,the available pore pressure transducers and piezometers provided pore pressure values developed near the interface of two different layers. The transparent walls of the apparatus facilitated the clear visual of water films and live failure paths. The depiction of failure stages has been presented for different stratification patterns. Additionally, the generated sand boils correlated with the key parameters of the silt layer used. The diameter and location of sand boils were assessed to understand the mechanics involved. The observation of water films and sand boils have been presented and strongly correlated with the related variables. Additionally, the numerical simulation of any experimental setup or real site is valuable to understand a few unseen failures, which are sometimes impossible during experimentation. In this thesis work, a numerical simulation of the shaking table has been carried out using the UBC3D-PLM model under plane strain conditions. Different models have been prepared to compare the liquefied soil profiles of the homogeneous and stratified conditions. Several nodes were selected at the sand-silt interface and other useful depths. In the stratified models, the critical role of silt thickness and location in governing liquefaction triggering and failure paths has been observed. The response of the pore pressure curve and contours also suggested the presence of silt layer hindrance in pore water movement.