Interpretation of soil behavior from laboratory specimens subjected to non-uniform loading conditions

Current laboratory testing requires uniform stress and strain distribution in a specimen for convenient data reduction and interpretation of soil behavior. Available laboratory devices provide information on soil behavior over a very limited range of shearing modes while understanding of soil behavior is needed under general shearing modes to improve constitutive modeling and the solution of engineering boundary value problems. This study implements an innovative inverse analysis framework, Self-learning Simulation (SelfSim), to interpret soil behavior from laboratory specimens subjected to non-uniform loading conditions.;A general incremental strain probe is introduced to examine rate-independent constitutive model response under all possible strain loading conditions. The probing procedure reveals that the symmetry of a Neural Network (NN) based constitutive model is determined by the symmetry of training datasets. This result led to use of a 3-D finite element model for SelfSim learning in this study.;SelfSim learning is demonstrated using two simulated laboratory tests, a triaxial compression shear test with frictional loading platens, and a triaxial torsional shear test with frictional ends. SelfSim successfully extracts the diverse stress-strain behavior from within the specimens. A NN based constitutive model is developed using extracted soil behavior from both laboratory tests and used successfully in the forward prediction of the load-settlement behavior of a simulated strip footing. The results demonstrate that SelfSim establishes a direct link between laboratory testing and constitutive modeling. SelfSim is employed to interpret Racci sand drained shear behavior. The "short" sand specimens cover three different relative densities, and were tested under three different confining pressures in a triaxial cell with fully frictional loading platens. SelfSim analysis extracts stress strain behavior from within each specimen using load and displacement measurements, and reveals that both principal stress rotation and variation of intermediate stress ratio are important features of the extracted "element level" stress paths. Mobilized friction angles are interpreted for the 3-D loading conditions.;Preliminary designs of next-generation laboratory testing devices are introduced and numerically demonstrated to multitudes of generating general shearing modes from a single laboratory test through integration with SelfSim.*.;*This dissertation is a compound document (contains both a paper copy and a CD as part of the dissertation). The CD requires the following system requirements: Microsoft Office.