Inverse analysis-based interpretation of sand behavior from triaxial compression tests subjected to full end restraint

Current laboratory testing often imposes or assumes uniform stress and strain distribution in a specimen for convenient data reduction to interpret soil behavior. This paper presents an inverse analysis framework, Self-learning Simulations (SelfSim), to interpret the drained behavior of sand from triaxial compression tests with fully frictional loading platens. The frictional platens result in significant bulging of and nonuniform stresses and strains within sand specimens. SelfSim treats the specimen as a boundary value problem (BVP) and extracts these nonuniform stresses and strains from within each specimen using external load and displacement measurements. The extracted behavior shows significant principal stress rotation, variation of intermediate principal stress, and nonuniform volume change throughout the specimen. Mobilized friction angles are interpreted on the two-dimensional slip surface associated with the Mohr-Coulomb failure criterion, on the octahedral plane associated with the Drucker-Prager failure criterion, and on the spatially mobilized plane (SMP) associated with the Matsuoka-Nakai failure criterion. The extracted stress-strain behavior is used to examine the sand's stress-dilatancy characteristics. Proposed integration of SelfSim inverse analysis with laboratory testing opens the way for new and efficient approaches to soil behavior characterization under general loading conditions, needed for the solution of general geotechnical boundary value problems, from readily available laboratory tests.