Calibration of an elastoplastic model of sand liquefaction using the swarm intelligence with a multi-objective function

According to post-seismic observations, spectacular examples of engineering failures can be ascribed to the occurrence of sand liquefaction, where a sandy soil stratum could undergo a transient loss of shear strength and even behave as a “liquid”. Therefore, correct simulation of liquefaction response has become a challenging issue in geotechnical engineering field. In advanced elastoplastic models of sand liquefaction, certain fitting parameters have a remarkable effect on the computed results. However, the identification of these parameters, based on the experimental data, is usually intractable and sometimes follows a subjective trial-and-error procedure. For this, this paper presented a novel calibration methodology based on an optimization algorithm (particle swarm optimization (PSO)) for an advanced elastoplastic constitutive model. A multi-objective function was designed to adjust the global quality for both monotonic and cyclic triaxial simulations. To overcome computational problem probably appearing in simulation of the cyclic triaxial test, two interrupt mechanisms were designed to prevent the particles from wasting time in searching the unreasonable space of candidate solutions. The Dafalias model has been used as an example to demonstrate the main programme. With the calibrated parameters for the HN31 sand, the computed results were highly consistent with the laboratory experiments (including monotonic triaxial tests under different confining pressures and cyclic triaxial tests in two loading modes). Finally, an extension example is given for Ottawa sand F65, suggesting that the proposed platform is versatile and can be easily customized to meet different practical needs.