Parameter identification for Cam-clay model in partial loading model tests using the particle filter

Abstract

Data assimilation is a versatile methodology, developed in the earth sciences, such as geophysics, meteorology, and oceanography, for estimating the state of a dynamic system of interest by merging sparse observation data into a numerical model for the system. In particular, the data assimilation method referred to as the particle filter (PF) can be applied to nonlinear and non-Gaussian problems, and it holds the greatest potential for application to geotechnical problems. The objective of this study is to demonstrate the theoretical and the practical effectiveness of the PF for a geotechnical problem, i.e., applying the methodology to numerical experiments and actual model tests to identify the parameters of elasto-plastic geomaterials. Since the mechanical behavior of soils depends on both the current stress and the recent stress history of the soil, the sampling method called SIS, which can take into account the stress history experienced by soils, identifies the parameters of elasto-plastic geomaterials remarkably well. The results of the numerical tests have shown that the parameters identified by the PF based on the SIS have converged into their true values, and the approach presented in this study has shown great promise as an accurate parameter identification method for elasto-plastic geomaterials. Moreover, the simulation results using the identified parameters were close to the actual measurement data, and long-term predictions with high accuracy could be achieved, even though short-term measurement data were used. The PF approach produces more information about the parameters of interest than simple estimated values obtained from optimization methods. Namely, the identification comes in the form of probability density functions.