Model identification and parameter estimation of elastoplastic constitutive model by data assimilation using the particle filter

Data assimilation, using the particle filter and incorporating the soil-water coupled finite element method, is applied to identify the yield function of the elastoplastic constitutive model and corresponding parameters based on the sequential measurements of hypothetical soil tests and an actual construction sequence. In the proposed framework of the inverse analysis, the unknowns are both the particular parameter within the exponential contractancy model, n(E), which parameterizes various shapes for the yield function of the competing constitutive models, including the original/the modified Cam-Clay models and in-between models and the parameters of the corresponding constitutive model. An appropriate set, consisting of the yield function of the constitutive model and the parameters of the constitutive model, can be simultaneously identified by the particle filter to describe the most suitable soil behavior. To examine the validity of the proposed procedure, hypothetical and actual measurements for the displacements of a soil specimen were obtained for consolidated and undrained tests through a synthetic FEM computation and for consolidated and drained tests, respectively. After examining the applicability of the proposed procedure to these test results, the present paper then focuses on the actual measured data, ie, the settlement behavior including the lateral deformation of the Kobe Airport Island constructed on reclaimed land.