Investigation into MOGA for identifying parameters of a critical-state-based sand model and parameters correlation by factor analysis

Adding refinement and accuracy to constitutive models of soil results in the introduction of complexities along with more model parameters. These parameters (such as hardening-/softening-, dilatancy-/contractancy-related parameters and critical state parameters) are usually not easily obtained in a straightforward way. How to identify these key parameters and estimate their correlations of advanced soil models is a particular issue for geotechnical engineering. This paper was aimed to investigate multi-objective genetic algorithms for identifying parameters of advanced sand models based on standard laboratory tests, followed by the correlation analysis of parameters. A critical-state-based sand model has been developed to simulate three triaxial compression tests performed on loose and dense Hostun sand. Two widely used genetic algorithms with two initialisation methods are examined. The performance of the two genetic algorithms is assessed by comparing their simulation performance using optimal parameters, the convergence speed and the distribution of solutions on the Pareto front. The optimal parameters can then be classified into two factors by their correlation relationship.