LAUSR

In the numerical simulation of geotechnical engineering, correctly determining rock parameters is one of the keys to ensure the accuracy of calculation. However, in order to obtain high-precision rock parameters under complex conditions, the inverse analysis method is required to characterize the high nonlinearity and global searching optimization. In this paper, considering that the differential evolution (DE) algorithm has strong robustness in solving the non-convex, multi-peak and nonlinear function optimization problems, it is applied to inversion of rock parameters. At the same time, considering that single DE algorithm may fall into the local optimal problem, a simulated annealing (SA) algorithm which can accept the worsening solution with certain probability is introduced in the search process, so that the hybrid algorithm (SADE) can jump out of the local optimum and find the global optimal solution. By selecting two typical basic test functions to compare the performance of a single difference evolutionary algorithm (DE) and a differential evolution-simulated annealing algorithm (SADE), it is proved that the SADE has the capacity of jumping out of the local optimum, which makes up for the deficiency that the DE may fall into the local optimum. The numerical simulation on triaxial creep experimental models validates that the proposed SADE has faster convergence rate, better global convergence and the strong robustness for nonlinear function optimization. Application of the SADE for inversing of rock creep model parameters illustrates high accuracy and efficiency, which demonstrates the promising of the improved SADE for recognizing creep parameters of rocks under complex stress state.