Back analysis of frozen soil thermal properties based on simulated annealing optimization algorithm

The thermal parameters are vital to frozen soil temperature field prediction and stability analysis in cold regions engineering. According to the characteristics of the permafrost temperature field, the paper established permafrost one-dimensional phase change parameters estimation model. The model used nonlinear finite element method to compute the soil temperature field and based on a very fast simulated annealing (VFSA) algorithm to optimize the thermal parameters. VFSA is an artificial intelligence algorithm which is of high search efficiency, easy to realize and does not require derivative information of objective function. Besides, in order to overcome the difficulty of optimization of too many parameters, thermal parameters are grouped and optimized step by step considering the condition of parameters sensitivity and soil layers. As an example, the above approach is applied to a monitoring section of the Qinghai-Tibet Highway in permafrost regions to get frozen soil thermal parameters. Then, these parameters of the inversion results are used to predict the frozen soil temperature field. The analytical results of posterior variance test show that the predicted soil temperature is well in line with actual temperature field in shallow and deep layers, and not bad near permafrost table zone. It can demonstrate that the back analysis results are reliable and good accuracy. The method can be used in practical engineering.