模拟退火优化算法的冻土热传导参数反分析

土体热参数是寒区工程温度场预测和稳定性分析至关重要的参数。根据冻土温度场计算的特点，建立冻土一维相变热参数估计模型，采用非线性有限元法得到土体温度场，基于一种非常快速的模拟退火（VFSA）算法对热参数进行反分析。考虑各个参数的敏感度因子和土体分层情况，将热参数分组、分步进行反演，克服参数较多反演困难的缺陷。以青藏公路多年冻土区监测断面下土体反分析为例，反演分析冻土路基下部不同土层的导热系数、体积热容量和孔隙率等参数。最后，运用参数反演结果对温度场进行预测，经检验和与实测地温对比，表明参数反分析结果可靠，有较高的精度，该方法可用于实际工程。

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.