深基坑工程土性参数的位移反分析

Back Analysis of Displacement of Soil Parameters in Deep Foundation Pit

针对应用试验中力学参数模拟计算的结果与实际情况偏差较大的问题,根据现场实测结果,采用有限元与BP神经网络相结合的位移反分析方法进行计算,得到了较符合工程实际的力学参数;将所得的力学参数输入到建立的BP神经网络训练模型中进行训练后,再用实测的桩体侧向位移进行反分析得到相应的土性参数值,从而避免了在数值计算中参数取值的相对随意性,提高了计算精度。

To solve the difference of calculation results between mechanics parameter analysis and practical situation and attain mechanics parameters fitted for the practical engineering situation, an approach of displacement back-analysis combined with BP neural networks is applied based on the field measured results. When the training of BP neural network is finished, corresponding parameters are attained through the network simulation of actual measured displacement of pile. Thus, the relative arbitrariness is avoided, and the calculation accuracy is promoted.