基于变形监测资料的边坡智能优化反分析研究

Slope Intelligent Optimization Inversion Research Based on Deformation Monitoring

利用粒子群算法优化改进BP神经网络,通过对BP权值、阈值的修正来提高收敛速度,防止网络陷入局部极值,进而建立起待反演参数和效应量间的非线性映射关系,通过训练网络模型,搜索满足实测效应量值的待反演参数值。基于某高边坡工程算例的实测变形监测数据资料,对边坡体结构的力学参数进行了智能优化反演分析计算,并将PSO-BP智能优化算法与仅使用BP神经网络算法的拟合成果进行了对比分析,进而验证了PSO-BP智能优化算法的优越性与合理可行性。

This paper,by using particle swarm optimization algorithm to optimize and improve the BP neural network,and through the modified BP weight value and threshold value to improve convergence speed,has prevented network into local extremum,then set up the nonlinear mapping relationship between the inversion parameters and effect quantity,through the training of network to search stand-by inversion parameter value satisfied the actual effect of the value.Based on the example of one hydropower station engineering and the measured displacement monitoring data,the optimization inversion calculation and analysis are carried out about the landslide structure mechanics parameters,and the PSO-BP intelligent optimization algorithm and fitting result of BP neural network algorithm are compared and analyzed,so as to verify the advantages and feasibility of PSO-BP intelligent optimization algorithm.