摘要
利用位移实测资料对大坝地基力学参数进行反演时,常规反演算法存在收敛速度慢、精度差等问题。基于GRU神经网络代理模型和狼群算法,提出了一种新的反演方法GRU-WPA法,该方法在保证计算精度情况下能提高计算效率。将此方法应用于某碾压重力坝工程中的地基力学参数的反演,以典型断面中测点的观测数据为依据,反演分析寻优得到地基弹性模量,同时将反演弹性模量的有限元计算位移与实测位移相比较以分析精度,并与常规粒子群算法的反演分析结果进行对比。计算结果表明,GRU-WPA法计算速度更快,反演弹性模量有限元计算位移与实际位移的整体误差更小,表明GRU-WPA法在大坝地基力学参数反演分析中具有良好应用效果。
Aiming at the problems of slow convergence speed and poor accuracy in conventional inversion process of dam foundation mechanical parameters using measured displacement data, a new inversion method(GRU-WPA method) is constructed based on GRU neural network proxy model and wolf pack algorithm. The GRU-WPA method can improve calculation efficiency under the condition of ensuring calculation accuracy. The GRU-WPA method is applied to the inversion of foundation mechanical parameters in a roller-compacted gravity dam project, and the ground elastic modulus is obtained by inversion optimization analysis based on the measured data of typical dam section. At the same time, the finite element calculation results of displacement based on the inverted elastic modulus are compared with the measured displacements to analyze the accuracy, and compared with the inversion analysis result of conventional particle swarm(PSO) algorithm. The calculation results show that, compared with the PSO algorithm, the calculation speed of GRU-WPA method is faster, and the overall error between the displacements calculated by finite element based on inversion elastic modulus and actual displacement is smaller, which verifies the good application effect of GRU-WPA method in the inversion analysis of dam foundation mechanical parameters.
作者
徐喆成
刘晓青
季威
王雪红
XU Zhecheng;LIU Xiaoqing;JI Wei;WANG Xuehong(College of Water Conservancy and Hydropower Engineering,Hohai University,Nanjing 210024,Jiangsu,China;Nanjing Yangtze River Administration Office,Nanjing 210011,Jiangsu,China)
出处
《水力发电》
CAS
2022年第7期39-43,74,共6页
Water Power
基金
国家重点研发计划(2018YFC0407102)。
