基于PSO-GRNN的大跨桥梁有限元模型修正方法  

Finite element model updating method for long span bridges based on PSO-GRNN

作　　者：周红利 周广东[1] 刘凯凯 奚佳欢 Zhou Hongli;Zhou Guangdong;Liu Kaikai;Xi Jiahuan(College of Civil and Transportation Engineering,Hohai University,Nanjing 210098,China)

机构地区：[1]河海大学土木与交通学院,南京210098

出　　处：《东南大学学报（自然科学版）》2024年第6期1489-1495,共7页Journal of Southeast University：Natural Science Edition

基　　金：国家自然科学基金资助项目(51978243)。

摘　　要：为了对大跨桥梁的有限元模型进行高精度修正,提出了一种基于粒子群算法-广义回归神经网络(PSO-GRNN)的方法.该方法采用广义回归神经网络(GRNN)来描述有限元模型输出与待修正参数之间的复杂非线性映射关系,利用粒子群(PSO)算法对GRNN的光滑因子进行优化.采用一座大跨钢箱梁悬索桥的有限元模型对提出的修正方法进行了验证.研究结果表明:经过PSO优化后的GRNN能够更加准确地描述频率-待修正参数之间的非线性关系,预测误差显著减小;相比于误差反向传播(BP)神经网络方法,GRNN方法和PSO-GRNN方法修正后的频率误差更小;由于PSO的优化,PSO-GRNN方法修正后的频率误差进一步减小,最大误差不超过5%;基于PSO-GRNN的修正方法可广泛用于各种大跨桥梁有限元模型的修正.A method based on particle swarm optimization algorithm-generalized regression neural network(PSO-GRNN)was proposed for high-precision updating of the finite element model of large-span bridges.In this method,the generalized regression neural network(GRNN)was employed to describe the complex nonlinear relationship between the output of the finite element model and the parameters to be updated,and the particle swarm optimization(PSO)algorithm was adopted to optimize the smoothness factor of GRNN.The proposed updating method was verified using the finite element model of a long-span steel box girder suspension bridge.The results indicate that the GRNN optimized by PSO can more accurately describe the nonlinear relationship between frequencies and the parameters to be updated,and the prediction errors are significantly reduced.Compared with the error back propagation neural network method,the updated frequency errors of the GRNN and PSO-GRNN method are smaller.Due to the optimization of PSO,the updated frequency error of the PSO-GRNN based updating method is further reduced,and the maximum error is less than 5%.The updating method based on PSO-GRNN can be used for updating finite element models of various large-span bridges.

关 键 词：大跨桥梁 有限元模型 模型修正 广义回归神经网络 粒子群算法 

分 类 号：U446[建筑科学—桥梁与隧道工程]