BP神经网络在大跨斜拉桥性能检测中的应用研究  

Research on the application of BP neural network in performance detection oflong-span cable-stayed bridges

作　　者：王凯[1] 陈韵 汤建林 WANG Kai;CHEN Yun;TANG Jianlin(School of Civil Engineering and Architecture,East China Jiao Tong University,Nanchang,330013,China;Zhejiang University of Technology Testing Technology Co.,Ltd.,Hangzhou 310014,China;Zhejiang Traffic Detection Co.,Ltd.,Hangzhou 311215,China)

机构地区：[1]华东交通大学土木建筑学院,江西南昌330013 [2]浙江浙工大检测技术有限公司,浙江杭州310014 [3]浙江交科工程检测有限公司,浙江杭州311215

出　　处：《浙江工业大学学报》2023年第2期171-179,共9页Journal of Zhejiang University of Technology

摘　　要：大跨斜拉桥工程的迅速发展对给该类桥梁的工作性能评估提出了更高的要求。针对大跨度斜拉桥的复杂情况,以浙江省某大跨度斜拉桥为检测对象,利用MIDAS/CIVIL有限元建模预测和荷载试验相结合的方法对桥梁的实际工作性能进行评测。采用Python软件构建BP神经网络,利用MIDAS/CIVIL有限元模型提供的训练样本对算法进行训练,将荷载试验实测值作为算法输入,得到桥梁模型参数和桥梁挠度应变改良值。由结果对比可知:MIDAS/CIVIL模型参数取值和真实值相差较大,误差最大可达27%。BP神经网络算法优化后的桥梁应变预测值更接近实际,优化后理论值误差可减少10%~20%。研究结果表明BP神经网络算法在桥梁工作性能评估中具有实践意义,该研究为大跨斜拉桥工作性能评估提供了可靠的方法借鉴。The rapid development of long-span cable-stayed bridges engineering has put forward higher requirements for its performance evaluation.Aiming at the complex state of long-span cable-stayed bridges,a long-span cable-stayed bridge in Zhejiang Province was taken as the test object.The actual working performance of the bridge was evaluated by the method of combining MIDAS/CIVIL finite element modeling prediction and load test.The BP neural network algorithm was built with Python software,and the algorithm was trained by the training samples provided by MIDAS/CIVIL finite element model.The measured value of load test is used as the input of the algorithm to obtain the parameters of bridge model and the improved value of bridge deflection strain.The comparison of the results shows that the parameter value of MIDAS/CIVIL model is significantly different from the real value,and the maximum error can reach 27%.The bridge deflection and strain prediction value optimized by neural network algorithm is closer to the reality,and the error between theoretical value and actual value after optimization is reduced by 10%~20%.The results indicate that BP neural network algorithm has practical significance in bridges performance evaluation,and provides a reliable method for long-span cable-stayed bridges performance evaluation.

关 键 词：BP神经网络 大跨斜拉桥 工作性能评估 荷载试验 MIDAS/CIVIL建模 

分 类 号：U441[建筑科学—桥梁与隧道工程]