摘要
为实现大跨度斜拉桥索梁锚固区钢锚箱的结构优化,依托某大跨度斜拉桥索梁锚固区结构实际工程,提出了一种基于NSGA-Ⅱ算法与IGA-BP神经网络模型的结构参数优化方法。首先基于BP神经网络确定了钢锚箱响应数据预测的拓扑结构,采用自适应交叉变异改进的遗传算法对钢锚箱结构响应神经网络预测模型的权值阈值调参,得到满足拟合精度要求的IGA-BP神经网络预测模型。然后建立考虑结构平均应力和主要板件上峰值应力的数学优化模型,采用改进交叉、变异算子的NSGA-Ⅱ算法设计了钢锚箱结构参数优化流程。最后联合改进NSGA-Ⅱ算法和IGA-BP模型实现了钢锚箱结构参数的优化求解。结果表明:自适应遗传算法对BP神经网络权值与阈值调参的效果良好,相较于标准BP神经网络,IGA-BP神经网络的拟合精度和训练效率均更高;改进NSGA-Ⅱ算法可以实现对钢锚箱结构参数的寻优求解,根据Pareto协调最优解的结果,钢锚箱支撑板与承压板厚度有一定增加,加劲板和锚垫板厚度略微降低;优化后的结构上平均应力降幅约为2.7%,其中承压板应力峰值由200.9 MPa降低至178.1 MPa,降幅约为11.3%,支撑板应力峰值由199.6 MPa下降至179.5 MPa,降幅约为10.07%。优化后结构高应力区域峰值应力明显降低,中等应力区域分布较优化前更大,一定程度上改善了结构应力集中现象,验证了该方法的可行性。
In order to achieve structural optimization of the steel anchor box in the cable beam anchorage zone of a large-span cable-stayed bridge,a structural parameter optimization method based on the NSGA-Ⅱalgorithm and IGA-BP neural network model is proposed based on the actual engineering of the cable beam anchorage zone structure of a large-span cable-stayed bridge.Firstly,the topology structure for predicting the response data of the steel anchor box was determined based on the BP neural network.An adaptive cross mutation improved genetic algorithm was used to adjust the weight threshold of the neural network prediction model for the response of the steel anchor box structure,and an IGA-BP neural network prediction model that meets the fitting accuracy requirements was obtained.Then,a mathematical optimization model was established considering the average stress of the structure and the peak stress on the main plates.The NSGA-Ⅱalgorithm with improved crossover and mutation operators was used to design the optimization process for the parameters of the steel anchor box structure.Finally,the NSGA-Ⅱalgorithm and IGA-BP model were jointly improved to optimize the structural parameters of the steel anchor box.The results show that the adaptive genetic algorithm has a good effect on adjusting the weights and thresholds of the BP neural network.Compared to the standard BP neural network,the IGA-BP neural network has higher fitting accuracy and training efficiency;The improved NSGA-Ⅱalgorithm can optimize the structural parameters of the steel anchor box.According to the Pareto coordinated optimal solution,the thickness of the steel anchor box support plate and pressure bearing plate increases to a certain extent,while the thickness of the stiffening plate and anchor pad plate slightly decreases;The average stress reduction on the optimized structure is about 2.7%,with the peak stress of the bearing plate decreasing from 200.9 MPa to 178.1 MPa,a decrease of about 9.5%.The peak stress of the support plate decreasing
作者
胡翌刚
何博文
袁庆
尹俊宇
刘国坤
郭伟奇
HU Yigang;HE Bowen;YUAN Qing;YIN Junyu;LIU Guokun;GUO Weiqi(Hunan Provincial Traffic Construction Quality and Safety Supervision and Administration Bureau,Changsha,Hunan 410116,China;China Communications Second Highway Survey,Design and Research Institute Co.,Ltd.,Wuhan,Hubei 430000,China;Hunan Provincial Road Transportation Administration Bureau,Changsha,Hunan 410000,China;Hunan Lianzhi Technology Co.,Ltd.,Changsha,Hunan 410299,China;School of Architectural Engineering,Hunan Institute of Engineering,Xiangtan,Hunan 411100,China;Hunan Communications Research Institute Co.,Ltd.,Changsha,Hunan 410015,China)
出处
《公路工程》
2024年第2期31-38,115,共9页
Highway Engineering
基金
湖南省教育厅优青项目(22B0737)。
关键词
索梁锚固区
钢锚箱
结构优化
BP神经网络
非支配排序遗传算法
cable beam anchorage zone
steel anchor box
structural optimization
BP neural network
non dominated sorting genetic algorithm
