基于ANN-GA协同寻优的大跨度双曲桁架拱钢闸门结构优化设计  

Study on Structural Optimization of Long-Span Steel Gate with Double-curved Truss Arch Based on ANN-GA

作　　者：王皓臣 张燎军[1] 张汉云[1] 章寰宇 林润丰 宋琰 WANG Hao-chen;ZHANG Liao-jun;ZHANG Han-yun;ZHANG Huan-yu;LIN Run-feng;SONG Yan(College of Water Conservancy and Hydropower Engineering,Hohai University,Nanjing 210098,China)

机构地区：[1]河海大学水利水电学院,江苏南京210098

出　　处：《水电能源科学》2025年第1期145-149,共5页Water Resources and Power

基　　金：国家自然科学基金项目(52479122);中国水利水电科学研究院水利部水工程建设与安全重点实验室项目(IWHR-ENGI-202303)。

摘　　要：针对大跨度双曲桁架拱钢闸门结构的优化设计,采用拉丁超立方随机抽样方法建立试验抽样点,通过对抽样点的训练建立人工神经网络(ANN)预测模型;同时协同遗传算法(GA)的全局搜索能力,基于ANN模型构造相应的适应度函数,提出了一种ANN-GA协同优化的结构优化模型,并对某拟建60 m大跨度双曲桁架拱钢闸门关键构件进行结构优化设计。结果表明,ANN模型可有效应用于结构尺寸与闸门总质量及最大折算应力的非线性建模,训练后的ANN-GA模型可根据结构尺寸准确预测该结构尺寸下所对应的闸门总质量及最大应力值;通过建立基于ANN模型构建的适应度函数,GA可实现在ANN模型预测的基础上快速全局寻优并快速收敛,基于ANN-GA的协同优化方法对于闸门结构尺寸优化切实有效。研究成果可为闸门结构优化设计提供参考。In order to optimize the structure of long-span double-curved truss arch steel gate,the experimental sam-pling points were established by using Latin hypercube random sampling method,and artificial neural network(ANN)prediction model was established by training the sampling points.In conjunction with the global search ability of genetic algorithm(GA),a structural optimization model of ANN-GA implementation was proposed.The structural optimization design was carried out on the key parts of a proposed 60 m long-span double-curved truss arch steel gate.The results show that the ANN model can be effectively applied to the nonlinear modeling of the structure size,total mass and maxi-mum equivalent stress of the gate,and the trained ANN model can accurately predict the total mass and maximum stress of the gate under the structure size.By establishing the fitness function based on ANN,the GA can achieve fast global optimization and fast convergence on the basis of ANN model prediction.The ANN-GA collaborative optimization method is effective for the size optimization of gate structure.The results can be used as reference for structural optimization design of gate.

关 键 词：钢闸门 结构优化设计 人工神经网络 遗传算法 

分 类 号：TV663.4[水利工程—水利水电工程]