东南大学学报（自然科学版）2025，Vol.55Issue(5)：1388-1398,11.DOI:10.3969/j.issn.1001-0505.2025.05.020
基于BP神经网络的扁平钢箱梁涡振性能预测
Vortex-induced vibration performance prediction of flat steel box girder based on BP neural network
摘要
Abstract
Taking the flat steel box girder commonly used in long-span bridges as the research object,the data-base of torsional vortex-induced vibration(VIV)response of the flat steel box girder section under different dy-namic characteristics and aerodynamic shapes was established through wind tunnel tests and numerical simula-tions.The backpropagation(BP)neural network was trained using this database.A method to determine the optimal number of hidden layer nodes was proposed.The cross-validation and genetic algorithm were used to optimize initial weights and thresholds of the BP to predict the torsional VIV performance of the section.The results show that the BP neural network optimized by the genetic algorithm achieves high accuracy in predict-ing the VIV characteristics of flat steel box girders.The average relative error of the two randomly selected samples is 8.18%.Furthermore,parametric analysis reveals that the smaller the web angle of the flat steel box girder section is,the more streamlined the box girder section is,and the smaller the torsional VIV response will be.Adding the wind fairing can reduce torsional VIV response.However,increasing the wind fairing angle counteracts this beneficial effect,leading to larger vibration amplitudes.关键词
扁平钢箱梁/涡振/BP神经网络/遗传算法/交叉验证Key words
flat steel box girder/vortex-induced vibration/backpropagation(BP)neural network/genetic al-gorithm/cross-validation分类
交通工程引用本文复制引用
白桦,杨光,杨鹏瑞,杨鑫,高广中..基于BP神经网络的扁平钢箱梁涡振性能预测[J].东南大学学报（自然科学版）,2025,55(5):1388-1398,11.基金项目
国家自然科学基金面上资助项目(52278478) （52278478）
陕西省自然科学基础研究计划面上资助项目(2023-JC-YB-408) （2023-JC-YB-408）
广西科技重点研发计划资助项目(桂科AB23026125). （桂科AB23026125）
