神经网络驱动的桥梁主动气动翼板颤振智能控制优化  

Active Aerodynamic Flap Flutter Control Optimization for Bridges Driven by Neural Network

作　　者：王子龙 赵林[1,2] 崔巍 方根深[1] 李珂 葛耀君[1,2] WANG Zi-long;ZHAO Lin;CUI Wei;FANG Gen-shen;LI Ke;GE Yao-jun(State Key Lab of Disaster Reduction in Civil Engineering,Tongji University,Shanghai 200092,China;Key Laboratory of Transport Industry of Wind Resistant Technology for Bridge Structures,Tongji University,Shanghai 200092,China;School of Civil Engineering,Chongqing University,Chongqing 400045,China;Key Laboratory of New Technology for Construction of Cities in Mountain Area,Ministry of Education,Chongqing University,Chongqing 400045,China)

机构地区：[1]同济大学土木工程防灾国家重点实验室,上海200092 [2]同济大学桥梁结构抗风技术交通运输行业重点实验室,上海200092 [3]重庆大学土木工程学院,重庆400045 [4]重庆大学山地城镇建设与新技术教育部重点实验室,重庆400045

出　　处：《中国公路学报》2023年第8期32-41,共10页China Journal of Highway and Transport

基　　金：国家重点研发计划项目(2022YFC3004105);国家自然科学基金项目(52078383,52008314,52108469)。

摘　　要：通过风洞试验和数值模拟获得主动气动翼板优化控制参数需要庞大的试验和计算成本,并且难以得到最优的翼板控制参数。基于流线箱梁主动气动翼板颤振控制的风洞试验数据,以翼板与主梁扭转运动相位差为输入,颤振临界风速变化比例为输出建立BP人工神经网络模型,对神经网络进行训练得到了主动气动翼板颤振临界风速预测关系。结果表明:预测输出值和实际值之间误差为5%左右,相关系数为0.965;使用训练得到的人工神经网络模型以1°增量对0°~360°范围内的气动翼板相位差进行遍历计算,得到了两侧翼板相位差对主梁-翼板系统颤振性能的影响规律,当迎风侧翼板相位差位于180°~360°内时系统颤振性能得以提高,最优参数组合为迎风翼板相位差231°,背风侧翼板相位差63°;利用获得的最优气动翼板相位差参数组合,建立了主梁-翼板系统流固耦合模型,对试验和神经网络模型的最优参数的颤振控制效果进行验证,证明了神经网络对颤振控制预测的准确性。提出的通过数据量较少的试验数据训练构建人工神经网络模型,构建预测主梁-翼板系统颤振性能的理论框架,显著改善了颤振控制效果,实现了高精度主动气动翼板颤振的优化控制。Determining the optimal parameters for active aerodynamic flaps through wind tunnel tests and numerical simulations incurs massive experimental and computational costs;moreover,it remains difficult to obtain the optimal flap control parameters.Based on the wind tunnel test data from the active aerodynamic flap flutter control of a streamlined box girder,a backpropagation artificial neural network model was established and trained in this study.The torsional motion phase difference between the flap and deck was the input,and the change ratio of the critical wind speed of the flutter was the output.A flutter-critical wind speed prediction relationship between the active aerodynamic flaps and oncoming wind velocities was built by training the neural network.The error between the predicted and actual values was approximately 5%,and the correlation coefficient between the predicted and actual values was 0.965.The artificial neural network model was used to traverse the phase differences of the aerodynamic flaps in the range of 0°-360°at 1°intervals.Correspondingly,the trend of the development of the effect of the phase difference of the two side-mounted flaps on the flutter performance of the deck-flaps system was obtained.The results show that the flutter performance of the system improves when the phase difference of the windward flap is within 180°-360°.The optimal phase difference of the windward flap is 231°and that of the leeward wing is 63°.The optimal phase difference for the mounted flaps was obtained according to the optimized results.A fluid-structure interaction model for the deck-flaps system was established to verify the flutter control effect of the optimized parameters from the tests and neural network model.Thus,a framework was proposed for training an artificial neural network model and predicting the flutter performance of the active deck-flaps system despite sparse experimental data.The improvement effect on the flutter performance can be obtained by inputting the optimized phase difference

关 键 词：桥梁工程 颤振控制 人工神经网络 主动翼板 大跨桥梁 智能控制 

分 类 号：U441.3[建筑科学—桥梁与隧道工程]