摘要
对于气动性能,钝体断面的气动外形非常重要,采用传统风洞试验及CFD模拟计算得到钝体断面气动性能需消耗大量时间,大大影响钝体断面气动外形的气动性能评估效率。通过卷积神经网络深度学习技术实现对气动性能的快速预测,深度学习模型训练完成后,输入形状信息和与形状相关的流场信息,即可输出不同几何形状下的阻力系数,进而得到钝体断面的气动性能。为寻找性能最优的深度学习模型,通过综合判定误差和参数量大小对卷积神经网络结构的深度和宽度进行优化。对深度学习模型输出阻力系数与CFD计算结果进行对比发现,误差符合预期要求,并且相较于传统方法,基于深度学习网络的预测所需时间达到数量级的提升,未来可作为钝体断面气动外形优化的关键方法。
The aerodynamic shape of the bluff body section is very important to the aerodynamic performance.However,it takes a lot of time to obtain the aerodynamic performance of the bluff body section using traditional wind tunnel tests and CFD simulation calculations,which greatly affects the aerodynamic performance evaluation efficiency of the bluff body section,s aerodynamic shape.This paper proposes to use the deep learning technology of convolutional neural networks to realize the rapid prediction of aerodynamic performance.After the deep learning model is trained,the shape information and the shape-related flow field information can be input to output the drag coefficients under different geometric shapes,then the aerodynamic performance of the bluff body section.However to find the best deep learning model,this paper optimizes the depth and width of the convolutional neural network structure through comprehensive judgment error and time performance.The output resistance coefficient of the deep learning model is compared with the CFD calculation results.It is found that the error meets the expected requirements,and the prediction time based on the deep learning network is an order of magnitude improvement compared with the calculation time required by the traditional method.It can be used as the bluff body section aerodynamic shape optimization in the future.
作者
李少鹏
李海
李珂
LI Shaopeng;LI Hai;LI Ke(School of Civil Engineering,Chongqing University,Chongqing 400045,P.R.China)
出处
《土木与环境工程学报（中英文）》
CSCD
北大核心
2024年第1期122-129,共8页
Journal of Civil and Environmental Engineering
基金
国家自然科学基金(51978108)
重庆市自然科学基金(cstc2020jcyj-msxmX0773)。
关键词
桥梁静风力
钝体断面
气动性能
深度学习
卷积神经网络
bridge static wind force
bluff body section
aerodynamic performance
deep learning
convolutional neural network
作者简介
李少鹏(1986-),男,博士,副教授,主要从事大跨度桥梁结构风特性研究,E-mail:lisp0314@cqu.edu.cn;通信作者:李珂,男,博士,副教授,E-mail:keli-bridge@cqu.edu.cn。
