基于机器学习方法的流体机械气动优化设计研究现状及展望  

A Review on Optimal Design of Fluid Machinery Based on Machine Learning Method

作　　者：汪逸然 赵文军 梁连国 吴琦 金晗辉[1] 王灿星[1] Yi-ran Wang;Wen-jun Zhao;Lian-guo Liang;Qi Wu;Han-hui Jin;Can-xing Wang(School of Aeronautics and Aerospace,Zhejiang University;Hangzhou Dunli Electric Appliance Co.,LTD.;Zhejiang Zhefeng Science and Technology Co.,LTD.)

机构地区：[1]浙江大学航空航天学院流体工程研究所 [2]杭州顿力电器有限公司 [3]浙江浙风科技有限公司

出　　处：《风机技术》2020年第5期77-90,共14页Chinese Journal of Turbomachinery

基　　金：国家自然科学基金(91741103)。

摘　　要：随着机器学习方法的不断发展,形成了各类算法,已有学者将其应用于流体机械的流动计算、流动控制以及优化设计。以深度学习为代表的机器学习方法具有强大的归纳学习能力,能够自主进行特征学习,不仅可以直接利用实验和数值模拟数据,从中挖掘出潜在的流场信息,更可以在优化设计过程中实现自适应的设计参数探索,以实现快速、鲁棒、全局且高效的优化,因此可以作为一种全新的流体机械优化设计方法。本文概述了流体机械气动优化的现状、机器学习方法的研究现状、机器学习在流动计算、流动控制以及流体机械优化设计中的应用现状,并指出了深度学习在流体机械优化设计上的应用前景。With the continuous improvement of machine learning method,various algorithms have been developed,and some algorithms have been applied to the flow simulation and optimal design of fluid machinery.The deep learning method,which is representation of recent machine learning method,have powerful capabilities of inductive and autonomic learning,not only can this method directly use datum of experiment and numerical simulation to obtain potential flow information,but also can adaptively explore design parameters during the optimal design process,so it has the property of fast,robust,global and efficient optimization,and can be used as a new optimal design method for fluid machinery.For such purposes,an overviews on aerodynamic optimization of fluid machinery,machine learning method,and the application of machine learning method for flow numerical simulation,flow control,and optimal design of fluid machinery are summarized,furthermore,the application prospect of deep learning in optimal design of fluid machinery is discussed.

关 键 词：机器学习 流体机械 优化设计 流动计算 流动控制 深度学习 

分 类 号：TH432[机械工程—机械制造及自动化] TK05[动力工程及工程热物理]