A Review of Machine Learning Application in Aircraft Aerodynamics Development: Methods, Challenges and Prospects

Abstract

Recently, with the development of AI technology and increasing requirement of Fluid Mechanics analysis, the utilization of machine learning (ML) in engineering is becoming an increasingly prominent research topic. In these engineering subjects, Fluid Mechanics is the one of the most discussed. Some researchers have already done diverse investigations in fusion of ML with Fluid Mechanics calculations. However, due to various limiting factors, the problems of relatively low accuracy, high price and large need of training data still exist in ML prediction results compared to Computational Fluid Dynamics (CFD) calculations. In this review, the present research in Fluid Mechanics and ML is presented. Achievements of ML application in aircraft optimization are briefly introduced with several examples. Then 3 well-developed ML aircraft aerodynamic performance analysis and optimization methods, which are neural networks (NNs), regression and clustering, are introduced. Their basic concepts and methodologies are analyzed with examples of applications. The benefits and drawbacks are compared, revealing that ML is tend to become another powerful tool of aerodynamics calculations in the near future. The review will provide the readers with a clearer understanding of ML development in aerodynamics.