A Comparative Study of RANS and Machine Learning Techniques for Aerodynamic Analysis of Airfoils 2024-26-0460
It is important to accurately predict the aerodynamic properties for designing applications which involves fluid flows, particularly in the aerospace industry. Traditionally, this is done through complex numerical simulations, which are computationally expensive, resource-intensive and time-consuming, making them less than ideal for iterative design processes and rapid prototyping. Machine learning, powered by vast datasets and advanced algorithms, offers an innovative approach to predict airfoil characteristics with remarkable accuracy, speed, and cost-effectiveness. Machine learning techniques have been applied to fluid dynamics and have shown promising results. In this study, machine learning model called the back-propagation neural network (BPNN) is used to predict key aerodynamic coefficients of lift and drag for airfoils. Here we provide the model with parameters like the airfoil's name, flow Reynolds number and angle of attack of the airfoils with respect to the incoming flows as input parameters. The BPNN model is trained on a dataset obtained by performing CFD simulations using the RANS-based Spalart–Allmaras turbulence model on three different NACA series airfoils under varying aerodynamic conditions. The data that are obtained from the CFD simulations are divided into two subsets: 70% is used as training data and the remaining 30% is used as validation and testing data. The results showed that the BPNN achieved a high level of accuracy in predicting these coefficients, with a low root mean square error (RMSE) and a close fit between predicted and actual values. This suggests that machine learning can be a valuable tool for aerodynamic predictions.
Author(s):
Lochan M N, Rakshitha N, B K Swathi Prasad, Jayahar Sivasubramanian
Affiliated:
Ramaiah University of Applied Sciences
Event:
AeroCON 2024
ISSN:
0148-7191
e-ISSN:
2688-3627
Related Topics:
Machine learning
Computer simulation
Rapid prototyping
Neural networks
Wind tunnel tests
Design processes
Computational fluid dynamics (CFD)
Mathematical models
Aerodynamics
SAE MOBILUS
Subscribers can view annotate, and download all of SAE's content.
Learn More »