CNN-based vane-type vortex generator modelling

Abstract

The simplicity and accuracy of Computational Fluid Dynamics (CFD) tools have made them the most widely used method for solving fluid dynamics problems. However, these tools have some limitations, being the most significant the required computational resources. This fact, added to the growth of the Artificial Intelligence, has led to an increasing number of studies using data-driven methods to solve fluid dynamic problems. Flow control devices are a very popular research topic, since their implementation can significantly improve the behaviour of the flow. Among these devices, Vortex Generators (VGs) can be highlighted for their simplicity, efficiency and numerous applications. In this study, a Convolutional Neural Network (CNN) is proposed to predict the flow behaviour on the wake behind VGs. In order to obtain data for training the network, 20 different CFD simulations were conducted, considering the same inflow conditions but different vane heights and angles of attack of the VGs. The results show that the CNN is able to accurately predict the velocity and vorticity fields on the wake of the VG, being the most conflictive cases the ones with tall VGs, large angles of attack and close distances to the VGs. Additionally, the vortical structure, vortex path and velocity profiles on the vortex core of the main vortex are evaluated, showing good agreements with CFD results.