Artificial neural networks applied to fluid mechanics for predicting fluid flow around a moving obstacle

The analysis, understanding, and prediction of fluid behavior around a fixed or moving structure are of major importance in fluid mechanics and, more generally, in engineering, particularly for the design and optimization of aircraft components. The study of fluid flows largely relies on high-fidelity simulations performed using codes that solve the Navier-Stokes equations. However, while these codes are essential tools for understanding the fundamental physics behind fluid behavior, their use remains expensive for optimization studies and real-time computation. 

With the recent development of Artificial Intelligence (AI) techniques, deep neural networks are increasingly being applied to problems studied in fluid mechanics.

This internship is divided into three parts. First, the candidate will need to become familiar with high-fidelity fluid simulations to study the flow around a fixed obstacle in 2D geometry. Second, a small numerical dataset will be constructed to train neural networks to predict fluid dynamics around the obstacle. This will be achieved by combining unsupervised and supervised deep learning techniques. Finally, these methods will be extended to the dynamics of a moving obstacle.