Investigations on Physics-Informed Neural Networks for Aerodynamics

TL;DR

The ability of physics-Informed Neural Networks to construct parametric surrogate models, to achieve multiphysic couplings and to infer turbulence characteristics via data assimilation is demonstrated.

Abstract

Physics-Informed Neural Networks (PINNs) have recently emerged as a novel approach to simulate complex physical systems on the basis of both data observations and physical models. In this work, we investigate the use of PINNs for various applications in aerodynamics and we explain how to leverage their specific formulation to perform some tasks effectively. In particular, we demonstrate the ability of PINNs to construct parametric surrogate models, to achieve multiphysic couplings and to infer turbulence characteristics via data assimilation. The robustness and accuracy of the PINNs approach are analysed, then current issues and challenges are discussed.

Figures (20)

• Figure 1: Overview of PINNs principle
• Figure 2: Parametric simulation for the heated cavity case
• Figure 3: Sampling for the cavity case.
• Figure 4: Predictions for different parameter values (parametric simulation): $\mu=0.1$ and $k^f=0.1$ (top), $\mu=0.05$ and $k^f=0.05$ (middle), $\mu=0.01$ and $k^f=0.01$ (bottom)
• Figure 5: Prediction for $\mu=0.05$ and $k^f=0.05$ (single-parameter simulation )

Theorems & Definitions (10)

• Remark 1
• Remark 2
• Remark 3
• Remark 4
• Remark 5
• Remark 6
• Remark 7
• Remark 8
• Remark 9
• Remark 10