Physics-informed deep-learning applications to experimental fluid mechanics

Abstract

High-resolution reconstruction of flow-field data from low-resolution and noisy measurements is of interest due to the prevalence of such problems in experimental fluid mechanics, where the measurement data are in general sparse, incomplete and noisy. Deep-learning approaches have been shown suitable for such super-resolution tasks. However, a high number of high-resolution examples is needed, which may not be available for many cases. Moreover, the obtained predictions may lack in complying with the physical principles, e.g. mass and momentum conservation. Physics-informed deep learning provides frameworks for integrating data and physical laws for learning. In this study, we apply physics-informed neural networks (PINNs) for super-resolution of flow-field data both in time and space from a limited set of noisy measurements without having any high-resolution reference data. Our objective is to obtain a continuous solution of the problem, providing a physically-consistent prediction at any point in the solution domain. We demonstrate the applicability of PINNs for the super-resolution of flow-field data in time and space through three canonical cases: Burgers' equation, two-dimensional vortex shedding behind a circular cylinder and the minimal turbulent channel flow. The robustness of the models is also investigated by adding synthetic Gaussian noise. Our results show excellent capabilities of PINNs in the context of data augmentation for experiments in fluid mechanics.

Enhancement of PIV measurements via physics-informed neural networks

“…A similar study using PINNs for velocity-field reconstruction from sparse data can be found in [27]. Another application of PINNs was presented by Eivazi et al [3] and Eivazi et al [4,5], who applied the PINN algorithm to solve the Reynolds-averaged-Navier-Stokes (RANS) for incompressible ZPGTBLs without any prior assumptions for turbulence. Along with supervised learning from the domain boundary, mean-flow quantities and their coordinates at the domain boundary were used as input data.…”

“…Note that the laws governing fluid-dynamics problems are non-linear partial differential equations (PDEs). Eivazi and Vinuesa [5] showed the potential of using PINNs to reconstruct experimental data with noise and other measurement errors, including vortical flows and wall-bounded turbulence. A similar study using PINNs for velocity-field reconstruction from sparse data can be found in [27].…”

Velocity Reconstruction in Puffing Pool Fires with Physics-Informed Neural Networks

“…A similar PINN-based framework was also applied to the problem of super-resolution, i.e. inferring higher resolution flow features from low resolution ones in space and time 31,32 . It was shown that the PINN could reconstruct the higher resolution flow characteristics in space and time for a series of canonical non-reacting flows both from simulation data 31 or from experimental data 32 .…”

“…inferring higher resolution flow features from low resolution ones in space and time 31,32 . It was shown that the PINN could reconstruct the higher resolution flow characteristics in space and time for a series of canonical non-reacting flows both from simulation data 31 or from experimental data 32 . Therefore, the HFM approach has only been applied to flows of constant density or small density variations that can be described by a Boussinesq approximation.…”

Data reconstruction for complex flows using AI: Recent progress, obstacles, and perspectives

“…The objective of this paper is to examine some of the ML tools that have been applied with promising results to reconstruct data from incomplete observations of complex systems, including idealized turbulent [34][35][36][37][38][39][40][41], engineering [42][43][44][45], and geophysical flows [32,33], and to discuss the possible future directions for quantitative advancements in fluid mechanics.…”