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

Abstract

In recent years the fluid mechanics community has been intensely focused on pursuing solutions to its long-standing open problems by exploiting the new machine learning, (ML), approaches. The exchange between ML and fluid mechanics is bringing important paybacks in both directions. The first is benefiting from new physics-inspired ML methods and a scientific playground to perform quantitative benchmarks, whilst the latter has been open to a large set of new tools inherently well suited to deal with big data, flexible in scope, and capable of revealing unknown correlations. A special case is the problem of modeling missing information of partially observable systems. The aim of this paper is to review some of the ML algorithms that are playing an important role in the current developments in this field, to uncover potential avenues, and to discuss the open challenges for applications to fluid mechanics.

Can Artificial Intelligence Accelerate Fluid Mechanics Research?

“…A great number of reviews meant that an even greater number of original research articles had been published. If we limited the results of the Scopus search to reviews related to fluids, three major fields of application emerge that attracted research interest in applying ML methods to (a) theoretical and industrial oil and gas [33][34][35][36][37][38][39][40][41][42][43][44][45][46][47][48][49][50][51][52], (b) CFD simulations [53][54][55][56][57][58][59][60][61][62][63][64][65][66][67][68], and (c) health and medical applications [69][70][71][72][73][74][75][76][77][78][79][80]…”

“…Nevertheless, generative models are often incorporated because DNS data are hard to obtain. In the review of Buzzicotti [53], three prevalent models are discussed: variational autoencoders (VAE), GANS, and diffusion models. These models are based on CNNs, and their implementation is focused on reproducing the multiscale and multifrequency nature of fluid dynamics, which is more complex than classical image recognition applications.…”

From Sparse to Dense Representations in Open Channel Flow Images with Convolutional Neural Networks

“…VAEs consist of an encoder, which takes input data and maps it to a probabilistic latent space, and a decoder, which generates output data and reconstructs the input data as accurately as possible [33]. VAEs work on the fundamental premise of acquiring a mapping between a straightforward and constant distribution and the probability distribution of the data [34]. On the other hand, GANs are an alternative class of AI algorithms used in unsupervised learning (UL) and consist of two parts.…”

Multi-Scale Reconstruction of Turbulent Rotating Flows with Generative Diffusion Models

“…In the context of turbulent flows, this implies that POD-like methods primarily emphasize large-scale structures [22,23]. In recent years, machine learning has led to an increasing number of successful applications in reconstruction tasks for simple and idealized fluid mechanics problems (see [24] for a brief review). We refer to super-resolution applications (i.e., finding high-resolution flow fields from low-resolution data) [25][26][27], inpainting (i.e., reconstructing flow fields having spatial damages) [23,28], and inferring volumetric flows from surface or two-dimensional (2D)-section measurements [29][30][31].…”