Assimilation of statistical data into turbulent flows using physics-informed neural networks

• Published in: The European physical journal. E, Soft matter and biological physics Vol. 46; no. 3; p. 13
• Main Authors: Angriman, Sofía, Cobelli, Pablo, Mininni, Pablo D., Obligado, Martín, Clark Di Leoni, Patricio
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.03.2023; Springer Nature B.V
• Subjects: Biological and Medical Physics; Biophysics; Boundary conditions; Complex Fluids and Microfluidics; Complex Systems; Condensed matter physics; Fluid dynamics; Nanotechnology; Neural networks; Physics; Physics and Astronomy; Polymer Sciences; Quantitative AI in Complex Fluids and Complex Flows: Challenges and Benchmarks; Regular Article - Flowing Matter; Soft and Granular Matter; Surfaces and Interfaces; Synchronism; Thin Films; Velocity distribution
• ISSN: 1292-8941; 1292-895X
• DOI: 10.1140/epje/s10189-023-00268-9

When modeling turbulent flows, it is often the case that information on the forcing terms or the boundary conditions is either not available or overly complicated and expensive to implement. Instead, some flow features, such as the mean velocity profile or its statistical moments, may be accessible through experiments or observations. We present a method based on physics-informed neural networks to assimilate a given set of conditions into turbulent states. The physics-informed method helps the final state approximate a valid flow. We show examples of different statistical conditions that can be used to prepare states, motivated by experimental and atmospheric problems. Lastly, we show two ways of scaling the resolution of the prepared states. One is through the use of multiple and parallel neural networks. The other uses nudging, a synchronization-based data assimilation technique that leverages the power of specialized numerical solvers. Graphic Abstract