Machine-Learning for Stress Tensor Modelling in Large Eddy Simulation

Abstract The accurate modelling of the unresolved stress tensor is particularly important for Large Eddy Simulations (LES) turbulent flows. This term affects transfer energy from largest to smallest scales and vice versa , thus controlling evolution flow field-in reacting flows, field transports scalar fields such as mass fractions temperature both which control species production destruction rates. A large number models have been developed in past years incompressible non-reacting common characteristic majority classical that simplifying assumptions are typically involved their derivation limits predictive ability. At same time, various tunable parameters appear relevant closures whose value depends on geometry/configuration/spatial location, require careful regularisation. Data-driven methods an emerging alternative approach may help circumvent above issues, recent studies several were evaluated. chapter discusses problem, presents some most popular algebraic models, reviews advances data-driven methods.