|Speaker(s)||Adrian Lozano-Duran Massachusetts Institute of Technology|
|Date||27 April 2023 – 14:30 to 15:30|
|Venue||INI Seminar Room 1|
|Session Title||Machine-learning building-block-flow model for large-eddy simulation|
|Event||[DDEW03] Computational Challenges and Emerging Tools|
A wall/SGS model for large-eddy simulation (LES) is proposed by devising the flow as a collection of building blocks whose information enables the prediction of the wall stress. The core assumption of the model is that simple canonical flows contain the essential physics to provide accurate wall-stress predictions in more complex flows. The model is constructed to predict wall-attached turbulence, favorable/adverse pressure gradient turbulence, separation, statistically unsteady turbulence, and laminar flow. The approach is implemented using two interconnected artificial neural networks: a classifier, which identifies the contribution of each building block in the flow; and a predictor, which estimates the wall stress via combination of the building-block units. The training data are directly obtained from wall-modeled LES with exact modeling for mean quantities to guarantee consistency with the numerical discretization. The output of the model is accompanied by the confidence in the prediction. The model is validated in two realistic aircraft-like configurations: High-lift Common Research Model and NASA Juncture Flow Experiment.