Assessment of Machine Learning Methods for State-to-State Approach in Nonequilibrium Flow Simulations

State-to-state numerical simulations of high-speed reacting flows are the most detailed but also often prohibitively computationally expensive. In this work, we explore usage machine learning algorithms to alleviate such a burden. Several tasks have been identified. Firstly, data-driven regression models were compared for prediction relaxation source terms appearing in right-hand side state-to-state Euler system equations one-dimensional flow N2/N binary mixture behind plane shock wave. Results show that, by appropriately choosing regressor and opportunely tuning its hyperparameters, it is possible achieve accurate predictions full-scale simulation significantly shorter times. Secondly, several strategies speed-up our in-house solver investigated coupling with best-performing pre-trained algorithm. The embedding into ordinary differential solvers may offer orders magnitude. Nevertheless, performances found be strongly dependent on interfaced codes set variables onto which realized. Finally, solution was inferred means deep neural network by-passing use while relying only data. Promising results suggest that networks appear viable technology task.