Machine learning methods for modeling the kinetics of combustion in problems of space safety

Abstract

Combustion is a complex physical and chemical process, which is considered both in the modeling of new propulsion systems with high energy efficiency and sufficient safety, and in the modeling of explosion safety and fire extinguishing problems. Fundamental research of this process is one of the key factors responsible for the safety of current and future space flights. Modeling the behavior of chemically reacting systems is computationally complex problem. It is necessary to take into account many details and processes, such as multicomponent structure, diffusion, turbulence, chemical transformations, etc. The modeling of chemical kinetics is the most computationally complex stage. In this paper, we consider the problem of approximating chemical kinetics for modeling the detonation of a hydrogen-air mixture using neural networks. The dataset for training the neural network were prepared using the principal component analysis from the results of numerical modeling of detonation in a narrow channel. The results of the obtained neural network showed that the presented model is capable of approximating chemical kinetics processes without significant restrictions on the range of pressure, temperature or the choice of the used time step.