The capability of a deep learning based ODE solution for low temperature plasma chemistry

Abstract

A deep learning-based solution is proposed to resolve the highly non-linear ordinary differential equation (ODE) system of the plasma chemistry model. A feed-forward neural network (FNN) is built and trained based on the data generated by the existing global plasma kinetics code. Good agreement is achieved between the results obtained from the deep learning-based method and the traditional plasma kinetics solver for both argon and air discharge conditions. The results demonstrate that the temporal evolution of O-atom density predicted by both the FNN and the 0D model aligns closely with the measurements obtained from the fast ionization wave discharge. Furthermore, the differences in O-atom density between the predictions and measurements are the same order of magnitude. The computational costs of the ODE solver and the FNN model are compared and discussed in this work. The feasibility of using deep learning methods to resolve low temperature plasma chemistry systems is demonstrated through the tests shown in this study.