Dynamic model and deep neural network-based surrogate model to predict dynamic behaviors and steady-state performance of solid propellant combustion

Abstract

Understanding the dynamic combustion behavior of a solid propellant with reaction mechanisms is essential for designing rocket engines and safely disposing of expired propellants. In this study, a rigorous mathematical model was applied to predict the dynamic behavior of propellant combustion (HMX/GAP). Because accurate and fast prediction can contribute to saving design time and enhancing operational safety, a deep neural network surrogate model was also developed to predict propellant combustion characteristics, such as burning rate, gas phase temperature, and mole fraction. After the mathematical model with a moving boundary approach was validated with reported data, the DNN surrogate model was trained and tested with data generated through a stochastic simulation using the mathematical model. The prediction accuracies of the developed surrogate model were 96.46%, 99.72%, and 97.62% for the burning rate, gas phase temperature, and mole fraction, respectively. Compared with the dynamic simulation, the computational time of the surrogate model was 596 times faster for the prediction of the burning rate and gas phase temperature and 590 times faster for the mole fraction. The developed framework can be applied to the design of the optimal composition of solid propellants for the desired combustion behavior. It will also help in the safe incineration of expired propellants in real systems.