Optimization of a Solid Rocket Motor Design Through Neural Network-Assisted Performance Prediction

This article aims to utilize neural networks (NNs) for the performance prediction and design optimization of solid rocket motors (SRMs). The optimization of SRM design parameters is crucial for achieving desired performance. Herein, an NN-based method is proposed to achieve optimized design that improves efficiency. To train the NN models, datasets were generated by considering various design parameter values and erosive burning effects on grain combustion. Multiple NN architectures were trained, and their training accuracies were compared. The trained NNs were then employed to predict target values and determine the optimal coefficients and design parameters for an SRM. Furthermore, a sensitivity analysis was performed to assess the influence on the trained NNs. Subsequently, the obtained results were compared with target performance characteristics through simulations using the predicted values acquired from the trained NNs. By leveraging NNs, this study aims to improve the performance prediction and design optimization of SRMs, thereby providing insights into the influence of design parameters and erosive burning on the SRM performance.