Inverse design framework for optimizing solid propellant grains toward target performance profiles

Abstract

This study presents a computational optimization framework for the design of solid rocket motor (SRM) propellant grains, which are significant in shaping the thrust-time characteristics of SRMs. Conventional grain design methods predominantly depend on heuristic approaches and iterative trial-and-error processes, which are not only time-intensive but also likely to result in suboptimal designs. To address these limitations, the proposed methodology integrates an artificial neural network (ANN) with a genetic algorithm (GA) to enable inverse design of grain geometries that achieve prescribed thrust profiles. The process begins with a design of experiments (DOE) strategy to systematically explore the design space. Burnback simulations are then conducted to model the regression behavior of the grain over time, generating a dataset for training the ANN. The trained ANN serves as a surrogate model, predicting performance metrics from input geometries with reduced computational cost. The GA subsequently iterates over candidate designs to minimize the deviation between the predicted and target thrust profiles. The optimization framework specifically targets axisymmetric grain configurations, which are advantageous due to their manufacturing simplicity and inherent capability to minimize sliver formation. Validation is conducted through a series of case studies, demonstrating the framework’s capacity to derive optimal grain geometries that satisfy various target performance profiles. Notably, the proposed method effectively identified an axisymmetric grain configuration that replicates the performance of a Finocyl grain, traditionally considered more complex in shape. This highlights the potential of the method to generate simpler, manufacturable designs that achieve comparable performance outcomes. In light of these findings, the proposed framework constitutes a systematic and computationally efficient methodology for grain geometry optimization, effectively reducing dependence on manual iterations and expert intuition. Consequently, it holds substantial potential for direct application as a general design process for solid rocket motors, supporting the systematic development of propulsion systems across various mission requirements.