Aging prediction in single based propellants using hybrid strategy of machine learning and genetic algorithm

Abstract

This study proposes novel approach that combines optimization and machine learning to predict the aging of single-base propellants in alignment with the high-performance liquid chromatography (HPLC). To obtain aging and reduce HPLC experimentation, the proposed algorithm provides an efficient procedure. A hybrid strategy of genetic algorithm and ML models including support vector machine, ensemble trees, Gaussian process, and regression tree is combined to optimize SBPs aging. The correlation map provides an interdependence table of aging with initial composition, caliber, and environmental factors. The coefficient of determination and RMSE will determine predictive capabilities of ML models. ET-GA is an optimum-performing ML model with a 0.89 coefficient of determination. Partial dependence plots give an overview of the impact on aging by several variables with maximum impact by temperature, humidity, and diphenylamine. ET-GA shows 95% agreement with experimental data. This results in an economically viable interface that reduces experimentation and provides real-time aging prediction.