Machine Learning modelling for nitrocellulose-based propellant ageing

Abstract. Predicting the ageing behaviour of nitrocellulose-based propellants is a key issue in Munition Health Management, particularly in the context of long-term storage. In the present study, the ageing behaviour of propellants from three ammunition calibres-7.62 mm, 12.7 mm and 30 mm-was investigated using High-Performance Liquid Chromatography (HPLC) measurements obtained after controlled accelerated ageing conditions, in accordance with NATO AOP-48 guidelines. Propellant samples were exposed to several isothermal ageing temperatures, and stabiliser concentration was monitored over time to build degradation profiles representative of long-term storage conditions. Due to the limited size of the experimental dataset, a data augmentation strategy was adopted to improve model robustness and reduce the risk of overfitting. This approach enabled the practical training of several machine-learning models, including regression trees, support vector machines, kernel regression, linear regression, Gaussian process regression, and neural networks, to predict stabiliser depletion. The performance of these models was evaluated with particular attention to their stability and suitability under small-data conditions. The analysis conducted during propellant ageing aims to characterise stabiliser depletion over time and compare degradation trends among the three propellant types studied. This analysis seeks to assess differences in stabiliser consumption rates across the three nitrocellulose-based propellant types and determine whether a correlation exists between the ageing behaviours of the three propellant classes. By relying on empirical data, this study investigates the feasibility of developing predictive models using machine learning rather than traditional methods for shelf-life assessment. Overall, this work supports the application of Machine Learning, adapted to small experimental datasets through data augmentation, within future Munition Health Management frameworks, with relevance for storage monitoring, operational planning and disposal decision support.