Data-driven deep learning approach for thrust prediction of solid rocket motors

Thrust is a crucial performance parameter of solid rocket motor. Accurately predicting the thrust is very important for rocket engineering monitoring and design. Traditionally, the study of solid rocket motor thrust has relied on model-based methods, which involve building complex models and requiring specialized knowledge. To address this challenging problem, this paper proposes a data-driven approach using deep learning to predict the thrust of solid rocket motor. The method utilizes raw data obtained from ground thrust tests and undergoes preprocessing steps, including segment-wise down sampling and sliding window sample preparation, to extract data features and enhance prediction accuracy. The approach employs a deep neural network architecture based on ConvNeXt to extract and learn advanced features from the preprocessed thrust data, resulting in an end-to-end framework for predicting solid rocket motor thrust. A significant number of experiments conducted on field data from solid rocket motors have confirmed the validity and superiority of the proposed method. The results indicate that both the proposed data processing and prediction methods perform well, with MSE, MAE, and MAPE values of 0.0047, 0.0170, and 1.1283 % on Dataset NO.1, respectively. On Dataset NO.2, the corresponding values are 0.0032, 0.0141, and 0.9846 %. Additionally, comparisons with other commonly used methods demonstrate that the proposed method generally exhibits lower errors. The experiments also investigated the key parameters that affect the performance of the model. The method provides robust support for rocket engineering monitoring and design, showing promising prospects in the field.