【摘要】 采用多种集成学习算法构建了NEPE推进剂的燃速预测模型，包括bagging、boosting及stacking三类集成学习框架，并通过贝叶斯优化算法实现模型超参数的全局寻优；通过制备推进剂样品并开展燃速测试验证模型可靠性，实验结果表明该模型具有较高的预测准确性；采用SHAP方法对GBR模型进行了可解释性分析，明确了影响NEPE推进剂燃速的关键特征。结果表明，梯度提升回归(GBR)模型预测性能最优，其决定系数(R~2)达0.983,均方根误差(RMSE)为0.319mm/s,平均绝对误差(MAE)为0.217mm/s; SHAP分析显示，压力是影响燃速的关键因素；GBR模型预测的燃速与实验值接近，观察到的差异较小，表明GBR模型可作为预测NEPE推进剂燃速的可靠方法。更多还原

【Abstract】 Various ensemble learning algorithms were employed to construct the burning rate prediction model of NEPE propellants, including three ensemble learning models—bagging, boosting, and stacking, and global optimization of the model′s hyperparameters was achieved through a Bayesian optimization algorithm. The SHAP method was employed to conduct an interpretability analysis of the GBR model, identifying the key features influencing the combustion rate of NEPE propellant. To validate the model, the propellant samples were prepared and burning rate tests were conducted. The comparison between experimental results and model predictions demonstrated that the established model can accurately predict the burning rate of NEPE propellants. The results indicated that the Gradient Boosting Regression(GBR) model exhibited the best performance, achieving an R~2 value of 0.983, a root mean squared error(RMSE) of 0.319mm/s, and a mean absolute error(MAE) of 0.217mm/s. The SHAP analysis revealed that pressure is the most critical factor influencing the burning rate. The burning rates predicted by the GBR model were close to the experimental values, with only minor observed discrepancies, indicating that the GBR model can serve as a reliable method for predicting the burning rate of NEPE propellants.更多还原