Rapid mechanical evaluation of the engine hood based on machine learning

With the development of machine learning and data mining, rapid design, verification manufacturing have become mainstream in machinery industry. In this paper, mapping function between mechanical properties hood its 11 dimensional parameters was mined using algorithms. By combining XGBoost LightGBM algorithms with bagging method, we proposed a hybrid model hyperparameters optimized by grey wolf algorithm. Subsequently, several models were trained tested on dataset 6959 simulation samples, found to excellent predictive performance for torsional stiffness ( $$R^{2}$$ 0.9969 root-mean-square error 1.32185) first-order modal frequency (0.9977 0.00989) hood. Moreover, SHAP method (Shapley additive explanations) used as interpretation explain process performance. The results show that has great potential interpretation. This paper aims develop mathematical hood, which can quickly predict based each key parameter. Therefore, engineers designers apply approximate their design space exploration directly without training extra low-dimensional surrogate models.