Intelligent prediction model of mechanical properties of ultrathin niobium strips based on XGBoost ensemble learning algorithm

Abstract

Ultrathin niobium strips with different thicknesses are prepared by an accumulative rolling process. The tensile test of the ultrathin niobium strips is carried out, and the microstructure of each niobium strip is characterized by electron backscattered diffraction (EBSD). The process parameters, mechanical properties and microstructure characterization data of rolled ultrathin niobium strips with different thicknesses are collected, analyzed and sorted. A data-driven intelligent prediction model for the mechanical properties of ultrathin niobium strips is established by deeply integrating the mechanical properties of ultrathin niobium strips with the microstructure evolution mechanism and combining the integrated data with the XGBoost ensemble learning algorithm. The optimal parameters of the XGBoost model are determined by a grid search and used for mechanical performance prediction. The overall generalization performance of the model is evaluated by the coefficient of determination (R2), mean absolute error (MAE), root mean square error (RMSE) and mean absolute percentage error (MAPE). To reflect the advancement of the proposed model, the prediction results of this model are compared with those of the random forest (RF), multi-layer perceptron (MLP) and gradient boosting decision tree (GBDT) prediction model. The research results show that, on the test set, the R2 in terms of the tensile strength and yield strength predicted values based on the XGBoost algorithm were higher than those of other models, reaching 0.944 and 0.964, respectively. The three error indicators corresponding to the XGBoost model were also at the lowest level. This means that the model based on the XGBoost algorithm has the optimal generalization performance and can thus realize the accurate prediction of the mechanical properties of ultrathin niobium strips. This research provides a new method and idea for the optimal design of rolling process and the prediction of their mechanical properties.