Research On Prediction Of Mechanical Properties Of Hot Rolled Strip Steel Based On Ensemble Learning

The development of strip steel products is the same as the intensive economic growth mode in China,and the essence is to improve product quality and economic benefits as center.At present,the research on the quality index of hot rolled strip steel products is mainly divided into three aspects: surface quality,dimensional accuracy and mechanical properties.Among them,the main mechanical properties of steel: tensile strength,yield strength and elongation.To improve the mechanical properties of steel,the only way is to carry out appropriate heat treatment.Appropriate heat treatment can significantly improve the strength of steel and maintain good plasticity and toughness.For example,normalizing can refine the grain,improve the tensile strength without reducing the elongation.In addition to pass design and other conditions in hot rolling production,there are its own internal factors(chemical composition)and external conditions such as heating(related process parameters)that affect the hot plastic deformation of the metal.In the process of predicting the mechanical properties of steel rolling,four different statistical learning methods are used to predict the tensile strength of steel rolling,namely,principal component analysis coupled with Gradient Boosting Decision Tree(PCA-GBDT),Random Forest(RF),e Xtreme Gradient Boosting(XGBoost)and Rule fitting(Rulefit).The data about the tensile strength of steel rolling comes from a large hot continuous rolling mill in China.The data with 2489 sample values are randomly divided into two groups according to the ratio of 7:3,which are used as the training set and the test set respectively.Finally,the data in the training set are used to establish regression models with the above four statistical methods,and the data in the test set are used to predict the mechanical properties of steel rolling,and the model is tested.At the end of this paper,the comparison and of analyze these regression models from two aspects of model precision and model interpretability,and find that PCA-GBDT model has the highest precision,but when considering the interpretability of the model,it is found that Rulefit model has much higher characteristic interpretation than the other three models,and it is more suitable for the actual production process of steel rolling.