Research On Prediction Of Mechanical Properties For Hot Rolled Strip Based On Big Data And XGBoost

With the improvement of the overall strength of my country’s industry and the rapid development of related industries,the quality level of steel materials in industrial production is becoming higher and higher.Therefore,when the steel material is shipped from the factory,the mechanical properties of the finished product must be tested to determine whether it meets the mechanical performance requirements of the relevant use scenario.In addition,some new steel materials that meet specific working conditions are expected to be developed faster through mechanical property testing.The testing methods of mechanical properties of metal materials can be roughly divided into three categories,namely traditional methods,expert system methods and artificial intelligence methods.However,due to the shortcomings of the above methods,such as low accuracy,insufficient knowledge acquisition,non-associative deduction and self-learning ability,it is difficult to meet the growing demand for testing the mechanical properties of hot-rolled strip,and new theories and methods are urgently needed.This subject studied a method for intelligent prediction of mechanical properties of hot strip rolling based on big data and machine learning,and used nearly 20,000 sets of hot continuous rolling production process data from a steel plant to verify the effectiveness of the method.The main work and innovations of this topic are:(1)A BP neural network-based prediction model for the mechanical properties of hot-rolled strip steel with 20 input attributes(15 steel component elements and 5 hot rolling process parameters)and three mechanical performance output parameters was constructed,and a genetic-based prediction model was proposed.The algorithm’s BP neural network parameter global optimization method,through the use of 20,000 sets of hot rolling strip production process data of a steel mill,the above model was trained and verified,and a good prediction effect was obtained.(2)On the basis of the above research,in order to further improve the prediction accuracy,a prediction model for the mechanical properties of hot strip mill based on XGBoost was constructed,and a global optimization method of XGBoost parameters based on genetic algorithm was proposed,and 20,000 The production process data of hotrolled strip steel in a certain steel plant verified the validity of the model.The results show that the prediction accuracy(R2 value)of the tensile strength,yield strength and elongation of the model are 0.99895,0.99576 and 0.96260,respectively.For BP neural network model.(3)According to the specific situation of the data set used in this article,the interface framework of the hot-rolled strip mechanical performance prediction system was constructed.The Python programming language was used to design the mechanical performance prediction interface with friendly human-computer interaction,including user management,data set import,Model parameter selection settings and the main console of the mechanical property prediction system and other interfaces,and the simulation realizes the one-key prediction function of the mechanical properties of hot strip rolling.