Thickness Prediction Of Hot Strip Rolling Based On Machine Learning And Feedforward Correction Of Regulations

With the prosperity of the country and the progress of the society,from the spacecraft of the country to the electronic equipment of the home,all of them are advancing with the times.At the same time,the requirements of the equipment on the quality of industrial rolled products have been improved.In the rolling process of hot strip rolling,the outlet thickness is an important index to evaluate the quality of the strip.Among the many factors that affect the strip thickness,the influence of the finishing mill is particularly important.The second level parameters set are high.Coupling,the traditional strip thickness prediction model simplifies some actual factors in production in mathematical modeling,resulting in unsatisfactory thickness in production.When traditional prediction methods are difficult to meet the status quo,intelligent control is applied to The prediction of strip thickness is the primary problem to be solved urgently in the rolling industry.This paper proposes a strip thickness model based on machine learning,uses the novel algorithm LightGBM as the subject framework,and takes the data collected from a 1700 mm actual production line of a domestic steel plant as the main body,creates a strip thickness prediction model based on LightGBM,and proposes A feed-forward correction algorithm based on the factory gives suggestions for changes in production.The main contents of this article are as follows:(1)Constructed a hot strip thickness prediction model based on BP neural network with 53-dimensional(both secondary control settings in actual production)input and onedimensional output(actual strip export thickness).After that,after using principal component analysis to reduce the dimensionality of the input and reducing the input data to 20 dimensions,it is found that the overall performance has improved.(2)On the basis of the above research,in order to further improve the prediction accuracy,a light GBM-based hot strip thickness prediction model is constructed,and the grid search algorithm is used to optimize the selection and optimization of hyperparameters,and linear regression,XGBoost,Cat Boost Several algorithm models are compared and tested,and it is found that the prediction accuracy of the thickness of the model based on LightGBM is significantly better than that of other models.(3)Aiming at the prediction model of LightGBM,an interface design for hot stripthickness prediction was constructed,and based on the actual production data of the factory,a small algorithm framework was proposed to correct the deviation of the predicted thickness to avoid wasting materials.