Prediction Of Dynamic Mechanical Parameters In The Rolling Deformation Zone And Dynamic Characterization Study Of Hot Rolling

As technological barriers are constantly being broken by human beings,the fields of automobile manufacturing,household appliances,housing construction and aerospace have also ushered in great development.The strip is also facing the challenge of high quality,high precision and high output required by all walks of life.However,during the production of hot continuous rolling mill,the problem in mill vibration has been an important factor that limits the quality and yield of the strip.Aiming at the problem of rolling mill in vibration,this thesis took hot continuous rolling mill of a steel plant as the research object and quantitative analyzed the relationship between rolling technological parameters and rolling mill vibration by means of theoretical analysis,data mining and experimental verification.In order to solve the problem of a large number of assumptions and simplifications in the mechanism model,a data-driven CNN-BN-LSTM model for dynamic mechanical parameter prediction in the rolling deformation zone under the deep learning perspective is proposed.This model uses deep learning techniques to deeply mine the relationship between mill vibration and process parameters on the basis of production process monitoring data with rich information on the operating status of rolling equipment.In this model,firstly,CNN is used to extract data features effectively,then,BN algorithm is used to improve the training speed,inhibit the disappearance of gradient and reduce overfitting,and finally LSTM is used to extract the data signals of time series,which realizes the study of time series analysis algorithm for predicting dynamic mechanical parameters of rolling deformation area from the perspective of multi-feature deep learning.The experimental results showed that the CNN-BN-LSTM model for dynamic mechanical parameter prediction form the perspective of rolling deformation is featured by the highest prediction accuracy compared with the conventional mechanistic model,KNN model,BP neural network model,and CNN model.The horizontal and torsional coupling vibration dynamics model was established,and the numerical simulation experiment was carried out with the measured industrial data in MATLAB to verify the effectiveness of the dynamic model.Numerical simulation experiments were carried out in the study combined with the prediction model of CNN-BN-LSTM dynamic mechanical parameters in rolling deformation field from the perspective of deep learning established in this thesis.By analyzing the influence of technological parameters,such as,rolling speed,inlet and outlet thickness on rolling mill vibration,the relationship between rolling process parameters and rolling mill vibration was quantitatively analyzed,and the effective vibration suppression measures were also proposed in this thesis.