Research On Prediction Of Rolling Mill Vibration During Cold Rolling Based On Data-Driven

Abnormal vibration of rolling mill is one of the most common problems in cold rolling.The abnormal vibration of the rolling mill will not only reduce the rolling production efficiency,but also cause product quality defects.In severe cases,it will even damage equipment and endanger personal safety,causing enterprises to bear huge economic losses.Therefore,it is of great research significance to predict the rolling mill vibration to realize the early warning of abnormal rolling mill vibration and to provide effective vibration suppression measures when the rolling mill abnormal vibration occurs.In the past,scholars mainly used the mechanism model method to study rolling mill vibration.However,this method cannot fully explain the internal mechanism of rolling mill vibration,and it is difficult to meet the needs of real-time monitoring and rapid diagnosis of the rolling process in the field.With the improvement of modern rolling mill automation level,a large amount of production process data is generated in the rolling process,and behind these data are information closely related to rolling mill vibration.How to effectively use these data to realize the intelligent upgrade of the steel production process is the focus of future research.In this paper,a data-driven method is used to study the rolling mill vibration problem,the purpose of predicting rolling mill vibration by using the process parameters of the rolling process is achieved,and measures for suppressing abnormal rolling mill vibration are given.The main research contents of this article are as follows:(1)Analyze process parameters related to rolling mill vibration,design vibration data and process parameter acquisition and processing scheme.First,fourteen rolling process parameters related to rolling mill vibration were obtained according to the rolling mill vibration mechanism model.Then,according to the characteristics of cold-rolling vibration,work rolls,intermediate rolls,support rolls,and stand arches were selected as the measuring points of rolling mill vibration signals,and rolling mill vibration data was collected.Meanwhile,production process monitoring data(PDA data)is collected at the same time as the rolling mill vibration.Finally,in order to restore the authenticity of the vibration data,the vibration data under different working conditions were cleaned and smoothed.In order to eliminate the magnitude difference between the data caused by different dimensions,the data is also normalized.(2)A rolling mill vibration prediction model of artificial neural network,support vector machine regression and random forest regression was established.The artificial neural network,support vector machine regression and random forest regression models were built respectively.The vibration process parameters were used as input and the rolling mill vibration data was used as output to train the model.Based on the correlation coefficient and mean square error between the predicted and true values of the model as evaluation indicators,the structure and parameters of the three models were optimized.Analysis of the results shows that the prediction model of random forest regression is more accurate than the other two models.(3)In order to achieve the purpose of optimizing the vibration process parameters to suppress the rolling mill vibration,the random forest method is used to quantitatively analyze the contribution rate of each vibration process parameter to the rolling mill vibration.The results show that the rolling speed,the cumulative rolling mileage of the work rolls,and the average deformation resistance of the rolling stock are the main vibration process parameters.Then,the direction of the rolling speed,the cumulative rolling mileage of the work rolls,and the contribution rate of the average deformation resistance of the rolling stock were analyzed using the rolling mill vibration data and the measured data of the induced vibration process parameters.Finally,according to the direction of the contribution rate of the main vibration process parameters,the measures to optimize the process parameters to suppress the rolling mill vibration are given.