Research On Vibration Diagnosis Of Tandem Cold Rolling Mill Based On Data Drive

The rolling mill vibration problem has a long history,and the previous researches focused on mechanism modeling and vibration signal analysis.Due to the nonlinearity and large hysteresis of the rolling mill system,the modeling is difficult and the accuracy is low.Traditional signal processing methods are also not applied.This problem can be solved by a data-driven method,enabling vibration prediction and fault diagnosis.Considering the nonlinear and strong coupling characteristics of vibration and its correlation with historical input and output,a model of ensemble empirical mode decomposition-long short-term memory is established.Empirical mode decomposition can decompose the signal into multiple simple components to reduce its complexity.Long shortterm memory can establish a model,and the accuracy is improved by introducing historical information.The results show that the accuracy reaches 96%.Considering the harsh rolling environment and the noise in the collected signals,to improve the prediction accuracy,a model of variational mode decomposition-long shortterm memory is established.The signal is denoised by the optimal variational model decomposition of the Babbitt distance,and a long short-term memory model is established to reduce the influence of noise.The results show that the accuracy is increased by 2%,and the mean absolute error and root mean square error are reduced by 0.13 and 0.17 respectively.A reference is provided for suppressing rolling mill vibration.Considering the difficulty in identifying vertical vibration faults caused by abnormal fluctuation of process parameters,a model of variational mode decomposition-back propagation is established.Variational modal decomposition is used to extract features of vibration signals under normal and abnormal fluctuations,and the essential features of the signals are extracted;the extracted feature vectors are labeled and input into the back propagation multi-classifier to achieve vibration diagnosis.The results show that the diagnosis rate reaches 99%,which provide a reference for rapid faults diagnosis.Figure 37;Table 16;Reference 48...