Research On Running State Prediction And Fault Diagnosis For Hydraulic System Of Vertical Roller Mill

Vertical roller mill is a large mechanical equipment with the ability to crush,dry,grind,select and convey.With its high grinding efficiency,small occupied area and long service life,it is widely used in cement,chemical industry,electricity and other industries.The running state prediction and fault diagnosis of the vertical roller mill can reduce the maintenance cost of the equipment and increase the service life of the equipment.It can also realize the accurate identification and positioning of the faults,thus ensuring the safety of the vertical roller mill during the production process.In this paper,the hydraulic system of vertical roller mill is taken as the research subject.Firstly,an improved EEMD de-noising method is proposed to preprocess the collected FBG sensor signal,and an echo state network prediction model optimized by fruit fly optimization algorithm is put forward for the prediction of pressure signal of the hydraulic system of the vertical roller mill.Then the feature extraction method based on the relative wavelet packet energy and the fault recognition method based on the multi-class relevance vector machine are used to identify the four states of the hydraulic system of the vertical roller mill,including normal state,abnormal pressure of the nitrogen air bags,the fracture of the hydraulic system and roller connecting rod and the body injury or oil leakage of the hydraulic cylinder.Finally,a running status prediction and fault diagnosis system of hydraulic system of vertical roller mill is developed based on the study of running state prediction method and fault diagnosis method.The main contents of this paper are organized as follows:(1)Aiming at the problem that the sensor collected by the vertical roller miller often contains a large amount of uncorrelated noise,a noise de-noising method is proposed based on the improved ensemble empirical mode decomposition.According to the traditional EEMD algorithm,we will ignore the effective part of high frequency if we reconstruct the low frequency component to achieve the goal of noise reduction.In this method,the IMFs are obtained after decomposing the signal by EEMD and the low-frequency components are smoothed by median filter,then the high frequency parts are handled by threshold denoising,and finally the processed IMFs are reconstructed to achieve the goal of noise reduction as well as ensuring integrity of the signal.At last,the method is used to reduce the noise of acceleration and pressure signals collected on the hydraulic system of the vertical roller mill,which will lay the foundation for the following running state prediction and fault diagnosis.(2)Because the collected signal from vertical roller mill has nonlinear and non-stationary character,a kind of trend prediction algorithm based on the improved echo state network state is put forward to improve forecast accuracy of running state of vertical roller mill in this condition.The selection of the key parameters of the echo state network reserve pool affects the prediction accuracy and the performance of the model,so fruit fly optimization algorithm is combined to select the key parameters of reserve pool dynamically.Finally,the proposed model is verified and analyzed by experiment and simulation.(3)The feature extraction method based on the relative wavelet packet energy and the fault recognition method based on the multi-class relevance vector machine are studied.Through the wavelet packet decomposition of the acceleration signals of the hydraulic system of the vertical roller mill,the relative energy of each band is combined into the eigenvector and input into the multi-class relevance vector machine model for fault identification.The experimental results show that this method has a high recognition accuracy for the hydraulic system of vertical roller mill.(4)A running status prediction and fault diagnosis system of hydraulic system of vertical roller mill is developed.The system can acquire and display the acceleration and pressure signals collected by sensors of the hydraulic system of vertical roller mill,and can predict the future development trend of the signal and judge the fault type of the hydraulic system of vertical roller mill by historical data.