Research On Fault State Prediction And Diagnosis For Hydraulic Loading Subsystem Of Vertical Roller Mill

As a new type of grinding equipment with high efficiency,less energy consumption and low operating cost,vertical roller mill has become the main equipment in the process of cement production.Due to the severe operating environment as well as the complicated mechanical and electrical structure of the vertical roller mill,it's inevitably that a variety of faults will occur.Using the fault prediction and diagnosis technology,early warning can be made in the early stage of failure or when the fault trend is presented,then we can find fault features in time and maintain the equipment according to the result of diagnosis,in order to improve the stability.However,with the development of modern industry,the composition and structure of vertical roller mill become more and more complex,so do the function and links between each part,thus high requirements are put forward for the reliability.Therefore,the research on the state prediction and fault diagnosis technology of vertical roller mill is extremely necessary.In this paper,the hydraulic loading subsystem of vertical roller mill is researched,the main research contents are listed as follows.(1)Using modern digital signal processing technology to treat the pressure and acceleration signal of vertical roller mill hydraulic loading subsystem,remove the noise and extract useful information which can significantly distinguish the fault status of equipment.Firstly the wavelet analysis method is used to purified the collected signal,after noise reduction,the sample entropy is extracted as the required feature information by the method of ensemble empirical mode decomposition.(2)The fruit fly optimization algorithm is used to optimize the generalized regression neural network.Based on FOA-GRNN algorithm,the prediction of pressure and acceleration signal for the vertical roller mill hydraulic loading subsystem is proceeded according to the characteristic value obtained by the signal pretreatment,and simulation tests are carried out using MATLAB.The simulation results show that FOA-GRNN neural network algorithm has good performance for fault state prediction of vertical roller mill hydraulic loading subsystem,by using historical data it can accurately predict the future trends.(3)The classical data mining algorithm is studied and optimized.According to the special needs of vertical roller mill fault diagnosis,the optimization is carried out from aspects of the efficiency and anti-noise performance on the basis of the original RIPPER algorithm,by comparing with the performance of BFEA algorithm and IDMK-SVM algorithm,the efficiency and accuracy of this algorithm are proved.Then the improved RIPPER algorithm is used to further diagnose the machinery fault of vertical roller mill hydraulic loading subsystem.When the monitoring objects show abnormal tendency,the type of fault can be judged,thus realize the fault diagnosis of the vertical roller mill hydraulic loading subsystem.