Mechanical Fault Diagnosis Of Cnc Lathe Based On Vibration Signal Analysis

CNC machine tools are complex mechanical and electrical products that combine mechanical,electrical and cutting processes.During its operation,any part of the failure will affect the normal operation of the machine tool.At present,the CNC system has the ability of automatic diagnosis,which can detect electrical faults.Therefore,the fault diagnosis of mechanical components based on the online monitoring information of CNC machine tools is of great significance for reducing the probability of failure and reducing economic losses.This paper takes the mechanical components of CNC lathes as the research object.In view of the characteristics that non-stationary vibration signals will be generated during operation,and the frequency statistical characteristics of the vibration signals will change when a fault occurs,multi-channel sensors are used to collect vibration signals.The signal is decomposed,denoised,and the feature information of the operating state is extracted,and fault diagnosis is realized based on information fusion.The main research contents are as follows:(1)Decomposition and denoising of mechanical components vibration signal based on improved LMD-wavelet packet.In order to improve the problem of endpoint effect in the process of decomposing signal by traditional local mean decomposition(LMD)method,an extreme value positioning waveform continuation method is proposed to improve the decomposition effect of LMD.By analyzing the whole waveform,the method determines the variation rule and the interval rule of the extremum point to locate the target extremum point,and use it as the extremum extension scheme of the left and right endpoints.In addition,the effective Product Function(PF)components generated by the improved LMD decomposition are denoised by the new threshold wavelet packet method.Finally,the improved LMD-wavelet packet method was used to decompose and de-noise the vibration signals of mechanical parts of CNC lathe under two states.(2)Feature extraction of operating state of mechanical components based on LMD-FastICAIn order to separate the disturbance signal in the vibration signal of the mechanical components,the signal separation is carried out by means of Fast Independent Component Analysis(FastICA)algorithm.The effective PF component decomposed by the improved LMD method is used to expand the number of observed signals.The effective PF component after denoising and the original signal after denoising are used as input signals,and signal separation is carried out through FastICA algorithm.The frequency spectrum is analyzed for the separated components,and the frequency is used as the description object of the state feature,and the running state features existing in each separated component are extracted respectively.First,the effectiveness of the entire signal separation is verified by the simulation signal,and then the vibration signal of the mechanical components collected on site is analyzed to extract the operating state features of the mechanical components.(3)Mechanical fault diagnosis based on information fusionIn order to avoid the diagnostic bias caused by single evidence diagnosis,the multi-source evidence information is fused by means of information fusion,and the mechanical fault diagnosis is made according to the fusion results.The operating state feature matrix is constructed by using the extracted operating state features of mechanical components.Through the analysis of historical information,combined with expert experience,the fault state feature matrix of CNC lathe is obtained.The basic trust distribution function is obtained by analyzing the distance between the operating state feature and the fault state feature.In order to avoid the possible influence of false evidence in the basic trust distribution function on the diagnosis results,a normalized Euclid distance is proposed,and the curve trend analysis is used to identify false evidence.Substitute evidence can be obtained by constructing contradictory evidence of false evidence,and the basic trust distribution function can be updated by using the substitute evidence.The D-S evidence synthesis rule is used for multi-evidence fusion,and the fault diagnosis of CNC lathe mechanical parts in two operating states is carried out,and the support vector machine algorithm is used for diagnosis verification.