Research On Weak Fault Feature Extraction Of Transmission Under Compound Fault

Transmission faults are mainly divided into two categories.One is failures such as pitting and wear that occur during long-term operation,and the other is failures caused by processing and assembly errors during production and manufacturing.The second type of fault not only affects the occurrence and development of the first type of fault,but with the development of high-speed,ultra-quiet transmission,effective detection of such faults is of great significance for improving the safety and reliability of the transmission.However,the structure of the transmission is complex,and parts interact during operation,especially when the faults of various positions and degrees are compounded.In addition,due to various background noises,weak fault features are difficult to extract.In view of the above problems,this article takes a certain type of transmission as the research object,analyzes its abnormal noise problem,and studies the transmission fault weak feature extraction method under compound fault condition by deconvolution and denoising.The main research contents of the paper are as follows:By analyzing the vibration response mechanism of transmission gears and bearings and the characteristics of different fault signals,a transmission composite fault simulation signal containing the components of the impact fault is established.A transmission test bench was set up to collect sound and vibration signals under different working conditions in different gears of the faulty transmission.The sound pressure signals were used to preliminarily analyze the sources of transmission abnormal sounds and to arrange measurement points.The time-domain statistical features of the faulty transmission based on the vibration signal are studied with the change of the speed.At the same time,the evaluation of the measurement point is combined with the permutation entropy,and the appropriate measurement point is selected to pick up the vibration signal to provide reliable data for the extraction of weak fault features.Based on the in-depth study of deconvolution theory,the fault feature extraction effects of three methods,MED,MCKD,and MOMEDA,are studied through composite fault simulation signals.MED has good effect on the extraction of strong impact faults,but fails to extract the weak impact faults under compound faults.MCKD can extract weak shocks in compound faults.Based on this,MCKD combines autoregressive models to extract transmission gear fault characteristics.However,MCKD needs to set proper parameters to generate local optimal filters by iteration,and the computational efficiency is low.MOMEDA obtains the optimal filter in a non-iterative manner,which has high computational efficiency.By calculating the multipoint kurtosis spectrum of the vibration signal,MOMEDA can track the source of the impact fault,but the trace of the fault source is not obvious in the strong noise.This paper further combines the spectral kurtosis to carry out the band filter to improve the tracking effect.In order to solve the problem that the deconvolution period is difficult to be determined and the deconvolution method is invalid,the convolution period of the solution is determined by the order resampling conversion of the non stationary signal in the time domain as the stationary signal in the angular domain.The effectiveness of the method is verified by simulation and transmission data.