Research On Fault Feature Extraction Method Of Gearbox Based On Maximum Overlapping Discrete Wavelet Packet Transform

With the development of modern production and the progress of science and technology,the mechanical equipment is developing towards the direction of structural complexity and automation.The working intensity of the equipment is increasing,and the production efficiency and automation degree are getting higher and higher.At the same time,the equipment is more complex,and the connection between various parts is becoming more and more intensive.As a key part of mechanical equipment used to connect and transmit power,gearbox fault will directly affect the safety and reliability of the equipment,and may also cause economic losses or even casualties.Fault feature extraction is one of the core contents of mechanical equipment fault diagnosis.And how to extract the weak fault information hidden in the signal from the background of strong noise is the hotspot and difficulty in the field of fault diagnosis.In this paper,the gearbox is taken as the research object,and the maximum overlap discrete wavelet packet transform(MODWPT),bilinear time-frequency distribution and spectral kurtosis are used as the analysis tools.The feature extraction method of gearbox multistage gear transmission low-speed gear fault is mainly studied,and the effectiveness and feasibility of the proposed method are verified by simulation and experimental analysis.The main contents are as follows:(1)The basic theory of modwpt is introduced.Based on the introduction of modwpt algorithm,this paper focuses on the advantages of modwpt compared with EMD in the processing of complex multicomponent signals.The simulation results show that the modwpt method is superior to the EMD method in terms of anti end-point effect and modal aliasing,as well as the accuracy of frequency expression in Hilbert spectrum and marginal spectrum.(2)In order to solve the problem that multi-stage gear transmission is easy to be disturbed by noise and it is difficult to extract low-frequency weak fault features,a method of gear low-frequency fault feature extraction based on modwpt and Choi Williams distribution is proposed in combination with the respective advantages of modwpt and bilinear time-frequency distribution.Firstly,the complex vibration signal is decomposed into several components with instantaneous frequency and amplitude by modwpt method.Then,the appropriate components are selected according to kurtosis criterion.Finally,the remaining components are analyzed by CWD,and the fault information of gear is extracted successfully.(3)In order to solve the problems of low frequency weak fault feature extraction in gearbox multi-stage gear transmission,and the filter parameters of traditional resonance demodulation method need to be determined artificially in advance,a low frequency fault feature extraction method based on modwpt and spectral kurtosis is proposed.Firstly,the vibration signal is decomposed into several components by modwpt method,then the appropriate components are selected for fast spectral kurtosis calculation and band-pass filtering.Finally,the envelope demodulation analysis of the filtered signal is carried out to effectively extract the low-frequency weak fault features.