Research On Gear Fault Diagnosis Based On Local Mean Decomposition

Gear is one kind of general part of the common mechanical devices, which is responsiblefor the connection and transfering power. Due to factors such as poor working conditions, thegear is very prone to failure, and which directly affects the operational status of machineryand equipment. So, gear fault monitoring and diagnosis is of great significance.The key of gear fault diagnosis is to extract the fault feature, and to make a judgment onthe fault location and the damage degree. As gear vibration signal is closely related to the gearstate, and which often contains a great deal of fault information, for the non-stationarity,modulation and multi-component properties of gear fault signal, in this paper, an approachbased on LMD (local mean decomposition) is applied to gear fault diagnosis, and the conceptsof the energy operator demodulating,cycle frequency and spectrum kurtosis based on LMDare proposed according to the characteristic of gear fault vibration signal. And further more,local characteristic-scale decomposition based on LMD has been proposed. The main researchcontents of this paper are as follows:1. Aiming at the characteristic that gear fault vibration signal is composed of severalAM(amplitude modulation)-FM(frequency modulation) components, the LMD method isapplied to gear fault diagnosis. LMD is a new adaptive time-frequency analysis method. Thegear vibration signal can be decomposed into a set of single-component signals by usingLMD, so LMD is very suitable for the gear fault diagnosis. From the analysis on thesimulation signal and actual gear signal, it can be concluded that better time-frequencydistribution can be obtained by LMD method because of its better adaptability andtime-frequency clustering.2. Aiming at the modulation characteristic of gear fault signal, and the fault characteristiccan be extracted from the modulation signals, the energy operator demodulating and cyclefrequency approach based on LMD is applied to gear fault diagnosis. By using LMD methodthe gear signal can be decomposed a series of product functions (PF), then the amplitudemodulation and frequency modulation information of the relative PF can be got by using theenergy operator demodulating and cycle frequency demodulating respectively, which canextracted the fault characteristics. From the comparison with the direct LMD method, it canbe conclude that the energy operator demodulating and cycle frequency approach based onLMD is effective and superior for gear fault diagnosis.3. Aiming at the defect that the filter parameters, for seeking the best band whichincludes the fault information, can’t be selected objectively and effectively before the demodulation analysis of signal, an spectrum kurtosis approach based LMD is applied to gearfault diagnosis. In this approach, firstly the kurtogram can be obtained according the signaltime-frequency distribution, then the best filtering band can be selected by the max kurtosisand the fault characteristics frequency of the signal can be extracted by envelope analysis forfiltered signal. The analysis results show that the spectrum kurtosis approach based on LMDcan be effectively used to gear fault diagnosis.4. Aiming at that LMD is not intended for on-line monitoring because it iscompute-intensive. Based on the definition of the non-component signal, namely intrinsicscale component (ISC) whose instantaneous frequencies own physical sense, a newlyself-adaptive signal decomposition method, the local characteristic-scale decomposition (LCD)is proposed in this paper. By using LCD, any complicated signal can be decomposed into anumber of ISC whose instantaneous frequencies own physical sense, it’s an adaptive method.And the LCD method is compared with the empirical mode decomposition (EMD) and thelocal mean decomposition (LMD) method; then the results show the superiority of the LCDmethod, this method has no envolop error while has faster operation. And in this paper, theLCD method has been applied to gear fault diagnosis successfully.