The Research Of Gear Fault Diagnosis Based On Delayed Autocorrelation And LMD

When gear suffers fault, its fault signal is multicomponent signal which has both AM and FM feature. The fault feature information can be extracted through demodulation of the signal. However, traditional demodulation approaches are only suitable for the demodulation of single modulated signal. When they are used for the multicomponent modulated signal, the result will not precise. Meanwhile, gear fault signal often has lots of noise, which makes a huge influence on the accuracy of fault diagnosis. Therefore, aiming at how to extract fault feature information from gear fault signal with noise disturbance precisely, this paper makes following studies:First, this paper makes some study to the delayed autocorrelation function of multicomponent AM-FM signal. And the result shows that the delayed autocorrelation function got from the original multicomponent AM-FM signal is still multicomponent AM-FM signal whose each single component still possesses the modulation information of original signal and that after the process of delayed autocorrelation the power of modulation sideband of its spectrum is effectively diminished so that we can distinguish the carrier frequency of original signal from its spectrum more precisely.Second, it makes some comparisons for common signal decomposition approaches including Empirical Mode Decomposition and Local Mean Decomposition and does some research about their applications to the decomposition and purification of multicomponent AM-FM signal. And the result shows that Local Mean Decomposition is much more suitable for the decomposition of multicomponent AM-FM signal and its decomposition and purification result of multicomponent AM-FM signal is much better. In addition, the result also clarifies that compared with the imf component got by EMD, the PF component got by LMD has a better fit with Hilbert envelope demodulation. What’s more, the accuracy of Local Mean Decomposition can be easily influenced by noise. In order to get a precise decomposition result, it is necessary to combine Local Mean Decomposition with other effective noise reduction approach.Then, it combines delayed autocorrelation with Local Mean Decomposition and proposes an effective fault diagnosis approach. Firstly, this approach executes delayed autocorrelation to the gear fault signal in order to reduce the disturbance of the noise. And then, it executes Local Mean Decomposition to the multicomponent signal to acquire a series of purified PF components. After that, it implements Hilbert envelope demodulation and FFT analysis to each purified PF component and combines each divided spectrum to get united envelope demodulation spectrum and united amplitude spectrum so that precisely extract gear fault feature information.In the end, this paper makes some test to the approach by simulated experiment and takes some comparisons among this approach, Hilbert envelope demodulation and delayed autocorrelation demodulation. The result shows that this approach can successfully extract the gear fault feature information from multicomponent gear fault AM-FM signal with noise disturbance and the diagnosis result is more exact.