Study On The Extraction Method Of Gearbox Fault Based On EMD

With the development of the industry, more and more attention is given for gear box. Automobile, locomotive and marine operations are inseparable from the gear box. Gear box is usually composed of shaft, gear, bearing and gearbox parts. The health of gear, shaft, and bearing play a key role in the normal operation of machinery and equipment.It raises the main job by introducing the traditional vibration method of this paper. First introduces the time-frequency analysis method, On the basis of this method, apply the wavelet decomposition method into the signal de-noising, Compared with the Fourier and wavelet decomposition, the later has obvious advantages. According to the characteristics of mechanical fault data of the large amount of information, this paper introduces the feature extraction method of EMD. The fourth chapter and the fifth chapter respectively introduce the PCA and LPP method into the actual fault diagnosis.The outline of the work is as follows:1. This paper analyses the application of traditional time-frequency analysis method in the vibration signal, and it is pointed out the key problem that it cannot handle the non-stationary and nonlinear signal. A new theory of time frequency analysis method, Hilbert-Huang Transform (HHT), is introduced. Do the jobs of the EMD for analog signal, at the same time, do the job of Hilbert-Huang transform; it gets the Hilbert marginal spectrum and the HHT spectrum.2. The paper introduces the SVM into the fault diagnosis of gear box, constructs the SVM network model. Introducing the definition of the intrinsic energy entropy of EMD, when the gear box fault occurs, the vibration signal energy will change. The different frequency component leads to the different fault forms, so it can make the SVM network to classify the fault according to the different energy entropy.3. Because the fault data is difficult to solve, this paper introduces a feature extraction method for extracting time domain features based on EMD—The principal component analysis method. The PCA is based on the idea of finding the linear minimum variance to simplify the data. In engineering practice, often requires the fault monitoring and tracking system. Therefore, this paper introduces the intelligent analysis theory—Neural network. This paper Combine the PCA with spectrum to analyses the fault of the gear box. After the analysis, get the twelve components. Making the first4groups of vibration data (1-4vibration data) as the subset samples, the rest8groups of vibration as the training subset sample. Then train and predict the BP network. The result of the correlation of target output next to actual is0.913. The prediction is consistent with the actual.4. Introducing the manifold method of locality preserving projection (LPP) method into the gear box fault diagnosis. To make the extracted HHT spectrum matrix decompose of singular value, and diagnose the fault. Then use LPP to classify the fault overlapping signals for dimensionality reduction, it is feasibility of this method used for the fault pattern recognition problem, Using this method into the fault diagnosis of bearing can effectively classify the fault of the gearbox bearing. At the end of this paper, according to the overall arrangement of the article, several vibration analysis modules are put forward.