AR Model Based LMD Applied For Diagnosising Fault Of Rotating Machine

As the development of science and technology, the work intensity of the rotating machine is increasing. It is necessary for rotating machine to do fault diagnosis and condition monitoring. In this way it can ensure safe, reliable and efficient of this machine. What’s more, it also can avoid great economic losses and catastrophic. This paper focuses on the vibration signal of rotating machine. Non-stationary signal based on the local mean decomposition, AR model and prediction method of AR model based on the local mean decomposition are important parts of this paper.This paper introduces the form of rolling bearing fault and diagnosis method of vibration signal. Then it expounds natural frequency and characteristic frequency of rolling bearing. All of this is the foundation of analysising rolling bearing.Then it introduces Fourier transform, Wigner-Ville distribution, Wavelet transform and Hilbert-Huang transform method, and compares the advantages and disadvantages of each method, then proposes that it needs a new time-frequency analysis methods to deal with rotating machine signal.The local mean decomposition is a new adaptive time-frequency analysis method. Then LMD is improved with mirror extension and cubic spline method. It gets two PFs and surplus component after analyzing the FM-AM signal. Compared with the original one, the accuracy of improved LMD is increased by30.56%,10.02%and28.32%. The validity of this method is demonstrated by diagnosing rolling bearing outer ring of rotating machine with fault. Through the analysis:In rotating machine, there will have the outer ring fault characteristic frequency and rotational frequency (or multiple of rotational frequency) when it occurs fault in rolling bearing outer ring.After the study of AR model, MA model and ARMA model, the AR model is applied to linear predict the running state of rotating mechanical equipment. The method estimates AR model under the least square method and fixes the model order through the FPE, AIC and BIC. The application has been successfully applied to the field of the operating condition prediction in rotating mechanical.Based on the LMD method and AR method, it establishes AR model with product function from LMD. Then it establishes the template eigenvector with model parameters and the mean square deviation from AR model. The operating condition of the rotating machine is successfully identified by the Mahalanobis distance function.