EMD Based on Time-Sequence and Window Function and its Application in Diagnosis of Machinery Faults

Abstract

To solve the end effect occurring in empirical mode decomposition adopted in the course of decomposition, we propose an improved method on the basis of time-sequence analysis and cosine window function. First, the ARMA (Autoregressive Moving Average) of time-varying parameter is adopted to extend signals, and thus the extended data can be smoothly connected with the original signal at the end. Second, the extended signals are processed with cosine window, so that the extended errors will exert no impact on the existing data. Finally, the signals processed as above mentioned will be decomposed with EMD to confine the end effect to the ends of the signal. The simulation and fault signal analysis prove that the proposed method can effectively reduce the impact of the end effect and be applied in rotating machinery fault diagnosis.