Diagnosis of bearing fault signals based on empirical standard autoregressive power spectrum signal decomposition method

Abstract

Signal decomposition is an essential tool for the time–frequency analysis of bearing fault signals. Methods for extracting effective fault characteristic information from bearing vibration signals have received increasing attention from researchers. This paper proposes a novel signal decomposition method, called empirical standard autoregressive power spectrum decomposition (ESARPSD), to diagnose bearing faults. First, the normalized autoregressive power spectrum of the bearing fault signal is obtained and its bounds are derived using the lowest minima principle. The decomposed component signals are then filtered through a zero-phase filter bank. Each decomposition component is then demodulated and the respective envelope spectrum is observed to determine the corresponding fault frequency. Zero-phase filter banks are used to deal with the problems of noise interference, which makes decomposition difficult, and frequency aliasing, which occurs when the signal-to-noise ratio is low. Moreover, through normalized autoregressive power spectrum and resonance demodulation techniques, adaptive signal decomposition can accurately separate the target high-frequency vibration signals and detect the fault frequency. The accuracy and performance of the proposed ESARPSD method were validated using simulated signals and actual experimental data. The results demonstrate that this method can effectively decompose bearing fault signal and identify all fault characteristics.