A Simulation Based Band-Pass Filter to Improve the Polynomial Chirplet Transform in Fault Detection

Abstract

Bearings are the basic components in mechanical or hydraulic power transmission systems. Defects in bearings will lead to a sudden stoppage of machines. In this paper, we develop a new fault detection method based on the numerical simulation determined band-pass filter and polynomial chirplet transform (PCT). Using band-pass filter, the sub-signal containing the information of fault feature is preserved and further decomposed by PCT to obtain time-frequency spectrum with higher signal to noise ratio (SNR). To design the efficient band-pass filter, the input center frequency is predetermined by finite element simulations. To improve the performance of the traditional chirplet transform (CT), a polynomial kernel function is employed to construct one of the parameterized time-frequency analysis methods, i.e., PCT. PCT reconstructs the time-frequency spectrum along the frequency trajectory so that the energy distribution on the spectrum is concentrated around the real instantaneous frequency of the signal. Thus, the PCT generates high time-frequency resolution, which is suitable for weak fault detection. The effectiveness of the proposed method in fault diagnosis is demonstrated by applications for vibration signal collected from bearings in both mechanical and hydraulic power transmission systems.