Novel 2D representation of vibration for local damage detection

Abstract

In this paper a new 2D representation for local damage detection is presented. It is based on a vibration time series analysis. A raw vibration signal is decomposed via short-time Fourier transform and new time series for each frequency bin are differentiated to decorrelate them. For each time series, autocorrelation function is calculated. In the next step ACF maps are constructed. For healthy bearing ACF map should not have visible horizontal lines indicating damage. The method is illustrated by analysis of real data containing signals from damaged bearing and healthy for comparison.

Keywords

• local damage detection
• bi-frequency plane
• partial autocorrelation function
• fault pattern surface