Use of information embedded in vibroacoustic signal to crack evolution tracking of gear failure

摘要

The paper describes a concept of signal processing which is used to develop new failure tracking ability. Fault identification is achieved by applying proper preprocessing. Three analytical forms of raw vibration signal were investigated: synchronous averaging signal with rotation compensation, differential and residual. To define a vector that characterize trend of failure evolution, two techniques were confronted: empirical mode decomposition (EMD) and blind equalization (BE). Performance those methods comparatively to preprocessing technique were considered. Presented method is validated by data collected during simulation where gear tooth fatigue cracks grew over the time and data from real machine. Both experimental vibrations were preprocessed and parallel decomposition on finite sum of signal components called intrinsic mode functions IMF's and blind equalization were processed. Kurtosis value of selected IMF and equalized signal were calculated for early detection of fault. Damage identification on both experimental signals yield low energy information appears as a monotonic trend describing evolution of tooth defect.