Bearing Fault Diagnosis Based on Weighted Phase Space Reconstruction

The vibration signals are often non-linear and non-stationary when the bearings are in failure. This paper will adopt phase space reconstruction algorithm and make bearing vibration time series extend to high-dimensional phase space. Some characteristics are difficult to identify in one-dimensional time series, but we can transform them into many different attractors in high-dimensional phase space, in this way, the background signal and noise would be decomposed into different sub-space. Through the phase space projection method and after the window operation, the different phase points in phase space could be projected to the one-dimensional space, so it can be effective to separate the background signal and noise components. Weighted combination of noise reduction algorithm for phase space reconstruction and the Hilbert transform to extract the low-frequency characteristics of the background signal realize bearing fault identification. The experimental data shows that the algorithm has an effective result in bearing fault identification.