Early Fault Detection of Rolling Element Bearings Using Wavelet Packet Energy Ratio (WPER)

Abstract

Early detection of defects in high-speed ball bearings is critical for preventing machinery failures and unplanned downtime. This study introduces a wavelet-packet–energy-based diagnostic method for identifying incipient bearing faults. Five real wavelets were evaluated, and the reverse Biorthogonal 5.5 wavelet was selected as the optimal basis according to the maximum energy-to-Shannon-entropy ratio criterion. Vibration signals from healthy and faulty bearings were acquired and decomposed using Wavelet Packet Decomposition (WPD) into narrow frequency sub-bands. An energy-ratio metric—defined as the energy of the faulty condition relative to the healthy condition for each sub-band—was computed to quantify fault severity. The method was experimentally validated using vibration data collected under localized single-defect conditions, including outer-race, inner-race, and rolling-element faults, at multiple rotational speeds. Results demonstrate that the proposed WPD-based energy-ratio approach offers superior sensitivity and accuracy for early fault detection compared with conventional vibration-analysis techniques.