Spindle Bearings Fault Diagnosis Technique Based on Integration of Zero Resonator Frequency Filter and Discrete Wavelet Packet Transform

Spindle bearing is one of the machine elements in the spindle that is mostly vulnerable to failure. Its failure may result into total machine tool breakdown and other associated catastrophic consequences. An early identification of the failure is emphasized for reducing extreme damages of the machine tools. This study develops a novel hybrid algorithm combining the Zero Resonator Frequency Filter (ZRFF) and the Discrete Wavelet Packet Transform (DWPT) for early spindle bearing fault detection and diagnosis. The integrated method uses the ZRFF as the first level of de-noising the vibration signals and the DWPT for clear extraction of crucial periodic impulse features that are not easily visible from the first de-noising. The obtained frequency spectrum gives a dominant peak line which corresponds to the fault frequency of interest. An optimum wavelet decomposition level is also determined using the minimum Shannon entropy criteria. The experimental datasets from Case Western Reserve University (CWRU) and simulated signal were used to test the validity of the proposed algorithm. The proposed algorithm had superior performance in terms of computational efficiency (45s) and high classification accuracy of the bearings faults when compared with other methods.