Rolling Bearing Fault Analysis by Interpolating Windowed DFT Algorithm

This paper focuses on the problem of accurate Fault Characteristic Frequency (FCF) estimation of rolling bearing. Teager-Kaiser Energy Operator (TKEO) demodulation has been applied widely to rolling bearing fault detection. FCF can be extracted from vibration signals, which is pre-treatment by TEKO demodulation method. However, because of strong noise background of fault vibration signal, it is difficult to extract FCF with high precision. In this paper, the improved algorithm of rolling bearing fault diagnosis is analyzed. Based on the envelope analysis by TKEO demodulation, it combines zero padding technique and the Improved Iterative Windowed Interpolation DFT (IIWIpDFT) algorithm to correct demodulated signal. Experimental result shows that the proposed algorithm decreases Root Mean Square Error (RMSE) of FCF(inner race) form about 2Hz~5.5Hz to about 0.5Hz for short data length, the same treatment also decreases RMSE form about 1.1Hz~3Hz to about 0.4~0.5Hz for longer data length in most cases. Meanwhile, the RMSE of FCF (outer race) improved 2.3 to 84.5% as compared to the application of traditional TEKO demodulation alone.