Bearing Condition Monitoring via Multiple Instantaneous Frequency Path Extraction from Enhanced Time Frequency Representation

Short Time Fourier Transform (STFT) has been widely adopted for instantaneous frequency (IF) ridge extraction in bearing fault diagnosis under variable speeds. However, the readability of STFT-resulting time frequency representation (TFR) depends on how close the frequencies matching between the transforming bases and signal components. However, frequencies of STFT bases are time-invariant and thus do not match IFs of the vibration signal under variable speed conditions. Then, energy scattering may occur, resulting in IF ridge smearing on TFR for the multi-component vibration signal. To address this problem, transforming bases with time-varying frequencies are developed and a more energy concentrated TFR can be obtained. To get bases with time-varying frequencies, fast path optimization (FPO) is adopted to pre-estimate the IF of the analyzed signal. The TFR is then enhanced, from which multiple IF can be extracted and bearing health condition can be diagnosed. The effectiveness of the proposed approach is validated by experimental signals.