Adaptive Fast Chirplet Transform and Its Application Into Rolling Bearing Fault Diagnosis Under Time-Varying Speed Condition

Tacholess order tracking (TLOT) is a commonly used technique for the fault diagnosis of bearing with time-varying speed, in which time-frequency analysis (TFA) is a key step for estimating the rotation frequency. The current TFA methods are either weak in energy concentration or have high computational complexity. To this end, an adaptive fast chirplet transform (AFCT) based on the adaptive optimal search angle band is proposed in this article. The proposed method uses the modulation operator of synchrosqueezed transform to adaptively optimize the search band of the frequency modulation (FM) parameters, which overcomes the problem of low computational efficiency in chirplet transform (CT) and its variants. Moreover, a novel time-frequency energy concentration evaluation index mean-energy-to-peak ratio (MEPR) is proposed. The simulation result shows that the proposed method has both good time-frequency concentration performance and low computational complexity. In particular, its calculation speed is much faster than other TFA methods with high time-frequency resolution. The proposed method is successfully applied to estimate the rotation frequencies from the fault vibration signals from a test rig and a civil aircraft engine. The comparative results verify the comprehensive advantage of the proposed method in ridge extraction accuracy and computational efficiency compared to the existing classical TFA methods. With the proposed method, the fault features of rolling bearings under time-varying speed can be effectively extracted.