Proportional Chirplet basis transform for rotating machinery vibration signal analysis without prior knowledge

For the task of multicomponent nonstationary signal analysis in mechanical fault diagnosis, this article provides a new time-frequency analysis method with high energy concentration and resolution. The method takes full consideration of the synchronous time-varying property among the vibration signal components. To ensure the adaptivity for nonstationary signals, theoretical parameter estimate is first derived from the phase model and further optimized based on ridge energy weighting. And then, the adaptive Chirplet basis transform is proposed to generate optimal time-frequency results for each synchronous component. For the purpose of retrieving all those components from noisy background, the proportional extraction operator is constructed and effectively enhances the time-frequency readability. By simulating gearbox fault signal and multicomponent nonstationary signal, numerical validation evaluates the performance in energy concentration and noise robustness. Applying the proposed algorithm to three real cases, the experimental results demonstrate the effectiveness in rotating machinery fault diagnosis.