A Novel Adaptive Synchrosqueezing Transform Based on Enhanced STFT Scheme for Wind Turbine Bearing Fault Diagnosis

Accurate estimation of the instantaneous frequency (IF) of multi-component non-stationary signals is crucial for the fault diagnosis of wind turbine transmission systems under variable speed. Current methodologies exhibit several key limitations, such as unsatisfied accuracy of instantaneous parameter estimation due to the inadequate time–frequency representation, poor time–frequency energy concentration and multimode separation effects caused by a single post-processing technique, and the inability of non-adaptative window parameters to dynamically accommodate the local time-varying frequency characteristics of signals. These deficiencies degrade time–frequency plane interpretability and the accuracy of extracting time-varying fault features. To address these challenges, this paper proposes a novel Improved Short-Time Fourier Transform-based Adaptive Synchrosqueezing Transform (ISTFT-ASST) and demonstrates its application in wind turbine bearing fault diagnosis. Firstly, to achieve the precisely parameter estimation of multi-component non-stationary signals, an improved Short-Time Fourier Transform (STFT) sparse observation model integrated with the Finite Rate of Innovation (FRI) is innovatively constructed. Secondly, Synchrosqueezing Extraction Transform (SET), a time–frequency post-processing technique, is incorporated to further enhance the time–frequency concentration. Finally, drawing on the idea of local Rényi entropy-based window parameter estimation, an adaptive Synchrosqueezing algorithm is designed to dynamically accommodate the time-varying frequency characteristics. Through the analysis of numerical simulation signals and measured wind turbine vibration signals, the effectiveness of the proposed ISTFT-ASST in extracting time-varying features is verified, establishing a novel approach for wind turbine fault diagnosis.