A feature-guided adaptive frequency band optimization framework for fault diagnosis of wind turbine pitch bearings

• A novel Feature-Guided Adaptive Frequency Band Optimization (FAFBO) method is proposed. • Encoder zero-position-based signal reconstruction eliminates nonstationary disturbances. • The Refined Fault Harmonic Index (RFHI) enhances fault feature detection under low SNR. • A two-stage coarse-to-fine grid search optimizes the demodulation frequency band. Ensuring the reliability of pitch bearings is critical for the safe and efficient operation of wind turbines within modern power systems. However, fault diagnosis under low-speed reciprocating conditions remains extremely challenging due to weak fault-induced impulses and non-stationary dynamics. This paper proposes a Feature-Guided Adaptive Frequency Band Optimization (FAFBO) for fault diagnosis of low-speed pitch bearings. In the proposed diagnostic strategy, a signal reconstruction method is developed to eliminate the effects of reciprocating dynamics and extract relatively stationary signals for further analysis using the encoder zero-position alignment. The reconstructed signals are then analyzed by the Refined Fault Harmonics Index (RFHI) to suppress reversal impacts and stitching artifacts and extract effective fault features through two-stage coarse-to-fine grid search optimization algorithm. Experimental studies demonstrate that the FAFBO is effective and robust for pitch bearing fault diagnosis at low signal-to-noise ratios. The proposed framework provides a reliable and computationally efficient tool for condition monitoring and preventive maintenance of wind turbine pitch systems, contributing to improved reliability of wind energy generation.