Order components extraction of wind turbine gearbox main shaft based on SET joint NLMS filtering

The wind turbine gearbox is a critical transmission component in wind power generation systems, with its operational condition directly affecting the safety, reliability, and stability of the entire system. Given that wind turbine gearboxes operate continuously under complex conditions, including variable rotational speeds, high-speed operations, and coupled alternating loads, the risk of failure significantly increases. Therefore, developing efficient and accurate condition monitoring (CM) technologies is essential to ensure safe wind turbine operation and power generation stability. Order analysis (OA), a widely-used technique in rotating machinery fault diagnosis, plays a pivotal role in fault feature extraction under these complex conditions, thanks to its capability to mitigate speed fluctuation interference. However, traditional OA methods face challenges, such as order confusion during unsteady processes like sudden speed changes and start-stop operations, primarily due to inadequate anti-interference performance. Additionally, target order signals are often overwhelmed by background noise, leading to distorted fault feature extraction that fails to meet practical engineering standards. To address these issues, this paper proposes a combined OA method based on Normalized Least Mean Square (NLMS) filtering and Synchroextracting Transform (SET). First, NLMS filtering is applied to reduce noise in the vibration signals, effectively suppressing interference. Then, SET is used on the filtered signals to generate high-resolution time-frequency (TF) distributions. Finally, accurate separation and extraction of order components is achieved through a multi-ridge detection algorithm. This combined NLMS-SET approach satisfies the dual requirements of noise suppression and time-frequency feature preservation: the NLMS filter adapts to eliminate random noise and background interference, avoiding the attenuation of useful order components seen with traditional filters, while the SET algorithm generates high-aggregation time-frequency distributions without cross-term interference. To validate the effectiveness of the proposed method, comparative analyses are conducted using simulated signals, semi-physical simulation signals, and experimental signals. Results show that the method exhibits excellent performance in order feature extraction under strong noise and unsteady working conditions. Based on this, a LABVIEW-based OA system is developed to automate order feature extraction and visual display, providing reliable technical support for CM and fault diagnosis of wind turbine gearboxes.