Lanczos decomposition based sub-Gaussian component extraction method for wind turbine transmission chain's health...

Transmission chains are an essential part of wind turbines, and their vibration signals often carry a wealth of health information that can be used for condition monitoring. When the signal-to-noise ratio (SNR) is high, existing monitoring methods always demonstrate good performance. Unfortunately, the operating conditions of transmission chain are often extremely harsh, implying that its vibration signals carry a significant amount of noise, resulting in a low SNR. This poses challenges for health monitoring. During this article, a health monitoring method based on Lanczos decomposition for extracting sub-Gaussian distribution components from signals is proposed. Taking into account that fault signal indicative of system health generally approach to a sub-Gaussian distribution, the suggested method employs Lanczos decomposition-based filters to iteratively decrease the signal variance, thereby isolating components in the signal that tend towards a sub-Gaussian distribution. As a result, fault characteristics can be obtained. Both simulation and experimental outcomes confirm that the proposed technique demonstrates remarkable resilience to white noise and harmonic interference signals, and its effectiveness in monitoring exceeds that of current techniques.