Planetary gear faults detection in wind turbine gearbox based on a ten years historical data from three wind farms
Dynamic principle component analysis, Fault detection, Feature extraction, Support vector machine, Wind turbine gearboxes
Gear faults contribute to a significant portion of failures in wind turbine system. As such, condition monitoring and fault detection of these components assist in maintenance scheduling; hence, preventing catastrophic failures of the gearbox. This paper introduces a new hybrid fault detection approach to detect gear faults in wind turbines. to accomplish this task, vibration signals are collected and used to extract various time-domain features. Next, a Dynamic Principle Component Analysis (DPCA) is adaptively employed to identify failure dynamics by reducing the time-domain feature dimension. Following that, a Support Vector Machine (SVM) is implemented to detect and isolate gear faults. Experimental test studies with ten-year historical data of three wind farms in Canada are conducted. Test results indicate that the proposed hybrid approach performs superior compared to DPCA using Multilayer Perceptron (MLP) Neural Networks (NNs).
Kordestani, M.; Rezamand, M.; Orchard, M.; Carriveau, R.; Ting, D. S.K.; and Saif, M.. (2020). Planetary gear faults detection in wind turbine gearbox based on a ten years historical data from three wind farms. IFAC-PapersOnLine, 53, 10318-10323.