Presentations

Abstract

This study elaborates on developing an automated pipeline for the health monitoring of offshore wind turbine drive trains and fault detection on a wind turbine gearbox. Vibration signals acquired via custom hardware are employed to track the health condition of the gearboxes. Due to the complex nature of the vibration signals emitted from a gearbox equipped with numerous components and its dependence on the varying operating conditions of the wind turbine, condition monitoring remains a challenging task. A hybrid pipeline with signal processing and data-driven techniques is proposed to address this complexity. The quality of the vibration signals was assessed using the SCADA measurements. Signals are further processed to estimate the wind turbine's instantaneous angular speed for angularly resample the vibration signals for speed independence. Pre-processing is performed on the speed-independent signals to evaluate time and spectral indicators for the vibration signals and their envelopes. The indicators undergo a machine learning algorithm to distinguish the healthy state of the machine from a faulty one. A faulty turbine is discernable compared to other turbines on the spectral coherence maps estimated from the speed-independent signals as higher amplitude frequency bins are detected around the characteristic fault frequency. Furthermore, the machine learning algorithm can detect the faulty turbine based on the statistical indicators. Results demonstrate that the fully automated hybrid pipeline can effectively be used for fleet-based health tracking of the offshore wind turbines' gearboxes.