Vision-Based Monitoring of Wind Turbine Tower-Blade Coupling Dynamics for Fault Diagnosis

Wind turbine blades frequently experience unexplained structural failures, contributing significantly to operational downtime and maintenance costs. Conventional vibration monitoring techniques, dependent on manual method or contact-based sensors, exhibit inherent limitations for rotating blade applications due to slip-ring reliability issues and installation constraints. Furthermore, monitoring these large-scale rotating blades poses substantial technical challenges. This study proposes a remote vision-based method for fault diagnosis by analyzing blade-dynamics information manifested in tower vibrations. A non-contact diagnostic framework exploits tower-blade coupled dynamics, utilizing a simplified 5-DOF model to demonstrate how mass imbalance and torsional anomalies induce quantifiable tower oscillations. The methodology employs subpixel-accurate edge localization, incorporating Gaussian filtering, gradient-based edge detection, and signed-gradient centroid refinement, to extract vibration signatures from video data. Experimental validation confirms effective detection of critical faults including mass imbalance and large-angle torsional deformation (identified via spectral shifts and rotational frequency reduction). Results correlate strongly with reference sensors, providing a cost-effective in-situ solution that overcomes contact-sensor limitations for rotating blade monitoring.