Manas Ranjan Sethi and Sudarsan Sahoo
Wind turbines are increasingly deployed in remote onshore and offshore locations due to abundant wind resources and the benefits of mitigating land use with visual impact concerns. The crucial role played by a wind turbine's rotor blades in converting wind energy into electricity cannot be overstated. Any damage to these blades directly impacts power generation and can lead to turbine shutdowns. In addition to the ongoing efforts to reduce the cost of wind energy, there is a growing emphasis on condition monitoring as a promising solution to address maintenance issues. Regular detection of blade faults can reduce downtime and enhance overall efficiency. Machine learning approaches, particularly pattern recognition systems, prove effective in identifying and diagnosing faults in wind turbine blades. This research study aims to demonstrate the effectiveness of machine learning models in detecting blade faults by analyzing filtered and unfiltered vibration signals. Among the tested models, the logistic regression model utilizing resample filter-based vibration signals achieved the highest classification accuracy, reaching an impressive 99.75% within a mere 0.69 seconds.
