Analysis of Vibration Signals Based on Machine Learning for Crack Detection in a Low-Power Wind Turbine

Area of Science:

• Engineering
• Renewable Energy Systems
• Materials Science

Background:

• Wind turbines (WTs) are critical for renewable energy but vulnerable to environmental damage.
• Blade damage, particularly cracks, reduces efficiency, increases costs, and necessitates early detection.
• Condition monitoring is essential for maintaining WT operational integrity and longevity.

Purpose of the Study:

• To develop and validate a machine learning-based method for detecting and assessing the severity of cracks in wind turbine blades.
• To analyze vibration signals under various wind conditions to identify distinct crack signatures.
• To provide a reliable tool for proactive maintenance scheduling in wind energy infrastructure.

Main Methods:

• Vibration signals from WT blades with healthy, light, intermediate, and severe cracks were analyzed.
• Feature extraction involved statistical and harmonic indices, followed by feature selection using analysis of variance (ANOVA).
• Classification was performed using the k-nearest neighbors algorithm, with comparisons to neural networks, decision trees, and support vector machines.

Main Results:

• The proposed machine learning approach achieved a classification accuracy exceeding 99.5% for detecting and assessing crack severity.
• The method effectively differentiated between various crack levels (healthy, light, intermediate, severe) based on vibration signal analysis.
• ANOVA proved effective for feature selection, enhancing the classification performance of the k-nearest neighbors model.

Conclusions:

• Machine learning analysis of vibration signals offers a highly accurate method for detecting and grading wind turbine blade cracks.
• This technique enables precise condition monitoring, crucial for optimizing maintenance strategies and reducing operational costs in wind farms.
• The study demonstrates the potential of AI-driven diagnostics for ensuring the reliability and efficiency of wind energy systems.