Wind Turbine Bearing Anomaly Detection using CMMS data and Machine Learning

Abstract

The need to anticipate failures in wind turbines has become increasingly urgent. The exponential increase in the number of installed turbines, coupled with the aging of the generation fleet, has intensified the competition to reduce operation and maintenance costs, which means minimizing unplanned downtime and costly major repairs. The aim of this study is to utilize the vibration data available in the Condition Monitoring and Management Systems (CMMS) to identify turbines with significant condition deviations that pose a high risk of failure. The data processing approach using CNN and PCA in the pre-processing stage, along with SVM for health state classification, demonstrated excellent accuracy, above 90 %, for both single turbine and multi-turbine tests, making it suitable for managing wind farms with a large number of turbines.