Scaling Operational Modal Analysis Through Delayed DBSCAN Tracking: Lessons from Widespread Offshore Wind Turbine Monitoring

The OWI-Lab has a long history in employing Operational Modal Analysis (OMA) to monitor the resonance frequencies of (offshore) wind turbines and other infrastructure. However, with a growing number of monitored structures, existing Automated OMA tools face challenges in handling the increasing volume and workload in configuring new OMA campaigns efficiently. This contribution outlines lessons learned, and solutions implemented to address these issues.

A significant change involves removing the real-time tracking from the Automated OMA process. Previously, OMA was followed by mode tracking based on predefined settings, discarding all untracked before monitoring, which is impractical for large-scale projects like farm wide monitoring. The new approach involves performing OMA once and storing all results, including spurious ones. Tracking occurs separately and can be adjusted at any project stage.

This alternative strategy allows to update the tracking settings at any stage of the project. Automation using DBSCAN simplifies generating tracker configuration, allowing for the inclusion of modes dependent on SCADA parameters like rotor speed, even at later project phases. These advancements extend to training machine learning models for predicting mode behavior under specific environmental and operational conditions, enabling data normalization and smart tracking in a single step after reliable models are established.