Extraction of Environmental and Operational Conditions of Wind Turbines using Tower Response Data for Structural Health Monitoring

Abstract

Proper normalization of structural response data that accounts for the environmental and operational conditions (EOCs) of a structure is a key step in structural health monitoring (SHM) analyses. Normalizing data based on EOCs enables a more effective comparison of damage sensitive features extracted from structural response data derived from a system operating under a wide variation in its operations. In this paper, structural response data from an operational wind turbine is used for both damage detection as well as for EOC-based data normalization in a damage detection framework. The structure under consideration is the tower of a 3-kW wind turbine located at Los Alamos National Lab. A wireless monitoring system was installed in the turbine tower to record the tower acceleration response when the tower was undamaged and intentionally damaged. Gaussian Process Regression is used to find tower response features that correlate with EOCs, specifically rotor angular velocity and nacelle yaw angle, yet are insensitive to structural damage. The features are then used in the damaged detection framework to enhance the performance of clustering algorithms used for EOC normalization. Damage-sensitive features are then used as condition parameters for damage detection. The efficiency of the proposed EOCs normalization process is evaluated by comparing the Receiver Operating Characteristic (ROC) curves of a threetier damage detection strategy previously proposed for wind turbine systems