Structural health monitoring is a multi-disciplinary engineering field that should allow the actual wind turbine maintenance programmes to evolve to the next level, hence increasing safety and reliability and decreasing turbines downtime. The main idea is to have a sensing system on the structure that monitors the system responses and notifies the operator when damages or degradations have been detected. However, some of the response signals that contain important information about the health of the wind turbine components cannot be directly measured, or measuring them is highly complex and costly. Thanks to the advanced system identification methods, the majority of these signals can be indirectly measured by assuming a realistic sensor scenario.
This thesis addresses the problem of using system identification techniques on monitoring time-varying signals that direct measuring is prevented due to the expensive and impractical nature of the required measurement equipment. The thesis consists of two parts. The first part is focused on monitoring the structural loads and presents a novel approach for low-cost monitoring of the wind turbine structural loads from the standard turbine measurements. As test cases are considered, two practical problems from the wind industry are studied, i.e. monitoring of the gearbox shaft torque and the tower root bending moments. The second part of the thesis is focused on the influence of friction on the health of the wind turbine and on the nonlinear identification techniques for time-varying system identification. The test case chosen hereto concerns blade bearing friction estimation. Different nonlinear system identification algorithms are considered and their performances are benchmarked on problems of time-varying parameter estimation in a blade bearing friction model.
- Wind Turbines
- Structural Loads
- Sensing System
- Case Studies