Rotational speed and vibration measurement of wind turbine blades using image processing techniques

The high cost of operation and maintenance of wind turbines makes structural health monitoring an important front of research. Accurate and real-time measurement of wind turbine blades' rotational speed and vibration is key to ensure operating safety and reduce the cost of maintenance. Existing methods mostly rely on contact measurement on the blades or sophisticated feature extraction. This thesis proposed new methodologies to measure the rotational speed and vibration frequency of wind turbine blades using image processing techniques, improving the response time of the measurements. The methods are tested under various conditions to evaluate their performance and feasibility in simulation as well as an experimental test rig.

A literature review is conducted to understand the advantages and limitations of existing techniques and methods for rotational speed and vibration measurement of wind turbine blades. Consequently, marker-based image processing techniques were selected due to their easy applicability. A marker tracking method is proposed to measure the wind turbine blades' rotational speed and a moiré pattern analysis method is proposed to measure a wind turbine blade's vibration frequency.

The marker tracking method is evaluated mainly by experimentation through conditions including camera parameters, camera positions, and environmental brightness. Results show that higher gain generally yields better results. The marker tracking method works well even in low light conditions with a mere increase of 0.1% in the error range. Camera position greatly affects the measurement accuracy, and is recommended that the camera is positioned within 15° perpendicular to the rotational plane. The marker tracking method is more instantaneous compared to correlation-based methods, especially at low rotational speeds. However, the marker tracking method is less resilient to noise, with the marker detector failing at a Gaussian noise with 0.07 variance.

The moiré analysis method was tested under various camera positions, camera distances, blade angles, and vibration frequencies. The experimental results showed that the moiré analysis method achieves 0% between 14 and 18 Hz at a camera distance between 1 to 2 m. Simulation results further indicate that the measurement range is up to 6 m with a maximum pitch of 60°. Simulating the vibration frequencies of a large wind turbine between 0.8 Hz to 21.9 Hz resulted in an error range from -2.33% to 0.46%. The moiré analysis method's accuracy is not affected by Gaussian noise below 0.3 variance and salt-and-pepper noise below 0.3%.