Vibration-Response-Only Structural Health Monitoring for Offshore Wind Turbine Jacket Foundations via Convolutional...

Area of Science:

• Engineering
• Mechanical Engineering
• Structural Health Monitoring

Background:

• Offshore wind turbines require robust structural health monitoring (SHM) for safety and efficiency.
• Jacket-type foundations are critical components susceptible to various damage types.
• Traditional SHM methods may be limited in detecting subtle structural changes.

Purpose of the Study:

• To propose a vibration-response-only methodology for SHM of jacket-type offshore wind turbine foundations.
• To develop a signal-to-image conversion technique for accelerometer data.
• To enhance the performance of deep convolutional neural networks for damage classification.

Main Methods:

• Utilized accelerometer data from a laboratory experiment on a steel jacket-type foundation.
• Developed a signal-to-image conversion process, creating multichannel grayscale images from sensor data.
• Implemented a data augmentation strategy to improve the deep convolutional neural network's accuracy.
• Employed deep convolutional neural networks for image classification to identify damage scenarios.

Main Results:

• Achieved a classification accuracy exceeding 99% in damage detection and identification.
• Demonstrated the effectiveness of the signal-to-image conversion for SHM.
• Validated the data augmentation strategy's ability to reduce test set errors.

Conclusions:

• The proposed vibration-response-only methodology shows significant promise for SHM in jacket-type offshore wind turbine foundations.
• The integration of signal-to-image conversion and deep learning offers a highly accurate approach to damage detection.
• This method can contribute to improved safety and maintenance strategies for offshore wind energy infrastructure.