Enhancing Wind Turbine Quality Assurance With AI Techniques

An international collaboration between EPFL and the University of Glasgow has developed an advanced machine-learning algorithm for effectively detecting concealed manufacturing defects in wind turbine composite blades prior to their deployment.

Detecting faulty wind turbine blades early is crucial for avoiding significant costs for companies. Quality assurance is a strategic concern for global wind turbine manufacturers, as defects can lead to costly repairs or replacements if not identified in time. Inspections are limited to surface checks of specific areas as blades are manufactured.

However, a new approach, developed by researchers from EPFL and the University of Glasgow, introduces a patented radar technology and an AI assistant to identify potential anomalies beneath the surface, enhancing the inspection process.

The newly developed approach offers several advantages. It is non-destructive, non-contact, supports agile and rapid data acquisition and analysis, and requires very little power to operate. The research detailing this approach has recently been published in Elsevier Mechanical Systems and Signal Processing (MSSP).

Merging Signal Processing and AI

The research draws on earlier work by both institutional partners. This research was led by Olga Fink, a tenure-track Assistant Professor of civil engineering and Head of the Intelligent Maintenance and Operations Systems Laboratory (IMOS) within EPFL’s School of Architecture, Civil and Environmental Engineering (ENAC).

In prior research, she developed methods for detecting anomalies by analyzing the sounds produced by faulty machines, suppressing background noise in audio recordings, and classifying bird songs by incorporating learning capabilities into established signal-processing approaches.

Today, Fink is looking at new applications for AI-driven systems.

Wind turbines are made from several different composite materials like fiberglass and carbon fiber. Manufacturers are also building them bigger, with more complicated designs. All that increases the chances of a defect occurring during the manufacturing stage.

Olga Fink, Assistant Professor and Head, Civil Engineering, Intelligent Maintenance and Operations Systems Laboratory

Measurement Technology

The University of Glasgow team, led by Prof. David Flynn from the James Watt School of Engineering, has been at the forefront of research in prognostics and health management, focusing on how Robotics and Artificial Intelligence (RAI) can contribute to achieving net zero infrastructure. The Glasgow researchers utilized a patented Frequency Modulated Continuous Wave radar with a robotic arm to inspect industrial wind turbine blade samples at distances 5, 10, and 15 cm from the sample.

Using advanced signal processing methods, the researchers were able to isolate features and precursors to future failures in these complex composite samples.

Improve Data Representation

When the experimental data was provided to the IMOS team, the challenge was to enhance the information content of the features embedded within this raw data. The signals obtained by the radar varied depending on the inspection distance and the blade's surface and core materials.

At IMOS, we used a complex-value representation of the signals to better separate the information they contain and to adapt the AI model accordingly. As a result, the algorithm they developed can distinguish anomalies from uniform turbine parts.

Gaëtan Frusque, Postdoc and Study Lead Author, Intelligent Maintenance and Operations Systems Laboratory

The Glasgow researchers are now focusing on collecting additional data to validate the results obtained by the IMOS team further. Their ultimate goal is to test their method on existing turbines, which they plan to achieve by attaching the sensor to a robotic arm or a drone. This approach would allow them to identify manufacturing defects in turbines before they are put into service or to inspect operational turbines. Since defect-free turbines can operate for approximately 20 years, this technology could significantly enhance turbine reliability and lifespan.

Video Credit: EPFL

Journal Reference:

Frusque, G., et al. (2024) Non-contact sensing for anomaly detection in wind turbine blades: A focus-SVDD with complex-valued auto-encoder approach. Mechanical Systems and Signal Processing. doi.org/10.1016/j.ymssp.2023.111022.

Source: https://www.epfl.ch/en/