AI-based Defect Detection on Wind Turbine Surfaces with Reasoning and Continual Learning Capabilities

PhD Student: Clifford Zhang Academic Supervisor: Dr Georgina Cosma

Industry Supervisor: Jason Watkins, Railstons & Co Ltd

Funders: Jointly funded by Railstons & Co Ltd and Loughborough University

Paper: Image Enhanced Mask R-CNN: A Deep Learning Pipeline with New Evaluation Measures for Wind Turbine Blade Defect Detection and Classification

Demand for wind power has grown, and this has increased wind turbine blade (WTB) inspections and defect repairs.

The project will develop an innovative tool for detecting and classifying wind turbine blade defects; provides reasoning behind the predictions; and has capabilities to continue to learn when deployed. The project is conducted in close collaboration with Railston & Co LTD who have provided data and expert support throughout the project.

Railston & Co LTD Collabaro – Field service management (FSM) software helps companies organise and optimise their business activities performed by field-based service professionals. These solutions are primarily used by companies that need to inspect or maintain a piece of industrial hardware under warranty or as part of a planned maintenance schedule.

We have empirically investigated the performance limitations of state-of-the-art deep learning algorithms, namely, YOLOv3, YOLOv4, and Mask R-CNN for detecting and classifying defects by type, and proposed a new AI tool that is 85% accurate for recognizing and classifying wind turbine blade defects.

Given that data can be limited and it is often difficult to have enough samples to suitably train AI algorithms. Therefore, our tool benefits from suitable image enhancement and augmentation methods to increase the amount of images by adding slightly modified copies of already existing images to the dataset.

One of the limitations we identified during evaluations was lack of suitable measures for evaluating the performance of defect detection systems. Hence we also implemented and tested new performance evaluation measures suitable for defect detection tasks and embedded those in the tool.

Experiments were carried out using a dataset, provided by the industrial partner Railston & Co Ltd, that contains images from Wind Turbine Blade inspections. We wrote a paper discussing the findings.

The paper proposes a new defect detection tool that includes image enhancement and augmentation techniques for pre-processing the dataset, and a state-of-the-art AI model tuned for the task of wind turbine blade defect detection and classification.

Selected media Articles:
Loughborough University New AI tool, developed by Loughborough experts, 85% accurate for recognising and classifying wind turbine blade defects
The EngineerAI tool locates and classifies defects in wind turbine blades
The Civil EngineerArtificial Intelligence tool identifies wind turbines defects
Azo robotics AI-Based System Inspects Defects in Wind Turbine Blades More Accurately
TechXploreNew AI tool 85% accurate for recognizing and classifying wind turbine blade defects
connectivity4IR – AI tool recognises wind turbine blade defects