Norway-based DNV, an independent expert in assurance and risk management, has launched a research project to develop a framework for data verification and validation in fully automated drone inspections of wind turbine blades.
With the University of Bristol and Perceptual Robotics joining in as partners, DNV wants to address the need for reliable, fully automated processing of drone data in the UK inspection industry.
Right now, this process largely remains semi-autonomous with dependence on visual inspections of image data by trained experts. DNV and its partners, however, are looking to leverage machine learning technologies to create artificial intelligence (AI) models capable of autonomously classifying and segmenting drone data and detecting blade defects.
Now, before you point out that automated windfarm inspection is nothing new, note that Perceptual Robotics and the University of Bristol have also been working on this concept since 2017, thanks to a generous UK Innovate grant.
While Perceptual Robotics has been acquiring drone data in commercial environments, researchers at the Visual Information Lab at the University of Bristol have been utilizing their expertise in 3D computer vision and image processing to create algorithms for automated localization of inspection images and defect detection. What DNV adds to the mix is much-needed industry standardization and conformity.
Broader acceptance of autonomous drone inspections
As an energy industry expert, DNV’s role is to provide inspection expertise, verify the drone data collected, validate the methodology and performance of the AI algorithms, and provide guidance as to existing DNV and IEC recommend practices, regulations, and industry networks.
Explaining how this research project will provide a stepping stone to the growth of the fully automated drone inspection industry, Dr. Elizabeth Traiger, a DNV senior researcher in digital assurance, says:
This collaboration will develop and demonstrate an automated processing pipeline alongside a general framework with the aim of generating broader acceptance across the industry and informing future regulations.
The research project, which began in April 2021, is expected to run for 12 months. As Pierre C. Sames, group research and development director at DNV, sums up:
With the number of installed wind turbines worldwide increasing, including those in remote and harsh environments, the volume of inspection data collected is quickly outpacing the capacity of skilled inspectors who can competently review it. This research project will develop means to tackle this challenge through machine learning algorithms and process automation.
FTC: We use income earning auto affiliate links. More.