Detecting Corrosion on Steelwork

Case Study

Keen AI is using deep learning and image processing to detect corroded steelwork on National Grid towers in high resolution imagery. Using machine vision, Keen AI’s technique processes images and automatically identifies towers that do not contain any corrosion and therefore do not require further inspection.

The technique is being developed further to grade towers based on the severity of the corrosion found.

The Problem Statement

  • National Grid are responsible for maintaining over 20,000 electricity transmission towers across England and Wales.
  • They capture high resolution images of their towers to identify corroded steelwork.
  • Analysts spend hours reviewing the images, assigning a grading to each tower based on the severity of corrosion found.
  • The majority of towers contain little to no rust, and are considered clean.
  • The process is time-intensive and laborious.

25.3%

Successful Outcomes

  • The developed algorithm is able to accurately identify clean towers, removing them from the inspection process.
  • OPEX Saving: Reduction in time taken to review towers.
  • Self Sufficiency: Analysts are able to train AI models to identify and extract relevant components (Spacers, Joins, Dampers, Insulators).
  • Risk Reduction: Reduced the risk of corrosion reaching a severe state by removing the backlog of images to be processed.
  • Process Consistency: Identification of clean towers is independent of the individual analysing the images.
Identifying distinct regions of a tower that are corroded
Rapid detection of corrosion achieved with deep learning and image processing