New world first for WRc with Ofwat-funded AI sewer assessment toolThe Water Research Centre (WRc) has supplemented its 1980 industry essential g...
Published on by Lisa Templeton
The Water Research Centre (WRc) has supplemented its 1980 industry essential guide, the Manual of Sewer Condition Classification (MSCC), with another world first, a free-to-download library of sewer images that can be used to instruct artificial intelligence (AI) software solutions to assess the condition of drains and sewers.
WRc, part of the RSK Group, worked on the Ofwat Innovation project with United Utilities as the lead water company partner and Spring, which hosts the library on its platform. It is the first of Ofwat’s open data projects and has been designed to accelerate the development of AI software and to assist in the inspection of sewers, helping to prevent sewer collapses, flooding and overflow spills.
WRc Principal Consultant Peter Henley said: “The WRc manual has been an industry standard around the world since we introduced it in 1980, providing the first coding structure for sewer inspection. It has served the industry well, and we are justifiably proud of the significant contribution it continues to make. However, we recognised that the increasing demands on the water sector and its supply chain called for innovation to enable sewer condition assessment to become more consistent and effective. With the development of AI as a tool in identifying sewer defects, it was clear that an obstacle to its deployment was an absence of data to enable the AI software to reliably train its AI algorithms.
“It’s crucial that we in the water sector use all the skills at our disposal to help the water companies direct their teams effectively and efficiently to sewer issues requiring attention as quickly as possible, arming them with detailed and reliable information. We believe that this library could accelerate that process by making the inspection of sewers more accurate, efficient and effective through the use of AI software, enabling engineers to focus on developing solutions to proactively tackle issues to prevent failure. Studies have shown that the accuracy of CCTV surveys can be as low as 30%, so innovation is clearly needed to improve this.”
Peter said that significant data was required to effectively train the AI software to recognise features found within sewer pipes.
The collaborative project between seven water companies (United Utilities, Thames Water, South West Water, Dŵr Cymru Welsh Water, Scottish Water, Severn Trent Water and Yorkshire Water), which operate 72% of the UK’s sewers or 560,000 km, saw WRc use previously coded CCTV survey footage. Accompanying metadata were also provided, which included a video reference, the sewer material and diameter, a defect code, a defect description and the location of a defect in the frame.
Peter said: “This process gave us a total of 726,290 images, and from this we successfully developed a library of 27,262 images to act as the single benchmark dataset by checking and categorising each of the images for accuracy and clarity. For each defect, an optimum target of 1,000 images was sought from the original total of 726,290 images. This target was selected after discussion with current AI software providers concluded that it was the ideal number to train an AI solution. However, providing even a limited number of images for a defect type is beneficial in the improvement of AI accuracy.
“We were able to identify sufficient images for 17 defect codes (with 1000 images each) and a further 55 defects having images identified and classified, which will still be useful in the development of AI solutions. The library now contains images of 72 defect codes – significantly more than the expected 60 defect codes identified at the onset of the project.”
WRc had 15 water engineers and technical consultants from its technical consulting and catchment modelling teams assigned to this project over a period of three months completing the defect classification process to select the best images for the training library. These were then checked by colleagues to ensure the selected images were of sufficient quality. WRc hopes that this image library can be improved with the addition of more data to extend the range and quality of images to incorporate even more defects and, following initial feedback, even release the images not selected for initial inclusion so that the AI software can be trained using not only positive examples but also negative ones.
United Utilities Head of Innovation Kieran Brocklebank said: “I am very pleased with the outcome of this collaborative project which is one of the first delivered projects from the Ofwat Innovation Fund.
“I would like to congratulate WRc for their resilience in helping us deliver this great solution for the sector. Water companies are active in reviewing the dataset and seeing how they can use it effectively; for United Utilities, our AI provider has already uploaded the new data into their AI model and we’re seeing the benefits.”
Spring Innovation Knowledge Manager Chloe Tooth said: “We were delighted to work with United Utilities and WRc as a knowledge sharing partner on this project. It was great to see how open and honest this group of project partners were in sharing their lessons with the sector. It is critical that knowledge amassed from projects like this is shared across the sector to accelerate learning and ensure the Ofwat Innovation Fund is delivering innovation insights to benefit all customers.”
WRc said that feedback from the launch had been hugely positive, with ongoing discussions with current AI software providers demonstrating that the library has already been used by data specialists to improve their work.
The library is free to download from the Spring platform.
Here is the link to the WRC website: https://www.wrcgroup.com/resources/case-studies/ofwat-innovation-challenge-ai-and-sewer-defects-analysis/
Media
Taxonomy
- Treatment
- Technology
- Infrastructure
- Water, Waste Water Chemical & Treatment
- Domestic Water Use
- Water Treatment Equip