Using AI to ensure TotalEnergies remote corrosion detection process is optimised and safe
TotalEnergies UK faced the challenge of a manual and time-consuming process when reviewing their data to detect areas in need of maintenance. The corrosion inspection process involved walking around the site, observing different areas and taking photographs to share with the rest of the inspection team.
Merkle worked on the artificial intelligence component of an architecture designed to optimise remote monitoring and Digital Twin capability, where the complete system is always working with the latest information available that reflects the ground truth.
The purpose of the solution is to leverage image/video data to enable remote monitoring through automated inspection using image recognition predictive techniques deployed in Azure. Drone surveys enabled collection of large volumes of image and video data. Our solution processed this data to provide a report to highlight severe areas of corrosion. Data is also collected through regular automated drone surveys, which avoids humans having to climb ladders/use rope access, etc. to collect data and perform assessment.
Furthermore, predictive asset maintenance enabled the client to identify when maintenance needs to be performed on equipment, rather than having to fix equipment after it has broken down. This allows the business to pivot towards being more proactive, rather than reactive, so they can put a plan in place to make repairs before the situation becomes significant enough to urgently require it.
The immediate benefit to the business is a more efficient and safe inspection process and quicker turnaround of reports to support evidence-based decision making. Not only does the process reduce the risk of injury caused by manual inspections, but it also means that corrosion is identified and resolved much quicker than with the previous manual process.
Standardising the process ensures consistency and objectivity in the way the level of corrosion is detected and interpreted. Human error is reduced and the volume of images that can be assessed is greatly increased.
With less time spent on inspecting for corrosion and assessing damage, more time can be spent on fixing the damage and performing other crucial maintenance tasks.
This reduced the detection phase of the process from several weeks down to just 24 hours.