Smart tool finds corrosion with 99% accuracy

The ABS AI Corrosion Detection tool uses machine learning to train its algorithm and continually improve its accuracy.

Corrosion is estimated to cost the shipping industry around US$50 million to US$80 million per year. “This application of advanced machine learning technology to assist with tank coating and corrosion management is unique to ABS, and now we are able to demonstrate that it is highly effective in real world applications,” commented Kash Mahmood, ABS senior vice president, digital solutions.

The ABS AI Corrosion Detection digital solution uses an algorithm that analyses images and video to deliver prediction scores and percentages for tank coating breakdown and corrosion. Clients receive proactive alerts of structural defects such as coating failures, fractures and deformations without the costly and intrusive aspects of traditional structural inspections.

A recent deployment on two vessels revealed that the tool graded the imagery with 98.9% accuracy. The findings are based on the tool’s analysis of more than 600 images of a client’s ballast water tanks. The more images are uploaded and evaluated, the more refined and precise the tool becomes.