Artificial Intelligence as a Means to Automate Detection of Defects in Inspection of Hull Structures

This paper reports on the status from further development from the outcome of an R&D project developing a digital twin of a physical asset. The asset in this case is a ship tank, and the digital twin is populated with both new and historic data representing the condition of the asset, such as video, point clouds from laser scans, and thickness measurements. The data are collected by autonomous drones equipped with video camera, LIDAR, and UT sensor. AI algorithms then "inspect" the data and detect anomalies such as corrosion, cracks, and deformations. Use of the automated drone improves safety by reducing the need for manned entry into confined spaces. Use of the AI-enabled analysis algorithms improves efficiency by automatically identifying and flagging potential defects for evaluation by a surveyor, thus obviating the need for human review of many hours of video footage. This paper is intended to be a follow-on to the work presented at OTC in May 2024, reference OTC-35308-MS ‘Risk-Based Inspections using Digital Twins for Hulls and Structures of Floating Assets’ (Stensrud et al. 2024). Thus, more extensive motivation and other background information are not repeated in this paper. The functionality in the digital twin includes semi-automated inspection planning, automated generation of the drone's flight plan from the inspection plan, automated and autonomous drone flight, and ML algorithms for automated detection of defects such as cracks, corrosion, coating damage, and deformations.