Linking the Weld Geometry and Fatigue Life Using Optical Scanning and Computer Vision for Reliable Maritime Structures

Kurzfassung

Welded joints are critical to the safety and reliability of maritime structures, where fatigue failure is a major concern affecting the design of engineer structures. This study examines the correlation between weld geometry quality (classified by standards such as ISO 5817 or Volvo STD 181-0004) and fatigue life using a data-driven approach. Optical scanning is used to capture the local weld geometry which then are linked to fatigue test results. Additionally, convolutional neural networks are applied to identified crack initiation sites on fatigue fracture surfaces. The aim is to identify which local geometric features act as the most relevant stress concentrators which would enable the prioritization of inspection points and potential repairs as post-processing after manufacturing. Preliminary results and recent literature indicate that current quality levels do not reflect the fatigue behavior supporting the need of new quality definitions.