Surface defects such as scale deposits, notches, indentations, or cracks on rolls present a significant problem in the rolling process of a steel mill, as these defects are transferred from the roll to the produced steel strip, potentially causing substantial damage to the product. Therefore, regular inspection of these rolls is essential to detect defects early and take appropriate countermeasures.
In the course of this bachelor's thesis, different methods and approaches for defect detection were examined, based on extensive literature research and existing measurement data from a defect-prepared metallic test roll and a plastic roll. The measurement data come from a laser line triangulation sensor, which assigns each measured point (x,y) a scalar distance value (z) between the sensor and the object being measured, as well as an intensity value (i) of the laser beam reflection. Several algorithms were developed that are capable of detecting defects, visually representing or highlighting them, and possibly classifying them.   

Keywords: Defect detection, Surface defects, Lasertriangulation, OpenCV, YOLO

July 4, 2024