Crack Detection in "As-Cast" Steel Using Laser Triangulation and Machine Learning

We describe a high-accuracy inspection system designed to automatically detect cracks in "as-cast" steel slabs. Real-time slab inspection requires instrumentation capable of withstanding high temperatures above the steel surface as well as coping with the dirty and dustyenvironment present in a steel mill. Crack detection is also challenging due to the presence of oxidation scale on the slab surface. A bespokelaser triangulation system has been developed, providing images at 250 fps with a calibrated surface resolution of 97 µm from a 1m standoff distance. Cracks are detected using a combination of morphological detection and SVM classifier. Results are reported from laboratory testing and from extended trials at a production steel mill.