Real-time characterization of non-metallic inclusions by optical scanning and milling of steel samples

Abstract

ABSTRACT The cleanliness of steel is described by the amount, size, composition, morphology, and distribution of non-metallic inclusions (NMIs). These nonmetals are present because of natural physical-chemical eects, and becauseduring continuous casting steel is accidentally contaminated with slag, refractories, and materials from castingmoulds. NMIs in”uence the properties of steel. Therefore, in this paper, a combined milling and image processingsystem is proposed that mills and scans slices of steel samples to retrieve volumetric information about NMIs.The system is capable of scanning steel samples of 300 × 100 × 90 mm 3 in size at spatial resolutions of either3,5,10, or 20 µm and a volumetric resolution of 10 µm within a few hours. After each milling operation thesteel surface is captured by a moving large-area CCD image sensor. The optical system further consists of adistortion-free macro lens and diuse coaxial lighting for bright“eld illumination. Additional results using domelighting are also presented. The interaction of an NMI with the milling cutter results in non-homogeneous NMIre”ectances which carry information about the NMIs mass density and chemical compound. Although the steelsurface is highly re”ective, the milling cutter creates a periodic pattern of moldings which is accentuated bypatterns of shadow and light. An adaptive wedge “lter in the Fourier space dampens those artifacts. NMIs arebinarized separately in every image by local thresholding. In order to reduce segmentation artifacts neighboringslices in the volumetric stack of images are “ltered using morphological operators. A statistical analysis of thesegmentation results estimates the macro cleanliness. Furthermore an interactive 3D visualization enables theexploration of NMIs and their distribution within the sample. Dierent viewing, “ltering and sorting capabilitiesare implemented, like ordering NMIs with regard to their shape factor. It is expected that the study of theseattributes will lead to information about the composition and formation of NMIs.Keywords: Anomaly detection, adaptive parameterization, texture analysis, interactive visualization