Prediction of Inclusion State in Molten Steel by Morphology and Appearance of Inclusions in Liquid Steel Samples. Issue 2 (30th October 2019)

• Record Type: Journal Article
• Authors: Alatarvas, Tuomas; Vuolio, Tero; Heikkinen, Eetu-Pekka; Shu, Qifeng; Fabritius, Timo
• Is Part Of: Steel research international. Volume 91:Issue 2(2020)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2019-10-30
• Subjects: calcium treatments -- classifications -- inclusion analyses -- inclusions; Steel -- Periodicals; Steel -- Metallurgy -- Periodicals; 669.142
• DOI: 10.1002/srin.201900424
• Languages: English
• ISSNs: 1611-3683

Abstract

Abstract : Inclusions are unwanted but to some extent inevitable in molten and solid steel. Usually solid inclusions are considered to be the most harmful. Inclusions can be converted into a less detrimental form with calcium treatment. The success of calcium treatment can be evaluated by analyzing the state of the inclusion population. The state of inclusions is usually determined by computational thermodynamics making use of the chemical composition of inclusions and system conditions. In this process, liquid and solid inclusions are usually distinguished. Herein, a classification procedure which combines computational thermodynamics and data‐driven reasoning is presented. The objective of this work is to study the predictability of the inclusion state based on its appearance and morphological properties. As a result, Al2 O3 –CaO–MgO–CaS inclusions are classified as liquid and solid ones based on their aspect ratio, equivalent circle diameter, and mean gray value with a recall of 82.7% and precision of 84.9%, by making use of a logistic regression‐based classifier. Abstract : An inclusion classification procedure is proposed, combining computational thermodynamics and data‐driven reasoning. The state of inclusions in molten steel is predicted based on their appearance and morphological properties. Al2 O3 –CaO–MgO–CaS inclusions are classified as liquid and solid ones based on their aspect ratio, equivalent circle diameter, and mean gray value by making use of logistic regression classifier.