TY - JOUR

T1 - Classification of non-metallic inclusions in steel by data-driven machine learning methods

AU - Ramesh Babu, Shashank

AU - Musi, Robert

AU - Thiele, Kathrin

AU - Michelic, Susanne Katharina

PY - 2022/9/29

Y1 - 2022/9/29

N2 - Non-metallic inclusions have strong influence on final steel properties. An important characterization tool to make a comprehensive analysis of non-metallic inclusions is the scanning electron microscope equipped with electron dispersive spectroscopy (SEM-EDS). A major drawback which prevents its use for online-steel assessment is the time taken for analysis. Machine learning methods have been previously introduced which circumvents the usage of the EDS for obtaining chemical information of the inclusion by classifying inclusion based on their back scatter electron images. This paper introduces a method based on a simpler tabular data input consisting of morphological and mean grey values of inclusions. A Naive Bayes and Support Vector Machine classifier models were built using the R statistical programming language. Two steel grades were considered for this study. The prediction results were shown to be satisfactory for both binary (maximum 89 %) and 8-inclusion class (maximum 61 %) categorization. The input dataset was further improved by optimizing the image settings to distinguish the different types of non-metallic inclusions. It was shown that this improvement resulted in a higher rate of correct predictions for both binary (maximum 98 %) and 8-class categorization (maximum 81%).

AB - Non-metallic inclusions have strong influence on final steel properties. An important characterization tool to make a comprehensive analysis of non-metallic inclusions is the scanning electron microscope equipped with electron dispersive spectroscopy (SEM-EDS). A major drawback which prevents its use for online-steel assessment is the time taken for analysis. Machine learning methods have been previously introduced which circumvents the usage of the EDS for obtaining chemical information of the inclusion by classifying inclusion based on their back scatter electron images. This paper introduces a method based on a simpler tabular data input consisting of morphological and mean grey values of inclusions. A Naive Bayes and Support Vector Machine classifier models were built using the R statistical programming language. Two steel grades were considered for this study. The prediction results were shown to be satisfactory for both binary (maximum 89 %) and 8-inclusion class (maximum 61 %) categorization. The input dataset was further improved by optimizing the image settings to distinguish the different types of non-metallic inclusions. It was shown that this improvement resulted in a higher rate of correct predictions for both binary (maximum 98 %) and 8-class categorization (maximum 81%).

KW - Non-metallic inclusions

KW - machine learning

KW - chemical analyses

KW - characterization

KW - steel

U2 - https://doi.org/10.1002/srin.202200617

DO - https://doi.org/10.1002/srin.202200617

M3 - Article

VL - 2022

JO - Steel research international

JF - Steel research international

SN - 1869-344X

IS - September

ER -