MULTISCALE MODEL FOR NON-METALLIC INCLUSIONS/STEEL COMPOSITE SYSTEM USING DATA SCIENCE ENABLED STRUCTURE-PROPERTY LINKAGES Akash Gupta 1 , Ahmet Cecen 2 , Sharad Goyal 1 , A. K. Singh 1 , S. R. Kalidindi 3 1 TRDDC-TCS Innovation Labs, Tata Consultancy Services Ltd., 54 B, Hadapsar Industrial Estate, Pune, Maharashtra, 411013, India 2 Department of Mechanical Engineering and Mechanics, Drexel University, Philadelphia, PA 19104, USA 3 Woodruff School of Mechanical Engineering, Georgia Institute of Technology, Atlanta, GA 30332, USA Keywords: Steel, Inclusions, Multiscale model, Data science, 2 point statistics, ICME Abstract Cleanliness is a major concern for steel manufacturers. Therefore, they constantly strive to modify and reduce non-metallic inclusions in the final product. Performance and quality of final steel sheet is strongly influenced by composition, morphology, type, size and distribution of inclusions in steel sheet. The aim of current work is to critically evaluate the versatility of a new data science enabled approach for establishing objective, high fidelity, structure-property correlations that are needed to facilitate optimal design of the processing path to realize enhanced performance of the final product. A 2-D finite element based micro-mechanical model was developed to simulate, the effect of various spatial configurations and geometries of hard and soft inclusions in a steel matrix system, on the final properties of processed sheet. From each microscale simulation macroscale parameters such as yield strength, effective hardening rate, localization propensity, and plasticity index, were extracted. A large number of microstructures were evaluated using the micro- mechanical model. A reduced-order representation was extracted for the selected ensemble of microstructures using principal components of their 2-point statistics. These objective measures of the microstructure were then linked with the macroscale parameters listed above using regression methods. The extracted structure-property correlations are presented on this paper. Introduction Steelmakers constantly strive to modify and reduce the non-metallic inclusions to meet the ever- increasing demands on the quality of steel. Non-metallic inclusions are foreign substances, (defects) mainly oxides, sulphides and oxy-sulphides, which are embedded in steel matrix. The inclusions are generated during upstream steelmaking operations and get embedded in the steel matrix during solidification in a caster. They disrupt the homogeneity of structure, so their influence on the mechanical and other properties can be considerable. Performance and quality of final steel sheet is strongly influenced by composition, morphology, type, size and distribution of inclusions. For example, during rolling, hard inclusions lead to interface delamination, voiding and crack formation while soft inclusions get elongated. Many authors have studied the inclusions present in a matrix. In the classical model of Eshelby [1], elastic field in and around elastic ellipsoidal inclusion in an elastic matrix was determined. Using the finite element method (FEM), a few authors have attempted to study void formation and cracks originating from inclusion/matrix interfaces during the process of hot rolling [2,3] and cold rolling [4]. 57 2 nd World Congress on Integrated Computational Materials Engineering Edited by: Mei Li, Carelyn Campbell, Katsuyo Thornton, Elizabeth Holm, and Peter Gumbsch TMS (The Minerals, Metals & Materials Society), 2013