Development of a data driven approach for inverse design of microstructure-property linkages of austenitic steels and recycling caused variation in chemical composition

• Applicants: Professorin Dr.-Ing. Lisa Scheunemann; Professorin Dr.-Ing. Martina Zimmermann
• Subject Area: Mechanics ; Mechanical Properties of Metallic Materials and their Microstructural Origins
• Term: since 2025
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 562153065

Austenitic stainless-steels(ASS) used as sheet-metal in many industry sectors, obtain their final shape through metal forming. Among them, met-astable ASS show deformation-induced austenite-to-martensite transfor-mation, the amount of which strongly varies depending on the stacking fault energy (SFE). The volume fraction of austenite and martensite in the microstructure, especially in the interplay with nonmetallic inclusions, plays a major role for mechanical behavior, static and cyclic strength. The SFE is, however, strongly influenced by the chemical composition of the mate-rial, with ASS having a distinct valid window of alloying elements. Classical sorting techniques during recycling cannot ideally distinguish minor differ-ences in chemical composition, leading to variations in alloy elements and thus SFE. The assessment of the relationship between process parame-ters (deformation history), microstructure (austenite-martensite microstruc-ture with non-metallic inclusions) and property (static and cyclic mechani-cal behavior) is realized in this project through a data-based approach, combining expertise from material science and mechanics. The goal is to identify microstructure-property linkages to be used in an inverse ap-proach. Based on experimental micrographs and strength measurements, microstructural and material data for numerical simulation is provided. Through quasistatic numerical simulation of different microstructure repre-sentations, a training data-base is gained. Suitable in the case of sparse data, a physics-enhanced neural operator framework is trained to describe the relation between microstructural descriptors and microstructurally re-lated stress response. These stress maps are analyzed regarding indica-tors for static and fatigue strength. The inverse approach identifies micro-structural descriptors of distinct indicators through optimization and cre-ates a reversed structure-property link. A further link to process parame-ters and -history can be obtained through knowledge based evaluation of the microstructure. Thereby, loss of properties due to repeated recycling and poor scrap sorting can be counteracted.

• DFG Programme: Priority Programmes
• Subproject of: SPP 2489: DaMic - Data-driven alloy and microstructure design of sustainable structural metals
• Major Instrumentation: Arc Melter
• Instrumentation Group: 8420 Spezielle Oefen (Induktions-, Lichtbogenheizung, Vakuumöfen)