Thermal–microstructural coupling in laser DED 316L: Predictive modeling of δ-ferrite fraction as a microstructural indicator

Directed Energy Deposition (DED) of 316L stainless steel typically yields an austenitic matrix containing retained δ-ferrite, which forms during rapid solidification and partially dissolves during reheating-induced recrystallization. As a result, the fraction and morphology of δ-ferrite are strongly influenced by local thermal history, which is governed by process planning and reflects cumulative heat exposure within the build. Because δ-ferrite retention is closely coupled to recrystallization behavior, its spatial distribution provides a measurable indicator of microstructural evolution, making it central to understanding processing–microstructure relationships in DED 316L. This work establishes quantitative relationships between δ-ferrite variation, microstructure, and local thermal history within the deposited volume, and develops a predictive framework capable of estimating δ-ferrite fraction from simulated or in-situ temperature–time profiles. δ-ferrite distributions were quantified from etched cross-sections and these measurements were linked to site-specific thermal histories generated from a validated finite-element thermal model. Thermal descriptors representing heating, cooling, and repeated reheating cycles governing recrystallization-driven δ-ferrite dissolution were used to train a machine-learning model that reproduced experimental trends with average deviations of 2–3%. Because toolpath sequencing and process parameters govern local heat flow, the predictive framework enables evaluation of processing conditions and their impact on microstructural uniformity. The integrated framework establishes δ-ferrite fraction as a quantitative metric linking processing conditions and microstructural evolution in laser DED 316L, enabling thermal-history-driven prediction and evaluation of microstructure. Predicted δ-ferrite maps derived from thermal simulations or in-situ temperature data therefore provide a basis for assessing processing–microstructure relationships within the build.