Thermodynamics-based modification of the Koistinen-Marburger model for martensitic transformation in metal additive manufacturing

Additive manufacturing (AM) processes result in complex phase transformations due to rapid thermal cycles and steep localized thermal gradients. Predictive modeling of these phase transformations is essential since the microstructural evolution and the resulting mechanical properties are extremely sensitive to processing conditions. This study focusses on predicting site-specific austenite-to-martensite transformation in two steel alloys, Maraging M300 and H13 tool steel, during laser powder bed fusion (LPBF) AM. These alloys are selected for their industrial relevance and distinct martensitic transformation characteristics. The Koistinen-Marburger (K-M) equation is commonly used to estimate the formation of martensite during cooling. This equation employs an empirical k-parameter that does not account for alloy chemistry and the non-equilibrium nature of AM processing. To address this limitation, a thermodynamically-informed modification of the K-M model is proposed, in which the k-parameter is reformulated as a function of the entropy change (ΔS) associated with the austenite-to-martensite transformation. The ΔS term is extracted from CALPHAD-based Thermo-Calc® calculations. Location-specific thermal history is obtained using a Finite Element Method (FEM) simulation performed using Simufact Additive®. This enables alloy-specific, temperature-dependent modelling of transformation kinetics under AM conditions. Experimental validation was conducted through EBSD and XRD characterization of LPBF fabricated samples to validate the predictive capability of the modified model. By enhancing the K-M parameter based on thermodynamics of martensitic transformation, this formulation aims to capture more realistic transformation kinetics encountered in LPBF. The flexibility of the modified equation allows for its extension to other alloy systems with minimal reparameterization. Overall, the proposed model advances the quantitative prediction of phase transformations during rapid solid-state cooling, enabling more reliable microstructural control in AM components.