The Importance of Materials Property Data for Finite Element Simulations of AM: Why Handbook Data Are Unsuitable

Conference Abstract: Finite Element (FE) simulations are commonly used to predict the melt pool shape, solidification behavior and thermal history of the build part during additive manufacturing. Such simulations can be used to better optimize the process parameters including laser power, beam shape and scan strategies. Microstructure predictions can also be obtained which form the basis of localized property predictions using other tools in an ICME (Integrated Computational Materials Engineering) framework. However, the accuracy of these simulations are strongly reliant on the suitability of the material properties used as input to the simulations. Such properties are often obtained from handbooks which do not capture the composition and temperature dependence of the properties of interest or by experiments which are specialized, costly and take a long time to perform.
This talk will highlight an approach using computational thermodynamics along with a FE code to more accurately predict melt pool behavior during laser powder bed fusion (LPBF) processes. This methodology takes a unified treatment of both process parameters and chemistry-dependent thermophysical properties when solving the multi-physics problem of a moving heat source that melts and solidifies metal powder. Using the CALPHAD (Calculation of Phase Diagrams) approach, temperature and chemistry dependent physical properties such as specific heat, density, thermal conductivity, viscosity, and surface tension of liquid are calculated from evaporation temperature down to room temperature. Using an extended Scheil-Gulliver model these are transferred to the FE code where the multi-physics simulation concurrently treats thermal conduction, fluid flow, evaporation-, radiation- and convective- heat loss. For validation, simulations have been made comparing this approach to benchmark data from the NIST AM-Bench test series on alloy 718. Good agreement is found when comparing calculated vs. measured melt pool shape. The talk will conclude with a discussion on how such an approach can be extended to predicting printability maps.