Additive manufacturing of metallic alloys have become a disruptive technique allowing to design and fabricate very complex components with a near neat shape. This is specially relevant for the aerospace industry with high interest in producing complex shape components using Selective Laser Meling (SLM), in high performance superalloys as Hastelloy-X. However, in SLM fabrication the resulting parts present strong anisotropy and their response have large dispersion. These features are due to the special microstructures developed as a result of rapid heating and cooling during the SLM process. The mechanical behavior of these alloys is determined by the microstructures developed during SLM (namely grain size, aspect ratio, and orientations) and defects such as surface roughness. For similar fabrication parameters, these microstructures also vary depending on build orientation and sample thickness. The experimental characterization of the quasi-static and fatigue response of SLM-fabricated specimens necessitates a large experimental campaign for a wide variety of microstructures.

Complementary, computational modelling approaches aims to predict the mechanical response of SLM alloys as a function of the actual microstructure thus reducing the number of experiments required for material characterization. However, the existing computational methods and their application to predict the quasi-static and fatigue response have certain limitations. a) Regarding the models for mechanical behavior, the majority are empirical and based on large experimental datasets. b) There is no study which analyzes (and quantitatively predicts) the anisotropic mechanical response of SLM components that results from microstructure, sample thickness, and surface roughness defect. c) There is a lack of fatigue models that takes into account the effects of microstructure, build direction, and surface roughness in the life prediction.

In this work, a microstructure sensitive virtual testing framework for SLM-fabricated Hastelloy-X is presented, which overcomes the aforementioned limitations. Regarding a) numerical models that take microstructural details into account have been developed to predict the quasistatic and cyclic response, reducing the number of experiments for material characterization. These models are based on a computational homogenization approach that links the macroscopic response of an SLM part with its polycrystalline microstructure. A crystal plasticity (CP) model is used to represent the single crystal response which is ftted using limited macroscopic data. Finally, the behavior is obtained by simulating the deformation of Representative Volume Elements (RVE) of the actual microstructure using an effcient FFT solver, which allows for the resolution of large synthetic RVEs, with complex and detailed microstructure, built from data obtained using EBSD. Furthermore, the effect of sample geometry (sample thickness) and surface roughness has also been considered in the simulations.

Regarding b) the proposed model is used to predict the quasi-static response of SLM-fabricated Hastelloy-X as a function of the SLM resulting microstructure (for different building directions) and also considering the effect of temperature and sample thickness. The CP model was ftted using the tensile response of one of the specimens considered. The temperature dependency was added into the CP model via the initial and saturation critical resolved shear stresses. The proposed framework correctly predicted (within error 4%) the Young’s modulus and plastic stress-strain response of various specimens across a wide temperature range. The work demonstrates that the anisotropic mechanical response found in SLM parts can be entirely attributable to the polycrystalline microstructure.

With respect to the study of fatigue, the proposed CP-FFT framework is utilized to forecast the fatigue performance of SLM-fabricated Hastelloy-X at 750°C while taking into account the infuence of the microstructures produced during the SLM process. A phenomenological CP model with a back stress term is utilized, and it is calibrated using inverse ftting to the experimental hysteresis curve of one experiment under strain-controlled loading. The technique is based on extracting fatigue indicator parameters (FIP) from simulations, and different fatigue life prediction laws are utilized for both strain- and stress-controlled, which are based on cyclic and total dissipated work, respectively. The fatigue life is calculated for each case using the FIP distribution obtained from proposed power law which is calibrated using two fatigue experiments. For the different SLM microstructures and loading conditions considered, there is a very good agreement with the experimental data on fatigue life. Furthermore, the model can capture the effect of fabrication direction on anisotropic fatigue performance, highlighting the importance of microstructure on the fatigue performance of Hastelloy-X manufactured using SLM.

Finally, the fatigue model is extended to predict the fatigue life of fat samples as a function of microstructure and surface roughness. The surface roughness layer, which is generated based on the experimental mean surface roughness Ra, is added to the RVE that imposed by free boundary conditions. It is shown that including in the simulation free surface with roughness allows to accurately predict the fatigue life of fat un-machined specimens which is considerable lower than polished bulk specimens.