In this work, we use a large-strain elasto-viscoplastic fast Fourier transform (LS-EVPFFT) code enhanced with a continuum damage mechanics model to predict failure response of a subcontinuum mesoscale tensile specimen in the context of the National Institute of Standards and Technology (NIST) 2022 Additive Manufacturing Benchmark (AM-Bench) Challenge. In the Challenge, participants were provided with data from X-ray computed tomography and electron backscattered diffraction (EBSD) for an AM IN625 sample and asked to predict stress and strain response and locations of necking and fracture. To account for uncertainty in the subsurface microstructure, we instantiated 10 semi-synthetic microstructures using a Potts model in a modified version of the open-source software SPPARKS. While all 10 models maintain identical surface grain structure, surface roughness, and internal porosity, their subsurface grain structures vary due to randomness in the microstructure-generation procedure. Results from the blind predictions using the LS-EVPFFT framework are compared to the experimental results. Lessons learned are discussed.
EXTENDED ABSTRACT