Abstract
A data-driven framework is developed and examined for creating spatially-varying crystallographic textures over component-scale Computer-Aided Design (CAD) models. Here, a set of three orthogonal 2D micrographs of an Additively-Manufactured (AM) specimen are first obtained experimentally through Electron Backscatter Diffraction (EBSD) and subsequently converted to a 3D representative unit cell using the Markov Random Field (MRF) technique. Features such as grain size, crystallographic orientation, and grain boundary misorientation distributions are used to validate the reconstructed 3D microstructure against input experimental EBSD images. The variations of microstructural features during a powder-based additive manufacturing process are subsequently modeled by merging patches from the 3D snapshot of AM microstructural unit cell in a part-scale geometry using a tensor-based optimization process. The optimization algorithm repeatedly pastes microstructural elements from the reconstructed MRF unit cell onto the geometrical CAD domain until it is entirely covered. Here, through a simple Graphical User Interface (GUI), the user specifies a tensor field over the volumetric CAD model, defining the local control over grain-scale, anisotropy, and crystal growth orientation. This new approach provides a workflow for reconstructing global maps of AM microstructures in real-time by embedding site-specific images based on known AM microstructural patterns seen in experimental characterization techniques. The numerical results are helpful specifically for the visualization of process-microstructure relationships in metal additive manufacturing techniques.
NSTL
Elsevier