By applying computer vision and machine learning methods, we develop a system to characterize powder feedstock materials for metal additive manufacturing (AM). Feature detection and description algorithms are applied to create a microstructural scale image representation that can be used to cluster, compare, and analyze powder micrographs. When applied to eight commercial feedstock powders, the system classifies powder images into the correct material systems with greater than 95\% accuracy. The system also identifies both representative and atypical powder images. These results suggest the possibility of measuring variations in powders as a function of processing history, relating microstructural features of powders to properties relevant to their performance in AM processes, and defining objective material standards based on visual images. A significant advantage of the computer vision approach is that it is autonomous, objective, and repeatable.

Abstract Three-dimensional crystal orientation fields of a copper sample, characterized before and after shock loading using High Energy Diffraction Microscopy, are used for input and validation of direct numerical simulations using a Fast Fourier Transform (FFT)-based micromechanical model. The locations of the voids determined by X-ray tomography in the incipiently-spalled sample, predominantly found near grain boundaries, were traced back and registered to the pre-shocked microstructural image. Using FFT-based simulations with direct input from the initial microstructure, micromechanical fields at the shock peak stress were obtained. Statistical distributions of micromechanical fields restricted to grain boundaries that developed voids after the shock are compared with corresponding distributions for all grain boundaries. Distributions of conventional measures of stress and strain (deviatoric and mean components) do not show correlation with the locations of voids in the post-shocked image. Neither does stress triaxiality, surface traction or grain boundary inclination angle, in a significant way. On the other hand, differences in Taylor factor and accumulated plastic work across grain boundaries do correlate with the occurrence of damage. Damage was observed to take place preferentially at grain boundaries adjacent to grains having very different plastic response.

A fast Fourier transform (FFT) based spectral algorithm is used to compute the full field mechanical response of polycrystalline microstructures. The field distributions in a specific region are used to determine the sensitivity of the method to the number of surrounding grains through quantification of the divergence of the field values from the largest simulation domain, as successively smaller surrounding volumes are included in the simulation. The analysis considers a mapped 3D structure where the location of interest is taken to be a particular pair of surface grains that enclose a small fatigue crack, and synthetically created statistically representative microstructures to further investigate the effect of anisotropy, loading condition, loading direction, and texture. The synthetic structures are generated via DREAM3D and the measured material is a cyclically loaded, Ni-based, low solvus high refractory (LSHR) superalloy that was characterized via 3D high energy x-ray diffraction microscopy (HEDM). Point-wise comparison of distributions in the grain pairs shows that, in order to obtain a Pearson correlation coefficient larger than 99%, the domain must extend to at least the third nearest neighbor. For an elastic FFT calculation, the stress—strain distributions are not sensitive to the shape of the domain. The main result is that convergence can be specified in terms of the number of grains surrounding a region of interest.