Development of statistical models for porosity from digital optical micrographs with application to metal additive manufacturing microstructure

Using techniques widely available in digital image processing, machine learning and spatial statistics, this paper proposes a novel workflow that generates two dimensional spatial models using objects extracted from digital micro-graphs of material micro-structure that can be used in statistical reconstruction modeling within a numerical procedure, such as finite element analysis. This paper also reviews the relevant image processing techniques, spatial statistical theories and reconstruction (modeling) algorithms with unique contributions. As an end-to-end illustration, the workflow is applied to a two dimensional, digital micro-graph of hydrogen porosity taken of as-fabricated AlSi10Mg manufactured using laser powder bed fusion adopted from the literature.