TRACR: a software pipeline for high-throughput materials image analysis--an additive manufacturing study

Abstract

In high-dimensional materials design spaces such as additive manufacturing, elucidating processing-property relationships is a prerequisite for intelligently controlling structure and tailoring behavior in fabricated components. However, relationships between processing control and resulting properties are not typically straightforward, and often require large volumes of data to develop a sampling that spans the relevant processing space su ciently. For the present study, this comes in the form of high-throughput screening, which demands automated, standardized, procedures for e ciently characterizing large volumes of data. With a significant part of materials data generated in the form of images (e.g. optical microscopy, X-ray computed microtomography, scanning electron microscopy, etc.), rapid characterization becomes an image processing problem. Describing the geometry of visually discernible regions of interest in materials image data is an essential step in describing prop- erties that relate to processing history. In this manner, correlations are identified through standardized image data analysis approaches that would not otherwise be feasible by man- ual methods. This motivates development of the Tomography Reconstruction Analysis and Characterization Routines (TRACR) pipeline—a scalable, open source, Python based col- lection of image processing and statistical tools intended for image feature characterization in various forms of 2D and 3D visual data. Morphological distinctions between virgin and recycled Inconel 718 powders for additive manufacturing are explored. Porosity profiles in selectively laser melted Inconel 718 compression cylinders are investigated through the lens of several printing parameters and post-processing regimes.