Analysis of Porosity in Additively Manufactured Parts using 3D Metallographic Analysis Compared to Conventional 2D Stereology Estimates

Porosity is a typical defect in additively manufactured (AM) parts. Such defects limit the properties and performance of AM parts, and need to be characterized accurately. Current methods for characterization using microstructure rely on classical stereological methods that extrapolate information from two dimensional images. Newer three dimensional (3D) microstructural analysis techniques provide an opportunity to precisely and accurately quantify porosity in materials. In this work, we analyzed the porosity of an additively manufactured alloy (Ti 6Al 2Sn 4Zr 2Mo (Ti6242)) using Robo-Met.3D, an automated system for 3D materials characterization. The results from this 3D serial sectioning analysis were then compared to classical 2D stereological methods (Saltikov analysis). Mean pore diameter within 2D slices analyzed increased from early to later slices, and varied by over 20μ from minimum to maximum, as well as from the 3D measurement. The results suggest that acquiring 3D experimental data to measure porosity in a volume may yield a more complete analysis. Introduction and Motivation of the Study It is important to understand the size, distribution and morphology of porosity that helps in predicting the mechanical performance of a material [1]. Until the advent of 3D analysis techniques, analysis was limited to classical stereological methods that extrapolate information from two dimensional images. There is always a level of uncertainty associated with classical approaches 2D in overestimating or underestimating the true feature size and shape. In this study, we use Robo-Met.3D, an automated system for 3D characterization of materials, to acquire and analyze a series of 2D images through the section of a Ti6242 component fabricated through additive manufacturing [2]. Conventional 2D analysis, involving single images, is compared to the analysis output of a 3D reconstruction of the material section. Experimental Methods Material Composition and State: An additively manufactured Ti 6Al 2Sn 4Zr 2Mo (Ti6242) sample, produced by electron beam melting, was excised from a larger part. The sectioned piece had the dimensions of 20mm x 20mm x 18mm in length, width and depth respectively. The sample was conventionally metallographically mounted in a powdered thermoset compound. Image Acquisition: The experiment set up was predicated on capturing all porosity with a feature size greater than 5μ. After automated grinding and polishing at 9, 6, 1 and 0.05 μ, optical images were automatically acquired with the Zeiss Axiovert microscope built into the Robo-Met.3D system. A 10 X objective with 1X optivar was used, with bright field illumination. Seventy five (75) slices of 16 image tiles each were used for 3D reconstruction. Post Processing: The 2D image tiles from each layer were stitched into 4 x 4 montages, registered with the images from the next layer using Fiji and Image J. Binary images for 2D analysis were made by selecting a pixel intensity threshold using Fiji/Image J (Fig 1, Left). These images were stacked and aligned using Image J, which was used to calculate the Feret and equivalent sphere diameter of the pores visible in this two-dimensional image. A distribution of pore size versus cumulative fraction % was created (2D measured, single image). Saltikov’s area analysis was used to extrapolate a 3D equivalent pore diameter distribution from a set of 2D measurements of pore intersections with a plane [3,4]. Next, 3D datasets were reconstructed, visualized and analyzed in 3D using Dream.3D and Paraview (Fig 1, right). Key Results and Conclusions An analysis was undertaken to compare Robo-Met.3D direct measurements of the pore size distribution in a volume of material with the 2D to 3D extrapolation calculation using classical stereology methods. The 3D analysis yielded a total of 571 pores, detected with an overall range 8.3 158.9μ. We also compared the mean pore diameter for each 2D section in comparison to the mean pore diameter for the 3D reconstruction. Our analysis suggests that mean pore diameter varied significantly from slice to slice, and could be quite dissimilar from the true distribution. The variation in mean pore diameter within 2D slices analyzed was over 20μ from minimum to maximum (Fig. 2). Histogram analyses in feature-tagged datasets using MIPAR software revealed that, 2D analyses tended to underestimate smaller pores, and underestimate larger pores. Reliance on 2D analyses alone could mis-estimate true porosity in a component. A 3D reconstruction could therefore provide higher fidelity than projection techniques. References: [1] Y. Zhu, et al., Materials Science and Engineering: A, 607, (2014) p427. [2] J. Madison, et al., (2015). 3D RoboMET Characterization (No. SAND2015-8957R). Sandia National Laboratories (SNL-NM), Albuquerque, NM (United States). [3]S.A. Saltikov, The determination of the size distribution of particles in an opaque material from a measurement of the size distribution of their sections in Stereology (Springer Berlin, 1967), p163. [4] Shen, H., et al., Mechanics of materials, 38(8), (2006), p933-944. Figure 1. Conventional 2D Micrograph of Porosity vs 3D Reconstruction Figure 2. Large variation in mean Feret diameter of porosity between slices in a sample volume, compared to the 3D reconstruction mean.