Efficacy of X-ray Computed Tomography for Flaw Detection in Laser Powder Bed Fusion Additive Manufacturing

Flaws in additively manufactured (AM) material are thought to contribute to variability in both static and dynamic mechanical properties, limiting use of these components in fatigue applications. X-ray computed tomography (XCT) is often used to quantify flaw populations in AM material since traditional means of non-destructive inspection (NDE) are challenged by the complex geometry and rough surfaces of AM components. However, we show that the flaw detectability of XCT data can be lower than 5%, even with seemingly high quality data and a voxel size of 10 µm. A custom automated defect recognition (ADR) algorithm was used to identify flaws in both XCT and automated, optical serial sectioning (SS) data of a 0.63 mm thick, cylindrical section of additively manufactured Ti-6Al-4V. The cylinder was built on a commercial, laser-based powder bed fusion (L-PBF) system using standard processing parameters, and over 1000 flaws ranging from 5-70 µm in diameter were identified in the SS data. In the corresponding region of the CT data, only 52 flaws were identified following manual verification of ADR-identified flaws in both data sets. Our results show that while many of the spatial and morphological characteristics of the flaw populations, such as the location, orientation, sphericity, elongation, etc., are similar between the two data sets, XCT underpredicts the overall part porosity in the region analyzed. Furthermore, flaws absent from the XCT data do not correspond to small flaws below the voxel size of XCT. Instead, they are comprised of randomly distributed flaws spanning a range of sizes. We conclude that the voxel size alone is not a sufficient parameter for determining the adequacy of XCT data for detecting AM flaws and recommend two additional, quantitative metrics for determining CT image quality: (1) the Modulated Transfer Function (MTF) at 10% modulation and (2) the Contrast Discrimination Function (CDF).