Additively manufactured (AM) components contain discontinuities, indications and defects that can change the component’s mechanical performance during high energy impact events. X-ray computed tomography (XCT) reconstructions of AM metastable titanium disks (Ti-5Al-5V-5Mo-3Cr or Ti-5553) were generated on an industrial micro-focus system. Each sample was scanned before and after high velocity impact testing. The porosity resulting from the direct metal laser sintering (DMLS) powder bed fusion machine was detected and characterized. The samples were placed in a gas gun configuration to induce a high-rate tensile load (shock test). The post-test results on the recovered disks contained incipient spall cavities. These features were identified by standard volume segmentation techniques. This inspection data is also evaluated with machine learning (ML) algorithms. A comparison between ML segmentation of the pores/cavities to standard commercial segmentation algorithms will be presented. Improvements using ML were specifically seen in the identification of pores and spall planes in regions of low x-ray attenuation (brightness and contrast). Common XCT artifacts, which include beam hardening and systematic noise, were overcome by the applied ML methods. Porosity and spall plane regions identified by the machine learning analysis were then compared to serial sectioning and scanning electron microscope (SEM) data to judge the precision and accuracy of the machine learning technique. Results show that the CT reconstructed porosity aligned well with the serial sectioned and SEM data. The only discrepancies were in the small pores near the detectability limit and other metrics dependent on the XCT reconstruction resolution.
 J. Schindelin, I. Arganda-Carreras, E. Frise, V. Kaynig, M. Longair, T. Pietzsch, S.
Preibisch, C. Rueden, S. Saalfeld, B. Schmid, J.-Y. Tinevez, D. White, V. Hartenstein, K.
Eleceiri, P. Tomancak and A. Cardona, "Fiji: an open-source platform for biologicalimage
analysis," Nature Methods, vol. 9, pp. 676-682, 2012.
 I. Arganda-Carreras, V. Kaynig, C. Rueden, K. Eliceiri, J. Schindelin, A. Cardona
and H. S. Seung, "Trainable Weka Segmentation: a machine learning tool for microscopy
pixel classification," Bioinformatics, vol. 33, no. 15, pp. 2424-2426, 2017.
68 Page Views
0 PDF Downloads
0 Facebook Shares