Understanding Powder Feedstock-Part Performance Relationships in Additive Manufacturing

Abstract

In this dissertation, the relationships between the powder feedstock and the mechanical performance of additively manufactured (AM) parts were investigated. In this regard, different aspects of powder feedstock were considered. First, the effects of the particle size distribution (PSD) and shape on tensile and fatigue performance of AM Ti-6Al-4V parts via a laser powder bed fusion (L-PBF) were investigated. Two plasma atomized Grade 23 Ti-6Al-4V powder batches were used with PSDs of 15-45 µm (fine) and 15-53 µm (coarse) with highly spherical particles. Both batches showed comparable powder flowability due to very spherical powder particles. The fine powder, however, showed some improvement in packing state due to the narrower PSD resulting in fewer empty spaces between the powder particles. The powder characteristics resulted in negligible changes in tensile strength, which were attributed to the similar microstructures in the specimens manufactured from both powders. The tensile ductility, however, improved by using the coarse powder ascribed to the smaller defects in coarse powder specimens. Although the overall number of defects was considerably lower in fine powder specimens, some large defects were noted in the case of fine powder specimens, which was explained by more spattering during fabrication with this powder caused by higher O content. Such large defects were found detrimental to both tensile ductility and fatigue performance. In addition, no variation was observed along with the powder delivery system, related to the excellent flowabilities of both powders. Second, the plasma atomized Ti-6Al-4V powder changes throughout L-PBF fabrication and with continuous reuse were traced. It was found that continuously reusing the powder can improve the powder packing state and flowability. However, the O content also increased with powder reuse. Although reusing the powder up to 3 times improved fatigue performance, further reuse up to 7 times resulted in adverse effects compared with the unused powder. The lower fatigue performance and tensile ductility of specimens manufactured using 7-times reused powder were correlated with the higher O content in 7-times reused powder batch, which resulted in more spattering during the fabrication and formation of large defects. This hypothesis was justified by observing larger and more frequent defects downstream of the argon flow, which is commonly used as the shielding gas during AM fabrication. The specimens in the downstream region also had slightly higher tensile strength owing to the higher O content, lower fatigue performance, and higher critical energy release rate. This study was repeated with gas atomized 17-4 precipitation hardening (PH) stainless steel (SS) with a nominal PSD of 15-45 µm to investigate the build location dependency with a powder consisting of also less spherical particles and track the changes with continuous powder reuse. It was realized that the powder behavior, including flowability and packing state, did not change the tensile strength and the microstructure of the L-PBF 17-4 PH SS specimens even though they improved with continuous reuse. Fatigue performance in the high-cycle-fatigue (HCF) regime, however, significantly improved with powder reuse, ascribed to the smaller gas-entrapped pores in the specimens manufactured using 14-times reused powder. It was also noted that when the powder does not contain highly spherical powder particles, the lower powder flowability may jeopardize the powder bed layer uniformity and result in some variation along with the powder delivery system. For instance, more frequent and larger lack-of-fusion defects, which are detrimental to fatigue performance, were observed in the specimens placed west of the build plate (i.e., farther away from the feedstock bin). Owing to the powder flowability improvement by continuously reusing the powder, this location dependency across the build plate decreased. Observing such variations across the build plate signified the importance of developing standardized powder atomization and reuse practices to ensure consistent build quality. The build-to-build variability due to the effects of powder reuse was also investigated in different AM technologies, including laser powder directed energy deposition (LP-DED) using NASA HR-1 superalloy with a nominal PSD of 45-105 µm. Multiple sets of thin walls were deposited using the unused NASA HR-1 powder as well as 1- to 5-times reused powder. The powder flowability did not change with continuously reusing the powder. The chemical composition also remained constant throughout all depositions. The PSD, however, coarsened with repetitive powder reuse resulting in fine particle reduction and mean particle size enlargement. Investigating the powder particle shape did not reveal any changes with reuse. Therefore, it was assumed that highly spherical powder particles resulted in comparable flowability in all powder batches, although fine particles (typically known for lower flowability) decreased with powder reuse. The tensile strength and ductility were also investigated. No change in tensile strength was noticed due to the similar microstructures and O content in the LP-DED NASA HR-1 specimens. All deposited specimens had relative densities larger than 99.997%, and the maximum defect size in all specimens was almost comparable, which was correlated with the similar flowability of all powders. On the other hand, the similar defect content also resulted in identical tensile ductilities and fatigue performance. Moreover, the effects of specimen orientation with respect to the building direction (i.e., perpendicular or parallel) was investigated to better correlate the specimen property in the laboratory to part performance. The tensile strength did not change with the orientation owing to the similar microstructures, defect content, and O concentration. The ductility, however, was somewhat lower in parallel specimens due to the more critical effects of surface undulations when subjected to tensile loading. The lower ductility of parallel specimens also showed slightly lower resistance to fatigue failures in the low-cycle-fatigue (LCF) regime. In this regime, the majority of fatigue life is spent in crack growth, and more ductile materials are typically favored. Therefore, the lower fatigue performance could be correlated with the lower ductility of parallel specimens. In the HCF regime, however, negligible changes with a greater scatter in data were observed owing to the dominant effects of micro-notches on the surface resulting from the layer-by-layer deposition process. Lastly, a single powder quality metric was identified to correlate several powder characteristics. Such quality metrics and similar ones could contribute to faster adoption of AM technologies by indicating the effects of powder on the final part performance without the need for the experiment. The proposed powder quality metric could satisfactorily predict the powder behavior and correlate with observed mechanical properties in this dissertation. In addition, an image processing method was used to obtain the powder PSD and shape information (e.g., sphericity) from the 3D reconstructed images via X-ray computed tomography (CT). Subsequently, a machine learning algorithm was implemented via a feed-forward artificial neural network (ANN) multi-layer perceptron. Due to cost-effectiveness, developing statistical tools such as machine learning algorithms is essential for the faster adoption of AM technologies. The proposed ANN model was trained using the PSD and sphericity as well as the powder rheological characteristics. It was observed that the ANN model could effectively predict the powder quality score with a low mean absolute percentage error (MAPE), indicating the model's robustness.