Session: 02-01-02: 7th Annual Conference-Wide Symposium on Additive Manufacturing: Metals II

Paper Number: 95145

95145 - Dependency Evaluation of Defect Formation and Printing Location in Additive Manufacturing 

Laser powder bed fusion (LPBF) additive manufacturing fabricates objects by spreading a thin coat of metal powder across a build plate and then selectively fusing particles in a layer-by-layer manner. The process not only offers a promising route to the manufacture of tight-tolerance and near-net-shape components, but also brings the ability to operate under extreme loads in environmentally challenging conditions. Despite the transformative capability of LPBF, various process factors ranging from design to material may lead to inconsistent repeatability and quality issues from build to build, which eventually degrade their functional and mechanical properties. Recent advancements in in-situ imaging and data-driven learning have shown a great potential to characterize the impact of process parameters on build quality. However, the current analysis overlooks the effect of the print location and scan strategy. In fact, the printing location in build plate and scan strategy can significantly influence density and distortion of the final object. In addition, the shape and intensity distribution of the laser spot varies if builds are positioned near the edge of the platform because major scanner deflection angles can convey different energy densities than the nominal setting. In addition to the printing location, scan strategy governs the melting, remelting, and solidification phenomena, thereby affecting porosity formation in LPBF build. The goal of this work is to understand the correlation between process parameters (i.e., location and scan strategy) and quality performances (i.e., density and distortion) in the LPBF process. In pursuit of this goal, the objectives of this work are two-fold: First, a systematic image-guided solution is proposed to characterize the influence of process parameters on meltpool signatures (i.e., meltpool length and width and the number of spatters). Second, a self-supervised learning model is designed to relate the meltpool characteristics with the quality performances of final builds. Four identical components are fabricated in four different locations of build plate. Each component has the size of 9 mm × 5 mm × 5 mm rectangular prism and a 45-degree overhang feature as well as a horizontal cylindrical cutout. Each build has a layer thickness of 20 microns and has 250 layers printed from nickel superalloy 625 (IN625). All components have a hatch space of 100 microns and a scan strategy of 90-degree rotations between each layer. A co-axial meltpool monitoring camera acquires meltpool images during the fabrication, and Zeiss Metrotom 800 captures post-build X-ray computed tomography (XCT) scan data. To study the effect of process factors on in-situ process signatures, we first extract the meltpool characteristics (i.e., length, width, and the number of spatters) from binarized images from five consecutive layers for each build. Then quantify the impact of print location and scan strategy on these features. The experimental results demonstrate that as components locate further away from the center of the build plate, the meltpool length and width decrease while the number of spatters increases. Also, the short-length scan path leads to the longer-length meltpool. The correlation result between meltpool characteristics and quality performances shows that the proposed self-supervised learning model has a unique capability to map the signature-quality relationship compared to the other state-of-the-art models (i.e., ARIMA and XGBoost). The results also show that the length of meltpool is the most conducive feature for the analysis of quality performances. As the length of meltpool shrinks, the layer density decreases by 47.99%, and distortion increases by 15.86%. In addition, meltpool features extracted from components closer to the center of the build plate with a shorter scan path have a smaller variation than objects located further with a longer scan path. In summary, the proposed methodology helps identify a comprehensive insight into the impact of the location of the component printed on the build platform and scanning strategy on the LPBF process.

Presenting Author: Kosar Safari University of Connecticut

Presenting Author Biography: Kosar Safari is a Ph.D. student in the Department of Mechanical Engineering at the University of Connecticut (UConn). She received her bachelor's and master's degrees in Aerospace Engineering from Sharif University of Technology and K. N. Toosi University of Technology, respectively. Her research focuses on developing novel sensing, artificial intelligence, and control in advanced manufacturing.

Authors:

Kosar Safari University of Connecticut
Shihab Khalfalla University of Connecticut
Farhad Imani University of Connecticut