<p>The ability of metal additive manufacturing (AM) to produce parts with complex geometries from hard-to-process materials manifests its potential to revolutionize manufacturing. However, repeatability and product quality remains imposing barriers towards scaling metal AM to production environments. The significant challenge of metal AM is the occurrence of various defects such as cracks, delamination, distortion, rough surface, lack of fusion, porosity, foreign inclusions, process instability (keyhole and balling) that deteriorate the build strength, hardness, and fatigue life. The recent advancement in imaging technology leads to a data-rich environment in AM and provides a unique opportunity to enhance the understanding of defect formation and its interactions. Our prior studies, as well as the-state-of-the-art image-guided methods, have been concentrated on analyzing the impact of process settings (e.g., scanning velocity, laser power, and hatch spacing) and design parameters (e.g., height, width, and build orientation) on a single type of anomaly (e.g., porosity, discontinuity, and geometry distortion). Therefore, the simultaneous occurrence of multiple defects and their correlations on quality characterization have been ignored. Few, if any, previous works have incorporated a data-driven model to unveil the relationship between various anomalies and characterize their relationship according to design parameters (orientation, thickness, and height). This study introduces a novel data-driven model to simultaneously quantify interactions between design parameters and heterogeneous anomalies. Specifically, we propose a multi-response modeling and analysis methodology for design-quality interactions in the metal AM build based on a real-world case study on complex thin-wall structures. In our experiment, three thin-wall parts were built from Spherical ASTM B348 Grade 23 Ti-6Al-4V powder with a size distribution of 14-45&nbsp;µm,&nbsp;on an EOS M280 PBF machine. The laser power and velocity settings are 340 W and 1250 mm/s, respectively. Each part was built vertically with a layer thickness of 60 microns in three orientations (i.e., 0, 60, and 90) with respect to the travel direction of the recoater blade. In addition, we also varied the thin-wall width, and thin-wall height, and hatch spacing in our design. These thin walls are inspected using novel computed tomography (CT) based measurement techniques to assess their build quality. The post-build XCT data is obtained on General Electric V|tome|X system with a voxel size of 15 µm3. First, we collect the post-processing X-ray computed tomography (XCT) data and register with the pre-processing computer-aided design (CAD) slices to convert data from different builds into the same coordinate system. Second, we extract quality features, namely thickness, roughness (LWR), and density, from each registered XCT slice. Third, we model the relationship between extracted design features and design parameters through a novel multi-output Gaussian process. The proposed model exploits the correlation between outputs such that one output can leverage information from another. In this way, the multi-output Gaussian process is able to provide a more accurate prediction in comparison to the single-output models. We further compare the experimental results from the proposed multi-output Gaussian process model and the classic single-output Gaussian process model. Results show that the proposed methodology improved the RMSE by 31.25%. Finally, we perform the design-quality quantification through the proposed model. The proposed research enables multi-task learning and processing in AM process monitoring and control.</p>

Image-Guided Multi-Response Modeling and Characterization of Design Defects in Metal Additive Manufacturing

Paper Proceedings: https://asmedigitalcollection.asme.org/IMECE/IMECE2021/volume/85543

ASME 2021 International Mechanical Engineering Congress and Exposition (IMECE2021)

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.