Additive Manufacturing: Qualification and Certification: Approaches and Challenges
Sponsored by: TMS Additive Manufacturing Committee, TMS: Mechanical Behavior of Materials Committee, TMS: Nanomaterials Committee
Program Organizers: Faramarz Zarandi, RTX Corporation; Jacob Hochhalter, University of Utah; Douglas Wells, NASA Marshall Space Flight Center; Richard Russell, NASA Kennedy Space Center; Mohsen Seifi, ASTM International/Case Western Reserve University; Eric Ott, GE Additive; Mark Benedict, Air Force Research Laboratory; Craig Brice, Colorado School Of Mines; J Hector Sandoval, Lockheed Martin
Monday 8:00 AM
November 2, 2020
Room: Virtual Meeting Room 8
Location: MS&T Virtual
Session Chair: Faramarz Zarandi, Raytheon Technologies Research Center; Douglas Wells, NASA / Marshall Space Flight Center
8:00 AM
Introductory Comments: Additive Manufacturing: Qualification and Certification: Faramarz Zarandi1; 1Raytheon Technologies
Introductory Comments
8:05 AM Invited
Physics-based Qualification for Laser Powder Bed Fusion AM: Anthony Rollett1; 1Carnegie Mellon University
A multi-institution team is investigating a physics-based approach to qualification of laser powder bed fusion (LPBF) machines. By printing Ti-6Al-4V under a variety of conditions of power, speed and hatch spacing, the dependence of porosity and fatigue strength is quantified. The same model of LPBF machine is used at different institutions, along with other makes. Process windows are defined by keyhole porosity on the high power side and lack-of-fusion porosity on the high speed side. 4-point bend fatigue provides an efficient method for quantifying variations in fatigue. Amongst many other factors, controlling raster length is crucial for avoiding defects. A related effort in wire-feed robotic welding of scandium-modified Al alloys also leads to a process window in P-V space bounded by a limit on porosity that provides an analogous basis for qualification. Contributions from the many team members and support from the NASA ULI Program are gratefully acknowledged.
8:35 AM
A Comprehensive Digital Platform for Additive Manufacturing: Luke Scime1; Alka Singh1; Daniel Robertson1; Brandon Mathis1; William Halsey1; James Haley1; Samuel Leach1; Kyle Saleeby1; Amir Ziabari1; Michael Sprayberry1; Derek Rose1; Ryan Dehoff1; Vincent Paquit1; 1Oak Ridge National Laboratory
For Additive Manufacturing to be used in critical applications, part certification is critical. Furthermore, the printer is only one step in any manufacturing chain, with every physical link in this chain requiring a corresponding “digital twin” as part of any robust certification process. In this presentation, the Digital Platform at Oak Ridge National Laboratory’s Manufacturing Demonstration Facility will be discussed. This platform integrates data from in-situ sensors, feedstock provenance, machine maintenance and calibration histories, and metrology characterizations. Each part’s digital twin is defined as a list of operations (e.g. additive prints, heat treatments, subtractive machining, tensile tests) each with associated in-situ and meta data. The Digital Platform enables not only classical certification schemas, but also the use of Artificial Intelligence for defect detection, process control, and prediction of part performance. Examples from the Transformational Challenge Reactor program will be explored, and selected data analytics tools will be described.
8:55 AM
Similarity Analysis and Clustering of Thermal History to Understand Process-structure Relationships: Sujana Chandrasekar1; Jamie Coble1; Amy Godfrey1; Serena Beauchamp1; Fred List III2; Vincent Paquit2; Sudarsanam Babu1; 1University of Tennessee; 2Oak Ridge National Laboratory
Part qualification is challenging in AM processes like selective laser melting (SLM) due to variation in microstructure and associated properties with changing thermal history. Thermal cycling is inherent to SLM and dictated by laser parameters, scan patterns, geometry and defects. In this work, we develop time series analysis methods to identify similarities in thermal history using data from infrared monitoring of SLM process. Similarity is defined using Euclidean distance between infrared data from individual points across layers and time. Clustering is done using results of similarity analysis. Clustering results indicate regions with similar thermal histories within different layers of the part. Clusters change with scan angle and part cross-section. Clustering results are correlated with characterized microstructures, which change with thermal history. This work demonstrates potential in time series analysis of infrared data for understanding process-structure relationships in AM, which can be complemented by thermal modeling.
9:15 AM
Effect of Sample Geometry and Orientation on Tensile Properties of Ti-6Al-4V Manufactured by Electron Beam Melting: Gitanjali Shanbhag1; Evan Wheat1; Shawn Moylan2; Mihaela Vlasea1; 1University of Waterloo; 2National Institute of Standards & Technology
Tensile samples are often used to qualify builds and materials, however, there is currently no single standard sample geometry for additive manufacturing that is widely adopted to achieve this. While there is already work demonstrating the difference in measured properties between tensile samples produced in different build orientations, this does not extend to different sample geometries. To study the extent of geometry on product qualities for tensile samples, a selection of standard coupon types across the range provided in ASTM E8 and E8M was prepared using an Arcam A2X EBM machine and Ti-6Al-4V as the representative material. These coupons were characterized to observe any porosity defects, dimensional deviations, and surface topography that could impact part performance. This study presents the results of the tensile testing and evaluations of any systematic correlations of mechanical performance with respect to the various sample geometries and sample properties.
9:35 AM
The Effects of Powder Particle Size Distribution on the Powder and Part Performance of Laser Powder Bed Fusion 17-4 PH Stainless Steel: Jordan Weaver1; Justin Whiting1; Carlos Beauchamp1; Max Peltz1; Thien Phan1; Vipin Tondare1; Jared Tarr1; Alkan Donmez1; 1National Institute of Standards and Technology
The powder particle size distribution (PSD) is an important precursor material specification; however, the effects of small changes in the PSD on powder and part performance are not fully realized. In this study, three Argon atomized 17-4 PH stainless steel powders with small shifts in their PSD’s (fine, medium, and coarse) were characterized and used to build samples for metallography and mechanical testing. Powder performance measurements (tapped density, Hall flowmeter, rheometer, pycnometer, and spreading) provide a ranking of the powders from best to worst as coarse, medium, and fine. However, microstructure quantification (density, grain size, and crystallographic texture) and mechanical testing (Rockwell hardness and uniaxial tension) reveal that all three powders produce nearly identical, acceptable parts with slightly higher hardness and strength for the medium powder. The results reveal the process is robust for static mechanical properties with regards to small changes in PSD with a range in powder performance.
9:55 AM
Connecting Metal Powder Morphological Characteristics with Flowability Properties Using Machine Learning: Srujana Rao Yarasi1; Andrew Kitahara1; Ryan Cohn1; Elizabeth Holm1; Anthony Rollett1; 1Carnegie Mellon University
Computer vision and machine learning techniques are used to quantify the morphological characteristics of several different metal powders used in Powder Bed Fusion AM and connect them with their flowability. A framework is constructed to understand their differences in flowability based on their powder morphology distributions (PMDs), which includes powder sizes. These PMDs are obtained from SEM images of the powder particles. This is accomplished by using pre-trained convolutional neural networks to generate a feature vector for each powder particle, sets of which are then clustered according to morphological similarity, leading to a powder morphology distribution (PMD) for each powder system. Several machine learning algorithms are tested to correlate the PMDs to their flowability properties. Powder size metrics are also explored as a way to predict flowability behavior in powder bed fusion machines.