Additive Manufacturing Modeling, Simulation, and Machine Learning: Microstructure, Mechanics, and Process: AM Processes
Program Organizers: Jing Zhang, Indiana University – Purdue University Indianapolis; Brandon McWilliams, US Army Research Laboratory; Li Ma, Johns Hopkins University Applied Physics Laboratory; Yeon-Gil Jung, Korea Institute of Ceramic Engineering & Technology

Wednesday 2:00 PM
October 12, 2022
Room: 303
Location: David L. Lawrence Convention Center

Session Chair: Brandon McWilliams, CCDC Army Research Laboratory; Li Ma, Johns Hopkins University Applied Physics Laboratory; Yeon-Gil Jung, Changwon National University; Jing Zhang, Indiana University - Purdue University Indianapolis


2:00 PM  
Tailoring Formation of TCP Phases during Additively Manufactured Inconel 625 by CALPHAD-based Simulations with Experimental Validations: Hui Sun1; Shun-Li Shang1; Shipin Qin2; Allison Beese1; Zi-Kui Liu1; 1The Pennsylvania State University; 2Align Technology
    Detrimental topologically close-packed (TCP) phases are frequently observed in additive manufacturing (AM) due to elemental segregation and fast-cooling rate. Here, we investigate the formation of the TCP phase δ in Inconel-625 fabricated using laser powder bed fusion AM through an integrated computational and experimental approach. The unwanted  δ phase grows much faster in the samples subjected to stress-relief heat treatments than those in wrought samples due to fast-cooling rates during AM, resulting in segregations of Nb and Mo. CALPHAD-based thermodynamic modeling was performed aided by first-principles calculations based on density functional theory, which provides energetics for proper modeling of δ. The new database is used to design the chemistry within the alloy specification tolerance and identify suitable processes of heat treatment to decrease the formation of δ phase. It is shown that the combination of finite element predictions and kinetic simulations enables the modification of elemental segregation by adjusting AM processing parameters.

2:20 PM  
Discrete Element Method Based Simulations of Metal Powder Pouring and Raking Processes in Additive Manufacturing: Michael Fazzino1; Ummay Habiba1; Rainer Hebert1; Serge Nakhmanson1; 1University of Connecticut
    Discrete Element Method based approach implemented in the LIGGGHTS package was used to construct ‘digital twins’ of Ti 6-4 powder pouring and raking processes. Adjustable parameters of the pouring process digital twin were validated by an ASTM B213 standard Hall Flowmeter Funnel experiment, with experimental data for the particle size distribution used as input. Validation included comparing the general shape of the particle pile, such as its diameter, height, and slope angles to the experimental results. Local particle size distributions were obtained for different areas within the pile (e.g., top vs. bottom area), indicating the dominance of larger particles at the top of the pile, akin to the Brazil nut effect. Validated material parameters were then utilized in construction of a digital twin of the particle pile raking process over a powder bed, with local porosity and particle size distribution after the raking evaluated and compared with the experimental results.

2:40 PM  
Planar and Full-Process Modeling of the Powder-Bed Fusion Ti-6Al-4V Columnar-to-Equiaxed Transition Behavior: Brodan Richter1; Joshua Pribe2; George Weber1; Edward Glaessgen1; 1National Aeronautics and Space Administration; 2National Institute of Aerospace
    Recent advancements in powder-bed fusion (PBF) microstructure simulation using the Monte Carlo technique have enabled increased length- and time-scale simulations. These larger simulations enable the generation of engineering-relevant pseudo-microstructures for use in Process-Structure-Property frameworks to relate processing parameter to final part performance. However, the interaction of solidification alongside the rapid melt pool motion of additive manufacturing creates challenges when validating and calibrating the technique to new alloy systems. This work presents a reduced-scale planar solidification version of the PBF Monte Carlo technique for comparing the simulated columnar-to-equiaxed transition (CET) to experimentally determined CET for Ti-6Al-4V. The influence of modifying the nucleation parameters and dendrite tip velocity on the resulting CET map, grain structure, and texture are shown and compared to simulations of the full-scale PBF model. The results of this work demonstrate a method for the calibration of the PBF Monte Carlo technique for application to new material systems.

3:00 PM  
Powder Bed Packing Density Dependence on Particle Size Distribution: Simulation and Experimental Analysis: Ummay Habiba1; Michael Fazzino1; Serge Nakhmanson1; Rainer Hebert1; 1University of Connecticut
    This study examines powder particle distributions, actual and model distributions for laser powder bed fusion additive manufacturing. The study aims to determine if a powder size distribution exists that maximizes the particle packing density after powder spreading. The size distributions selected for this study were monodispersed, bimodal, and trimodal distributions, and a size distribution used in standard additive manufacturing practice. Packing densities were determined using the Discrete Element Modeling method (DEM) and for the experimental powder size distribution based on an experimental approach developed by NIST. The results indicate an overall increase in packing density during the first layers, approaching constant values after about 10 layers. An effect of the size distributions on packing density is detected, but future melting studies will show if the differences affect microstructures and defect formation of additively manufactured samples.

3:20 PM  
Rapid Qualification of Wire Feed Direct Energy Deposition Process Builds Using ICME Approach: Amit Verma1; Andrew Huck1; Rajib Halder1; Anthony Rollett1; 1Carnegie Mellon University
    To leverage the potential benefits of AM, the role of geometry in the qualification of parts needs to be minimized such that any arbitrary geometry can be printed with confidence. This translates to mapping the build as it is deposited, i.e., qualification of a small volume rather than of the entire build at the end of the deposition process and reducing the post-process characterization as much as possible. Currently, a suite of ICME (Integrated Computational Materials Engineering) tools in conjunction with in-situ process monitoring supported by artificial intelligence (AI) / machine learning (ML) methods provide a range of opportunities. This talk focuses on the Wire feed Direct Energy Deposition AM process, applied to Ti-6Al-4V, for which we employed a host of modeling and simulation tools, along with in-situ process monitoring, to accelerate the qualification process.

3:40 PM Break

4:00 PM  
Towards Qualification and Certification of Laser Powder Bed Fusion Ti-6Al-4V with In-Situ Process Monitoring and Automated Defect Detection: Andrew Kitahara1; Samuel Hocker2; Brodan Richter2; Wesley Tayon2; Joseph Zalameda2; Edward Glaessgen2; 1National Institute of Aerospace; 2NASA Langley Research Center
    Qualification and certification of laser powder bed fusion (LPBF) parts are two challenges that must be answered to ensure suitability for critical applications. In-situ monitoring using high frame rate thermal and conventional optical imaging sensors is applied to the LPBF build process. Currently, the large volume of data from such sensors becomes untenable for manual inspection in production environments. This presentation serves to address this in-situ monitoring deficiency in two ways. First, a framework for managing data streams from LPBF process monitoring sensors is described. Second, two candidate image analysis techniques are presented: one is a set of heuristics that are easily interpretable, and the other is an uninterpretable convolutional neural network. These strategies are compared in terms of performance, computational expense, and speed. These methodologies represent platforms for connecting processing conditions to process modeling efforts aligned with the qualification and certification mission for LPBF Ti-6Al-4V components.

4:20 PM  
The Effect of the Process Environment on Gas and Particle Entrainment in Laser Powder Bed Fusion: Michael Stokes1; Saad Khairallah2; Alexey Volkov1; Alexander Rubenchik2; 1The University of Alabama; 2Lawrence Livermore National Laboratory
    The laser irradiation of metal powders induces strong evaporation forming a vapor microjet that entrains powder metal particles causing denudation and spatter defects. The present work theoretically investigates the effect of the type of the ambient gas, its pressure, and surface temperature on the vapor jet structure and corresponding gas and powder particle entrainment. Gas kinetic simulations of the microjet are performed using the direct simulation Monte Carlo (DSMC) method. Trajectories of metal powders are also simulated to predict spatter-related defects. The results indicate that the structure of the microjet heavily depends on the processing environment. Changes in the surface temperature, pressure, and surface geometry affect the transition from subsonic to supersonic flow. The results also show that the degree of particle entrainment, e.g., the thickness of the surface denudation zone, strongly depends on the molar mass of the ambient gas, and characteristic vapor microjet velocity. LLNL-JRNL-829362; This work was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under contract DE- AC52-07NA27344. Lawrence Livermore National Security, LLC. LLNL-ABS-829363. This work was also partially supported by NSF through awards CMMI-1554589 and CMMI-1663364.

4:40 PM  
Thermomechanical Modeling of Axisymmetric Geometries for Laser Hot Wire Additive Manufacturing: Elizabeth Chang-Davidson1; Brandon Abranovic1; Jack Beuth1; 1Carnegie Mellon University
    Large-scale laser hot wire additive manufacturing shows promise for production of parts difficult to manufacture using conventional methods, as well as unique benefits relative to other additive processes. However, as this is a relatively new process, modeling work is required for part deposition planning. Axisymmetric geometries are of particular interest due to their frequency in real part geometries. Two major considerations when printing are ensuring consistent material properties throughout the part, as well as excessive part deformation in the build process. Using semi-analytical as well as finite element methods, thermal histories as well as part deformations for different deposition strategies in axisymmetric geometries were modeled in order to select optimal print conditions. Once the prints were completed, experimental results were compared to model predictions and used to further refine the models. As a result, a methodical process for selecting deposition strategies for successful printing of axisymmetric geometries was developed.

5:00 PM  
Studying Melt Pool Variation and Its Effects on the Formation of Porous Defects via GPU-based Process Simulation: David Anderson1; Chaitanya Vallabh1; Shawn Hinnebusch1; Xiayun Zhao1; Albert To1; 1University of Pittsburgh
    The development of parts using Laser-Powder Bed Fusion (L-PBF) additive manufacturing processes face several challenges that stem from porosity formation. These porosities are typically generated from under-heating or over-heating the powders, also known as lack-of-fusion and keyholing, respectively. Although these two occurrences are closely linked to printing parameters such as laser scan speed, laser intensity, hatch spacing, and scan path orientation, the independent effects of each of these parameters on porosity formation is not well documented. Through GPU-based computational modeling, the effects of these process parameters on the melt pool geometry have been simulated. By comparing these simulations with serial cross-sectioning data of printed samples, these results aid in calibrating the simulations to predict melt pool geometries more accurately, that may lead to keyholing or lack-of-fusion, and thus porous defects.

5:20 PM  
In Situ Confocal Imaging and Quantification of Defects in Binder-Jet Printed (BJP) Steel Parts: Pooja Maurya1; P.Chris Pistorius1; 1Carnegie Mellon University
    The structural integrity of BJP steel parts is determined by defects like porosity/oxide inclusions. Sintering being the critical step for improving its mechanical strength, any in situ diagnostic tool to monitor it obviously helps. In the present work, Confocal Laser Scanning Microscopy (CLSM) integrated with dew point controller is utilized to monitor and understand evolution of porosities and oxide inclusions in 316L BJP parts during high temperature sintering (~1380℃). The real time imaging from a particular focal plane, with high sensitivity to the variation in its topographic features, along with control of sintering atmosphere effectively controls the process integrity. Influence of critical parameters like sintering temperature, duration and environmental conditions (Ar + H2) on surface properties like porosity and oxide inclusions will be comprehensively analyzed. The use of machine learning to co-relate the in-situ CSLM images with actual analysis (optical microscopy) helps in quantifying the strength of the printed part.