Additive Manufacturing of Metals:
ICME Gaps: Material Property and Validation Data to Support Certification: On-Demand Oral Presentations
Sponsored by: TMS: Integrated Computational Materials Engineering Committee, TMS Additive Manufacturing Bridge Committee
Program Organizers: Joshua Fody, NASA Langley Research Center; Edward Glaessgen, NASA Langley Research Center; Christapher Lang, NASA Langley Research Center; Greta Lindwall, KTH Royal Institute of Technology; Michael Sansoucie, Nasa Marshall Space Flight Center; Mark Stoudt, National Institute of Standards and Technology
Friday 8:00 AM
October 22, 2021
Room: On-Demand Room 1
Location: MS&T On Demand
Phase Field Informed Monte Carlo Texture Evolution Models for Additive Manufacturing Microstructure Simulation and the Need for Experimental Grain Competition Data: Brodan Richter1; Joseph Pauza2; Anthony Rollett2; Edward Glaessgen1; 1NASA Langley Research Center; 2Carnegie Mellon University
Additively manufactured parts frequently have an anisotropic and strongly textured microstructure containing large columnar grains due to interacting solidification and processing effects. The complexity of those interactions makes experimental approaches for understanding and optimizing microstructure development difficult and time-consuming. Computational materials modeling provides a promising parallel avenue for understanding microstructure development during the additive manufacturing process. In this work, phase field modeling is used to inform Monte Carlo based texture evolution modeling. Phase field and Monte Carlo simulations are presented and compared to experimental data for Nickel-based superalloys. This work demonstrates the role that the texture evolution methodology has on the resultant microstructure texture, presents a pipeline for foundational data generation for grain competition during solidification, and demonstrates the need for accurate experimental grain competition measurements for validating predicted grain overgrowth properties. Additionally, the linkage presented herein aims to provide a foundation for future integrations between the two modeling techniques.
CFD Modelling for AM Processes: Pareekshith Allu1; 1Flow Science Inc.
Simulating the laser-material interaction in an LPBF process requires implementing relevant physics models at relevant temporal and spatial scales. Process parameters such as laser power, scanning velocity, scanning path, and powder size distribution influence melt pool dynamics, playing a key role in process stability. We will look at underlying mechanisms behind the formation of defects such as balling, porosity and spatter using computational thermal-fluid dynamics models built in FLOW-3D AM. While low energy densities can lead to lack of fusion defects, high energy densities result in strong recoil pressure and unstable keyholes that lead to the formation of porosity and spatter. Additionally, the need for temperature dependent material properties such as surface tension, viscosity, density, thermal conductivity, specific heat, etc. and their influence on melt pool dynamics shall be discussed. Lastly, such models output thermal gradient and cooling rate data that can be used to predict microstructure evolution.
On Scan Path Knowledge for Model Informed Process Planning and Material Quality Predictions: Emil Duong1; Lukas Masseling1; Ulrich Thombansen1; Christian Knaak1; Mustafa Megahed2; 1Fraunhofer Institute for Laser Technology ILT; 2ESI Group
Scan paths have been shown to play a significant role in printing accuracy, defect generation and material quality. Open processes such as DED enable access to virtual and digital twins to path planning tools and files thus supporting a complete representation of the process and the component thermal history. Processes where path planning is not available, such as in commercial LPBF systems, pose a significant challenge to modelers limiting research efforts to trial and error and reverse engineering machine behavior before focusing on technological progress. In this presentation a DED hybrid twin consisting of physics- and data-based models will be presented. In contrast physics-based LPBF models are used to demonstrate the negative effect of inaccurate path planning knowledge on numerical predictions. As a result LPBF digital twins must represent several process unknowns as well as the material response to process parameters.
Predicting Melt Properties Using Atomistic Simulations with a Highly Accurate Physically Informed Neural Network Interatomic Potential: Vesselin Yamakov1; Yuri Mishin2; Edward Glaessgen3; 1National Institute of Aerospace; 2Geroge Mason University; 3NASA Langley Research Center
This presentation discusses the use of a recently developed machine learning (ML) interatomic potential for molecular dynamics simulations of aluminum melt properties. Such properties are critical for process modelling in additive manufacturing, including the melt pool size, solidification, and formation of solidification microstructures. Direct first-principles modeling of these processes is computationally prohibitive whereas simulations employing ML potentials combine the high accuracy of quantum-mechanical methods with high computational speeds. The physically-informed neural network (PINN) method used herein, integrates a high-dimensional regression implemented by an artificial neural network with a physics-based bond-order interatomic potential. PINN potentials can accurately reproduce many properties of aluminum in both crystalline-solid and liquid phases. We examine the accuracy of a PINN Al potential in predicting the density, self-diffusivity, viscosity, and the tension of the liquid surface and liquid-solid interfaces. Comparison with experimental data and ab initio molecular dynamics calculations shows very good agreement for all properties tested.
Capturing and Analyzing In-situ Data within the Directed Energy Deposition Process with DEDSmart: Michael Juhasz1; Romilene Cruz1; 1FormAlloy
Within the various AM processes, there are several process parameters that determine the resultant geometry and part quality of a build. In certain systems, the data from the builds remain locked away in a closed box. But can this data be useful for the end-users? At FormAlloy, we know so. Along with developing and integrating their fleet of in-process sensors and closed loop control features, FormAlloy internally developed their data logging capability known as DEDSmartTM for their directed energy deposition systems. All DEDSmartTM data is automatically generated post build and contains all the parameters and process signatures on a time scale. The data sets provided can be linked to post-process evaluations to enable machine learning possibilities and defect detection algorithms. Join FormAlloy as they discuss how their DEDSmartTM data was used to link defects that were observed post-build to anomalies that were found within the process parameters and their signatures.