Materials Processing Fundamentals: Additive Manufacturing and Materials First Principles
Sponsored by: TMS Extraction and Processing Division, TMS Materials Processing and Manufacturing Division, TMS: Process Technology and Modeling Committee
Program Organizers: Samuel Wagstaff, Oculatus Consulting; Alexandra Anderson, Gopher Resource; Adrian Sabau, Oak Ridge National Laboratory
Thursday 2:00 PM
March 23, 2023
Room: 29B
Location: SDCC
Session Chair: Samuel Wagstaff, Oculatus Consulting
2:00 PM Introductory Comments
2:05 PM
Automatic Process Mapping for Ti64 Single Tracks in Laser Powder Bed Fusion: Toby Wilkinson1; Massimiliano Casata1; Daniel Barba1; 1Universidad Politécnica de Madrid
Traditional process optimisation for metal laser powder bed fusion (LPBF) often involves significant labour time including machine set up, cleaning or microstructural quantification. This problem is even more acute in the development of new alloys, when the material wastage during process optimisation presents a serious barrier for new AM-orientated designed alloys. All these barriers are limiting the progress of LPBF, preventing the full exploitation of this technology. This work tackles this problem and investigates the feasibility of using machine learning and surrogate modelling methods combined with powder-free methodologies to significantly reduce the time and material required to discover the optimal processing window for new alloys. As well as optimising for the final density and microstructure of the 3D material, this method also quantifies the stability of the single tracks and its effect on the final part quality. The developed framework is validated using Ti6Al4V and a novel beta-titanium alloy.
2:25 PM
A High-fidelity Numerical Model Informed Machine Learning Framework for Melt Pool Prediction in Laser Additive Manufacturing: Shashank Sharma1; Mohammad Parsazadeh1; Zhaochen Gu1; Narendra Dahotre1; Song Fu1; 1Center for Agile and Adaptive Additive Manufacturing, UNT
The recent implementation of machine learning (ML) and artificial intelligence (AI) in metal additive manufacturing (AM) has proven to be a significant step toward the realization of its (AM) digital twin. However, the major bottleneck faced in the implementation of ML in AM is the need for an unprecedented amount of data-set (“big data”), which can be expensive if obtained using experiments. In this work, a physics-informed machine learning framework is proposed for laser-based additive manufacturing, in which, a high-fidelity Multiphysics single-track melt pool simulation is used to provide a sufficient set of input data-set for supervised machine learning models. The model accurately predicts significant process attributes such as melt pool geometry, and its transition from conduction to keyhole regime.
2:45 PM
A Mesoscale Thermo-mechanical Numerical Model for Residual Stress Prediction in Laser Powder Bed Fusion Process: Shashank Sharma1; Mangesh Pantawane1; Sameehan Joshi1; Narendra Dahotre1; 1Center for Agile and Adaptive Additive Manufacturing, UNT
Residual stress in laser powder bed fusion is one of the pivotal challenges in terms of its application in the energy industry. Impediments in residual stress measurements have created a bottleneck in cost-effective process design, paving way for numerical model-based predictions. In this work, a robust thermal model calibrated (in terms of melt pool geometry) using experiments and high fidelity thermo-fluidic model is constructed, which is coupled with an elastic-plastic (linear isotropic hardening) mechanical model to obtain thermally induced residual stress distribution at mesoscale. The developed model also considers the effect of solid-state phase transformation on residual stresses. To demonstrate model capabilities, the role of essential L-pbf design elements (for material: Ti6AlV4) such as energy density, scan length, and scanning strategy has been evaluated in terms of residual stress and material deformation.
3:05 PM
Investigation of the Keyhole and Molten Pool Stability in Laser Welding Process Depending on Intensity Distribution of Dual Beam: Juyeong Lee1; Jin-young Kim1; Junmyoung Jang1; Taehwan Ko1; Jaeheon Lee1; Geonmin Kim1; Seung Hwan Lee1; 1Hanyang University
In this study, a numerical model of laser welding process of aluminum alloys using dual beam laser was developed to investigate the keyhole and molten pool behaviors depending on the intensity distribution of the laser beam. In the developed model, the coupled study of heat transfer and fluid flow was conducted to analyze the keyhole and molten pool behaviors. The core and ring beams in the model were set by adjusting the intensity distributions. The model was validated by comparing the geometry of the molten pool calculated from the model and the weld cross-sections obtained from welding experiments. Based on the developed model, the keyhole and molten pool behaviors were analyzed according to the intensity distribution of the core and ring beams. Furthermore, the optimal condition of the intensity distribution which can stabilize the keyhole and molten pool was proposed.
3:25 PM Break
3:45 PM
Activation Energy of Simulated Surface Diffusion in Nanoporous Gold.: Conner Winkeljohn; Sadi Shahriar1; Erkin Seker1; Jeremy Mason1; 1University of California Davis
Nanoporous gold (np-Au) has applications in a wide variety of fields, from optical sensors to biomedical devices. Part of the reason it has found use in such diverse areas is because its morphology can be varied to accommodate the required length scale of each device. This morphological evolution is thought to be governed by the surface diffusion of gold atoms, and a thorough characterization of surface diffusion in np-Au would expand the range of attainable morphologies and allow accurate predictions of the conditions in which the morphology is stable. This talk explores the relationship between the curvature of the surface and the activation energy for surface diffusion. Specifically, molecular dynamics simulations are used to estimate the surface diffusion coefficient for a variety of surface curvatures. Given the high surface curvatures present in np-Au, this relationship will likely be critical in developing accurate models of np-Au morphology evolution.
4:05 PM
Machine Learning and Monte Carlo Simulations of the Gibbs Free Energy of the Fe-C System in a Magnetic Field: Ming Li1; Luke Wirth2; Stephen Xie3; Ajinkya Hire1; Michele Campbell4; Dallas Trinkle2; Richard Hennig1; 1University of Florida; 2University of Illinois Urbana-Champaign; 3KBR at NASA Ames Research Center; 4University of California-Merced
Modeling the thermodynamics and kinetics of steels for designing processes in high magnetic fields requires knowledge of the magnetic Gibbs free energy, G. To obtain G, we first accelerate the energy evaluation of magnetic and atomic configurations by training an ultra-fast force field (UF3) machine learning potential on density-functional theory calculations with an applied magnetic field for various atomic and magnetic configurations of the bcc and fcc Fe-C phases. We show that the UF3 models trained and validated on this database accurately reproduce the potential energy landscapes as a function of the applied field. Thermodynamic integration using grand canonical Monte Carlo simulations utilizing the resulting UF3 energy model predicts the magnetic Gibbs free energy. We combine the simulations at different temperatures and fields to obtain a comprehensive model of the Gibbs free energy for the two phases as a function of temperature, atomic fraction of carbon, and magnetic field.
4:25 PM
Carbon Diffusion in Bcc Fe Under Magnetic Fields From First Principles: Luke Wirth1; Ming Li2; Richard Hennig2; Dallas Trinkle1; 1University of Illinois Urbana-Champaign; 2University of Florida
Heat treatment of steels in the presence of external magnetic fields, such as during induction-coupled thermomagnetic processing (ITMP), offers the potential for greater control over transformation kinetics than traditional techniques allow. Understanding the atomic-scale mechanisms that determine this behavior will better enable engineering applications to make use of this control, thereby yielding increases in energy efficiency. Within a density-functional theory (DFT) framework, we apply a Zeeman splitting energy to impose an external magnetic field on Fe-C supercells. Calculations of activation energy barriers to interstitial carbon transitions within these environments inform diffusivity models at various splitting energies. Field-dependent changes in the vicinity of the C atom, such as redistribution of the magnetic moments of neighboring Fe atoms, provide insight into how external magnetic fields affect diffusion.
4:45 PM
Simulation of Fe Diffusion in Thermal Decomposition of γ’-Fe4N using Molecular Dynamics: Jianxin Zhu1; Jian-Ping Wang1; 1University of Minnesota
α”-Fe16N2 is a promising environmentally friendly rare-earth-free permanent magnet with ultra-high saturation magnetization. Recent research demonstrated experimentally through thermally quenching treatment using γ'-phase Fe4N as precursor to synthesize α”-Fe16N2 in bulk formation. Here we investigated γ’-Fe4N thin film thermal decomposition process and potential localized phase-transition from fcc phase to bct phase using Molecular Dynamics (MD) simulation. As nitrogen concentration is higher in γ’-Fe4N (5.9 wt.%) than that in α’-Fe8N or α’’-Fe16N2 (3 wt.%), increase of Fe may occur during thermal treatment to form possible low-Nitrogen bct-FeN solid solution. A localized “Fe-rich” grain boundary lattice defect model is constructed accordingly, and we studied Fe diffusion to neighboring lattice cells to increase local fe concentration with temperature effect. Modified Embedded Atom Method potential is applied. Atom displacement analysis and energy minimization are performed in simulated quenching. LAMMPS XRD method is used to detect new material phases.