6

Summary of Issues from Subgroup Discussions

During the third day of the workshop, participants met in subgroups to discuss some of the challenges and possible solutions to the questions set forth for each session. Summaries of the subgroup discussions are provided in the following subsections.

THEORETICAL UNDERSTANDING OF MATERIALS SCIENCE AND MECHANICS

Subgroup Members

Steve Daniewicz (Mississippi State University), Marianne Francois (Los Alamos National Laboratory), Edward H. Glaessgen (NASA), Neil Hodge (Lawrence Livermore National Laboratory), Saad Khairallah (Lawrence Livermore National Laboratory), Peter Olmsted (Georgetown University), and John Turner (Oak Ridge National Laboratory).

Summary

Several members of the subgroup mentioned the importance of developing a scientific methodology integrating theory, modeling and simulation, and experiments toward prediction and control, which could benefit the broader additive manufacturing (AM) community. They discussed fun-

damental scientific issues of AM to be addressed in the future and identified a number of topics for more research, including the following:

• Strongly non-equilibrium processes (e.g., high rates of flow, energy deposition rates, phase transformations),
• Mechanics of strongly heterogeneous materials (e.g., varying porosity, microstructure and orientation, composition),
• Multiple simultaneous processes (e.g., heat, fluid, particle flow, phase changes, stress-strain including crystal plasticity), and
• Experimental material design utilizing information science.

Subgroup members also made several observations about the state of fundamental science in AM. Some stated that the qualification cycle could be shortened by utilizing a scientific-based approach, while others noted that computational thermodynamics as a field needs more depth and breadth.

Microstructure variation was highlighted by many subgroup members as a major challenge. They commented that AM processes produce different microstructures from traditional manufacturing processes. There seems to be two length scales: the weld beads and the length scale of individual grains (smaller than the weld bead). The difference in microstructure (e.g., orientation, crystal phase) results in very different properties and damage processes. They also noted that certification is challenging given that durability and life cycle performance result from microstructural variability, residual stress, and defects.

Several opportunities for utilizing multiple materials with AM were discussed during the workshop. Several members of this subgroup noted that current methods, expertise, and machines are typically focused on a single material or class of material. In the future (10+ years) they hope that these approaches will be linked together to develop hybrid materials (e.g., organic/inorganic materials, multiple materials, aqueous and biological materials). One challenge to achieving this is minimizing abrupt interfaces between dissimilar materials. They suggested that simulations could be developed to account for and enable optimization of such gradient interfaces. They also commented that multiple metastable crystalline states arise from many-component alloys under strong nonequilibrium conditions.

The subgroup also discussed the mathematical models and state-of-the-art theoretical, computational simulation models that describe the different aspects of AM that exist or are needed to simulate the various stages of AM. The group began with a discussion of microstructure-aware continuum

dynamics code, both for material processing (e.g., modeling and simulation of solidification) and performance modeling (e.g., thermo-mechanical, elasto-plastic). A number of participants emphasized the need to develop new methods in phase field modeling to capture the microstructure in the nonequilibrium and dynamical environment prevalent during an AM process. Starting with some challenges, they highlighted that high-performance computing does not work well with the numerical methods for the stiffness matrix calculations. Others noted that many multiphysics (e.g., heat transfer, fluid flow, solid mechanics) and finite element structural analysis codes already exist but more efficient, accurate, and robust numerical methods with implicit solvers would be useful, as would additional reduced-order models and computational techniques for multiscale methods. They also emphasized that models are needed for melt pool dynamics (e.g., interface tracking method with phase change), as are phase-field models for microstructure evolution. Lastly, some of the subgroup noted that data-based analysis could to help elucidate trends and possibly provide process bounds for AM users.

Another topic discussed was the integration frameworks that currently exist for coupling modeling techniques together. Some subgroup members noted that the Materials Genome Initiative (MGI)¹ offers a paradigm that fits well with the needs of AM. MGI is built around identifying how specific materials lead to different end properties, via different processes. It links multiple scales, from quantum and atomistic to molecular mechanics and derived potentials, to mesoscale (nanometer) methods, and finally to continuum level. Process characteristics and effects play an integral role in the genome of a material. A similar approach to unite scales and methods is appropriate for AM, these individuals emphasized. With AM, the microstructure of the material could be tailored to specific requirements and needs, which provides for many opportunities of material design.

Many members noted that there are no software suites that contain all of the functionalities necessary to model all aspect of various AM processes. Even existing tools often do not consolidate well with each other.

Recent multi-institutional programs have demonstrated that collaborative development of simulation environments aimed at solving complex problems can successfully deliver capabilities that would otherwise not be possible, several participants explained. Examples include the following:

___________________

¹ The website for the Materials Genome Initiative is http://www.mgi.gov, accessed August 23, 2016.

• The Consortium for Advanced Simulation of Light Water Reactors (CASL),² which serves as the Department of Energy (DOE) Innovation Hub on modeling and simulation for nuclear energy;
• DOE’s Accelerated Climate Modeling for Energy (ACME) program;³ and
• DOE/EM Advanced Simulation Capability for Environmental Management (ASCEM).⁴

A number of subgroup members observed that a similarly coordinated effort for AM could benefit the development of common input and output formats, thereby reducing necessary end-user effort to perform analyses. It could also aid the development of common interfaces for physics components, which could facilitate the sharing of models. These individuals noted that such an effort would also leverage existing simulation tools, enable researchers to focus on their areas of expertise without having to create other required components, and provide the ability to explore coupled and multiscale interactions. Ideally, they said, such environments could also provide both reduced-order and high-fidelity models and implementations able to run on systems ranging from workstations to the largest HPC platforms.

Several subgroup members identified what they viewed as the most important open questions in materials and mechanics, including related scientific disciplines, engineering and mathematics, as well as the technical challenges to be addressed for predictive theoretical and computational approaches in order to enable widespread adoption of AM. In doing so, they listed some areas of fundamental research in theoretical and computational materials science, mechanics, and multiscale computation that could advance AM:

• Polymer FDM (P-FDM). Fundamental polymer science issues include the glass transition, the flow-induced crystallization, and the relationship between complex microstructure and mechanical properties.

___________________

² The website for the Consortium for Advanced Simulation of Light-Water Reactors is http://www.casl.gov/, accessed August 23, 2016.

³ The website for Accelerated Climate Modeling for Energy is http://climatemodeling.science.energy.gov/, accessed August 23, 2016.

⁴ The website for Advanced Simulation Capability for Environmental Management is http://esd1.lbl.gov/research/projects/ascem/, accessed August 23, 2016.

• Build region/melt pool level. Fundamental issues include orientation prediction, glassy and semi-crystalline morphology as appropriate, and resulting strength, weld properties, local moduli, and heat transfer and processing conditions (e.g., speeds, flow rates, temperature, geometry).
• Part level. Key challenges relate to predicting, designing, and controlling the overall mechanical and structural properties. Some members noted that it would be extremely helpful if part-scale models could run in two different modes: (1) high-resolution, suitable for detailed determination of the solids response, and (2) fast, which would be able to execute much more quickly, at the cost of solution accuracy. Once mode (2) is realized, it would be useful to link the simulations to process-informed topology and function optimization, which would also need to be able to handle material and geometric nonlinearities (which they currently cannot do) to be really useful.
• Metals. High-fidelity physics-based models face challenges with their processing, properties, and performance. Processing improvements would help to model melt pool dynamics, melting and solidification cycles, mass and energy deposition model, and alloy composition distribution for multiple components. It would also be beneficial to include G&R map in solidification modeling as well as model microstructure-aware solidification. These models cannot yet predict strength properties of AM metals (as discussed by Bishop and Francois), but several subgroup members stated that their development would be helpful. Performance challenges mentioned related to plasticity, damage, and durability models.
• Linkage. Linking scales and models is also a challenge, including connecting phase change models through thermodynamic calculations and linking microstructure information to macroscale models.
• Materials. Some subgroup participants emphasized that the future of AM eventually might be dominated by material questions because there are limitless ways of combining materials. Potential issues include (1) determining optimal powder materials and alloys to use for a given application, (2) understanding how physical properties change (e.g., hardness, strength) when different materials melt and solidify under AM-prevalent non-steady-state conditions, and (3) printing functional products that consist of printed or embedded electronic components or functional organics for biological application.

• Mathematics. Several relevant mathematical issues were discussed, including statistical methods, machine learning algorithms, pattern recognition, database and data science, applied mathematics (e.g., coupling of partial differential equations, time integration schemes, higher-order discretization methods, implicit solvers), and multiscale methods (e.g., coarse graining, discrete-continuum).

Multidisciplinary and related materials and mechanical sciences needed for AM were also discussed, including computational modeling, metallurgy, mechanical and chemical engineering, physics, chemistry, data analytics, and computer science. A number of subgroup members stated that the field will benefit from large teams working together to link different length and time scales, as well as different arenas of interest (e.g., material, deposition methodologies, metrologies, modeling, testing and validation, design).

Regarding verifying and validating theoretical and computational models for AM processes, many subgroup members noted that computational and AM-build benchmarks are needed and that AM-build benchmarks would be of great interest to standards organizations to address variability within AM processes. In the short run, they noted that validation of simulations could be done via classical methods such as creating and characterizing parts against model results. Ostensibly, they said this should start with fairly simple geometries and proceed to increasingly complex problems. In the context of the small scale, this could consist of single-track experiments. At the part scale, it could start with small parts (e.g., 1 mm³), and increase to a medium-sized part with a few features (e.g., cm³, with some holes, overhangs, and interior passages), and then finally to a large, complex part (e.g., 4-cylinder engine block).

Several group members also noted that in situ monitoring and diagnostics could be improved through increased access to real-time data, with data analytics, or by state-of-the-art characterization. This might include exploiting big data techniques to deepen the validation process, such as using reconstructed images and or diffraction data. They noted that there is a strong need for on-line in situ metrologies to be built in as standard on equipment (e.g., temperatures, molecular and microstructural information, microscopies). In the long term, they hoped to see in situ metrologies linked to high-performance computing and modeling to adjust and optimize the design and build on the fly in real time. These members also emphasized that the development of standard benchmark cases (simple to complex) to test computational methods and code would be helpful, as would open--

source and published computational algorithms and physics models. An example for simple test cases could be a single-track pass for a powder-bed machine and a LENS system, where the metrics of interest (e.g., temperature history and melt pool shape) are developed and can be used for comparison. Such comparison exercises would drive improved quality of computational methods and software.

A number of subgroup members suggested possible opportunities for public-private partnerships to advance theoretical and computational mechanics capabilities for AM, the first of which being internships and doctoral student networks to provide interdisciplinary training. They also noted that it is important to link materials producers (e.g., powders, FDM filaments, and grains), equipment manufacturers, metrologists, computational and theoretical modelers, end users and developers, and standards bodies to increase communication and collaboration while accommodating the groups’ various interests. In particular, they suggested potential focuses of partnerships, such as developing physics models and numerical methods, implementing models and methods in software, and performing validation. One concern these participants raised was whether creative partnerships could be developed to preserve individual intellectual property. They emphasized for the value of specific public-private partnership calls that span government organizations (e.g., National Science Foundation [NSF], Department of Energy [DOE], Department of Defense [DoD], and National Institute of Standards and Technology [NIST]), commenting that DOE’s newly-announced High-Performance Computing for Manufacturing (HPC4Mfg) program⁵ provides a possible model for industry-driven collaboration on shorter-term needs. A complementary longer-horizon program focused on R&D challenges in materials science, applied mathematics, and numerics, and on development of a community simulation environment would accelerate progress further, they observed, ideally influenced by lessons learned in HPC4Mfg.

Several subgroup members noted that partnerships would benefit from the strengths of the other groups and the fundamental scientific and engineering baseline would be raised so that all partners would be able to operate more efficiently and more flexibly. Academic researchers would bring expertise and advanced training opportunities to the partnerships, for example. National laboratory partners could provide high-performance computing capabilities

___________________

⁵ The website for High-Performance Computing for Manufacturing is https://hpc4mfg.llnl.gov/, accessed August 23, 2016.

and user experimental facilities. Industry partners could gain insight from their product experience and also provide funding opportunities.

Many group members emphasized the variability in processes and materials and the need for standards in measurement, process, and materials. The development of sophisticated predictive modeling that relies on sound metrology and materials could in turn lead to a positive feedback that would encourage new standards. They explained that theorists and modelers need a full understanding of life cycle process and well-characterized model materials. A push to understand properties based on these details could lead to a better understanding of how to design functionality more rapidly and efficiently. These members also stressed the importance of coordination and acceptance from certification authorities (e.g., the Federal Aviation Administration [FAA]).

In terms of policy, some subgroup members noted that visible and prominent challenges have the potential to capture the public’s imagination, analogous to the XPRIZE⁶ where a large monetary prize could focus the imagination and attract large investment. They commented that AM can be linked to societal needs such as energy, climate, personalized medicine, and sustainability, with potential for a wider engagement.

Future needs for AM were also discussed by the subgroup. Given the high speeds (especially of selective laser melting processes), feedback control may not be fast enough to make process corrections in real time, several observed. One possible way to improve in situ corrections would be to use small, idealized simulations to predict process and estimate process parameters ahead of time.

Computational and AM-build benchmark tests were also discussed. Specifically, some subgroup participants mentioned the need for a common test bed with multiple challenge problems to build confidence in simulations and AM builds, observing that AM-build benchmarks would be of great interest to the ASTM F42 Committee on Additive Manufacturing. They envisioned two classes of test problems:

1. Benchmark problems. These would be well-defined problems, ideally a progression from simple geometries with single materials with well-known properties designed to test a single physical phenomenon up to complex geometries, including diverse materials, multiple scales, and physical phenomena. Details of initial and boundary

___________________

⁶ The website for XPRIZE is www.xprize.org, accessed August 23, 2016.

2. conditions, along with characterization data, would be available to the community. These data would provide common verification and validation information against which models could be compared, enabling trade-offs between fidelity and computational resource requirements. The benchmark problems might also drive development of new experimental techniques to provide data that is currently unavailable. Data and results could be provided through a DOE national laboratory or other institution.

3. Challenge problems. These would typically be driven by real-world needs and might not be possible currently, either physically or virtually. They might include extremely complex geometries, materials that may not exist, and multiple physical phenomena, these participants said. Some of these problems might be similar to the XPRIZE concept mentioned above and would serve to both inspire and drive R&D activities. A few subgroup members noted that standards organizations such as ASTM E08 and F42 work commonly with coupons, but simulations are needed to bridge the gap between coupons and components.

Several subgroup members commented that workshop presentations repeatedly mentioned the various parameters describing melt pool solidification rate or heat input, which can be used to describe a successful build. They wondered if computational simulations could be used to define a nondimensional parameter that is more powerful than the currently known parameters.

COMPUTATIONAL AND ANALYTICAL METHODS IN ADDITIVE MANUFACTURING

Subgroup Members

Corbett Battaile (Sandia National Laboratories), Joe Bishop (Sandia National Laboratories), Wayne King (Lawrence Livermore National Laboratory), Anthony Rollett (Carnegie Mellon University), David Snyder (QuesTek Innovations), and Gregory Wagner (Northwestern University).

Summary

Several of the subgroup members discussed unique characteristics of AM including the capability for material design, specifically with respect to

alloy design. However, they noted a number of challenges including difficulties with CAD and surface representation that often rely on nonparameteric surfaces, point clouds, and implicit surfaces where tolerances can only be applied at interfaces. Concerns about corrosion and surface effects can also cause issues for AM build, tribology, and wear. Some subgroup participants noted that AM has several unique design constraints, such as overhangs and a top-down identification of processing. They described the lack of fundamental understanding of process, structure, and material properties, design of alloys specific to AM (accounting for nonequilibrium thermodynamics from high cooling rates), physics-based process control (both feed-forward and feed-back), and accounting for surface effects at all scales (chemistry and interfacial chemistry, surface finish). Some of the technical concerns include determining what combination of powders will result in desired material properties, how to control material grading (e.g., temperature gradients), and what relationships would scale across machines, alloys, laser heat inputs, and different beam sizes and energy density. These individuals stated that process control and length scale effects (e.g., exotic microstructures, high-temperature measurements, high-cooling rates, unique processing) and powder and laser interactions are not being well studied.

To move beyond these issues and advance the use of AM, members of the subgroup discussed the importance of designing parts that minimize residual stresses and snap back by design. Many participants stressed the value of integrating multidisciplinary optimization into a build, including topology, shape, material, manufacturing, uncertainty quantification, residual stress, build path, and feedback. They noted that advanced macroscale viscoplastic material models are also needed. These participants also stated that computational material science could be used to reveal mechanisms and for process design. Physics-based process control, whether feed-forward or feed-back (enables robust AM independent of machine), could allow for advances in microstructure-based process control and residual stress control.

They also commented on the importance of multidisciplinary optimization such as how to integrate different tools sequentially, considering topology, shape, material, manufacturing, uncertainty quantification, residual stress, and AM build path. These participants highlighted the merits of multiscale modeling (e.g., computational homogenization), predictive crystal-plasticity, and material thermodynamics for novel AM microstructures. Also, improved estimation and control of modeling errors on engineering quantities of interest (e.g., from material models) would be helpful, they explained.

Some subgroup members noted that many of the fundamental science questions reside in applied research areas, such as better understanding the melt pool physics and powder interactions. They also commented that HPC is transforming physics-based codes to take advantage of new architectures. Areas for opportunities include better use of big data, data compression (e.g., optimal representation, basis), and surrogate models.

Modeling and simulation for advanced and complex production would be helpful, members of the subgroup stated. Creating and using a fully predictive model is often too expensive but statistical methods could be used to gain more insight. Surrogate and reduced models derived from HPC models show promise for industrial use, especially those that focus on residual stress, prediction, control, and minimization. New models and software might have value, including advanced macroscale viscoplastic models with internal state variables capable of modeling processing history and new CAD representations for AM. These individuals also emphasized the benefits of HPC and exascale computing in support of uncertainty quantification and error estimation, Bayesian methods, optimal experimental design, data fusion in real time, and improved control of machine-to-machine variability. They commented on the importance of accelerating advanced software for industrial use.

Subgroup members also discussed the lengthy time currently required for part qualification. They believe that focusing on high-sensitivity parameters as well as utilizing advanced diagnostics and the processing history of each point in build (where statistics of each point over ensemble of builds can be used) could accelerate this process. These advances could help qualify each part by understanding the precise processing history of each point, while also improving first-run builds, these members explained.

Other subgroup members said that a science-based theory of AM capabilities could help researchers observe characteristics that are difficult to measure. They noted that data science for AM material systems discovery has potential, including data mining, discovery of emergent behavior, mechanistic-based data compression, high-throughput standardization, and quantitative material testing. The benefit of establishing an AM database for both experimental and modeling and simulation, similar to MGI, was also discussed by these participants. They noted that the community could benefit from more openly available data and software for enhanced use in data informatics and material models.

Many of the subgroup members emphasized the importance of developing interdisciplinary degree programs and fellowships for AM.

They stressed that AM processes rely on different design rules and require different ways of formulating problems. Lastly, they stated that it is challenging but important to bridge the gap between fundamental science and manufacturing.

MONITORING AND ADVANCED DIAGNOSTICS TO ENABLE AM FUNDAMENTAL UNDERSTANDING

Subgroup Members

Joseph Beaman (University of Texas at Austin), Jian Cao (Northwestern University), Feng Lin (Tsinghua University), Ade Makinde (GE Global Research Center), Z.J. Pei (National Science Foundation), Edwin Schwalbach (Air Force Research Laboratory), and Yu-Ping Yang (EWI).

Summary

Many subgroup participants began by emphasizing the importance of modeling and sensing for prediction and operation. However, they noted that several challenges were not discussed during the workshop, including the importance of the measurement matrix for different simulation tools and the fact that the closed nature of the machines makes it difficult to access the complete data history of machines. They also mentioned the challenges of testing and examining internal features of AM parts, such as identifying the voids in thin walls particularly when the tolerable pore size is just one order smaller than the feature size.

Several members of the subgroup offered short-, mid-, and long-term goals. In the short term, which they defined as up to the next two years, goals could be to (1) identify the correlation between defects and signals of interest, including developing algorithms for combining in situ sensing with post-built measurement and theory for signal processing and data mining; (2) understand and characterize thermodynamic behavior of metals and surface tension through thermal-mechanical modeling (high temp properties, high thermal gradient) and microstructure modeling; (3) develop national facilities with open-architecture and highly-instrumented machines; (4) study the fundamentals of AM processes with the help of advanced metrology such as in situ X-ray and neutron measurement and the development and integration of low-cost sensors for gas current measurement; (5) improve sensing for health monitoring of machines; (6) improve tem-

perature measurement accuracy (within 50°C) over the entire bed; and (7) improve simulation tools for powder spreading.

Mid-term goals (in the next two to five years) discussed by these subgroup members were to (1) further improve temperature measurements (within 10°C) over the area; (2) better integrate modeling with different sensing modalities (e.g., laser interferometer, inferred, pyrometer, microstructure, ultrasonic for porosity, X-ray) with statistical methods; (3) develop algorithms for integrating the sensing needs and capabilities into the design of the part; (4) further develop model-based reconstruction techniques for CT scan; (5) improve simulation tools for powder development; and (6) explore methods for defining function-driven metrology needs.

Long-term goals for the next five to ten years discussed by these individuals were to (1) develop a fully-coupled model; (2) improve in situ identification of gradient material structure, and (3) develop a digital thread for each part (e.g., data analytics, materials, process, and performance and failure in the field and application).

These subgroup members suggested that fundamental research to drive next-generation AM processes be a consideration as well as optimization of AM parts with functionally-gradient materials (e.g., local material design, topology design, re-separate and recycling of powders). Lastly, they emphasized the value of graduate fellowships and industrial summer internships for advanced AM. They stressed that the continuation of fundamental research for AM would continue to move the field forward.

SCALABILITY, IMPLEMENTATION, READINESS, AND TRANSITION

Subgroup Members

Anthony DeCarmine (Oxford Performance Materials), Tahany El-Wardany (United Technologies Research Center), Rainer Hebert (University of Connecticut), Lyle E. Levine (National Institute of Standards and Technology), Alonso Peralta-Duran (Honeywell International Inc.), and Yung C. Shin (Purdue University).

Summary

The subgroup members began with a discussion of the path for utilizing fundamental results for AM and scaling them for use in production.

Some members of the subgroup suggested that the accurate representation of feedstock, energy, environment, and kinematics of machines is important and could be accounted for, and the impact of energy inputs (e.g., laser beam profile, time-power behavior) and interactions of beams with materials could be better understood. The effect of the parameters on the interaction with materials and the resultant microstructure and phases, and the difference in melting and flow behavior between spherical and non-spherical (or satellite) particles, could also be better understood.

The main roadblocks some of these participants see in the scaling of AM technologies into production is the lack of hierarchical high-fidelity models, including hierarchical engineering models. They commented that too much time is often spent on one length scale without relation to higher level length scales. High-fidelity, physics-based simulations could be used to train surrogate reduced-order models. Also, prebuild engineering simulations could be run before each new build and the results could be used to identify and correct potential build problems. They noted that rapid simulations for in situ fine tuning of build parameters require a feedback loop with in situ process monitoring (e.g., temperature profile, melt pool width), possibly after each layer is deposited.

In terms of addressing these roadblocks through additional fundamental research, members of the subgroup stated that high-fidelity models could be developed for various AM processes as could reduced-order models that are more computationally efficient. They also emphasized that the establishment of mechanisms for longer-term collaborative efforts among researchers in academia, government, and industry would be helpful, as would the establishment of a repository or database for material selection, properties, or response surfaces. These subgroup members also suggested the development of a software test bed for benchmarking different AM software.

To leverage the fundamental research and scale it into production, other subgroup members suggested that the software companies become involved in developing open architecture and partnering topology optimization with AM machine design. Research would be valuable to better address support structures and design of geometrical features for use in ceramic applications and coatings.

Measurements of quality or systems that correlate computational and analytical methods to practical implementations were also discussed by several subgroup members. These members identfied geometry, microstructure, defects, mechanical behavior, environmental behavior (corro-

sion), part qualification, standards, and machine qualification or standards as aspects that would benefit from more input.

Software architecture and databases for AM model development were mentioned by subgroup participants. They suggested leveraging the existing efforts in the MGI for AM to develop databases of software and parameter data. They also noted that MGI works in a design space that is a subset of the design space where AM operates, but AM adds high heating and cooling rates. These participants suggested using scalable software with a hierarchical (modular) approach and a hierarchy of engineering simulations. Benchmark test series, they noted, could pull independent research groups together and identify effective modeling, measurement, and AM strategies.

Other subgroup members noted that careful design of validation experiments for model validation, uncertainty quantification, and in situ process monitoring is challenging because there is a lack of data for probabilistic modeling and error estimation. Test standards and test artifacts would help, as would an adequate suite of in situ monitoring to provide useful engineering data, they explained.

These subgroup members emphasized that there are drivers to integrate computational simulation and advanced optimization methodologies to enable unique AM design, and full life cycle simulations are important for qualifying AM-built parts. They noted that processing standards might change with an analytical and mechanistic model approach to implementation of full-scale AM.

Several other subgroup members described opportunities for public-private partnerships to advance HPC and other capabilities for AM. These partnerships are important to define and implement hierarchical models that are developed based on exchanges among industry, national laboratories, and universities. They emphasized that the proposed benchmark test series may present a mechanism for a public-private partnership. Partners could benefit from advancements within partnerships by having access to focused research with shared guidelines. Industry could focus on practical AM while national laboratories and academia could provide feedback on what to focus on and provide expertise on what can or cannot be measured experimentally.

This page intentionally left blank.