JOM, Vol. 68, No. 5, 2016
DOI: 10.1007/s11837-016-1867-4 Ó 2016 The Minerals, Metals & Materials Society
Insights from the 3rd World Congress on Integrated Computational Materials Engineering D. HOWE,1,5 B. GOODLET,2 J. WEAVER,3 and G. SPANOS4,6 1.—Drexel University, Philadelphia, PA 19104, USA. 2.—University of California, Santa Barbara, CA 93106, USA. 3.—Georgia Institute of Technology, Atlanta, GA 30332, USA. 4.—The Minerals, Metals & Materials Society, Warrendale, PA 15086, USA. 5.—e-mail: dhutchingshowe@gmail. com. 6.—e-mail:
[email protected]
The 3rd World Congress on Integrated Computational Materials Engineering (ICME) was a forum for presenting the ‘‘state-of-the-art’’ in the ICME discipline, as well as for charting a path for future community efforts. The event concluded with in an interactive panel-led discussion that addressed such topics as integrating efforts between experimental and computational scientists, uncertainty quantification, and identifying the greatest challenges for future workforce preparation. This article is a summary of this discussion and the thoughts presented.
INTRODUCTION As a discipline, integrated computational materials engineering (ICME) has continued to grow in recognition in the materials science and engineering community. Landmark studies such as the 2008 NRC report on ICME,1 the 2013 TMS study on ICME Implementation,2 as well as other key ICME publications,3–7 have served to illustrate the potential for ICME to transform how materials solutions are developed and deployed in engineering systems. The 1st World Congress on ICME* was held in 2011 and has since been repeated biannually, with the 3rd World Congress** being held in the summer of 2015 in Colorado Springs, Colorado. At this latest gathering, an interactive panel discussion was held on the final day of the Congress. This discussion provided a forum for thought-leaders in the field, including panelists and congress attendees, to consider the current progress of ICME, as well as to discuss potential paths for future directions. This article provides (I) a summary of representation at this most recent ICME World Congress by both
*see www.tms.org/Meetings/Specialty/ICME2011. **see www.tms.org/meetings/2015/icme2015.
1378
discipline and focus areas, (II) a synopsis of the process used to derive the facilitating questions for the closing dialogue, and (III) a summary of the commentary captured from the panel-led discussion. Congress Composition and Representation This article is an encapsulation of the dialogue, feedback, and guidance derived from the 2015 ICME World Congress. As such, a portrait of the interests and expertise of the Congress participants is provided in Figs. 1, 2, and 3 to help both orient the reader and serve as a contextual backdrop for the following commentary. The data in Figs. 1, 2, and 3 were prepared by word search within the 2015 ICME World Congress program to determine the relative emphasis of different topics. As such, the figures do not provide a direct demographic representation of the Congress participants and attendees but serve as a clear summary of the range of topics emphasized by the participants, as well as provide an overview of their backgrounds and expertise. As shown, the Congress topics encompassed a wide range of materials classes including both traditional and advanced materials. Additionally, the relative population frequency of certain words suggests the investigations focused on a wide range of properties and performance. Perhaps of greatest significance, this simple keyword search highlights a strong balance between an (Published online March 23, 2016)
Insights from the 3rd World Congress on Integrated Computational Materials Engineering
Fig. 1. Material types.
Fig. 2. Properties and performance.
Fig. 3. Experimental and computational methods.
1379
1380
Howe, Goodlet, Weaver, and Spanos
experimental emphasis and a variety of computational approaches, underscoring the integrated nature of the ICME discipline.8
(3)
TOPICAL PRIORITIZATION AND DISCUSSION CONSTRUCT The interactive panel-led discussion focused on the current ‘‘state of the art’’ with regard to ICME and how best to move the community to such a point. Furthermore, during the concluding session, the overarching challenge to all participants was to discuss the topics presented collectively with a view toward the following three fundamental arenas:
What are the areas in which the ICME community should have made significant progress by the 4th World Congress on ICME to be held in 2017? What is the emerging, collective vision for the status of ICME as a field? What does the community need to do to achieve the desired progress?
With these foci as an underlying basis for discussion, five prioritized topical questions were formulated by Congress organizers through synthesis of the feedback garnered throughout the duration of the Congress. This feedback was gathered through survey of each session chair, for identification of major themes, high interest areas, and significant scientific challenges within ICME as observed by the talks and questions arising during their session. The prioritization of topics occurred the evening prior to the panel discussion. The conference organizers and TMS staff parsed through all responses obtained to identify key themes and distill leading questions to drive the ensuing discussion. To stimulate more thought and engage the audience, four plenary speakers, who also served as the panelists, gave brief presentations within their respective subject areas, prior to the open discussion. These four panelists were as follows: Alexis Lewis, National Science Foundation (NSF); David McDowell, Georgia Institute of Technology (GT); James Warren, National Institute of Standards and Technology (NIST); and Georg Schmitz, MicressTM/ Aachen University. With George Spanos, TMS Technical Director, moderating, each leading question formed the kernel of a discussion wherein the panelists were first given the opportunity to respond before discussion was opened to the audience for comment, debate, and further questions. The five prioritized topical key questions were as follows: (1)
(2)
How can the community help foster the integration needed for ICME to be effective among experimentalists, theorists, computer scientists, etc.? What are examples of some key metrics to measure the progress of ICME and the Materials Genome Initiative (MGI) as the commu-
(4) (5)
nity moves forward (over the next 5 years, 10 years, 20 years)? How do we quantify uncertainty propagation through chained sets of models in complex applications (e.g., such as additive manufacturing)? How do we more effectively collaborate with the advanced manufacturing community at meetings such as this or in general? What are the greatest imperatives and challenges of future workforce preparation in ICME? DISCUSSION SUMMARY
The subsequent summary captures key content with respect to each key question and aims to relay the tone of discussion therein in a cohesive manner as opposed to a verbatim or unabridged record of events. Question 1: How can the community help foster the integration needed for ICME to be effective among experimentalists, theorists, computer scientists, etc. The discipline of ICME requires legitimate integration (the I in ICME) of a variety of skill sets to be implemented effectively. As described in the 2013 TMS ICME Implementation study,2 this includes experts such as mechanical engineers, computer scientists, theorists, and experimentalists in materials science and engineering. Thus, increasing the overall uptake of ICME in the science and engineering communities will require the fostering of close collaboration among these and other disciplines. One strategy for encouraging collaboration among those developing, verifying, validating, and ultimately using computational models and databases would be to encourage common file formats (e.g., HDF5) for materials datasets. Eliminating a barrier to integration within the community, a common file format would facilitate data exchange among simulations, experiments, and databases. In addition, agreeing on common metrics and descriptors for microstructures across different communities would build toward this end. This view was by no means universal among the panelists and the audience. Concerns were expressed regarding mandating file formats for the field. However, most were in agreement about the importance of setting minimum metadata requirements and guidelines for datasets and databases. This would allow data to be characterized and easily used by different parties, independent of file format. Although such strategies would not fully address the need for greater cross-disciplinary collaboration and integration of efforts, setting data requirements
Insights from the 3rd World Congress on Integrated Computational Materials Engineering
would significantly lower the barrier for integrated work flows since information could be more easily shared and used. Increased integration of efforts across multiple disciplines will require incentives of various kinds. One suggested mechanism offered was for funding agencies (e.g., NSF, Department of Energy (DOE), etc.) to support meetings and short courses designed to bring together individuals from disparate communities with a goal toward ICME. In addition, regional workshops to bring together individuals from different communities could be organized to share their efforts related to ICME and the MGI, as well as to explore future collaboration. Such collaborations cannot be forced, but they could occur where there is synergy between domains. An initial step could take the form of creating a context for cross-disciplinary communication. To contribute to this, collaborative attitudes should be instilled via undergraduate and graduate coursework focused on genuine problem solving in the field of study, with ICME methods being emphasized throughout the educational process. Other obstacles to cross-disciplinary collaboration in ICME include cultural barriers between discipline domains such as discipline-specific terminology. In addition, career development pressures drive individuals to become specialists, whereas, to some extent, ICME would seem to require a more generalized or integrative skill set. It was also acknowledged that some of the greatest barriers to integration and collaboration come from within the materials community itself. Question 2: What are examples of some key metrics to measure the progress of ICME and the Materials Genome Initiative (MGI) as the community moves forward (over the next 5, 10, 20 years)? It is not enough just to describe the theoretical potential of ICME to accelerate materials innovation, metrics are needed to identify tangible ways such efforts make an impact. This type of analysis is necessary for measuring the success of ICME initiatives. It was agreed that baseline values would be required to accompany any metrics used to measure the progress of the ICME and MGI goals. One such metric, as described in the MGI white paper,9 is the length of time needed for newly developed materials to be deployed in the market. Although this timeframe has been estimated to be on the order of 20 years on average,9 it could vary significantly, depending on the industry and the application. In particular, the community should seek out and publicize examples in which ICME has been used to accelerate materials development, and clearly quantify by how much. Another area of development to be measured could be the degree of software tool integration within a modeling or simulation ecosystem.
1381
Enhancing the capacity for existing software tools to use and pass information, or developing new software tools with such interoperability in mind, would significantly increase the ease and effectiveness of ICME efforts. One example of software interoperability is the software suite Dream3D, in which different members of the community have contributed to a single software platform to solve a variety of problems (in this case, analysis of 3D materials datasets). Within such an area, possible metrics for measuring progress would be the number of adoptees of a particular software suite, or perhaps the number of students trained in the use of the tool. For existing tools such as materials databases, metrics may include the number of new datasets created or the number of users. Specific examples of such efforts include the Materials Data Repository, developed by the National Institute of Standards and Technology (NIST), which serves as a repository for users to store valuable material property data that can be easily accessed well into the future. Seamless data storage and future access is a prime capability that may serve as incentive for the community to encourage use. Overall, the ICME community needs better metrics to demonstrate the value of the ICME approach. Metrics exhibiting the time or cost-savings of ICME through real-life examples would help others make a value decision about adopting ICME approaches. Additionally, it was suggested that a publication summarizing ICME case studies and success stories could help demonstrate the value and potential of ICME to the larger engineering and business communities. Question 3: How do we quantify uncertainty propagation through chained sets of models in complex applications (e.g., such as additive manufacturing)? Although uncertainty quantification (UQ) is a major challenge for the modeling and simulation community, UQ is critical for making engineering and design decisions. Measurements are often meaningless without knowledge of associated uncertainties. With the increasing utilization of ICME toolsets in industry, it is critical that UQ and its propagation through integrated models become better understood, and that development of understanding in this regard becomes ingrained in the community’s culture. It was also mentioned that UQ often tends to be better addressed by industry than academia. In an industry context, modelers are addressing UQ because of their experience with risk assessment and risk management inherent to the product design cycle and optimization processes. As materials modelers in industry often work from the design
see https://mgi.nist.gov/materials-data-repository.
1382
perspective, it is understood that quantifying and managing uncertainty is important. Designers often need rapid solutions, so managing and understanding uncertainty is critical to providing adequate information for design decisions. Education was also cited as critically important to developing a materials engineering culture that better addresses uncertainty quantification in modeling and simulation. During the open discussion, an audience member stated that undergraduate and graduate coursework addressing UQ is lacking in materials science and engineering (MS and E) programs. Although most university programs emphasize the importance of providing statistical distributions for experimental data, these distributions do not directly translate to bounds on uncertainty in simulated datasets. To address this lack of UQ integration in the ICME community, it was suggested that MS&E programs be adjusted to address such deficits in curriculum, while short courses be provided to professionals in the existing workforce. Funding agencies can also help encourage UQ to be addressed by stipulating it as a requirement on relevant projects. However, since UQ is a challenge in and of itself, requiring this as a project component could add significant complexity to projects. For these reasons, programs such as the NSF’s DMREF (Designing Materials to Revolutionize and Engineer our Future) program encourage cross-disciplinary collaboration. This encouragement is often guided toward interaction between materials modelers and mathematicians. Another suggested avenue relative to funding agencies was the inclusion of reviewer recommendations encouraging certain aspects of project proposals and manuscripts be more adequately considered with respect to uncertainty quantification during the peer review process of proposals or papers. Overall, there was a consensus that UQ must be better addressed in computational modeling and ICME projects. In some cases, the goal of a model may not be to provide a precise answer but simply to provide the broader understanding necessary to guide experimental work, or specify bounds of material phenomena. The requirements of uncertainty quantification are to a large degree contingent on the intended rigor of the model, and some UQ efforts may be greatly simplified by properly understanding the level of required detail. Question 4: How do we more effectively collaborate with the advanced manufacturing community at meetings such as this, or in general? A focus of ICME and the MGI is to accelerate the insertion of innovative materials technologies into everyday products and services. As such, it is imperative that the ICME community focus on coupling efforts with those from other disciplines to impact
Howe, Goodlet, Weaver, and Spanos
advanced manufacturing. To achieve this, software providers should be encouraged to provide solutions to industry via tailored software tools, particularly for complex alloy systems and other challenging materials classes. In addition, materials researchers (especially those in academia and at national laboratories) could prioritize making their software codes and materials datasets available to industry. The availability of such resources could be a significant asset for advanced manufacturing, and their optimization would enhance the usefulness of such tools and datasets even further. Furthermore, such an interaction could significantly drive productive collaboration between industrial and academic laboratory sectors. Since effective implementation of ICME in the advanced manufacturing setting, and industry at large, requires significant investment in ICME tools and personnel, it can be a challenge to make the business case for ICME. This can be partially overcome by a change in branding of ICME. Up to the present, ICME has often been characterized as a largely scientific endeavor to understand materials behavior better. However, it is more accurately described as the employment of collaborative tools and methods to solve industrial problems that will ultimately accelerate the deployment of materials technologies in products. As mentioned, a significant portion of the challenge is communicating the value of a successful ICME implementation in industry as the number and detail of success stories are limited. Increased publishing of ICME success stories to include assessments of return on investment (ROI) will help guide relevant decision-makers in industry and elsewhere. More broadly, stakeholders in industry, whether within advanced manufacturing or elsewhere, can share best practices for ICME integration, and extend efforts across their supply chains and product development processes. It was also pointed out that two barriers to coupling ICME with industry and the broader manufacturing community have not been fully addressed at this ICME World Congress. The first of these is that a decrease of 1/2 in the time required for the deployment of new materials technologies7 may be a positive big-picture metric to target, but it may not correlate to a high ROI for specific ICME projects in industry. This may be particularly true when taking into account the risks and costs associated with changing industrial production paradigms. Second, companies often lack personnel with the expertise to use the modeling and simulation tools currently in use or being developed today. In many cases, the limiting component is not the software tools themselves but the metallurgical or materials science knowledge required to interpret the model results. One potential solution for this would be to provide full-service consulting that couples ICME activities to produce directives and recommendations on specific courses of actions the industry partners should consider for implementation.
Insights from the 3rd World Congress on Integrated Computational Materials Engineering
Question 5: What are the greatest imperatives and challenges of future workforce preparation in ICME? As explored in several presentations at the ICME 2015 World Congress and in the final interactive discussion, preparing the next generation of the ICME workforce is essential for realizing the potential of the discipline. One of the greatest imperatives for preparing the future ICME workforce is the development of courses and curriculum for teaching ICME. Some initial progress has been made on this front, with at least one dedicated ICME text book being published,8 as well as some ICME-centric courses beginning to appear at universities, e.g., Northwestern University, Mississippi State University, and the University of Michigan. Additionally, numerous short courses on ICME have also been provided by professional societies such as the 2015 TMS short course on ICME implementation, which is taught in association with the 2015 ICME World Congress. However, there is much more work to be done in this arena. More broadly, it was observed that interdisciplinary graduate programs could encourage individuals to be effective ICME practitioners. For example, degrees that combine a background in materials engineering with physics and computer science could help prepare graduate students for the interdisciplinary environment of ICME. Discussion also included a variety of strategies for preparing graduate students to be effective users of ICME tools without requiring radical changes to established curricula or department programs. One such approach would be to bring together graduate students from different ICME subdisciplines for a summer research effort to work together to solve a single problem as a group. The program advisors could bring together experts from industry for regular discussions with students regarding the problems at hand, and they could provide background pertinent to the group’s problem. The team of students would all work on one foundational problem, in a location with both characterization and computing resources, and over a realistic timeframe such that measurable progress can be made toward developing a solution. This research effort could serve as a de facto ICME internship for students while being a valuable experience for faculty and providing a helpful result to industry partners. Another aspect to be addressed, particularly by the future workforce, is the overall understanding of the role of modeling in science and engineering. Traditionally, models are a fundamental entity on which science is built, which is why the ICME culture is so important. A more thorough understanding of the role that models serve would encourage all to think in ways that elevate model development and model implementation. However,
1383
seasoned experimentalists often express distrust in modeling results. This amounts to a barrier in community mindset, which will only be overcome with time and evidence. However, the global sentiment shared by discussion participants was that the community can accelerate this process by gathering and publicizing examples of successful ICME implementations that integrate both experimental and modeling approaches, as well as return real value to relevant stakeholders. CONCLUSION The topics discussed here can serve as a guide for future efforts in ICME, in both the short and the long term. The questions discussed also serve as a preview of issues that should be revisited at the 4th World Congress on ICME in 2017. At that meeting, the state of the art in ICME will once again be presented, progress made toward goals will be assessed, and a revised path forward will be charted. The information captured in this discussion can be taken as a call to action, to incorporate ICME concepts and tools within fundamental and applied research settings, and to provide guidance for further incorporation. As always, the community is encouraged to focus on the differentiating aspects of ICME that distinguish it from computational materials science. The ‘‘I’’ for integration signifies collaboration between experimentation and modeling in a multidisciplinary fashion, whereas the ‘‘E’’ for engineering signifies a focus on real-life applications and problems. Programs that focus on these aspects stand to gain the most from ICME toolsets, as well as to provide the greatest promise to accelerate the development and deployment of new material technologies at reduced cost. ACKNOWLEDGEMENT The authors thank Dr. Mark Tschopp, U.S. Army Research Laboratory, for his word frequency analysis from the ICME 2015 World Congress. REFERENCES 1. Integrated Computational Materials Engineering, A Transformational Discipline for Improved Competitiveness and National Security, 1st ed. (Washington, DC: The National Academies Press, 2008). 2. The Minerals, Metals, and Materials Society, Integrated Computational Materials Engineering: Implementing ICME in the Aerospace, Automotive, and Maritime Industries, 1st ed. (Warrendale: The Minerals, Metals, and Materials Society, 2013). 3. J. Allison, M. Li, C. Wolverton, and X.M. Su, JOM 58, 28 (2006). 4. D. Backman, D. Wei, and D. Whitis, JOM 58, 36 (2006). 5. S.M. Arnold and T.T. Wong, Models, Databases, and Simulation Tools Needed for the Realization of Integrated Computational Materials Engineering, Proceedings, 1st ed. (Materials Park: ASM International, 2010). 6. D. Furrer and J. Schirra, JOM 63, 42 (2011). 7. M.F. Horstemeyer, Integrated Computational Materials Engineering (ICME) for Metals: Using Multiscale Modeling
1384 to Invigorate Engineering Design with Science, 1st ed. (New York: Wiley, 2012). 8. M.A. Tschopp, Integrated Computational Materials Engineering: Accelerating Materials Discovery, Development, and Design by Merging Mechanical Engineering, Materials
Howe, Goodlet, Weaver, and Spanos Science, and Computational Science, conference presentation (ASME IMECE Annual Conference, 2015). 9. Materials Genome Initiative for Global Competitiveness (2011). https://www.whitehouse.gov/mgi.