Transforming Combustion Research through Cyberinfrastructure (2011)

The previous chapters in this report describe how modern information technology can be applied to speed the development of new combustion systems and alternative fuels, which would provide significant benefits to the nation’s economy, to the environment, and to U.S. national security. Such technology would also accelerate the advancement of science and increase the effectiveness of many government-sponsored research efforts.

This chapter focuses on the particularly urgent need for development of a cyberinfrastructure (CI) for combustion, to ensure a more rapid and reliable data flow connecting three communities (see Figure 4.1):

1. The fuel research community, which is focused primarily on molecules and chemistry;

2. The reacting-flow research community, which is focused on understanding flame structure and combustion dynamics; and

3. The industrial engine, combustion, and fuels research and development (R&D) community, so that it can rapidly put new understanding of combustion to practical use for society’s benefit.

Because of the different focus and organization of each of these communities, each poses different challenges for a CI.

For the fuels and chemistry communities, a CI must handle an extremely heterogeneous data set, including properties of individual molecules, reaction rates and transport properties, fuel compositions, and experimental data on the performance of different fuels under vari-

FIGURE 4.1 Overview of the needed technical components of a combustion cyberinfrastructure (CI), which will facilitate rapid data flow and improved coordination among various subcommunities contributing to research and development (R&D) in the areas of engines and fuels.

ous conditions. In this research area, which is dominated by chemists, information is usually organized by molecule. These data are being generated at a relatively slow rate by a very large number of researchers in a variety of disciplines scattered around the world. Much of the data has not yet been captured in usable electronic form, and extensive outreach to the combustion-chemistry community will be required to make the data more accessible and useful and to ensure that the data are up to date. Fortunately, in this area there is significant prior work and a history of cooperative community efforts to build on.

For the reacting flow community, the CI must handle very large volumes of data, but the data are more homogeneous, and the computational and experimental data of greatest interest are generated by researchers at supercomputer centers, national laboratories, and universities. Although there are still significant community-outreach and data-documentation issues impacting the flame and reacting-flow research community, they are not as severe as for the fuels and chemistry research community. Many of the flame and reacting-flow community’s challenges here are technical: the mechanics of how to share and use extremely large data sets, how to ensure access to the validation data without compromising security, and so on. In this field, information is organized according to flame type and computational fluid dynamics (CFD) methodology.

To be most valuable to industrial engineers developing new engines, combustors, and fuels, the combustion CI must allow efficient access

to a wide variety of data—from molecular properties to parameters of turbulence models, to data from a broad range of experimental and computational databases. The CI should also provide a variety of software tools for accomplishing common tasks, such as viewing and annotating the data, running simple simulations, and conducting comparisons with experimental data sets. In the industrial R&D community, information is organized by device type (e.g., jet engines, diesels, spark-ignited engines, and stationary turbines).

Despite the significant differences among the three user communities—in the required interfaces and in the data and tools that each will find most interesting—it is essential that a combustion CI smoothly connect the three. This is because data on the particular fuel of interest are the key input to any reacting-flow simulation, whereas outputs from the reacting-flow simulations can identify which fuel components or properties have the greatest effects on performance in different types of engines and combustors. If the CI is working correctly, it is certain that the industrial R&D community will draw heavily on both the chemistry-oriented fuel community and on the CFD-oriented reacting-flow community.

As discussed in Chapter 3, the CI should facilitate two-way communication so that researchers will more quickly become aware of which data are currently available and which additional data are most needed. Because many prospective CI users are located at industrial R&D sites and at U.S. government facilities and because data on efficient engines and fuels are so economically and militarily valuable, the combustion CI would have to be designed to facilitate secure password protection on selected data and on access to computer resources, but without impeding efficient data flow. It is already the case that proprietary hardware and design drawings have been made available to universities for simulation and testing. It may be possible that design information that is not exact but close enough for validation of computational methods can be developed and used by the combustion CI.

Recommendation 1: A unified combustion cyberinfrastructure should be constructed that efficiently and effectively connects with and enables the movement of data and the sharing of software tools among the different research communities contributing to engine and combustion research and development.

The committee believes that the CI should be a platform that makes it easy for the community to contribute, integrate, and share software. Its resources should be largely developed by the CI’s users and not entirely by its developers. Such an approach will also help create community buy-in and reduce costs.

In the sections below, the committee discusses and presents its recommendations with respect to each of the following areas:

• An organizational structure suitable for developing and maintaining the CI;

• The development of an implementation plan, including mechanisms for involving the broader combustion community;

• The educational role of the combustion CI; and

• Budgetary issues and sustainability.

The recommendations present detailed proposals for how this new CI should be created, introduced, and maintained. It is anticipated that building this CI will require an investment of tens of millions of dollars, and that the expected benefits are very much larger.

The proposed CI has three main requirements: (1) it must be unified and efficient, (2) it must interface effectively with the many subcommunities generating and using the data needed for combustion R&D, and (3) it must facilitate the validation of the submodels used in combustion simulations. These requirements and the experience of other CIs, discussed in Chapters 1 and 2, suggest a two-tiered organizational structure, with a central team and three individual outreach teams:

• A central CI team will be needed to devise the overall architecture and assemble and maintain the unified database and data-flow software that will connect all the communities.

• The first outreach team will be focused on the molecule-oriented fuel research community—developing data types, interfaces, and tools that will foster participation by each of the chemistry-oriented subcommunities and that will handle the highly heterogeneous chemistry data (from spectroscopy, calorimetry, kinetics, and so on).

• The second outreach team will be focused on the reacting-flow and turbulent-flame community—developing interfaces and tools for collecting combustion data and comparing them with full-model (or

submodel) predictions. This outreach team will also handle the specialized issues involving the very large data sets associated with petascale computer simulations.

• The third outreach team will ensure that the combustion CI is effective in meeting the needs of the industrial engine and fuels R&D community.

It is anticipated that many fuels researchers will interact with the first outreach team and that many mechanical and aerospace engineers involved in industrial combustion R&D will also interact with the second outreach team.

Recommendation 2: A centralized team will be needed to design and construct a unified, efficient combustion cyberinfrastructure in a timely fashion. At least three individual outreach teams should work closely with a central team: one outreach team connecting with the many chemistry-oriented subcommunities providing fuel data, one team connecting with the reacting-flow and turbulent-flame community, and one team ensuring that the cyberinfrastructure meets the needs of the industrial engine and fuels R&D community. These outreach teams will be responsible for interfaces, specialized software tools, and the development of formats and methods to handle different types of input data, and for the promotion of the new CI within their target communities.

The development and deployment of a CI for combustion that effectively meets all of the requirements discussed above involve a complex, multifaceted, multimillion-dollar project. Detailed planning is necessary to ensure this important project is carried out in an efficient and timely manner. The implementation plan should include detailed specification for the core CI hardware and software, including detailed plans and funding levels with timelines for building the system.

This planning should be accomplished through a study that will also determine what historical data will be included in the CI databases, how those data will be put into the system, and the level of effort required to do so. There have been serious efforts in the past to collect and disseminate some of the data needed for combustion R&D, most notably by the National Institute of Standards and Technology, but also by the Jet Propulsion Laboratory, the American Petroleum Institute, the Joint-Army-Navy-NASA-Air Force Interagency Propulsion Committee, the Process Information Model, and the Turbulent Nonpremixed Flame Workshop. The planning study should indicate how the new CI will build on this

prior work. Discussions with the curators of the data sets referred to above are needed from the outset.

Even more important than in the past, the planning study must determine how to interact most effectively with the various subcommunities involved in going forward. Early interaction with these subcommunities will help build a consensus on what is needed and will provide more detailed specifications on how the interfaces and software tools should work. Discussion groups and workshops may be an effective way to elicit input from the very beginning of the planning process. It is likely that some types of data could be most easily collected by cooperating with the journals publishing the original reports of the planning committee, but this might vary depending on the journal and the culture of the subcommunity involved. Ultimately, the success of the CI will depend on the level of engagement achieved with each community, so careful planning and effective execution of the CI’s outreach component are essential. The implementation plan should also include contingency options for alternative budgetary levels.

Recommendation 3: Because of the many issues involved in the development and deployment of a CI for combustion, experts in several areas—chemistry data, reacting-flow simulations, engine and fuels R&D, software development, CI maintenance, data curation, deployment, outreach, and education—all need to be involved in the planning, design, and construction of the combustion CI.

A critical issue in the implementation plan is the question of how to get researchers to use the new system. It is likely that separate outreach teams will be needed for each of the subcommunities mentioned in this recommendation. In addition, the committee anticipates that once the CI is established, the added value that it provides to researchers will naturally lead to broad participation. Initially an incentive structure will facilitate early adopters. Federal grant agencies can provide incentives in many ways. In other fields, such as biomedical research, some agencies require data to be deposited in certain electronic archives prior to submission for journal publication. Such a requirement has been imposed by the National Science Foundation (NSF) for research grants beginning January 18, 2011:

Information regarding how a researcher interacts with the CI system, both as a user and as a provider of data and software, could be included in project reports and grant applications. This information could also play a role in the evaluation of progress in incorporating the use of the CI into the combustion community, along with journal publications and other metrics.

Recommendation 4: Federal research agencies responsible for funding combustion research should incorporate specific policies regarding the use of the combustion cyberinfrastructure into their progress reports and their grant processes. The incorporation of such policies will provide incentives to the combustion community and related communities for making the transition to the new system for handling and archiving valuable data.

As demonstrated by the success of nanoHUB, a CI offers great opportunities to improve combustion education, including the following: better integration of combustion science and combustion engineering, providing stronger context and motivation for students; emphasis on the multiscale, multi-science nature of combustion, which positions it to be an excellent exemplar of state-of-the-art computational science and engineering; and better integration of combustion science with both computer and computational sciences, to expose students to a broader range of necessary tools and concepts.

It is noteworthy that instructors teaching combustion have a long history of working with open-source, cyber-based scientific data and software libraries, a history that started with the pioneering years of CHEMKIN (see Appendix B in this report). In combustion as in other engineering fields, the development of a cyber-based infrastructure drives profound transformations in the work environment and in professional practice; these transformations must be accompanied by corresponding changes in educational programs. Important ideas that should guide this transformation include the following:

• A renewed emphasis on establishing stronger pedagogical ties between fundamentals (i.e., thermodynamics, chemistry, fluid mechanics, heat transfer) and applications (e.g., engine design). By promoting unprecedented levels of integration, the CI can provide new ways to bridge the gap between different subcommunities (from commu-

nities of scientists who generate scientific ideas, to communities of engineers who generate design ideas and thereby express new scientific needs) and can enlist these subcommunities into a common framework.

• The promotion of combustion as a multiscale discipline (from quantum and nanoscales to engineering device scales). Combustion scientists and engineers will need to be exposed to an increasingly broad, cross-disciplinary, technical framework—for instance, to the concepts and tools of a multiscale approach to combustion. The framework of multiscale combustion includes molecular dynamics occurring at nanoscales, laminar-flame chemistry occurring at millimeter scales, turbulent-flame dynamics occurring at centimeter scales, and the overall engineering systems’ performance characterized by scales on the order of tens of centimeters or more.

• The integration of data science (a computer science topic) and scientific computing (a computational science topic) into the combustion curriculum. The future needs of a cyber-based work environment will require the development of stronger ties among computer science, computational science, and domain science. For instance, combustion science and engineering students will need to be educated in the concepts and tools of the CI. These include a variety of information technology (IT) methods, such as software design, data structures, data visualization, and network architectures, in a distributed and heterogeneous (grid-like) environment, as well as a variety of computational science methods, such as numerical methods, parallel software design, and parallel computing optimization.

The committee envisions two main educational components in the combustion CI:

1. The development of a combustion portal similar to nanoHUB.org that would be a CI-enabled resource to the combustion learning community at large (including students, scientists, engineers, and policy makers). This resource would host data and software libraries and a portfolio of pedagogical Web-executable tools for focused classroom activities. In addition, it would host general instruction material for a reference combustion curriculum (including introductory tutorials, advanced courses, topical lectures, and so on). Finally, it would host introductory material for the general public, explaining the role and place of combustion science and engineering in meeting today’s and tomorrow’s energy challenges.

2. The development of an advanced training program that would be a CI-enabled resource to the combustion research community. This training program could take the form of specialized workshops, summer schools, Web conferences, and so on, and it would be aimed at disseminating the tools, standards, and methodologies of CI-enabled combustion science and engineering to the broader combustion research community.

The combustion portal will help develop and articulate a community-wide vision for combustion, including an enhanced integration of combustion science and combustion engineering. The advanced training program will help promote combustion as a multiscale, multi-science discipline and a leading application field for an enhanced CI with strong ties to computer and computational sciences.

These aspects of the combustion CI will facilitate the sharing of tools, data, and practices. However, new techniques for accelerating the process of journal publication, discussion, and critical dialogue, perhaps through mechanisms embedded in the CI, will be needed.

Recommendation 5: The combustion cyberinfrastructure should be designed to serve the chemistry and education communities as well as the research community, and to integrate these communities with advances in computer science research and education.

In its report Revolutionizing Science and Engineering Through Cyber Infrastructure (NSF, 2003, p. 17), the NSF elaborated on the contribution that a CI would make to education in any scientific or engineering discipline:

This committee believes that a combustion CI shares this promise.

A precise estimate for the costs for implementing the proposed combustion CI cannot be determined until after the detailed implementation

plan is developed. However, comparison with similar recent or ongoing CI efforts such as the CI at the National Center for Supercomputing Applications² (see Chapter 2 in this report) allows an estimate of the number of full-time employees that would be required. The centralized information technology functions of developing and maintaining the core database and data-flow software are estimated to require approximately 12 full-time-equivalent (FTE) employees, mostly computer programmers. The initial development of the specialized user interfaces, data types, and software; the related outreach to the many disparate subcommunities involved in combustion research; and the electronic capture of the historical data in each area will require a much larger workforce, which is a mix of domain experts (i.e., scientists and engineers who can communicate well with each subcommunity and who are familiar with the science issues) and computer programmers and data curators. This initial effort could easily require approximately 50 FTE employees for several years, although it might be decided to roll this effort out in stages, in part owing to the difficulty in rapidly identifying and hiring a sufficient number of qualified domain scientists and computer scientists. A team of this size, beginning a completely new project, would require a leadership and management team of perhaps 8 more FTE employees, yielding an estimate of about 70 FTE total employees at the project peak.

It is expected that a much smaller number of FTE employees would suffice to maintain the CI after it is established, but this ongoing maintenance is key to the success of the whole effort, and a revenue stream must be identified that will support the CI after it is initially constructed.

Many prior efforts at developing combustion CI have had little longterm value to the R&D community because they collapsed for lack of funding after the initial grants expired. The committee believes that a portion of the required ongoing revenue stream to maintain the combustion CI should come from industry. However, additional funds will certainly be required from the federal government, in part because so many of the likely users of the CI are in universities and national laboratories doing federally funded research, but also because the CI will no doubt be viewed as a “commons” rather than as a source of competitive advantage for an individual firm. The CI will be expensive, requiring new hardware, newly developed software, and a continuing support staff, all of which must be kept current. It is essential that the funding agencies plan for how to meet this long-term funding requirement from the very beginning. Otherwise, it is likely that society will not capture the value associated with this significant project. However, the CI should seek additional funding from sources other than the federal government, both for its own sustain-

ability and to ensure that it is meeting the diverse needs of the combustion community. Such sources could include professional societies, industrial participants, and direct charges for the use of some of its facilities.

Recommendation 6: A fairly large short-term investment is required to achieve the benefits of a unified combustion cyberinfrastructure. Ongoing operations of this CI will require significant continuing funds. A failure to secure a continuing funding stream to maintain the CI will likely lead to the failure of the whole project.

The NSF, in its study Investing in America’s Future, identified four goals to meet the national scientific needs: discovery, learning, research infrastructure, and stewardship. The combustion CI proposed in this report contributes to all of these goals; but, in particular, it addresses NSF’s research infrastructure goal to “develop a comprehensive integrated cyberinfrastructure to drive discovery in all fields of science and engineering” (NSF, 2006, pp. 26-27). The development of a community-wide CI for combustion is a unique opportunity for the combustion community to reshape its structure and traditional modes of operation and thereby to achieve higher levels of integration and productivity. Such a community-wide CI is expected to drive a transformation of the combustion research community from a fragmented group of researchers and engineers characterized by light infrastructure and small research teams to an integrated community characterized by networked infrastructure and multidisciplinary research teams that function throughout the community. In the past, such a transformation toward a community-level integrated framework has been typically observed in research communities that share a large brick-and-mortar infrastructure (for instance, in the particle physics community, a community connected by the common need to use large-scale facilities such as particle accelerators). It is expected that a community CI will drive a similar transformation in the combustion research community; the drive and connections will, in that case, be provided by the common need to share data, software tools, computing resources, and personnel as well as the desire to bridge the gap between basic sciences and engineering applications.

NSF (National Science Foundation). 2003. Revolutionizing Science and Engineering Through Cyber Infrastructure: Report of the National Science Foundation Blue-Ribbon Advisory Panel on Cyberinfrastructure. Arlington, Va.: National Science Foundation. Available at http://www.nsf.gov/cise/sci/reports/atkins.pdf. Accessed December 10, 2010.

NSF. 2006. Investing in America’s Future. Arlington, Va.: NSF.