Revolutionizing Engineering Science

through Simulation
May 2006

Report of the National Science Foundation

Blue Ribbon Panel on
Simulation-Based Engineering Science
ii
A Report of the National Science Foundation
Blue Ribbon Panel on
Simulation-Based Engineering Science

Revolutionizing Engineering Science

through Simulation

May 2006

iii
iv
v
vi
 THE NSF BLUE RIBBON PANEL ON
SIMULATION-BASED ENGINEERING SCIENCE

J. Tinsley Oden, Chair

Associate Vice President for Research
Director, Institute for Computational Engineering and Sciences
Cockrell Family Regents Chair in Engineering
The University of Texas at Austin
Ted Belytschko
Walter P. Murphy Professor and McCormick Professor
Department of Mechanical Engineering
Northwestern University
Jacob Fish
The Rosalind and John J. Redfern, Jr.’33 Chaired Professor in Engineering
Department of Mechanical, Aerospace and Nuclear Engineering
Rensselaer Polytechnic Institute
Thomas J.R. Hughes
Professor, Aerospace Engineering and Engineering Mechanics
Computational and Applied Mathematics Chair III
Institute for Computational Engineering and Sciences
The University of Texas at Austin
Chris Johnson
Director, Scientific Computing and Imaging Institute
Co-Director, Center for Integrated Biomedical Computing
Distinguished Professor, Computer Science
Research Professor, Bioengineering
Adjunct Professor, Physics
University of Utah
David Keyes
Fu Foundation Professor of Applied Mathematics
Department of Applied Physics and Applied Mathematics
Acting Director, Institute for Scientific Computing Research at LLNL
Columbia University

vii
Alan Laub
Director, Institute for Digital Research and Education
Professor, Electrical Engineering / Mathematics
University of California – Los Angeles
Linda Petzold
Chair, Department of Computer Science
Professor, Department of Mechanical and Environmental Engineering
Professor, Department of Computer Science
Director, Computational Science and Engineering Program
University of California, Santa Barbara
David Srolovitz
Department Chair, Mechanical and Aerospace Engineering
Professor, Mechanical and Aerospace Engineering
Professor, Princeton Institute for the Science and Technology of Materials
Princeton University
Sidney Yip
Professor, Nuclear Engineering and Materials Science Engineering
Computational and Systems Biology (CSBi)
Massachusetts Institute of Technology

Panel Administrative Coordinator:

Jon Bass
Deputy Director, Institute for Computational Engineering and Sciences
The University of Texas at Austin

viii
PREFACE

This document is the final report of the findings and recommendations of the
Blue Ribbon Panel on Simulation-Based Engineering Science. The report
contains recommendations critical to the acceleration of advances in Simulation-
Based Engineering Science (SBES), and it identifies several areas in which SBES
can play a remarkable role in promoting developments vital to the health,
security, and technological competitiveness of the nation.
For over a decade, the nation's engineering and science communities have
become increasingly aware that computer simulation is an indispensable tool for
resolving a multitude of scientific and technological problems facing our country.
To define the field of computer simulation more precisely and to assess its
potential impact on important areas of engineering science, in April 2004 the
NSF organized a workshop on SBES. Encouraged by the widespread interest in
the results of the workshop, the Foundation appointed a Blue Ribbon Panel on
Simulation-Based Engineering Science. The purpose of the Panel was to explore
the challenges and potential benefits of SBES, as well as the barriers to its
development. Furthermore, the Panel was tasked with making recommendations
on how the discipline could be nurtured within academia, industry, national
laboratories, and government agencies. A second workshop on SBES was held in
September 2005, at which time the Panel received input on SBES from a broad
constituency.
Acknowledgements: The Panel has benefited from the advice and council of
many individuals. The enthusiastic support of Drs. Richard Buckius and Ken
Chong of NSF and Dr. John Brighton, formerly of NSF, is gratefully
acknowledged. Valuable advice from the over 100 workshop attendees has also
been factored into the findings and recommendations. In addition, a large group
of experts outside the workshops have provided valuable commentary on the
early drafts of this report.
The Panel gratefully acknowledges the following individuals and institutions
for providing the graphics that appear in this report: NASA’s Earth Observatory
(Cover Image); Charles Taylor, Stanford University (Figure 1); Y. Zhang and C.
Bajaj, University of Texas and the New York University School of Medicine,
respectively (Figure 2); SCI, University of Utah (Figure 3); SCI, University of
Utah, and J. Bell and V. Beckner, Lawrence Berkeley National Laboratory
(Figure 4); and the U.S. Department of Energy (Figure 5, a reproduction from the
SCaLeS report). Particularly helpful in the preparation of this final report were
the comments and advice of Drs. Chandrajit Bajaj, Larry Biegler, Mark

ix
Carpenter, Alok Chaturvedi, Weng Cho Chew, Frederica Darema, Omar Ghattas,
George Hazelrigg, Craig Henriquez, Anthony Ingraffea, Kirk Jordan, Chandrika
Kamath, Dimitri Kusnezov, Wing Kam Liu, Robert Moser, Habib Najm,
Anthony Patera, Mark Rashid, Mark Shephard, Charles Taylor, Elizabeth
Tornquist, Mary Wheeler, Daniel White, Jacob White, and David Young,.
Although this report was prepared by a Panel of the National Science
Foundation, all opinions, findings, and recommendations expressed here are
those of the Panel and do not necessarily reflect the views of the National
Science Foundation.

x
CONTENTS

PREFACE................................................................................................................ix
CONTENTS ............................................................................................................xi
EXECUTIVE SUMMARY .......................................................................................xiii
1.0 SBES: A National Priority for Tomorrow’s Engineering and Science .............1
2.0 THE PAYOFF: Driving Applications and Societal Benefits of SBES ................9
2.1 SBES in Medicine .............................................................................................9
2.2 SBES in Predictive Homeland Security ........................................................13
2.3 SBES in Energy and the Environment ..........................................................17
2.4 SBES in Materials ..........................................................................................18
2.5 SBES in Industrial and Defense Applications ...............................................22
3.0 CORE ISSUES: Challenges, Barriers, and Opportunities in SBES
Research ..........................................................................................................29
3.1 The Tyranny of Scales: The Challenge of Multiscale Modeling and
Simulation........................................................................................................29
3.2 Verification, Validation, and Uncertainty Quantification..............................33
3.3 Dynamic Simulation Systems, Sensors, Measurements, and
Heterogeneous Simulations ...........................................................................37
3.4 New Vistas in Simulation Software...............................................................40
3.5 The Emergence of Big Data in Simulation and the Role of
Visualization in SBES .....................................................................................44
3.6 Next-Generation Algorithms and Computational Performance...................49
4.0 THE CRISIS OF THE KNOWLEDGE EXPLOSION: SBES Education for
Tomorrow’s Engineers and Scientists ............................................................53
5.0 CONCLUSIONS................................................................................................57

xi
Appendix A: SBES Workshop Attendees ..............................................................61
April 2004 Workshop ...........................................................................................61
September 2005 Workshop..................................................................................62
BIBLIOGRAPHY.....................................................................................................63

xii
 EXECUTIVE SUMMARY

Simulation refers to the application of computational models to the study and

prediction of physical events or the behavior of engineered systems. The
development of computer simulation has drawn from a deep pool of scientific,
mathematical, computational, and engineering knowledge and methodologies.
With the depth of its intellectual development and its wide range of applications,
computer simulation has emerged as a powerful tool, one that promises to
revolutionize the way engineering and science are conducted in the twenty-first
century.

Computer simulation represents an extension of theoretical science in that it

is based on mathematical models. Such models attempt to characterize the
physical predictions or consequences of scientific theories. Simulation can be
much more, however. For example, it can be used to explore new theories and to
design new experiments to test these theories. Simulation also provides a
powerful alternative to the techniques of experimental science and observation
when phenomena are not observable or when measurements are impractical or
too expensive.

Simulation-Based Engineering Science (SBES) is defined as the discipline

that provides the scientific and mathematical basis for the simulation of
engineered systems. Such systems range from microelectronic devices to
automobiles, aircraft, and even the infrastructures of oilfields and cities. In a
word, SBES fuses the knowledge and techniques of the traditional engineering
fields—electrical, mechanical, civil, chemical, aerospace, nuclear, biomedical,
and materials science—with the knowledge and techniques of fields like
computer science, mathematics, and the physical and social sciences. As a result,

xiii
engineers are better able to predict and optimize systems affecting almost all
aspects of our lives and work, including our environment, our security and safety,
and the products we use and export.

Whereas the use of computer simulations in engineering science began over

half a century ago, only in the past decade or so have simulation theory and
technology made a dramatic impact across the engineering fields. That
remarkable change has come about mainly because of developments in the
computational and computer sciences and the rapid advances in computing
equipment and systems. There are other reasons. For example, a host of
technologies are on the horizon that we cannot hope to understand, develop, or
utilize without simulation. Many of those technologies are critical to the nation’s
continued leadership in science and engineering. Clearly, research in SBES is
quickly becoming indispensable to our country’s security and well-being.

This report was prepared by a Blue Ribbon Panel on Simulation-Based

Engineering Sciences, which was convened by the Assistant Director of the
Engineering Directorate of the National Science Foundation (NSF). The Panel
was directed to explore opportunities for and potential advances in SBES and to
make strategic recommendations as to how to structure programs to foster SBES.

The Panel developed its findings and recommendations from several

information sources. Among them, interactions with recognized leaders of the
computational engineering and science communities played an essential role.
Another important source of information was the work of previous panels and
committees. The results of those earlier efforts, which have accumulated over the
last decade, address major issues in the computational and computer sciences.
The Panel also relied on input from leaders in the computer simulation
community who participated in the NSF-supported workshops on SBES. Finally,
the Panel developed its findings and recommendations after thorough discussions
among its members.

This report explores the potential impact of advances in SBES on science and

xiv
technology and identifies the challenges and barriers to further advances in
SBES. For instance, we must overcome difficulties inherent in multiscale
modeling, the development of next-generation algorithms, and the design and
implementation of dynamic data-driven application systems. We must improve
methods to quantify uncertainty and to model validation and verification. We
must determine better ways to integrate data-intensive computing, visualization,
and simulation. Importantly, we must overhaul our educational system to foster
the interdisciplinary study that SBES requires. The payoffs for meeting these
challenges are profound. We can expect dramatic advances on a broad front:
medicine, materials science, homeland security, manufacturing, engineering
design, and many others.

For more than a decade, researchers and educators in engineering and science
have agreed: the computational and simulation engineering sciences are
fundamental to the security and welfare of the United States. The findings and
recommendations of this report strongly reinforce that contention.

xv
 Major Findings
1. SBES is a discipline indispensable to the nation’s continued leadership in
science and engineering. It is central to advances in biomedicine,
nanomanufacturing, homeland security, microelectronics, energy and
environmental sciences, advanced materials, and product development. There
is ample evidence that developments in these new disciplines could
significantly impact virtually every aspect of human experience.

2. Formidable challenges stand in the way of progress in SBES research. These

challenges involve resolving open problems associated with multiscale and
multi-physics modeling, real-time integration of simulation methods with
measurement systems, model validation and verification, handling large data,
and visualization. Significantly, one of those challenges is education of the
next generation of engineers and scientists in the theory and practices of
SBES.

3. There is strong evidence that our nation’s leadership in computational

engineering and science, particularly in areas key to Simulation-Based
Engineering Science, is rapidly eroding. Because competing nations
worldwide have increased their investments in research, the U.S. has seen a
steady reduction in its proportion of scientific advances relative to that of
Europe and Asia. Any reversal of those trends will require changes in our
educational system as well as changes in how basic research is funded in the
U.S.

xvi
 Principal Recommendations
1. The Panel recommends that the National Science Foundation and other Federal
research agencies make changes in their organizational structures to facilitate long-
range core funding of SBES. The new Cyberinfrastructure at NSF is envisioned as a
“portion of cyberspace” where scientists can “pursue research in new ways and with
new efficiency” by utilizing: 1) high-performance, global-scale networking, 2)
middleware, 3) high-performance computing services, 4) observation and
measurement devices, and 5) improved interfaces and visualization services. Serious
consideration should be given to the feasibility of developing a parallel program in
SBES that interfaces to multiple divisions of NSF in concert with
Cyberinfrastructure. Supporting SBES research should certainly be a goal of every
division within the Directorate of Engineering at NSF, but the realization of the full
potential of advances in SBES will require support across all directorates and from
other federal agencies as well.
2. To maintain our leadership in science and engineering, the Panel recommends a
minimum increase in NSF funding of $300 million per year over 2005 levels of
SBES-related disciplines. We cannot maintain our leadership in engineering and the
engineering sciences without substantial investments in SBES, because simulation is
a key element in accelerating progress in engineering. Advances in computing speed
alone or in measurement devices or in networking or interfaces cannot meet the great
challenges before us without advances in the basic components of SBES. Similar
observations have been made in the President’s Information Technology Advisory
Committee (PITAC) report, as well as in the results of several other related studies.
3. The Panel recommends a long-term program of high-risk research to exploit the
considerable promise of SBES. The Panel strongly supports the observation made in
the PITAC report and elsewhere that short-term investments and limited strategic
planning will lead to excessive focus on incremental research rather than on long-
range sustained research with a lasting impact. Progress in SBES will require the
creation of interdisciplinary teams that work together on leading-edge simulation
problems. The work of these teams should be sustained for a decade or more to yield
the full fruits of the investment.
4. The Panel recommends that NSF underwrite the work of an NRC committee to
explore the issue of interdisciplinary education in detail and to make
recommendations for a sweeping overhaul of our educational system. The problem
of education in SBES-component disciplines, and in multidisciplinary programs in
general, is large, pervasive, and critically important. The initiative for change will
not likely come from academia alone; it must be encouraged by the engineering and
scientific leadership and throughout the organizational structure of our universities.

Other important findings and recommendations of the Panel are given in the
body of the report.

xvii
xviii
 1.0 SBES: A National Priority for
Tomorrow’s Engineering and
Science

Today the field of computer simulation is on the threshold of a new era.

Advances in mathematical modeling, computational algorithms, the speed of
computers, and the science and technology of data-intensive computing have
prepared the way for unprecedented improvements in the health, security,
productivity, and competitiveness of our nation. To realize these advances,
however, we must overcome major scientific, engineering, sociological, and
educational obstacles. Progress will require significant breakthroughs in research,
changes in the research and educational cultures of our academic institutions, and
changes in the organizational structure of our educational system. For the
engineering fields, advances in computer simulation offer rich possibilities. Full
exploitation of the new capabilities, however, must await basic research into the
scientific components of modeling, simulation, and computing, among other
areas. We refer to the combination of these basic research activities as
Simulation-Based Engineering Science, or SBES.

In this report we describe this new discipline. We first identify some of the
remarkable benefits SBES brings to technologies that make our lives healthier,
safer, and better. Next, we survey the core issues of SBES, that is, the major
obstacles to and opportunities for its development. We then explore the impact of
SBES on our national research and educational resources, goals, and
organizations. Throughout the report, we highlight our findings and
recommendations. In making those recommendations, we attempt to reflect the
views prevailing within the nation’s scientific and engineering communities.

1
 This much is certain: there is overwhelming concurrence that simulation is
key to achieving progress in engineering and science in the foreseeable future.
Indeed, seldom have so many independent

Seldom have so many studies by experts from diverse perspectives

been in such agreement: computer simulation
independent studies been
has and will continue to have an enormous
in such agreement: impact on all areas of engineering, scientific
simulation is a key discovery, and endeavors to solve major
societal problems. This message is woven into
element for achieving
the principal findings of many key
progress in engineering
investigations. For example, the PITAC report
and science. [4] emphasizes the need to develop
computational science for national
competitiveness, and the SciDAC [26] and SCaLeS [18, 19] reports identify
opportunities for scientific discovery at the high end of today’s simulation
capabilities and call for a new scientific culture of interdisciplinary teamwork to
realize those capabilities. In addition, the Roadmap for the Revitalization of
High-End Computing [24] and the Federal Plan for High-End Computing [13]
both call for innovations in computer architecture to accommodate advanced
simulation. The Future of Supercomputing report [12] examines the role of the
Federal Government in sustaining the leading edge of supercomputing, and the
Cyberinfrastructure report [1] outlines a diverse program of interrelated research
imperatives stretching well beyond simulation into communication and data
technologies. Beginning with the Lax report of 1982 [20] and continuing with a
number of reports already appearing in this young century [3, 8, 9, 23], the
studies share a similar vision regarding the importance of simulation.

Consequently, the present report enters an arena already filled with voices
calling for more vigorous research and training in computation-based simulation.
The ideas in this report are in harmony with those voices, and in fact the report is
as brief as it is because others have already eloquently articulated the case for

2
simulation. Even so, this report addresses important elements of simulation that
have been overlooked. Moreover, it adds the voice of engineering to the
discussion, one that has not yet been fully heard.

Realizing the full potential of SBES will require a revolution in simulation

technology. Simulation-Based Engineering Science is not “simulation as usual”;
rather, it is research focused on the modeling
and simulation of complex, interrelated
SBES constitutes a new
engineered systems and on the acquisition of
results meeting specified standards of precision
paradigm that will be
and reliability. Indeed, the scope of SBES indispensable in meeting
includes much more than the modeling of the scientific and
physical phenomena. It develops new methods,
engineering challenges of
devices, procedures, processes, and planning
strategies. Not only does it draw on and the twenty-first century.
stimulate advances in our scientific
understanding, it capitalizes on those advances by applying them to challenges in
the engineering domain. For example, discoveries in SBES have direct
applications to optimization, control, uncertainty quantification, verification and
validation, decision-making, and real-time response. In short, SBES will lift
simulation to a powerful new level, a level where we hope to solve the most
stubborn problems of modeling, engineering design, manufacturing, and
scientific inquiry. In fact, so profound are the implications of advanced
simulation techniques that we can expect SBES to trigger the development of a
host of aggressive new technologies and to foster significant new scientific
discoveries.

Consider, for example, a few of the breakthroughs that SBES offers: (1) the
means to understand and control multiscale, multi-physics phenomena;
(2) fundamental developments in nanotechnology, biomedicine, materials, energy
and environment, and the earth and life sciences; and (3) dramatic enhancements
to the fidelity and utility of computational predictions. Clearly, SBES ushers in a

3
 technology that not only expands the reach and capability of every field of
engineering but also promises significant improvements in the health, security,
competitiveness, and wealth of our nation.

If we are to reap the benefits of SBES, however, we must first overcome the
obstacles. First, we must revolutionize the way we conceive and perform
simulation. This revolution requires that we learn to incorporate new discoveries
that simplify and enhance multiscale, multidisciplinary simulations. Second, we
must make significant advances in the supporting technologies, including large-
scale computing, data management, and algorithms. Third, we must overhaul our
educational institutions to accommodate the needs of SBES research and
training. Fourth, we must change the manner in which research is funded and
conducted in the U.S.

So far, developments in simulation have ridden the wave of advances in

hardware and software information technology. Because of these advances,
simulation has become an increasingly effective tool in traditional science and
engineering practices. Unlike most theory, which
Simulation has become posits restricted, idealized systems, simulation

indispensable in deals with real systems. For that reason,

simulation provides unprecedented access to
predictive methods for
real-world conditions. In addition, simulation has
weather, climate change, none of the limitations of experimental designs
and behavior of the and tests, which are often hampered by cost
constraints, unrealistic parameter ranges, and
atmosphere; and in
even restrictions imposed by treaties or health
broad areas of and environmental concerns. For these reasons,
engineering analysis and computer simulation is credited with numerous

design. triumphs in the twentieth century. It has become

indispensable, for example, in assessments of
vehicle crashworthiness. It is fundamental to the generation of predictive models
of the weather, climate change, and the behavior of the atmosphere. Its

4
importance in broad areas of engineering analysis and design are well known. It
has become essential to product manufacturing. Its achievements in biomedical
applications are widely discussed. Systems design in defense, communication,
and transportation also rely on computer simulation.

At the heart of these successes, however, are simulation methodologies that

are decades old, too old to meet the challenges of new technology. In many ways,
the past successes of computer simulation may be its worst enemies, because the
knowledge base, methods, and practices that enabled its achievements now
threaten to stifle its prospects for the future.

Our nation prides itself in being the leader in computational science and
simulation theory and technology. Unfortunately, many indicators suggest that
the United States is quickly losing ground. Particularly in SBES, the country is
no longer positioned to lead the world over the next few decades. Even today, the
consequences of falling behind are penetrating deep into our technology and
economy, as well as jeopardizing our position in the
global community.
The importance and
Our global competitors are well aware of the great
potential of computer simulation. Throughout Europe great potential of
and Asia, governments are making major investments simulation have not
in computing, mathematical and computational
gone unnoticed by
modeling, algorithms, networking, and generally in
computational engineering and science. Indeed, these
our competitors
nations are building on the technologies that the U.S. around the world.
pioneered in the twentieth century. We are in danger,
once again, of producing world-leading science but leaving it to our competitors
to harvest the technological and economic advantages.

Yet, even our traditional lead in basic research is under threat. According to
[4, p. 9]: “Since 1988, Western Europe has produced more science and
engineering journal articles than the United States and the total growth in

5
research papers is highest in East Asia (492 percent), followed by Japan (67
percent), and Europe (59 percent), compared with 13 percent for the United
States. Worldwide, the share of U.S. citations in scientific papers is shrinking,
from 38 percent in 1988 to 31 percent in 2001.” In Germany, 36 percent of
undergraduates receive degrees in science and engineering; in China, 59 percent,
and in Japan, 66 percent. In contrast; only 32 percent of undergraduates receive
such degrees in the United States [6, 25, 21].

The imbalance is beginning to reveal itself in international trade. “From 1980

to 2001, the U.S. share of global high-technology exports dropped from 31
percent to 18 percent, while the share for Asian countries rose from 7 to 25
percent” [4, p. 8]. Since 2001, the U.S. trade balance in high-tech products has
been negative.

The chief global economic competitor of the United States is China. In 2004
China graduated approximately 498,000 bachelor’s level engineers. By
comparison, India graduated 350,000 engineers, and the U.S. graduated 70,000
[6, 25]. The employment of an engineer in China costs roughly one-tenth to one-
sixth of what it costs in the United States. Some argue, however, that the U.S.
production of engineers, computer scientists, and information technology
specialists remains competitive in global markets when like-to-like data from the
representative countries are compared [11]. Nevertheless, even our competitors
in SBES believe that SBES research expenditures in Europe and Asia are rapidly
expanding while they are stagnant or declining in the U.S.

The key to offsetting those disadvantages is leadership in simulation. The

U.S. must be in the forefront of efforts to make simulation easier and more
reliable. We must extend the capabilities of simulation to the analysis of more
complex systems and the real-time acquisition of real-time data. Simulation must
no longer be relegated to the peripheries of an engineering student’s skill set;
instead, it should be a core part of the engineering curriculum, where it plays a
role in effective pedagogy.

6
 Computer simulation has become indispensable to the development of all
other technologies, including microelectronics, advanced materials,
biotechnology, nanotechnology, pharmaceuticals, medicine, and defense and
security. Many breakthroughs in these technologies derive from computer
simulation and simulation-based scientific discovery. Clearly, we must integrate
computer simulation into engineering education and practice. To do so, however,
requires great intellectual resources and a national commitment to SBES.

In Chapter 2, we describe some of the remarkable applications of SBES, all

of which represent enormous benefits to our society. If we are to realize such
great potential, however, we must overcome many technical, mathematical,
scientific, and computational challenges. So great are these challenges that they
will require a sea change in our approach to engineering and science education.
Chapter 3 surveys the greatest of those obstacles. In addition, the chapter
identifies opportunities for broadening the impact of SBES through
developments in visualization, sensors and image processing, and uncertainty
quantification.

Finally, in Chapter 4 we describe the impact of SBES on our educational

system. Our educational system must change if it is to equip future generations of
engineering scientists with knowledge of SBES and its value in science and
technology.

7
8
2.0 THE PAYOFF:
Driving Applications and
Societal Benefits of SBES

The applications and benefits of SBES are many. This chapter reviews
some of the important applications of SBES and explores the challenges and
benefits of each.

2.1 SBES in Medicine

Most diseases (such as heart disease, cancer, stroke, and respiratory

diseases) and their treatments (whether surgical, transcatheter, or pharmacologic)
involve complex physical responses and interactions between biological systems,
from the molecular to organism scales. Simulation methods can therefore
dramatically increase our understanding of these diseases and treatments, and
furthermore, improve treatment. Computational science has already made
significant inroads in many biomedical domains, most notably genomics and
proteomics. A current challenge is the application of SBES to clinical medicine
and to the study of biological systems at the cellular, tissue, and organ scales.

Medical practice and engineering have many

similarities. Both are problem-solving disciplines, and Both medical practice
both require an understanding of complex systems. and engineering are
Engineering design processes, however, are based
problem-solving
upon predicted outcomes. Often, engineering solutions
disciplines.

9
 require the satisfaction of numerous criteria simultaneously. Those solutions
often require sophisticated computer and analysis technologies. By contrast,
medical practice uses a “build them and bust them” approach. Historically, the
paradigm of medicine combines diagnosis and empiricism; that is, physicians use
various tests to diagnose a medical condition and then plan a treatment or
intervention based on empirical data and professional experience. Generally,
medical practice precludes any formal process to predict the outcome of a
treatment for an individual patient, although there may be some statistical data to
indicate the success rate of the treatment.

Figure 1: Example of a simulation-based medical approach applied to the design of a bypass

surgery procedure for a patient with occlusive cardiovascular disease in the aorta and iliac
arteries. Shown from left to right are magnetic resonance image data, a preoperative geometric
solid model, an operative plan, computed blood flow velocity in the aorta and proximal end of
bypass, and post-operative image data used to validate predictions. Advances contemplated in
SBES disciplines could greatly improve these approaches by allowing real-time adaptive
control of surgical procedures through the combination of simulation tools and imaging
technologies.

A program in SBES could lead to new

A program in SBES approaches to medical practice, approaches that

could also lead to could be collectively called Simulation-Based

Medicine. New SBES methodologies, already
new approaches to
leveraging tremendous advances in medical imaging
medical practice, and high-performance computing, would give the
collectively called practice of medicine the tools of modern
engineering. For example, doctors would be able to
Simulation-Based
use simulations—initialized with patient-specific
Medicine.

10
anatomic and physiologic data—to predict the outcomes of procedures and
thereby design optimal treatments for individual patients. This ability to predict
treatment outcomes and design procedures accordingly represents an exciting
new possibility for medicine.

Not only could physicians devise better treatments for individual patients, but
also manufacturers could use SBES methods to predict the performance of their
medical devices in virtual patients. The physical and animal testing procedures
currently used prior to human trials have significant limitations in their ability to
represent variations in human anatomy and physiology. With SBES methods,
manufacturers could conduct virtual prototyping of medical devices by
simulating the performances of alternate device designs for a range of virtual

Figure 2: The modeling of biomedical systems is becoming increasingly

sophisticated. Here is an example of a three-dimensional, tetrahedral-mesh,
heart model [27] developed from surface data obtained from the New York
University School of Medicine [22]. (a) Exterior view. (b) Boundary detection
represented in wire frame. (c) Cross-sectional view of the mesh. Patient-
specific modeling technologies need to be advanced significantly to make the
vision of Simulation-Based Medicine a reality. The benefits, however, are
impressive: dynamic models of deformation, blood flow, and fine-scale
capillary effects may greatly advance cardiovascular medicine.

11
patients. In this way, manufacturers would be able to refine their designs for
different patient conditions. As a result, these virtual clinical trials prior to
animal and human trials could lead to safer, more effective devices, reduce
development costs, and shorten time-to-market.

Manufacturers in pharmaceuticals and biotechnology could also benefit from

SBES methods. For example, targeted drug-delivery techniques are being used
increasingly to treat a range of diseases, including heart disease (for example,
drug-eluting stents), cancer (for example, local chemotherapies), and chronic
respiratory diseases (for example, therapeutic inhalants). In all those areas of
innovation, simulation-based methods could be used to model the transport of
drugs through the circulatory or respiratory systems and to determine the local
concentrations to use in pharmacokinetic models of drug metabolism.

Currently, important topics in cancer research are the mechanisms of cell

adhesion and invasion and signaling pathways. A better understanding of those
mechanisms is critical to advances in cancer research and neurobiology. The
developments of multiscale SBES technologies for investigations into cellular
structures and cell mechanics, as well as the development of novel cellular force-
measurement devices, will help explain dynamic cellular architectures and the
mechanism of cell motility.

Of particular significance is cell motility. To understand cell motility we

must first understand: (1) the mechanics of the cell interior, and (2) the
mechanics of the cell-substrate or cell-ECM (extra-cellular matrix) interaction
force. These complex mechanisms determine cell shape and migration, which in
turn allow the cell to perform its critical functions, such as wound healing and
embryonic morphogenesis. SBES technologies hold promise that we can improve
our understanding of those cellular functions and increase our ability to
differentiate normal cells from cancer cells. The stakes are high, because the
invasion of transformed cells into other tissues—the process called metastasis—
is believed to be the precursor of the development of cancer tissue.

12
2.2 SBES in Predictive Homeland Security

In the broadest sense, engineering design for security involves the

development of systems to protect human populations and the artificial and
natural infrastructures that support them. The systems protect us from a range of
threats, whether hostile (for example terrorists), environmental (for example, air
and water pollution), or natural (for example, earthquake or hurricane). The
protective systems must guard our entire support infrastructure: buildings;
transportation systems; food, water, and power distribution; communications; and
waste disposal.

The methods of SBES can play an important role in the design and
optimization of these protective systems. Most notably, SBES would allow our
emergency planners to predict not only the
consequences of threats (for example, the accidental SBES will allow the
or malicious release of a toxic chemical or prediction of the
biological agent), but also the effects of
consequences of
countermeasures. With the aid of predictive
simulations, engineers would be able to design and threats and
optimize infrastructures that would be impervious to countermeasure
a wide range of threats. With the ability to conduct
responses.
real-time simulations, moreover, an emergency team
would be able to identify the most rational response to a crisis. The World Trade
Center disaster of September 11, 2001 serves as a tragic example: if real-time
simulation had been available, the emergency response team would have realized
the importance of the immediate evacuation of the building complex.

SBES can give engineers and planners a remarkably large operational view
of the systems that make up our society. For example, SBES would give us the
ability to simulate the operation of a whole city as a single system. It is able to do
so because it integrates multiscale simulations of multiple subsystems and

13
processes, such as structural responses, fluid transport of contaminants, power
distribution, and transportation systems, as well as the response of the human
population.

This vision of a “Digital City” would require the acquisition of data of

unprecedented detail. Not only would the system have to acquire static data, that
is, data regarding the installed infrastructure, but it would also have to
continuously acquire dynamic data, that is, data undergoing constant change.
Dynamic data, for example, would include continuous measurements of air- and
water-contaminant concentrations; the flow rates of air, water, and effluents; the
locations and velocities of transportation and other movable assets (for example,
trains and heavy machinery); and the densities and movements of people and
automobiles.

A logical extension of the Digital-City concept is that of the Digital

Ecosystem, which may be artificial (such as a city) or natural (such as a forest,
watershed, continent, or even the entire planet). Whatever the scale, the benefits
are the same: we gain the ability to optimize human activity and infrastructures in
respect to adverse events or trends, and, through real-time simulation, we are able
to identify rational responses to crises.

The concepts and methods of SBES promise to revolutionize the practice of

urban planning, transportation, structural and environmental engineering, and
municipal and environmental management. To realize the visions of the Digital
City and the Digital Ecosystem, however, we must acknowledge that a great
amount of research must pioneer the way. The following are a few of the areas
requiring development:

• Quantitative models of the processes to be simulated must be developed. For

many of those processes, models of some level of fidelity already exist, or
they are being developed for narrower engineering purposes. Obvious
examples are structural models (of buildings and other structures), fluid-
dynamics models (air and water flows), combustion models (for example, for

14
 predicting the spread of fires), and transportation models (for example, to
analyze traffic flow). For other important processes, however, quantitative
models are rudimentary or nonexistent. For example, we lack sociological
models that can help us describe or predict the response of populations to
crises. In addition, we need better models for the evolution of natural
ecosystems such as forests or lakes.

• A comprehensive simulation system is required that integrates detailed

models of a wide range of scales. The comprehensiveness of the simulation
system is a requirement if SBES applications are to simulate multiscale
complex systems. Some of the issues are generic, but others are problem
specific.

• New models of exceptional fidelity are required. The development and

validation of such models entail the acquisition of data of extraordinary
detail. As a result, the development of the Digital City and Digital Ecosystem
will inevitably put pressure on the experimental sciences and theoretical
research to meet the demand for copious data. Furthermore, the real-time
simulation of some applications will drive developments in sensors and the
communication infrastructure, both of which must support streams of data. In
addition, we need to develop the simulation techniques that can
accommodate the data streams.

• A better understanding of the role of uncertainty is required. Some degree of

uncertainty is inevitable in the ability of a model to reflect reality and in the
data the model uses. We need to find ways to interpret uncertainty and to
characterize its effects on assessments of the probable outcomes.

The rewards of meeting those challenges are great: enhanced security,

safety, and convenience of life in the Digital City and the Digital Ecosystem; a
social infrastructure of unparalleled efficiency; rational responses to natural
events; and optimal interactions with natural environments. The following

15
summarizes some of the major applications.

• Protection Against Air Contaminants: SBES technologies will detect and

measure the presence of biological or chemical contaminants in the air and,
given detailed weather data, identify the likely release location and
magnitude of the release. The system will then design an optimal response
plan.
• Optimization of Infrastructures: SBES technologies will optimize the
designs of buildings and other infrastructural elements. Such designs would
be site specific, interact well with natural and man-made surroundings, and
blend with the urban system of which they are a part. In addition, the designs
would take into account the effects of the structures both on normal
operations and on operations during a wide range of large-scale emergencies.
• Prediction of Long-Term Environmental Impacts: SBES technologies
will predict the effects of effluents from existing and proposed facilities on
urban and natural environments. Such predictions would greatly increase the
reliability and usefulness of environmental impact studies. In addition, they
would allow planners to minimize the probability of unforeseen deleterious
events.
• Optimization of Emergency Responses: SBES technologies would
optimize emergency responses to fire and explosion (whether accidental or
intentional). Planning for emergency responses would consider how the
situation might evolve or escalate (for example, a fire might spread, building
collapse, or the response intervention itself might adversely affect the
situation).
• Optimization of Security Infrastructures for Urban Environments:
SBES technologies would assist in the design and placement of air- and
water-contaminant sensors and would help in the planning of
countermeasures, such as contaminant dispersal and flood abatement.

16
2.3 SBES in Energy and the Environment

The energy-related industries rely on modern simulation methods to monitor

the production of oil reservoirs, plan pollution remediation measures, and devise
control strategies. Recent advances in simulation-related technologies may raise
oil-reservoir management to a new level of sophistication. Those advances
include distributed computing, multi-physics and
chemistry modeling, parallel algorithms, and methods An instrumented
and devices for the dynamic use of well-bore and oilfield will result in
seismic data. In addition, the next generation of oilfield
more efficient, cost-
simulation tools could exploit developing technologies
for data-driven, interactive, and dynamically adaptive effective, and
strategies for subsurface characterization and reservoir environmentally
management. Soon we could see multi-resolution
safer reservoirs,
reservoir models that can be executed on very large,
with enormous
distributed, heterogeneous, computational environments.
Those models, moreover, could be fed data from sensors strategic and
embedded in reservoir fields (for example, permanent economic benefits.
downhole sensors and seismic sensors anchored on the
seafloor). Such a model-and-sensors system could provide a symbiotic feedback
loop between measurement data and computational models. This approach could
lead to an instrumented oilfield, one that is more efficient, cost effective, and
environmentally safe. The strategic and economic benefits are enormous:

• An increase in the volume of oil and natural gas produced from existing
reservoirs. With our better understanding of existing oil and gas reservoirs,
we can expect to deplete existing reserves more efficiently and to locate and
produce bypassed reserves. The additional production could help us reduce
our dependence on foreign oil.

• A better understanding of risks and uncertainties in exploration and lower

17
 finding costs. Better models of the subsurface would allow oil and gas
companies to focus on prospects that offer the best return. As a result, they
can allocate their capital much more efficiently.

The new SBES-related technologies have immediate application to other

areas as well, including environmental remediation and storage of hazardous
wastes. Again, these new application areas require an integrated and interactive
simulation framework with multiscale capabilities. The development and use of
such frameworks require the support of cross-disciplinary teams of researchers,
including geoscientists, engineers, applied mathematicians, and computer
scientists.

2.4 SBES in Materials

SBES-related technology may have its greatest societal impact where

innovations in modeling and simulation methodologies intersect with innovations
in materials. Multiscale modeling and simulation are transforming the science
and technology of new-material development and the improvement of existing
materials. This transformation is tantamount to
With SBES, materials a shift to a powerful new paradigm of

development becomes a engineering science. The new methods enable

the unprecedented ability to manipulate
unique opportunity for the
metallic, ceramic, semiconductor,
integration of funda- supramolecular, and polymeric materials. The
mental, interdisciplinary results are material structures and devices that
have remarkable physical, chemical,
knowledge, with techno-
electronic, optical, and magnetic properties.
logical applications of We can now anticipate the molecular design of
obvious benefit to society.

18
composite materials with undreamed-of functionalities. Moreover, to reap the
advantages that SBES technology brings to materials development, researchers
from many disciplines would have to integrate their knowledge in the materials
sciences. Such collaboration maximizes the possibilities for developing materials
of great technological value.

The principle of materials design is rooted in the correlation of molecular

structure with physical properties. From those correlations, models can be
formulated that predict microstructural evolutions. Such models allow the
researcher to investigate the mechanisms underlying the critical behaviors of
materials and to systematically arrive at improved designs.

The use of simulations to uncover structure-property correlations can be

superior to relying only on experimental data. The reason is that simulation
provides detailed information regarding the evolved microstructure, as well as
complete control over the initial and boundary conditions. Another significant
aspect of SBES, one that makes future materials development more robust, is that
it links simulation methods across different length and time scales. A great deal
of progress is being made in the first principles calculations of electronic
structure and in atomistic simulations. Now progress is also being made in
connecting these two powerful techniques of probing physical phenomena in
materials.

The benefits from new materials development are amply evident in the
current progress in nanoscience and technology, a world-wide enterprise that can
be compared to drug design. Because of the multiscale nature of materials
modeling and simulation, SBES is destined to play a key role in nanoscience.
SBES provides the capability of linking electronic-structure methods, which are
necessary for dealing with novel nanostructures and functional properties, with
atomistic and mesoscale techniques. That linkage ensures that the different
phases of materials innovation—from design to testing to performance and
lifetime evaluation—can all be simulated, examined, and optimized.

19
 The power of multiscale computation can be seen in a number of high-profile
applications involving the behavior of known materials in extreme environments.
For example, a problem that has occupied the attention of a sizable community of
researchers for several years is the characterization of the mechanical behavior of
plastic deformation in metals at high pressure and high strain rate. The challenge,
which is relevant to national security, is to conduct multiscale simulation that
links all of the following: calculation of the core of a dislocation using electronic-
structure methods; the modeling of dislocation mobility using molecular
dynamics simulation; and the determination of constitutive relations for
continuum-level codes. Multiscale simulation can also help solve problems in the
development of the structural components of nuclear power reactors. Such
materials must not only be radiation resistant, but they must have lifetimes of
over 40 years.

Even for materials that do not have to stand up to the extreme conditions of
high pressure and intense radiation, the field of materials innovation is rich with
challenges to our understanding of the underlying microstructures of the
materials. By meeting those challenges, we can reap enormous benefits. For
example, we could generate a molecular model of
Everywhere one cement, the most widely used substance made by
looks there are humans. Such a model would help us develop a new
cement with greater creep resistance and environmental
problems important
durability. Similarly, models would help us improve the
to society that
performance of catalysts for fuel-cell electric vehicles.
require optimizing We could also improve techniques in oilfield
the functional exploration, where instrumentation and digital
management of hydrocarbon reservoirs are issues. In all
properties of
these examples, improvements in materials performance
materials through would have great impact.
control of their Everywhere one looks there are problems important
microstructure.

20
to society that require optimizing the functional properties of materials through
the control of their microstructures. Clearly, SBES will have a long-term impact
on materials innovation. Three attributes of SBES in particular lead to this
conclusion.

• Exceptional Bandwidth: The conceptual basis of materials modeling and

simulation encompasses all of the physical sciences. It makes no distinction
between what belongs to physics versus chemistry versus engineering and so
on. This universality of SBES technology represents a scientific bandwidth
that is at least as broad as the entire range of multiscale applications in
science and engineering. In materials modeling and simulation, as in SBES
more generally, traditional disciplinary barriers vanish; all that matters is “the
need to know.”

• Elimination of Empiricism: A virtue of multiscale modeling is that the

results from both modeling and simulation are conceptually and operationally
quantifiable. Consequently, empirical assumptions can be systematically
replaced by physically-based descriptions. Quantifiability allows researchers
to scrutinize and upgrade any portion of a model and simulation in a
controlled manner. They can thus probe a complex phenomenon detail by
detail.

• Visualization of Phenomena: The numerical outputs from a simulation are

generally data on the degrees of freedom characterizing the model. The
availability of this kind of data lends itself not only to direct animation, but
also to the visualization of the properties under analysis, properties that
would not be accessible to experimental observation. In microscopy, for
example, researchers can obtain structural information but usually without
the energetics. Through simulation, however, they can have both. The same
may be said of data on deformation mechanisms and reaction pathways.

These three attributes of SBES, of course, are not restricted to materials

21
 development; they apply equally well to the other areas of SBES application. In
this section, however, the focus has been on the application of SBES to materials
development. The point that emerges is that, aided by SBES technology,
materials modeling and simulation, or computational materials, is becoming the
sister science of computational physics and computational chemistry.

2.5 SBES in Industrial and Defense Applications

Simulation is ubiquitous in industry. It plays an essential role in the design of

materials, manufacturing processes, and products. Increasingly, simulation is
replacing physical tests to ensure product

To increase U.S. reliability and quality. Fewer tests mean fewer

prototypes, and the result is a shorter design
competitiveness, short
cycle. Steady reductions in design cycles, in
design cycles are crucial if turn, are crucial to U.S. efforts to remain
we are to keep up with the competitive in a world where the pace at which
new consumer products are being developed is
rapid pace of new-
increasing every day. The need for shorter
products throughout the design cycles also applies to our national
world. defense and security. World events are often
unpredictable; our defense industry must be
able to design, modify, and manufacture equipment in quick response to military
and police exigencies. A case in point is the unanticipated need to reinforce
armored vehicles in Iraq after several such vehicles were destroyed by
improvised explosive devices.

The use of simulation has proved effective in some industrial applications.

For example, in crashworthiness studies the few instances that simulations

22
replaced testing are frequently cited as success
stories. Generally, however, simulation has yet to
play a central role in important industrial and
defense design applications. The reason is that
large-scale simulation does not enter into the
design cycle until its later stages. Model
preparation, after all, requires a substantial amount
of time and labor. Often it takes months to prepare
a model, and even then the model needs to be
calibrated with tests if the design is substantially
different from previous designs.

In addition, the preparation of a model usually

requires considerable knowledge of and skill in
finite-element analysis. For that reason, the
challenge of generating models normally falls on
engineers with advanced degrees, not designers.
This separation between simulation and design
activities results in delays and reduces the
effectiveness of the simulation.
Figure 3: An example of
The difficulties of simulation are compounded
contemporary visualization and
when new materials, such as composites and simulation capabilities. Shown is a
simulation of the explosion of a
metallics, are used in product designs. Before
steel tank filled with explosive
simulation can even begin, the new materials must subjected to a jet of heated air.
New methods are needed to
be tested extensively to determine their properties.
capture multiscale effects and
Such tests are time consuming, and they lengthen quantify uncertainties.
the time necessary to prepare and conduct useful
simulations.

In addition, simulation capabilities are currently limited in their ability to

model relevant phenomena. For example, in the simulation of a rear collision

23
 between two automobiles, it would be desirable to model the gasoline in the fuel
tank and the effects of any fracture of the fuel tank. To date, such multi-physics
simulations are not possible. Similarly, we are still unable to model the fracture
of the interior panels and trim, which is important in determining occupant
injury, or the fracture of sheet metal and structural members, which is common
with aluminum.

Thus, crashworthiness simulation technology

If an industry is to is usually used to check designs near their final
replace testing with stages. Optimization of a design for
crashworthiness in the early stages of the design
simulation, the
process is simply not yet feasible. The complexity
simulation tools must
of model generation and the uncertainties are too
undergo robust great. Even if an industry were to replace testing
verification and almost completely with simulation, the changeover
would still require robust verification and
validation procedures
validation procedures to ensure the effectiveness of
for effectiveness. the simulation tools.

Similar limitations are found in many other industrial applications where, at

least superficially, simulation has appeared to be a success. For example, in the
simulation of sheet-metal forming, important phenomena, such as the tribology
of the metal-die interface, are not modeled. Those phenomena are characterized
by much smaller scales than the overall process, and they are subject to great
variability because they depend on factors such as temperature and the age of the
lubricant. In the tire industry, simulation is usually limited to determining the tire
footprint; important performance characteristics such as pothole impact,
hydroplaning, and cornering performance cannot be simulated because they
involve multi-physics and many disparate length and time scales.

In the chemical processing industry, simulation would seem to have a

significant role. In fact, most petrochemical plants use steady-state process

24
simulation models to perform detailed real-time optimization. As a result,
chemical plants are more energy and environmentally efficient. Even in the
chemical industries, however, the use of simulation is limited. Current planning
models capture only nominal plant capacities; they do not address the real
performance of the plants and the processes at various scales, nor are they
sensitive to uncertainties.

Over the past two decades the integrated circuit industry has been a major
player in simulation-based engineering. Until now, the U.S. has been a leader in
the development of highly integrated, easy-to-use software for circuit analysis,
such as SPICE, and in the education of our engineering workforce in the
application of that software. As a result, the U.S. was able to maintain its
leadership in the high value-added segment of the market. As clock rates move
into the gigahertz range, however, circuit theory is no longer applicable. Future -
generation transistors, such as single electron transistors, low-threshold
transistors, and quantum computing devices, will be based on new physics that
links quantum mechanics and electromagnetics.

Overall, simulation in industry has yet to meet its full potential. The
following list is a summary of its current limitations:

1. The development of models is very time consuming, particularly for

geometries of complex engineering systems such as ships, automobiles, and
aircraft. Moreover, the determination of material properties often requires
extensive small-scale testing before simulation can be started, especially if
statistical properties are needed. This testing also lengthens the time to obtain
a simulation and hence the design cycle.

2. Methods are needed for linking models at various scales and simulating
multi-physics phenomena.

3. Simulation is often separate from the design optimization process and cannot
simultaneously deal with factors such as manufacturability, cost, and
environmental impact.

25
 Overcoming these barriers will require progress in our basic understanding
and in the development of powerful new methods. Among these challenges are
the following:

1. Multiscale methods that can deal with large ranges of time and spatial scales
and link various types of physics.

2. Methods for computing macroscopic phenomena, such as material properties

and manufacturing processes, in terms of subscale behavior.

3. Effective optimization methods that can deal with complex integrated

systems, account for uncertainties, and provide robust designs.

4. Frameworks for validation, verification, and uncertainty qualification.

5. Methods for rapidly generating models of complex geometries and material

properties.

Multiscale methods will provide extensive benefits. For instance, they will
enable us to understand relationships and interactions of phenomena at different
scales, which is crucial in the design of many products. In the design of products,
it is often necessary to couple diverse physics, such as fluids with solids or
electromagnetics with structures. Simulations of such couplings involve a large
range of length and time scales. For example, to model live fire testing of defense
products, it is necessary to incorporate phenomena of an immense range of scales
associated with combustion and structures. Similarly, designs involving different
components also have an enormous range of coupled time scales that determine
overall system behavior.

Multiscale methods may also make possible the prediction of material

properties in terms of basic building blocks, ranging from matrix and fiber
properties of composite materials to the properties of atoms in metals. Demand is
increasing for materials that reduce weight and cost; consequently, the
availability of tools that, through simulation, can predict the behavior of a
material from its basic building blocks would open tremendous possibilities for

26
quickly developing better, less costly, and safer products. Such tools would
eliminate the bottleneck of extensive materials testing, resulting in substantial
reductions in design-cycle times. The methods we envision would be able to link
models of different scales, such as models of micromechanics or even quantum
mechanics linked to models of macroscale behavior.

Of course, as design processes increasingly rely on computer simulation,

validation and verification procedures will become increasingly important.
Although some efforts have been made at providing validation benchmark
problems for linear analysis, nonlinear simulation software has not been
subjected to extensive validation procedures. In fact, we find considerable
controversy as to what appropriate validation procedures are, how broadly they
apply, and whether they are even feasible. Clearly, a basic understanding of
verification and validation procedures is urgently needed. After all, to be useful,
the simulation tools used by industry and defense agencies must provide reliable
results. Furthermore, since many real-world phenomena
are not deterministic, statistical methods that can
SBES has the
quantify uncertainty will be needed.
potential to deliver,
Design optimization is also in its infancy, and it too
within a short design
has many obstacles to overcome. The constraints on the
optimization of a product design relate to period, designs that
manufacturability, robustness, and a variety of other are optimized for
factors. Optimality often needs to be defined in terms of
cost performance
complex criteria, and the frameworks currently in use
are not readily amenable to that task. Moreover, to be
and total impact on
effective for engineering design, optimization methods the environment.
must be closely coupled with simulation techniques.
Generally, however, we still lack a fundamental understanding of what
constitutes an optimal design and how to find it in a complex multi-criteria
design environment. Once optimization methods are developed that can deal with
these complexities, we can expect to see chemical plants, automobiles, laptop

27
computers, and a host of other industrial and consumer products that feature
unprecedented efficiency at lower cost.

In summary, SBES has the potential to deliver designs that are optimized for
cost performance and their total impact on the environment (from production to
disposal or recycling), all within a short design cycle. This achievement is not
possible, however, simply by extending current research methods and taking
small, incremental steps in SBES development. The barriers to the realization of
SBES relate to our entire way of conducting research and development and
educating engineers. The next chapter discusses some of these core issues.

Finding
Because of the interdisciplinary character and complexity of SBES
challenges, incremental, short-term research efforts are inadequate to achieving
SBES goals. Instead, a long-term program of high-risk research will be needed to
resolve the numerous obstacles standing in the way of SBES developments. The
Panel agrees with the observation made in the PITAC report and elsewhere that
short-term investments and limited strategic planning will lead to an excessive
focus on incremental research rather than on the long-range, sustained research
necessary to have a lasting impact. Moreover, progress in such research will
require the creation of interdisciplinary teams that work together on leading-edge
simulation problems. The work of those teams should be sustained for a decade
or more for the investment to yield its full fruits.

28
3.0 CORE ISSUES: Challenges,
Barriers, and Opportunities in
SBES Research

All of the driving applications discussed in the preceding chapter share

common challenges, barriers, and requirements for research breakthroughs. We
elaborate on the major issues in this chapter.

3.1 The Tyranny of Scales: The Challenge of

Multiscale Modeling and Simulation

Researchers in the worldwide race toward miniaturization, nanoscience,

molecular modeling of drugs and biological systems, advanced materials, and
other applications, all of which involve events on atomistic and molecular levels,
have run into a formidable roadblock: the tyranny of scales. Virtually all
simulation methods known at the beginning of the twenty-first century were valid
only for limited ranges of spatial and temporal scales. Those conventional
methods, however, cannot cope with physical phenomena operating across large
ranges of scale—12 orders of magnitude in time scales, such as in the modeling
of protein folding [4, p. 4], or 10 orders of magnitude in spatial scales, such as in
the design of advanced materials. At those ranges, the power of the tyranny of
scales renders useless virtually all conventional methods. Confounding matters
further, the principal physics governing events often changes with scale, so that
the models themselves must change in structure as the ramifications of events

29
 pass from one scale to another.

The tyranny of scales dominates simulation efforts not just at the atomistic or
molecular levels, but wherever large disparities in spatial and temporal scales are
encountered. Such disparities appear in virtually all areas of modern science and
engineering, for example, in astrophysics, atmospheric science, geological
sciences, and in the design of complex engineering systems such as submarines,
commercial aircraft, and turbine engines.

In many ways, all that we know about the physical universe and about the
design and functioning of engineering systems has been partitioned according to
categories of scale. The designer manipulating the
electronic properties of materials sees the world as a
The development of
myriad of infinitesimal atoms with clouds of orbiting
effective multiscale
electrons. The atmospheric scientist sees the world as
modeling techniques the movement of great air masses that change
will require major climate conditions across thousands of miles of the
earth’s surface. Today, we are attempting
breakthroughs in
technological advances that cannot tolerate any view
computational of nature that partitions phenomena into neat
mathematics and new categories of scale. The modeling and simulation

thinking on how to tools we are seeking now must be commensurate in

their applications with the great breadth of the
model natural events
phenomena they must simulate.
occurring at multiple
The tyranny of scales will not be defeated simply
scales. by building bigger and faster computers. Instead, we
will have to revamp the fundamental ways we
conceive of scientific and engineering methodologies, long the mainstays of
human progress. Such a daunting challenge, historic in its significance, is beyond
the capability of single individuals and disciplines. The necessary breakthroughs
in computational mathematics and the development of new ways to model

30
natural events at multiple scales will require the efforts of interdisciplinary teams
of researchers and thinkers working in concert.

We can see instances already where preliminary and often ad hoc

methodologies have retarded technological progress. The design of nanodevices
is one example. Nanodevices are systems with tiny masses and relatively large
surface areas. The design of such devices is in urgent need of new simulation
tools, because they are at a scale too small to be captured by continuum
mechanics. Another example is any biological application that requires methods
that link diverse time scales. For instance, the sequence of events following a
medical implant is initiated by the interactions between individual water
molecules and the surface of the implant. This first set of interactions occurs on a
timescale of nanoseconds. The resulting water “shell,” in turn, has an influence
on proteins and other types of molecules that arrive later. This second set of
interactions has a timescale from microseconds to milliseconds. Thus, there is a
significant time gap in the behavior that is difficult to model. Similarly, temporal
gaps in the behavior of polymers cause problems for current simulation methods.

Numerous important applications of nanotechnology are being driven by

homeland security. As a result, the need and urgency for developing multiscale
tools has increased significantly. For example, we know that small concentrations
of chemical and biological agents can have lethal effects on large segments of the
human population. The recognition of that threat has prompted development of
new mitigation methods, such as miniaturized intelligent sensors, protective
clothing, and masks. The need for engineered nanostructured materials for
homeland-security applications, as well as for optical and structural applications,
has spurred much interest in the development of multiscale methods that can
accommodate diversity in spatial scales.

Recent work in multiscale modeling has emphasized the synthesis of theories

applicable to different scale ranges, such as quantum, molecular, and continuum
descriptions [5, p. 1504]. Nevertheless, enormously important technological

31
problems, such as turbulence modeling, remain unsolved. These problems
involve a very broad range of scales amenable to a single description, such as
continuum theory in the case of turbulence. In fact, turbulent-flow problems in
practical engineering involve such an enormous range of scales that they cannot
be currently solved on the world’s largest and fastest computers. If we assume
that progress continues at the rate of Moore’s Law, the turbulence-flow problems
will not succumb to solutions for many generations to come. The implications of
solving these problems are great; for example, they have to do with our
leadership in designing future generations of commercial and military aircraft.
Before we can lead, however, we must find the path to fundamental
developments in multiscale modeling.

The urgency of the development of multiscale simulation models has been

felt worldwide. Over the past five years, virtually every conference, symposium,
and international congress devoted to computational engineering and science has
listed multiscale modeling as an important theme. Multiscale modeling has often
been the subject of colloquia, study groups, or invited lectures. Three DOE
workshops were recently held on multiscale mathematics [8, 23, 10], and
programs at NSF and NIH are already in place for promoting the beginnings of
new research in this area. In recent years, a large and growing body of literature
in physics, chemistry, biology, and engineering has focused on various methods
to fit together simulation models of two or more scales, and this has led to the
development of various multi-level modeling approaches. To date, however,
progress on multiscale modeling has been agonizingly slow. Only a series of
major breakthroughs will help us establish a general mathematical and
computational framework for handling multiscale events and reveal to us the
commonalities and limitations of existing methods.

32
Finding
Formidable obstacles remain in linking highly disparate length and time
scales and in bringing together the disciplines involved in researching simulation
methods. These issues are common to many SBES applications. Fundamental
discoveries will be needed to surmount these obstacles.

3.2 Verification, Validation, and

Uncertainty Quantification

The ultimate goal of simulation is to predict physical events or the behaviors

of engineered systems. Predictions are the basis of engineering decisions, they
are the determining factor in product or system design, they are a basis for
scientific discovery, and they are the principal reason that computational science
can project itself beyond the realm of physical experiments and observations. It is
therefore natural to ask whether specific decisions can
rely on the predicted outcomes of an event. How What level of
accurate are the predictions of a computer simulation?
confidence can one
What level of confidence can one assign a predicted
assign a predicted
outcome in light of what may be known about the
physical system and the model used to describe it? outcome in light of
The science, technology, and, in many ways, the what may be known
philosophy of determining and quantifying the
about the physical
reliability of computer simulations and their
predictions has come to be known as V&V, or
system and the model
verification and validation. The methods of V&V are used to describe it?
fundamental to the success and advancement of SBES.

To appreciate the subtleties and goals of V&V, one must first dissect the

33
 process of simulation. Beginning with the conceptual understanding of certain
physical events of interest and with scientific theories that explain them (the
target physical phenomena or engineering system identified for study), the
analyst (the modeler, scientist or engineer) constructs a mathematical model of
the event. The mathematical model is a collection of mathematical constructions,
equations, inequalities, constraints, etc., that represent abstractions of the reality,
and are dictated by the theory or theories characterizing the events. The analyst
then develops a computational model of the event. The computational model is a
discretized approximation of the mathematical model, and its purpose is to
implement the analysis on a computer. Validation is the subjective process that
determines the accuracy with which the mathematical model depicts the actual
physical event. Verification is the process that determines the accuracy with
which the computational model represents the mathematical model. In simple
terms, validation asks, “Are the right equations solved?” while verification asks,
“Are the equations solved correctly?”

The entire field of V&V is in the early stage of development. Basic

definitions and principles have been the subject of much debate in recent years,
and many aspects of the V&V remain in the gray area between the philosophy of
science, subjective decision theory, and hard mathematics and physics. The
twentieth century philosopher of science Karl Popper asserted that a scientific
theory could not be validated; it could only be
The most invalidated. Inasmuch as the mathematical model of
confounding aspect of a physical event is an expression of a theory, such
models can never actually be validated in the strictest
V&V has to do with
sense; they can only be invalidated. To some degree,
uncertainty in the data therefore, all validation processes rely on prescribed
characterizing acceptance criteria and metrics. Accordingly, the

mathematical models analyst judges whether the model is invalid in light

of physical observations, experiments, and criteria
of nature.

34
based on experience and judgment.

Verification processes, on the other hand, are mathematical and

computational enterprises. They involve software engineering protocols, bug
detection and control, scientific programming methods, and, importantly, a
posteriori error estimation.

Ultimately, the most confounding aspect of V&V has to do with uncertainty

in the data characterizing mathematical models of nature. In some cases,
parameters defining models are determined through laboratory tests, field
measurements, or observations, but the measured values of those parameters
always vary from one sample to another or from one observation to the next.
Moreover, the experimental devices used to obtain the data can introduce their
own errors because of uncontrollable factors, so-called noise, or errors in
calibration. For some phenomena, little quantitative information is known, or our
knowledge of the governing physical processes is incomplete or inaccurate. In
those cases; we simply do not have the necessary data needed to complete the
definition of the model.

Uncertainty may thus be due to variability in data due to immeasurable or

unknown factors, such as our incomplete knowledge of the underlying physics or
due to the inherent nature of all models as incomplete characterizations of nature.
These are called subjective uncertainties. Some argue that since the data itself
can never be quantified with absolute certainty, all
uncertainties are subjective. Whatever the source of
The use of stochastic
uncertainty, techniques must be developed to
quantify it and to incorporate it into the methods and
models can represent
interpretation of simulation predictions. gigantic increases in
Although uncertainty-quantification methods complexity in data
have been studied to some degree for half a century, volume, storage,
their use in large-scale simulations has barely
manipulation, and
begun. Because model parameters can often be
retrieval requirements.

35
treated as random fields, probabilistic formulations provide one approach to
quantifying uncertainty when ample statistical information is available. The use
of stochastic models, on the other hand, can result in gigantic increases in the
complexity of data volume, storage, manipulation, and retrieval requirements.
Other approaches that have been proposed for uncertainty quantification include
stochastic perturbation methods, fuzzy sets, Bayesian statistics, information-gap
theory, and decision theory. The development of reliable methodologies—
algorithms, data acquisition and management procedures, software, and theory—
for quantifying uncertainty in computer predictions stands as one of the most
important and daunting challenges in advancing SBES.

Finding
While verification and validation and uncertainty quantification have been
subjects of concern for many years, their further development will have a
profound impact on the reliability and utility of simulation methods in the future.
New theory and methods are needed for handling stochastic models and for
developing meaningful and efficient approaches to the quantification of
uncertainties. As they stand now, verification, validation, and uncertainty
quantification are challenging and necessary research areas that must be actively
pursued.

36
 3.3 Dynamic Simulation Systems, Sensors,
Measurements, and Heterogeneous Simulations

One of the most challenging applications of SBES, but one that may yield the
greatest dividends, is the linkage of simulation tools directly to measurement
devices for real-time control of simulations and computer predictions. Some
preliminary investments in the research into this new idea have been made under
NSF’s program in dynamic data-driven applications systems (DDDAS) [7, 9].
The full development of this revolutionary and fundamentally important
discipline will take years of research and technological development.

The concept of DDDAS is envisioned as a new paradigm in computer

simulation, one involving a “symbiotic feedback control system” [7, 9] in which
simulations and experiments (or field data) interact in real time to dramatically
improve the fidelity of the simulation tool, its accuracy, and its reliability.

The document that originally put forth the idea [7], now over five years old,
described the goal of DDDAS as one of developing “application simulations that
can dynamically accept and respond to ‘online’ field data and measurements
and/or control such measurements. This
synergistic and symbiotic feedback control loop
among applications, simulations, and
This synergistic and
measurements is a novel technical direction that symbiotic feedback
can open new domains in the capabilities of control loop among
simulations with a high potential payoff, and
applications, simulations,
create applications with new and enhanced
capabilities. It has the potential to transform the and measurements has
way science and engineering are done, and the potential to transform
induces a major beneficial impact in the way
the way science and
many functions in our society are conducted, such
engineering are done.

37
 as manufacturing, commerce, transportation, hazard prediction/management, and
medicine, to name a few.”

A half-decade later, these words are still true, but we also better appreciate
the size of the challenge. To develop DDDAS, we must resolve issues involving
the complexity of the systems, the breadth of expertise and technologies required
to implement the systems, the new software infrastructures, and the efficiency
and capacity of the computational and data management systems required.
Success on all those technological fronts will mandate a sustained and well-
funded program of basic and applied research over possibly a decade or more.
The payoffs, however, are immense—so important, in fact, that the highest
priority should be given to developing and exploiting this fundamental SBES
discipline. DDDAS is a concept conceived, defined, and promoted in the United
States. To capitalize on our own initiatives, however, we must become aggressive
in our development of DDDAS. Our current complacency in this technology is
allowing our competitors to gain on us, once again.

The basic building blocks of DDDAS include the following: 1) a hierarchy of

heterogeneous simulation models, 2) a system to gather data from archival and
dynamic sources, 3) algorithms to analyze and predict system behavior by
blending simulation models and data, 4) algorithms to steer and control the data
gathering and model validation processes, and 5) the software infrastructure
supporting model execution, data gathering, analysis prediction, and control
algorithms.

In many ways, DDDAS will rewrite the

book on the validation and verification of
Dynamic data-driven
computer predictions. No longer will
application systems will
validation be a one-shot operation to judge the
rewrite the book on the acceptability of a simulation problem on the
validation and verification basis of a static data set. In DDDAS, validation
becomes a part of the dynamic control process
of computer predictions.

38
that identifies and assesses deficiencies of the computational model and upgrades
and improves the model on the fly. This incorporation of validation into the
dynamics of the model dramatically enriches the predictability of the model and
increases confidence in the predicted results.

DDDAS dynamically incorporates measurement data into a simulation as

that simulation is executing. The simulation, in turn, dynamically steers the
measurement process. To perform these operations, DDDAS integrates large-
scale numerical computing with data-intensive computing, sensors, imaging, grid
computing, and other measurement devices. The development of this technology
requires new concepts in software infrastructure, algorithms, control protocols,
and solvers. Once in place, however, the technology offers an endless list of
applications. Surgical procedures, homeland security, control of hazardous
materials, environmental remediation, drug delivery, manufacturing processes,
oil reservoir management, and vehicle flight control are just a few.

Finding
Research is needed to effectively use and integrate data-intensive computing
systems, ubiquitous sensors and high-resolution detectors, imaging devices, and
other data-gathering storage and distribution devices, and to develop
methodologies and theoretical frameworks for their integration into simulation
systems. Concomitant investments are also required in sensory-data computing,
the collection and use of experimental data, and the facilitation of interactions
between computational models and methods, all of which are necessary to
achieve dynamic adaptive control of the computational process.

39
 3.4 New Vistas in Simulation Software

Many contemporary engineering communities regard simulation software as

a commodity that vendors provide for well-defined, specific, and independent
domains of application. Occasionally, these long-lived codes for engineering
analysis receive incremental improvements, usually in the form of functional
extensions. This leisurely approach to software development will not support the
next generation of engineering problems—multiscaling with real-time data
interaction and abundant uncertainties in the data. As the PITAC report states
[14], “Today it is altogether too difficult to develop computational science
software and applications. Environments and toolkits are inadequate to meet the
needs of software developers in addressing increasingly complex
interdisciplinary problems. Legacy software remains a persistent problem
because the lifetime of a computational science application is significantly
greater than the three- to five-year life cycle of a computing system. In addition,
since there is no consistency in software engineering best practices, many of the
new applications are not robust and cannot easily be ported to new hardware.”

For those reasons, entirely new approaches are

needed for the development of the software that will
Entirely new
encapsulate the models and methods used in SBES.
approaches are Researchers must identify the methodologies that support
needed for the the interoperability of individual components of

development of the simulation software. Then they must develop those

methodologies and integrate them into the next
software that will
generation of engineering software. This search for new
encapsulate the methods and tools to support simulation software
models and development is fraught with difficulties. Not only do the
new simulation components require complex algorithms,
methods used in
they must also function efficiently on an evolving range
SBES.

40
of architectures designed for large-scale parallel computations.

Tomorrow’s SBES software requires extraordinary degrees of robustness,

efficiency, and flexibility. The new software must not only execute simulation
algorithms, but must also dynamically manage data throughput and model
adaptivity and control. It must steer observational and measurement systems to
optimize data collection and use. It must navigate efficiently across models of
multiple scales and accommodate multiple physical theories, and it must have
scalable methods that interact seamlessly with data-gathering devices.

Much of our contemporary software development tools—libraries (for

instance, linear equation solvers), language interoperability tools, component
coupling and data transfer tools, and simulation development frameworks—do
not meet the demands of SBES. To define the real requirements for the
implementation of SBES technologies, we require a new paradigm of software
development. Such a fundamental change calls for a great deal of “out-of-the-
box” thinking about the way we approach software
development and practice engineering. The change Not only will tomorrow’s
will even affect the way scientific computing is
software developers
taught and perceived in our universities. Not only
will tomorrow’s software developers have to cope have to cope with more
with more complex systems and heterogeneous complex algorithms,
hardware systems, but they will also have to
but they will also have
understand the important details of the applications.
to understand the
Whereas the future of SBES software is largely
important details of the
uncharted, some path-finding work is under way. A
Federal government group, for instance, has taken a applications.
similarly aggressive software philosophy and
developed software with which to bootstrap. This effort is called the Scientific
Discovery Through Advanced Computing (SciDAC) initiative [26] (see also
Chapter 2 of [18]). Organized in 2001, SciDAC is a highly interdisciplinary

41
 program tasked with finding methods by which state-of-the-art mathematics and
domain-specific application sciences can be embodied in robust codes that run
efficiently on current terascale supercomputers. The initiative is driven by
research into science applications relevant to the goals of DOE’s Office of
Science. Those applications include fields such as global climate modeling,
plasma fusion, quantum chromodynamics, accelerators, combustion, and
supernovae, as well as the mathematics and computer science relevant to those
disciplines. Most of this research portfolio involves the coordinated efforts of
dedicated domain scientists in collaboration with mathematicians, computer
scientists, and computational scientists throughout the nation. One of the
principal products of SciDAC is sharable software, and development of such
software takes the sustained effort of a permanent staff. For this reason, the
center of gravity for most of the work in this field is at national labs. In addition,
U.S. universities continue to play a key role in that research. At any rate, it is
clear that the questions are too large and complex for any single institution to
manage alone. Instead, we must encourage the formation of teams of researchers,
working in a collaborative software environment, where they can profit from
distributed resources and expertise.

The DOE’s National Nuclear Security Agency is also heavily involved in

computational science. In the mid-1990s, that agency embarked on an ambitious
program of Stockpile Stewardship. Part of the

The new paradigm for program was the Accelerated Strategic Computing
Initiative (ASCI, now ASC). With ten times the
SBES software
funding of SciDAC, the ASCI program, among other
research and goals, seeks to use simulation to manage our nuclear
development will stockpile. Simulation, in this case, would be a

allow for substitution for our actual testing of the devices,

which is prohibited by international treaty. Five
specialization with
centers were established at U.S. universities to
cross-accountability.

42
develop the computational science for ASCI.

As both DOE programs exemplify, if the engineering sciences are to realize

the full benefits of the rapid advances in computing technologies, we must
somehow integrate the knowledge and discoveries of mathematics, computer
science, engineering, and the domain sciences. We also need to recognize that
SBES is located at the intersection of those disciplines. In that sense, we can
think of SBES as a super-discipline.

The new paradigm for SBES software research and development will allow
for specialization with cross-accountability. As envisioned, mission-driven
teams, primarily practicing engineers, will define and model the engineering
systems that require breakthroughs in simulation methods (for example, artificial
organs or distribution networks). In addition, the engineers will tentatively
identify data interfaces and the computational tasks between those interfaces.
From there, teams of “enabling technologists,” primarily mathematicians and
computer scientists, will tackle the abstract requirements identified by the
mission-driven teams and develop software components that port across the
target architectures (for example, massively parallel distributed memory
computers). The component development will track the research frontier for each
algorithmic area of expertise (for example, error estimation or eigenanalysis)
while also conforming to mission requirements.

Finding
Much of our current software in computational engineering science is inadequate for
dealing with the multifaceted applications and challenges of SBES. New software tools,
paradigms, and protocols will need to be developed so that software is more transferable
between fields and not wastefully duplicated. In the multidisciplinary teams we establish
for SBES research, we must incorporate experienced software developers who will work
closely with engineering scientists to develop tomorrow’s SBES software.

43
 3.5 The Emergence of Big Data in Simulation and
the Role of Visualization in SBES

Since the advent of computing, the world has experienced an information

“big bang,” an explosion of data. Information is being created at an exponential
rate. Since 2003, digital information makes up
The era in which data- 90 percent of all information production,
vastly exceeding the amount of paper and film.
intensive computing and
One of the greatest scientific and engineering
large-scale scientific challenges of the twenty-first century is to
computing were understand and make effective use of this

essentially disjoint camps growing body of information.

is over. In the computational engineering and

science environment that existed near the end
of the previous century, data-intensive computing and large-scale scientific
computing were essentially disjoint camps. One transported, stored, and
manipulated large data sets, and the other implemented scalable parallel
computing strategies for resolving very large computational models of scientific
and engineering problems. That era of separation has passed. In all the
applications of SBES discussed in this chapter, the use and generation of
immense data sets are integral components. For example, uncertainty
quantification, a key component of SBES, will require data sets many orders of
magnitude larger than those of traditional deterministic computing. DDDAS, by
definition, will demand new methods that rapidly generate, store, access, and
transfer large data sets over computational grids or high-bandwidth networks.
Then there is the issue of interpreting the results of the simulation itself, a
problem that can involve gigantic data sets.

As we work to harness the accelerating information explosion, visualization

will be among our most important tools. Indeed, visualization capabilities will

44
have a dramatic impact on scientific, biomedical, and engineering research;
defense and national security; and industrial innovations.

The reason visualization is such a powerful tool is that it is fundamental to

our ability to interpret models of complex phenomena, such as multilevel models
of human physiology from DNA to whole organs, multi-century climate shifts, or
multidimensional simulations of airflow past a
jet wing. Visualization reduces and refines data Visualization is
streams rapidly and economically, thus enabling fundamental to our ability
us to winnow huge volumes of data – an ability
to interpret models of
important in applications such as the
surveillance of public health at a regional or complex phenomena,
national level in order to track the spread of such as multilevel models
infectious diseases. Visualization for solving
of human physiology from
problems in applications like hurricane
dynamics and homeland security are generating
DNA to whole organs,
new knowledge that crosses traditional multi-century climate
disciplinary boundaries. Finally, the use of shifts, or multidimensional
visualization is rapidly transforming business
simulations of airflow past
and engineering practices for the better [15],
thereby increasing the competitive edge of our a jet wing.
industry.

Visualization allows people to comprehend visual representations of data

much more rapidly than they can digest the raw numbers or text. The designers
of computer visualization tools exploit the high-bandwidth channel of human
visual perception. Software systems may provide either static or interactive
visual representations of data, depending on user needs and on whether the final
goal is the explanation or the exploration of the data.

Visual representation of information has a rich historical tradition, primarily

in manually created depictions such as anatomical drawings, spread sheets or

45
 basic graphics. Now, however, computer graphics has the scalability to handle
datasets much larger than any that could be manually depicted. In addition,
computer graphics offers new possibilities in animation and interactivity.
Visualization is useful for detecting patterns, assessing situations, and prioritizing
tasks. Computation alone does not lead to understanding. The end user also needs
a comprehensible interface with the computational output. Visualization provides
that interface, and in so doing becomes the key to the interpretation of the data.

Engineers need assistance in making complex decisions and analyses,

especially with tasks involving large amounts of data. Often, engineers have to
deal with over-specified situations, and visualization of the situation helps them
filter out the irrelevant data. Engineers can use visual analysis systems to explore
“what if” scenarios and to examine data under multiple perspectives and
assumptions. They can identify connections between any number of attributes,
and they can assess the reliability of any conclusions [15].

Visualization research must continually respond to

We need to create and address the needs of the scientific community. For

an SBES example, the ability to visualize measures of error and

uncertainty will be fundamental to a better
visualization
understanding of three-dimensional simulation data.
framework for This understanding will allow the validation of new
uncertainty and to theoretical models, improve the interpretation of data,

investigate and and facilitate decision-making. With few exceptions,

however, visualization research has ignored the need
explore new visual
for visual representation of errors and uncertainty for
representations for three-dimensional visualizations [16]. We need to
characterizing error create an SBES visualization framework for
uncertainty and to investigate new visual
representations for characterizing error and uncertainty.

Within DDDAS and SBES applications, visualization of time-dependent data

46
will be crucial. Currently, however, most interactive visualization techniques
make use of static data only. The prevailing method for visualizing time-
dependent data is first to select a viewing angle and then to render time steps off-
line and play the visualization back as a video. Whereas this approach is often
adequate for presentational purposes, the inability to engage in interactive
exploration undermines the effectiveness and relevancy of visualization for
investigative purposes. Thus, new methods for interactively visualizing large-
scale, time-dependent data are needed. In addition, we need methods for
visualizing vector and tensor fields, field data collected experimentally from
multiple sources, and the ability to visualize data from both a global and local
perspective.

Figure 4: State-of-the-art visualization of turbulence in combustible flow.

Tomorrow’s capabilities may include parallel, interrogative visualization tools,
integration of visualization with large-scale dynamics simulations of complex
multi-physics events, and zooming techniques to visualize events at multiple scales.

47
 New approaches to algorithmic visualization will be needed that focus on the
needs of SBES. One approach is interrogative visualization, which is another
way of saying that quantitative querying through analysis must be supported
hand-in-hand with fast rendering of domains and computed function fields. A
second approach is interpretive visualization. Interpretive visualization focuses
on informatics and techniques to interpret imaging data, as well as various
quantitative-analysis data. For example, going from imaging data to the
construction of a domain model is an arduous task. In particular, to capture
spatial domain realism at each of the desired scales of simulation is daunting, and
may in fact be impossible.

A third approach is repetitive visualization. Humans have several

biorhythms. We have the daily circadian rhythm of mental efficiency, and we
alternate between periods of work and rest. We should accommodate biorhythms
in our methodologies for quantitative or interpretive visualization. In other
words, we should develop a mathematical framework for visualizations that
allows our perceptions of information to change with repeated visualizations of
that information. Additionally, if we are simulating the same function using
models at varying scales, we would have a natural opportunity to revisit our
earlier visualizations and make comparisons at multiple scales [2].

Because of the complexity and the massive amounts of data from

simulations, researchers will turn to semi-automated techniques from the
multidisciplinary field of scientific data mining to extract useful information
from the data. To meet this purpose, data-mining techniques exploit ideas from
image and video processing, statistics, pattern recognition, mathematical
optimization, and other fields. Scientific data-mining techniques can be used to
quantitatively compare simulations to each other and to experiments, to extract
summary statistics from high-fidelity simulations for use in building models, and
to analyze experimental data.

Whereas data-mining techniques can be effective in the extraction of

48
information from simulation data, several open challenges remain. Those
challenges include the extraction of features of interest from adaptive mesh
refinements and unstructured grids, the processing and interpretation of
experimental images that are often of low quality, the definition of metrics used
in comparisons of simulations and experiments, and the analysis of distributed
data sets resulting from simulations on parallel systems [17].

Finding
Visualization and data management are key technologies for enabling future
contributions in SBES. In addition, they hold great promise for scientific
discovery, security, economic competitiveness, and other areas of national
concern. Computer visualization will be integral to our ability to interpret and
utilize the large data sets generated in SBES applications.

3.6 Next-Generation Algorithms and

Computational Performance

Algorithms, the recipes for turning mathematical into computational

processes, provides the bridge between the models describing physical and
engineered systems, on the one hand, and the computational devices that generate
the digital representations of simulations, on the other. Too often, only the speed
of a computational device is cited as the figure of merit for simulation
performance, and the impact of algorithms on reducing the time complexity
(number of operations) and space complexity (size of memory) is unappreciated.
For more than three decades, progress in microprocessor capabilities has been
described by Moore’s Law, the observation that the number of transistors per unit

49
area on a processor doubles every 18 months, with corresponding increases in
practical performance for a fixed algorithm. Faster and more cost-effective
hardware is a strong driver for simulation-based engineering. However,
algorithmic improvements have been far more important.

In tomorrow’s SBES environment, the computing performance of individual

microprocessors will be just one of many important factors. New metrics will be
needed to judge the effectiveness of systems based on SBES principles. More
fundamental metrics include: time to solution in a multiprocessor environment,
the wall-clock time that elapses from the initiation of the simulation process to
the predicted outcome, and a measure of the confidence level for the predicted
result. If the time to solution is short, but the quality of and confidence in the
solution are low, the prediction may be of little value.

Improved algorithms have resulted in significant performance gains as

measured by time to solution. Recent studies have noted remarkable progress in
this area [3, 13, 18]. Figure 5 shows an example of that progress. The figure
depicts improvements in performance for large-scale simulations of turbulent
gas-phase combustion [19, p. 79]. As the example shows, advances in simulation
algorithms have tripled the effective performance over that due to advances in
processor speed alone over a period of a couple of decades. Similar results have
been documented in many other domains, such as magnetohydrodynamics and
radiation transport, and should inspire efforts to obtain and document super-
Moore’s Law gains in all areas of SBES

Among the most challenging problems for new algorithms are optimization
and inverse problems. Simulation-based decision-making gives rise to complex
optimization problems, which are governed by large-scale simulations. These
optimization problems appear in engineering design (in which the decision
variables represent the configuration and constitution of the system) and in
manufacturing and operations (in which the decision variables represent control
parameters). Moreover, decision-making informed by predictive simulation

50
requires estimations of uncertain parameters that characterize the simulation. The
response to the resulting inverse problems is to seek estimates for those
parameters that minimize discrepancies with observations.

Figure 5. Increases in time to solution due to new algorithms, given in

effective gigaflops over a period of years during which Moore’s Law, the
bottom line on this log-linear plot, remains valid (from [19, p. 79]).

Unfortunately, simulation-based optimization—whether in the form of

optimal design, optimal control, or inverse problems—is notoriously more
challenging than the corresponding simulation. First, the optimization problem is
typically ill-posed, even though the simulation problems themselves are usually
well-defined. Second, optimization usually results in a four-dimensional space-
time boundary-value problem, despite the evolutionary nature of the forward
problem. Third, the optimization problem is often formulated in probabilistic
terms. Fourth, the simulation is merely a subproblem associated with
optimization, which can be orders of magnitude more computationally
challenging. Indeed, when the simulation problem requires terascale resources,

51
the optimization problem will be in the petascale realm.

Contemporary optimization methods are inadequate for those tasks. We need

entirely new classes of scalable, efficient, and robust optimization algorithms that
are tailored to the complex multiscale, multi-physics simulations engendered by
SBES. The resulting challenges are of the highest order; yet, they must be
overcome to fulfill the promise of SBES: to elevate decision-making from a
practice relying on simple interpolative models to a more rigorous science based
on high-fidelity predictive simulation.

Finding
Investment in research in the core disciplines of science and engineering at
the heart of SBES applications should be balanced with investment in the
development of algorithms and computational procedures for dynamic
multiscale, multiphysical applications.

52
4.0 THE CRISIS OF THE KNOWLEDGE
EXPLOSION: SBES Education for
Tomorrow’s Engineers and
Scientists

In Volume Two of the SCaLeS Report [19], one finds mention of the “crisis
of the knowledge explosion.” This expression refers to the dramatic expansion
of the knowledge base required to advance modern simulation. The expansion
ignores the traditional boundaries between academic disciplines, which have long
been compartmentalized in the rigid organizational structures of today’s
universities. The old silo structure of educational institutions has become an
antiquated liability. It discourages innovation, limits the critically important
exchange of knowledge between core disciplines, and discourages the
interdisciplinary research, study, and interaction critical to advances in SBES.

The PITAC report [4] lists the following as one its principal
recommendations [4, p. 9]: “Universities must significantly change their
organizational structures to promote and reward collaborative research that
invigorates and advances multidisciplinary science. Universities must implement
new multidisciplinary programs and organizations that provide rigorous,
multifaceted education for the growing ranks of computational scientists the
nation will need to remain at the forefront of scientific discovery.” The report
goes on to ask: Will research and educational study in the twenty-first century be
“medieval or modern?”

The Panel strongly supports the viewpoint of the PITAC Report. If

simulation is to become a discipline, an engineering tool, and a life-long learning

53
 opportunity, then the university-level engineering educational system in this
country must be restructured. The current system does not provide the broad
range of interdisciplinary knowledge that tomorrow’s engineers and scientists in
SBES require. To succeed, they must acquire substantial depth in computational
and applied mathematics, as well as in their specific engineering or scientific
disciplines. Graduate students, moreover, must be able to build foundations that
allow them to access quantum and molecular science; statistical and continuum
mechanics; biological science and chemistry; applied and computational
mathematics; computer science and scientific computing; and imaging, geometry,
and visualization. Participation in multidisciplinary research teams and industrial
internships will give students the broad scientific and technical perspective, as
well as the communication skills that are necessary for the effective development
and deployment of SBES.

The integration of SBES into the educational system will broaden the
curriculum for undergraduate students. Undergraduates, moreover, will have
access to educational materials that demonstrate theories and practices that
complement the traditional experimental and theoretical approaches to
knowledge acquisition. In addition, SBES will provide a rich new environment
for undergraduate research, in which students from engineering and science can
work together on interdisciplinary teams.

As in any entrenched culture, change is hard to

NSF will need to come by. To change the culture of separate

disciplines in U.S. universities will require well-
collaborate with other
directed, persistent, and innovative federal
federal agencies to initiatives. The NSF has already done much to
open the door for a encourage multidisciplinary research and education
through initiatives like the ITR and IGERT and DOE
new generation of
has the highly successful SciDAC program. Serious
multidisciplinary
consideration should be given to turning successful
research.

54
cross-cutting programs into permanent (but still cross-cutting) administrative
structures, just as the disciplinary divisions are permanent.

It is unlikely that the necessary changes in educational structure will come

without strong directives from leaders from academia, industry, and government
laboratories. Exactly what changes are needed and how they can best be
implemented are issues well beyond the scope of this Panel. A detailed study of
these issues, perhaps undertaken by an NSF-funded committee of the National
Research Council, could trace out the new educational framework needed for
effective interdisciplinary study and research.

The NSF needs to take the lead in “legitimizing” multidisciplinary research.

One possibility might be the introduction of CAREER awards in
multidisciplinary research areas. In some areas, NSF will need to collaborate
with other federal agencies to open the door for a new generation of
multidisciplinary researchers. A core base of funding should be provided that will
allow multidisciplinary research and education to flourish. Best practices in
multidisciplinary education should be identified and then encouraged. NSF
programs like IGERT or DOE’s Computational Science Graduate Fellowship
Program provide much-needed funding and guidance for multidisciplinary
graduate education, but they only have the resources to fund a very few such
efforts in a given area. More funding is urgently needed for multidisciplinary
graduate education programs that offer students an integrated approach of team
research and career development.

55
Finding
Meaningful advances in SBES will require dramatic changes in science and
engineering education. Interdisciplinary education in computational science and
computing technology must be greatly improved. Interdisciplinary programs in
computational science must be encouraged, and the traditional boundaries
between disciplines in higher education must be made pervious to the exchange
of information between discipline scientists working within multidisciplinary
research teams.

56
5.0 CONCLUSIONS

This report has documented the findings and recommendations of the Blue
Ribbon Panel on Simulation-Based Engineering Science, or SBES. As defined in
this report, SBES is a discipline that focuses on the computer modeling and
simulation of complex, interrelated engineered systems and on the acquisition of
data meeting specified standards of precision and reliability. SBES draws on
advances in scientific understanding and incorporates that understanding into
new approaches to problems in the engineering domain through computer
simulation.

The need for SBES as a distinct field of research comes at a crossroads in our
nation’s technological development. For almost half a century, developments in
mathematical modeling, computational algorithms, and the technology of data-
intensive computing have led to remarkable improvements in the health, security,
productivity, quality of life, and competitiveness of the United States. We have
now arrived at an historic moment. As described in this report, we are on the
verge of an enormous expansion in our ability to model and simulate an almost
limitless variety of natural phenomena. That expansion has profound
implications:

First, computer modeling and simulation will allow us to explore natural

events and engineered systems that have long defied analysis, measurement, and
experimental methodologies. In effect, empirical assumptions will be replaced by
science-based computational models.

Second, modeling and simulation will have applications across

technologies—from microprocessors to the infrastructure of cities. Not the least
of these new technologies will be effective systems for national security.

57
Moreover, new simulation methods will lay the groundwork for entire
technologies that are only now emerging as possibilities.

Third, modeling and simulation will enable us to design and manufacture

materials and products on a more scientific basis with less trial and error and
shorter design cycles.

Fourth, modeling and simulation will greatly improve our ability to predict
outcomes and optimize solutions before committing resources to specific designs
and decisions.

Fifth, modeling and simulation will expand our ability to cope with problems
that have been too complex for traditional methods. Such problems, for example,
are those involving multiple scales of length and time, multiple physical
processes, and unknown levels of uncertainties.

Sixth, modeling and simulation will introduce tools and methods that apply
across all engineering disciplines—electrical, computer, mechanical, civil,
chemical, aerospace, nuclear, biomedical, and materials science. For instance, all
engineering disciplines stand to benefit from advances in optimization, control,
uncertainty quantification, verification and validation, design decision-making,
and real-time response.

There is little wonder that independent studies into the future of the nation’s
technology are unanimous in their conclusions that computer modeling and
simulation are the key elements for achieving progress in engineering and
science. The challenges of making progress, however, are as substantial as the
benefits. We must, for example, find methods for linking phenomena in systems
that span large ranges of time and spatial scales. We must be able to describe
macroscopic events in terms of subscale behaviors. We need better optimization
procedures for simulating complex systems, procedures that can account for
uncertainties. We need to build frameworks for validation, verification, and
uncertainty quantification. Finally, we need methods for rapidly generating high-
fidelity models of complex geometries and material properties.

58
 We are not alone in recognizing the urgency of our need to find solutions to
these problems. Many of our international competitors are well ahead of us in
committing the necessary funding and intellectual resources to overcome the
technical problems described in this report. Indeed, the technological superiority
Americans have so long taken for granted seems to be slipping away.

To arrest that trend and to help restore the U.S. to its leadership role in this
strategically critical technology, the Panel has made four recommendations (see
page xiv of this report for details):

(1) The Panel recommends that the NSF change its organizational structures
to facilitate long-range core funding of SBES.

(2) The Panel recommends a minimum sixfold increase in funding over 2005
levels of SBES-related disciplines.

(3) The Panel recommends a long-term program of high-risk research to

exploit the considerable promise of SBES.

(4) The Panel recommends that NSF underwrite an effort to explore the
possibility of initiating a sweeping overhaul of our engineering
educational system to reflect the multidisciplinary nature of modern
engineering and to help students acquire the necessary modeling and
simulation skills.

These recommendations call for NSF to take decisive and aggressive action
to support SBES. Unfortunately, over the past decade, NSF and other agencies
have persistently funded far fewer simulation-related research projects than
recommended. The difference was sometimes a factor of three and occasionally
a factor of ten. Moreover, even projects receiving support were frequently under-
funded and the grant periods were so short that researchers could only hope to
achieve incremental advances in the development of key disciplines. By contrast,
over the same period, funding for SBES research in Europe and Asia increased
many fold. To overcome these combined shortcomings in funding and duration,

59
the Panel recommends at NSF, an annual fund of $300 million be made available
to advance the SBES components critical to the nation’s security, leadership, and
competitiveness.

The Panel recognizes that improvements in the speed and efficiency of

computers remain important components of advances in SBES. Nevertheless,
efforts in these improvements should not supersede efforts in other disciplines
underlying simulation. Instead, NSF should focus on initiatives in SBES that
promote interaction between multiple disciplines that fit naturally and
strategically in parallel with or within the Cyberinfrastructure framework. Within
NSF, SBES should represent a new and fundamental thread of the
Cyberinfrastructure theme, one that could well call for a parallel program that
interfaces every division within the Directorate of Engineering, if not across the
entire Foundation.

60
 Appendix A: SBES Workshop Attendees
April 2004 Workshop

Workshop Organizers
Ted Belytschko (Northwestern) Thomas J. R. Hughes (U Texas-Austin)
Jacob Fish (Rensselaer) J. Tinsley Oden (U Texas-Austin)

Universities
Narayan Aluru, (U Illinois-UC) Donald Millard (Rensselaer)
William Curtin (Brown U) Robert Moser (U Illinois-UC)
Leszek Demkowicz (U Texas-Austin) Alan Needleman (Brown U)
Charbel Farhat (U Colorado-Boulder) N. Radhakrishnam (N Carolina A&T U)
Omar Ghattas (Carnegie Mellon) Mark Shephard (Rensselaer)
Anthony Ingraffea (Cornell) Charles Taylor (Stanford)
Chris Johnson (U Utah) Mary Wheeler (U Texas-Austin)
David Keyes (Columbia)

Government Laboratories
Thomas Bickel (SNL) Walter Jones (AFOSR)
John Red-Horse (SNL) Raju Namburu (ARL)
Roshdy Barsoum (ONR) Noam Bernstein (NRL)
Luise Couchman (ONR) Jonathan B. Ransom (NASA)
Craig Hartley (AFOSR)

NSF
Kamal Abdali (NSF) George Lea (NSF)
John Brighton (ENG) Priscilla Nelson (NSF)
Ken Chong (ENG/CMS) Michael Plesniak (ENG/CTS)
Sangtae Kim (NSF) Galip Ulsoy (ENG)

61
 September 2005 Workshop

Universities
Richard Alkire (U Illinois-UC) Carl Timothy Kelley (NC State)
Narayana Aluru (U Illinois-UC) Yannis Kevrekidis (Princeton)
Kyle Anderson (UTC SimCenter) David Keyes (Columbia)
Chandrajit Bajaj (U Texas-Austin) Alan Laub (UCLA)
Jon Bass (U Texas-Austin) David Levermore (U Maryland-CP)
Ted Belytschko (Northwestern) Michael Levine (Pittsburgh)
Larry Biegler (Carnegie Mellon) David Littlefield (U Alabama)
Wei Cai (Stanford) Wing Kam Liu (Northwestern)
Alok Chaturvedi (Purdue) Raghu Marchiraju (Ohio State)
Leszek Demkowicz (U Texas-Austin) J. Tinsley Oden (U Texas-Austin)
Abhi Deshmukh (U Mass) Steven Parker (U Utah)
Craig Douglas (U Kentucky) Linda Petzold (UC-Santa Barbara)
Charbel Farhat (Stanford) Robert Pennington (U Illinois-UC)
Jacob Fish (Rensselaer) N. Radhakrishnan (NC A&T U)
Roger Ghanem (USC) Mark Rashid (UC-Davis)
Omar Ghattas (U Texas-Austin) Tom Russell (U Mass)
Michael Heath (U Illinois-UC) Mark Shephard (Rensselaer)
K. Jimmy Hsiah (U Illinois-UC) David Srolovitz (Princeton)
Thomas J. R. Hughes (U Texas-Austin) Fanis Strouboulis (Texas A&M)
Tony Ingraffea (Cornell) Andrew Szeri (UC-Berkeley)
Chris Johnson (U Utah) Brian Wirth (UC-Berkeley)
Ramdev Kanapady (U Minnesota) Jacob White (MIT)
Sidney Yip (MIT)

DOE, Government Laboratories, and NIH

Roshdy Barsoum (ONR) Chandrika Kamath (LLNL)
Peter Castle (INL) Dimitri Kusnezov (DOE)
Luise Couchman (ONR) Habib Najm (SNL)
Stephen Davis (ARO) Raju Namburu (ARL)
Evi Dube (LLNL) Tomas Diaz de la Rubia (LLNL)
Francois Gygi (LLNL) Daniel White (LLNL)
Terry Yoo (NLM)
Industry
Joshua Fujiwara (Honda/OSU) John Ullo (Schlumberger-Doll)
Kirk Jordan (IBM) David Young (Boeing)

NSF
Richard Buckius (ENG/OAD) Wen Masters (MPS/DMS)
Ken Chong (ENG/CMS) Michael Plesniak (ENG/CTS)
Frederica Darema (CISE/CNS) M. C. Roco (ENG/OAD)
George Hazelrigg (ENG/DMI) Celeste Rohlfing (MPS/CHE)
Deborah Lockhart (MPS/DMS) HenryWarchall (MPS/DMS)

62
BIBLIOGRAPHY

1. Atkins, D. (Chair); Revolutionizing Science and Engineering Through

Cyberinfrastructure; National Science Foundation Blue Ribbon Panel Report,
January 2003.
http://www.communitytechnology.org/nsf_ci_report/ExecSum.pdf
http://www.communitytechnology.org/nsf_ci_report/report.pdf
http://www.communitytechnology.org/nsf_ci_report/appendices.pdf

2. Bajaj, C., Input from the NSF Workshop on Simulation Based Engineering
Science, Arlington, VA, September 2005.

3. Belytschko, T., Fish, J., Hughes, T.J.R., and Oden, J.T., (Eds.); Simulation Based
Engineering Science; National Science Foundation Workshop Report, May 2004.
http://www.ices.utexas.edu/~bass/outgoing/sbes/SBES_Workshop_1_Report.pdf

4. Benioff, M. and Lazowska, E. (Chairs), Computational Science: Ensuring

America’s Competitiveness; President’s Information Technology Advisory
Committee (PITAC) Report, June 2005. http://www.nitrd.gov

5. Chong, K. P., ”Nanoscience and Engineering in Mechanics and Materials”,

Journal of Physics & Chemistry of Solids, vol. 65 (2004), pp. 1501-1506.

6. Colvin, G., “America Isn’t Ready”, Fortune Magazine, July 25, 2005.

7. Darema, F., “Engineering/Scientific and Commercial applications: differences,

similarities, and future evolution”; HERMIS, vol. 1 (1994), pp. 367-374,
Proceedings of the Second Hellenic European Conference on Mathematics and
Informatics.

8. Dolbow, J., Khalell, M.A., and Mitchell, J., (Eds.); Multiscale Mathematics
Initiative: A Roadmap; Department of Energy – Office of Science Roadmap,
December 2004. http://www.sc.doe.gov/ascr/mics/amr/Multiscale Math
Workshop 3 - Report latest edition. pdf

9. Douglas, C. and Deshmukh, A., (Eds.); Dynamic Data Driven Application

Systems; NSF Workshop Report, March 2000. http://www.cise.nsf.gov/dddas

63
10. Estep, D., Shadid, J., Simon, T., (Eds.); Final Report Second DOE Workshop on
Multiscale Problems; Department of Energy – Office of Science Workshop
Report, October 2004. http://www.sc.doe.gov/ascr/mics/amr/Multiscale Math
Workshop 2 version - Report.pdf

11. Gereffi, G., and Wadhwa, V., (Eds.); Framing the Engineering Outsourcing
Debate: Placing the United States on a Level Playing Field with China and
India; Master of Engineering Management Program, Duke University, December
2005.

12. Graham, S. and Snir, M., (Eds.); The Future of Supercomputing; National
Research Council – Computer Science and Telecommunications Board Interim
Report, May 2003.http://www7.nationalacademies.org/cstb/pub_supercomp.html

13. Grosh, J. and Laub, A., (Eds.); Federal Plan for High-End Computing; High-End
Computing Revitalization Task Force Report, Office of Science and Technology
Policy, May 2004. http://www.sc.doe.gov/ascr/hecrtfrpt.pdf

14. Joy, W. and Kennedy, K., (Chairs); Information Technology Research: Investing
in Our Future; Presidents Information Technology Advisory Committee Report,
February 1999. http://www.itrd.gov/pitac/report/pitac_report.pdf

15. Johnson, C., Moorhead, R., Munzner, T., Pfister, H., Rheingans, P., and
Yoo, T. S., (Eds.); NIH-NSF Visualization Research Challenges Report; IEEE
Press, ISBN 0-7695-2733-7, 2006. http://tab.computer.org/vgtc/vrc/index.html

16. Johnson, C.R., “Top Scientific Visualization Research Problems”, IEEE

Computer Graphics and Applications, pp. 2-6, July/August 2004.

17. Kamath. C., Input from the NSF Workshop on Simulation Based Engineering
Science, Arlington, VA, September 2005.

18. Keyes, D., Colella, P., Dunning Jr., T., and Gropp, W., (Eds.); A Science-Based
Case for Large-Scale Simulation - Volume 1; Department of Energy – Office of
Science Workshop Report, July 2003. http://www.pnl.gov/scales/

19. Keyes, D., Colella, P., Dunning, Jr., T., and Gropp, W., (Eds.); A Science-Based
Case for Large-Scale Simulation - Volume 2; Department of Energy – Office of
Science Workshop Report, September 2004. http://www.pnl.gov/scales/

20. Lax, Peter D., (Chair); Report of the Panel on Large Scale Computing in Science
and Engineering; Department of Defense and the National Science Foundation,
December 1982. http://www.pnl.gov/scales/docs/lax_report_1982.pdf

64
21. National Science Foundation, Division of Science Resources Statistics, Science
and Engineering Degrees: 1966-2001; NSF 04-31, Project Officers, Susan T.
Hill and Jean M. Johnson. http://www.nsf.gov/statistics/nsf04311/htmstart.htm

22. New York University School of Medicine, http://endeavor.med.nyu.edu/public/.

23. Petzold, L., Colella, P., and Hou, T., (Eds.); Report of the First Multiscale
Mathematics Workshop: First Steps toward a Roadmap; Department of Energy –
Office of Science Workshop Report, July 2004.
http://www.sc.doe.gov/ascr/mics/amr/Multiscale Math Workshop 1 - Report.pdf

24. Reed, D., (Ed.); The Roadmap for the Revitalization of High-End Computing;
National Coordination Office for Information Technology Research and
Development Report, June 2003. http://www.cra.org/reports/supercomputing.pdf

25. Rising Above the Gathering Storm: Energizing and Employing America for a
Brighter Economic Future, Committee on Prospering in the Global Economy of
the 21st Century, 2005. http://www.nap.edu/catalog/11463.html

26. Scientific Discovery through Advanced Computing; Department of Energy -

Office of Science Strategic report, March 2000.
http://www.sc.doe.gov/ascr/mics/scidac/SciDAC_strategy.pdf

27. Y. Zhang, and C. Bajaj, Finite Element Meshing for Cardiac Analysis, ICES
Technical Report 04-26, University of Texas, Austin, 2004.

65
66