Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
7 Using Supercomputing to Transform Thinking About Product Design Clifford R. Perry Eastman Kodak Company I am delighted to share with you why Eastman Kodak Company sci- entists are using the awesome power of supercomputers and associated visualization systems. I will also discuss some recent applications by Kodak scientists in multiple disciplines throughout the Kodak research labora- tories. But I will begin by briefly discussing the issue of communication and assimilation of the use of supercomputers within our industrial sector, because I believe that our failure to communicate how we assimilate the use of computer technology is principally responsible for the very slow rate of application of supercomputers to industrial R&D problems. UNDERSTANDING THE NEED FOR COMMUNICATION Our number-one priority, in my view, is to create better ways of com- municating not only how we use supercomputers but also how we encourage assimilation of their use by potential practitioners. Although I will be sharing with you what we have been doing at Kodak much more needs to be done. Communication and assimilation are industry-wide problems. We cannot continue to learn on our own in this highly competitive global marketplace which brings me to a very brief, personal story. Some 30 years ago when I was a college freshman, I accepted a job that required only that I "play" with the university's newly acquired IBM- 650 computer, a so-called first-generation computer. My job was to invent applications and to prove the computer's usefulness to the university's 81
82 CLIFFORD R. PERRY research community. They had been given the IBM-650 but had few ideas about what problems to apply it to. These people were scientists interested in science and not in tools that had not yet proved their direct and useful applications in helping them pursue their science. This was an example of computer technology preceding useful appli- cation, and there was no known way to assimilate its use into its intended environment. There was no form of communication and no way to learn from others. We each had to learn on our own. Some 7 years later I accepted a job at the General Motors Technical Center's Computer Technology Division, where I was again to play the role of a researcher using computers: I was to find useful applications within General Motors' R&D community for their newly installed IBM-360 computer, a second- generation computer. But the first-generation problem still existed: the latest computer technology again had preceded development of a process to assimilate its use into its intended environment. There was no form of communication and no way to learn from others. We each still had to learn on our own. Some 21 years after that I was asked to facilitate the use of supercom- puting within Kodak's R&D community. We were in the process of signing a 3-year contract with the National Center for Supercomputing Applica- tions (NCSA) with Larry Smarr at the University of Illinois. The readiness for supercomputing at Kodak, as I initially had surmised, was once again an example of availability preceding a process for complete implementation. There was still no known or established process to assimilate the use of supercomputing into its intended environment, and there was no form of communication and no way to learn from others. Would we still, after 30 years, have to learn on our own? In 30 years computing technology had advanced tremendously, but there was still no organizational process or procedural framework for making its applications clear. How then was supercomputing to fulfill its promise as a problem killer and as a tool with the potential to transform thinking? There was still no approach to rapidly and effectively making its enormous capacity understandable to its potential users, and hence there was little hone of making its use pervasive within its potential market. -r - - I believe that this past and present bumbling about is a direct result of our failure to communicate within our own organizations and with one another. Perhaps the greatest failure to communicate is between the practitioners and the laity, who fail to understand the promise, the payoff, the problems, and the limits of computational science and who, more importantly, fail to understand- because of the lack of effective ways to communicate the synergy achieved in attacking problems with the combined methods of theory, experiment, and computation. In this current era of intense worldwide competition, we cannot afford
SUPERCOMPIJTING AND PRODUCT DESIGN 83 to learn alone, organization by organization, one at a time. Although we may have communicated results obtained from applying supercomputer technology, we have not communicated approaches that deal with the soci- ology or social psychology of scientists who are the potential practitioners of supercomputing. We have been given a remarkable tool, the supercomputer, that holds great promise for us in industry, but we cannot assume that accommodating the supercomputer requires only minor changes in the way we assimilate new technology into our R&D activities. How do we help foster the cultural change that is required? I cannot speak of applications without also speaking of communications that are both internal and external to our organizations. We must, as supercomputer stakeholders, grow to understand the commonality of our endeavors through communication. USING SUPERCOMPUTERS TO INCREASE PRODUCTIVITY I think it is important to realize that supercomputers do not require visualization systems, and visualization systems do not require supercom- puters. However, we have found that when they are combined into a system that includes the scientist or engineer, we have new opportunities to trans- form the potential of our R&D opportunities for significant discoveries and breakthroughs. Visualization as a Stimulus for Creativity This synergistic system, visualization and the interpretation of what we visualize, can lead to new theories and scientific paradigms, that is, the set of beliefs, values, and techniques shared by members of the scientific community and new tools for the advanced engineering sciences. I think that the power of computer-generated scientific visualization is best summarized by Herbert Butterfield as quoted in Thomas S. Kuhn's The Structure of Scientific Revolutions (University of Chicago Press, 1970~. Butterfield described science's reorientation by a change in paradigm as "picking up the other end of the stick, a process that involves handling the same bundle of data as before but placing it in a new system of relations with one another by giving it a different framework" Supercomputer visualization is that system. It fosters different inter- pretations that can lead not only to insight and understanding but also to a shift in paradigms. It has been said that when Aristotle and Galileo looked at swinging stones, Aristotle observed a falling object that was constrained,
84 CLIFFORD R. PERRY whereas Galileo saw a pendulum. Priestley and Lavoisier both saw oxygen, but they interpreted their observations differently. The richness of visualization as part of the process of scientific inquiry is that it allows us to differ in our interpretations of what we have seen. These different interpretations can lead us to new theories and new scientific paradigms, and then to new technologies and new products. Supercomputing and visualization systems at Kodak have not caused shifts in paradigm as significant as those caused by the Newtonian or Ein- steinian revolutions, nor have they caused relatively small changes in the paradigms generated by the wave theory of light, the dynamical theory of heat, or Maxwell's electromagnetic theory. Supercomputing has, how- ever, allowed us to transform the way we think about problems involving polymers, crystalline compound dies, photographic imaging systems, and manufacturing technology development. For example, using supercomputer simulation, our engineers at the Kodak Park Division were able to test and then to redesign the delivery system used in a critical photographic film manufacturing process. This eliminated the need to build a prototype of the first flawed design and resulted in cost savings of hundreds of thousands of dollars. These new theories and processes will accelerate Kodak's ability to continually enhance the quality of its products. I will now discuss a few of the many examples that illustrate how the supercomputer has allowed us to change the way we think about specific problems and has led to new solutions that would not have been possible without supercomputing and visualization systems. I hope that these examples will serve as evidence that we are making progress through the effective use of supercomputing to enhance the quality of our products and our manufacturing processes. Visualization and Simulation of Physical Processes The following three examples of supercomputer-assisted science in- volve simulations of physical processes. The graphics allow us to visualize the gigabytes of scientific data generated by the supercomputer simula- tions. These computer-generated images illustrate physical qualities as well as quantities. The first example illustrates how we use supercomputers in fundamen- tal research. At Kodak, we synthesize polymers for many purposes. Thus we need to better understand the properties of these complex molecules in order to design new and better plastic materials. Supercomputers and visu- alization systems allow our researchers to study the physics and chemistry of basic polymers in new and different ways. For example, these researchers
SUPERCOMPUTING AND PRODUCT DESIGN 85 are investigating how a polymeric network behaves when stressed, the mech- anisms of self-diffusion, and the effects of polymer blending. Results from these studies will enable prediction of polymeric behavior in the design of materials with specific performance criteria. In a real sense, these polymer researchers are designers of new materials. The goal of another investigation is to understand the diffusion of a polymer chain when it is entangled with other chains. The interactions and intraactions of the chains significantly influence the ability of a polymer chain to diffuse. It is possible to see on videotape one polymer entangled in a network of other polymers. We can then follow the path the polymer takes while wandering about the material. The sooner the polymer is able to diffuse away from its initial configuration, the sooner it is able to relieve the stress. This important information will enable scientists at Kodak to infer the behavior of the material when it is under stress as well as its viscosity and other viscoelastic properties. Developing this knowledge is a challenge. Although the questions appear to be simple, obtaining answers requires well-crafted computer simulations involving many hours of computer time. Visualization is again required to gain insights from the gigabytes of data generated by the simulation, and it is often very helpful in determining the validity of the models. Visualization Applied to the Manufacture of Products A second example illustrates the use of supercomputers at Kodak in applied research at the engineering level. It is no surprise that Kodak has a very keen interest in the manufacturability of plastics. Plastics can be found in many of our products, from cameras and film spools to copiers and mass memory products. We also produce bulk plastics such as acetate fibers and PET products that are used extensively by the bottling industry. We are always looking for new plastics to enhance product quality and to reduce costs both the costs of materials and the costs of manufacturing. A simple example illustrates how supercomputing and visualization can help. Suppose we want to make an improved plastic part for one of our products, such as a camera body. We also desire increased productivity in the process used to manufacture the parts of the camera body. We assume that we will use a totally new plastic, but first we must familiarize ourselves with its characteristics. What happens when the plastic flows into the mold? How long will it take to fill the mold? How long will it take to cool the plastic and to eject the plastic part from the mold? We also need to know about the heat conduction the temperature, the pressure, and the velocity at every point of the mold. Prior to the use
86 CLIFFORD R. PERRY of supercomputers and visualization systems, we had little idea how to even think about these questions, other than to build a physical prototype that was extremely expensive and time-consuming. It was virtually impossible for mold designers to physically see into the mold and observe the physics at work. Now for the first time our designers can see these phenomena via supercomputing and visualization. A new vocabulary is emerging based on visualization of the computer simulation. More importantly, new in- sights are emerging. As a result, manufacturing productivity is increasing significantly. A third example of the application of supercomputing involves re- search on color, a vital element in many Kodak imaging products. Our photographic scientists are keenly interested in matching physics with per- ception. This is important because it can lead to increased flexibility in developing color reproduction processes. Color reproduction almost always involves creating color in a constrained way. Since only three dies are used in a photographic paper, or three phosphors in a cathode-ray tube, it is important that these choices be made so that the image quality and color rendition are not compromised in the eye of the observer. Color theory attempts to describe causal relationships between physi- cal color stimuli from the environment and psychological color sensations evoked by these stimuli. Color stimuli are radiations within the visible spec- trum and are described by radiometric functions, whereas color sensations are subjective and are described by words such as red, blue, or green. Kodak research scientists, in collaboration with faculty at the University of Illinois through Kodak's partnership in the NCSA, are exploring an elegant mathematical color theory to enable the computation of all possible ways of evoking a given color sensation. The key to this theory is the mapping of color stimuli to the color sensations they evoke. With the supercomputer, researchers can determine how different dies can be used in color reproduction by simulating a standard obse~ver's response to the colors produced by the die mixtures. This flexibility could lead to more cost-effective and higher-quality color reproductions. Although it is still in the early stages, this theory is showing much promise. ASSIMILATING SUPERCOMPUTING INTO THE R&D CULTURE Lacking an existing framework for assimilating supercomputing tech- nology into our scientific and industrial culture, a situation that I described earlier, we at Kodak created an approach or a process. While we believe it is a model and not the model, it is an approach that has served us well and one that we seek to continually improve.
SUPERCOMPUTING AND PRODUCT DESIGN 87 Obviously we realize that we cannot displace whole cultures overnight, so we initiated what might be called a supercomputer assimilation program. We started by organizing a program of on-thejob seminars on the latest applications and developments in supercomputing. In short, we communi- cate how other scientists use supercomputers. These computational science seminars have featured such distinguished experts as Kenneth Wilson, a Nobel laureate formerly from Cornell University and currently director of a supercomputing center in Ohio, and other leading scientists from super- computing centers such as NCSA, our valued partner. Donna Cox, for example, has visited Kodak three times in the last 18 months and has held seminars with literally hundreds of Eastman Kodak Company scientists and engineers. We are also taking advantage of a long-standing forum of exchange within the Kodak research laboratories, the Interplant Technical Confer- ence Series. Kodak scientists and technicians gather three times annually for these conferences to foster and to profit from a greater sharing of ideas, techniques, and applications across a wide variety of scientific disciplines and technological frontiers. The 77th Interplant Technical Conference, titled Supercomputing and Science and Eng~neenng, win focus on the de- veloping role of supercomputing technology in R&D at Kodak and will communicate how Kodak scientists and others use supercomputers. In addition to seminars and conferences, Kodak has another mech- anism to keep R&D personnel involved. We disseminate in the R&D community the many technical reports that have been written by those who have used the supercomputer, and we have had over 55 R&D personnel that have traveled from Rochester, New York, to Urbana, Illinois, who have written trip reports about their experiences at the center. We have since installed a T-1 link, but the trip reports have provided valuable insight into how we can facilitate the use of the supercomputer by other Kodak scientists. These reports document the results of the R&D activities and the techniques and computer tools used to generate those results. The terms supercomputing and v~sualuai'on are appearing more and more in our everyday vocabulary. We also have a supercomputer technology planning process, a partic- ipatory planning process that facilitates communication and involvement. We recently completed and disseminated to several hundred people, includ- ing top management throughout the company, a Kodak supercomputing requirements and provision plan that profiles our past needs and projects our future needs in specific areas of application. This report, which also details a plan to supply computing capacity and visualization systems to the worldwide Kodak R&D community, involved the direct participation of more than 50 Kodak scientists and engineers covering a broad spectrum
88 CLIFFORD R. PERRY of R&D from life sciences to photographic imaging systems to manufactur- ing technology. This Kodak R&D activity worldwide involves some 8000 scientists and technicians working in the United States, England, France, Germany, Australia, and Japan. By definition, the supercomputer represents the state of the art. As such, it is used by an elite set of pioneers, people willing to put in the tough work required to tame the tool in return for major payoffs. We call these pioneers computational scientists. They are role-model super- computer users who communicate and demonstrate the usefulness of the supercomputer. Thus as an additional step, Kodak research management also established the Computational Science Laboratory within the Informa- tion and Computing Technology Division. This laboratory provides one-on-one collaborative assistance to facil- itate the effective integration of supercomputing technology into R&D activities. While we do research using computational science methods, we also serve as in-house advocates for the use of supercomputing. Our mission is to be a catalyst. We promote advanced R&D computing technologies and systems that will help people increase their ability to do creative, cost- effective, business-focused research. And we follow up by keeping another of our important stakeholders, top management, involved. Because it is especially important to communicate supercomputing benefits to our management stakeholders, we formed a Supercomputing Technology Board, consisting of the directors of several research and en- gineering divisions, to inform management of the experiences gained and the successes realized through the use of our supercomputing program. As Kodak people become more involved with supercomputers, management is continually made aware of the tangible successes resulting from the company's investment in advanced computer technology. We have made extraordinary progress in supercomputing in a relatively short time. There have been both a remarkable symbiosis and a synergy among scientists and technologists in their applying of computational tech- niques to their respective disciplines. On balance what has come out of this symbiosis has been beneficial to each faction, be it academic or industrial. Yet there has not been a universal benefit. Such a benefit can be achieved only by an informed and self-confident scientific and technological popu- lation that can leverage the kind of cultural change necessary for growth. And that change in growth is predicated on improving communication at all levels. In closing, I leave you with the word communication, because com- munication is the keystone. We must communicate outside of our organi- zations, as we are communicating and learning from one another in this symposium, and we must orchestrate change within our organizations by creating new processes of internal communication. I believe that the means
SUPERCOMPUTING AND PRODUCT DESIGN 89 to achieving sound communications are forums such as this and the estab- lishment of processes and organizational advocates that facilitate the use of supercomputing technology within our organizations. Let us have more of these forums. The results will be the increased efficiency and enhanced effectiveness of supercomputing scientists and en- gineers in the American workplace.