Information Technology for Manufacturing: A Research Agenda (1995)

Page 12

1—
Vision and Recommended
Areas of Research

Introduction

The manufacturing sector is the crucible in which many technologies are refined and fused for the purpose of making things that people need or want. In 1993 manufacturing accounted for 18 percent of the $6.4 trillion gross domestic product and for nearly 18 million jobs in the United States (U.S. Department of Commerce, 1994; Council of Economic Advisors, 1994). Broadly defined, manufacturing includes all of the activities involved in determining the needs of potential customers, conceiving and producing products to meet those needs, and marketing and delivering those products to the ultimate customer. Money is made and needs are satisfied by meeting quality, cost, performance, and time-to-market goals for the product being manufactured. These attributes—quality, cost, performance, and time to market—may be taken to be the yardsticks against which any new advance must be measured.

Given that U.S. leadership in certain areas of manufacturing is no longer the rule, it is reasonable to ask what needs to be done to regain international leadership in manufacturing. Suggestions abound, but in the absence of a clear strategy in this area, federal decision makers have struggled to find the right mix of investment in manufacturing research.

This quandary has extended beyond the funding agencies to the universities and to academic research. There are few departments of manufacturing at U.S. universities because many academics do not believe that manufacturing is an academic discipline. Moreover, the disciplines basic to making progress in manufacturing belong not only in the "hard" sciences and engineering in physics,

Page 13

mathematics, mechanical and electrical engineering, industrial engineering, computer science and engineering, chemistry, and materials science but also in the "softer" sciences of sociology, psychology, management science, and economics. Even though these separate disciplines are individually supported by funding agencies and universities, there is a lack of focused attention on how to integrate basic knowledge from many disciplines into knowledge that furthers manufacturing goals.

Information Technology and the Increasing
Complexity of Manufacturing

At the same time that this lack of strategy is apparent, all dimensions of manufacturing (e.g., products, markets, processes) are becoming more complex, diverse, and international. Indeed, common products such as automobiles can have thousands of parts, and modern aircraft and integrated circuits include millions of parts or active elements. Each of these products takes years to design, requiring the effort of hundreds or even thousands of people worldwide. Complex new products based on information content and their accompanying information-dominated design and manufacturing methods already require us to deal with entirely new scales of complexity.1 Some products require such levels of precision, delicacy, or cleanliness that people can no longer make or assemble the parts; in some cases, they cannot even see them.

To realize these and other products, manufacturing firms must cope with design processes (e.g., converting customer requirements and expectations into engineering specifications, converting specifications into subsystems), production processes (e.g., moving materials, converting material properties or shapes, assembling products or subsystems, verifying process results), and business practices (e.g., converting a customer order into a list of required parts, cost accounting, and documentation of procedures). The illustration on the cover indicates the relationships among these various elements of manufacturing and the role of information technology (IT; Box 1.1) in integrating them (see also Figure 1.1). By providing ways to facilitate and manage the complexity of these information-intensive processes, as well as to achieve integration of manufacturing activities within and among manufacturing enterprises, information technology will play an increasingly indispensable role in supporting and even enabling the complex

1 A case in point is very large scale integrated (VLSI) chips. A single VLSI chip may have several million transistors with submicron feature sizes. A complex system may have hundreds of chips and tens of millions of transistors. Logic design, functional tests, fault tests, timing, placement, and wiring data run to gigabytes per chip. Validation of a design may involve many millions of simulated test cases. Finally, different aspects of chip design are coupled, so that changes required in the logic design (for example) often affect the analysis of derived fault, timing, and place and wire views of the logic. Similar observations apply to airplanes, ships, and cars.

Page 14

Page 15

FIGURE 1.1 Information technology as a means to integrate various basic manufacturing activities.

practice of manufacturing. In the decades to come, information technology may have an impact on manufacturing performance and productivity comparable to that of mass production.

Purpose, Scope, and Content of This Report

This study was conducted to identify areas of information technology-related research needed to support future manufacturing. The committee chose to define manufacturing broadly as the entire product realization process, from specification through design and production to marketing and distribution. Although it believes that information technology has important applications to both continuous and discrete manufacturing, the committee focused on discrete manufacturing as the type in which the problems of applying information technology are most pressing. It did not include in its deliberations such important dimensions of manufacturing as the study of physical processes in manufacturing, although it did address information technology as it might be applied to controlling these processes.

Page 16

Chapter 1 of this report outlines some of the technical and other challenges confronting the manufacturing enterprise at the outset of the 21st century, expresses a vision of future manufacturing based on what is known today and what might be expected from information technology-related R&D efforts in the future, and recommends a research agenda aimed at achieving this vision. Chapters 2 through 7 elaborate on the contents of Chapter 1. Chapter 2 presents the context for manufacturing. The R&D agenda implied by the vision of Chapter 1 is the subject of Chapters 3 (product and process design), 4 (shop floor production), 5 (factory modeling and simulation), 6 (information infrastructure issues), and 7 (nontechnology issues). Chapters 3 through 7 explore in more detail the research questions that must be answered successfully if the vision of a robust and internationally competitive 21st-century manufacturing enterprise is to be achieved. A list of contributors to the report, a description of an engine plant visited by the committee, and sketches of possible advanced long-range technology demonstrations are given in Appendixes A, B, and C, respectively.

Flexibility For The Future

In the manufacturing environment of the 21st century, several trends will place increasing pressure on manufacturers:

Page 17

Responding to these challenges will require unprecedented flexibility. Flexibility in manufacturing is associated with rapid responses at the appropriate level to new information and constraints, which may range from changes in consumer preferences or international trade regulations or union requirements to a temporary fault in a crucial piece of machinery on the factory floor. Whatever the source of change or constraint, information systems in the factory must enable an appropriate response. Information technology will enable better planning and organization as well, helping to control the events to which responses are needed.

Recognizing Information Technology's Increasing
Capability in a Changing World

The role of information technology in manufacturing can be seen in the increasing use of computers to underpin product design and fabrication processes and to support related business processes such as sales and distribution. To date, the primary uses of information technology in manufacturing have been to control machinery and tools on the shop floor, to assist with administration in areas such as accounting and bookkeeping, to speed the transfer of information, and to support the management of product and process complexity (e.g., through computer-aided design (CAD) or manufacturing resources planning). Table 1.1 compares past and present characteristics and roles of IT in manufacturing.

Although these roles will continue to be important, the committee believes that information technology will become an increasingly significant source of support for different types of decision making needed in manufacturing (Box 1.2). This belief in the benefits of information technology is based on three premises:

2A recent agreement between two European auto manufacturers establishes a recycling network aimed at decreasing waste. It has begun to increase awareness of the importance of creating environmentally correct products for sustainable development around the world. In other cases, the producers of tires and batteries have been challenged to produce a product that can be disposed of without landfills. Taxes have been imposed on the purchase of new batteries and tries to help offset the cost of disposal. Yet such efforts are still the exception given the scope of manufacturing worldwide.

Page 18

TABLE 1.1 Past and Present Information Technologies and Their Roles in Devices and Activities Integral to Manufacturing

Page 19

TABLE 1.1 Continued

continues

Page 20

TABLE 1.1 Continued

Page 21

Page 22

These premises have been articulated and tested before, with varying degrees of success. What is it that gives rise to the committee's belief that information technology will be the basis of the next paradigm shift in manufacturing and a source of enhanced productivity?

The most straightforward answer is that the world has changed. The cost-performance relationship for information technology has improved so much over the last decade that it now seems feasible to devote many more computational resources to problem areas that were previously starved for such resources. More importantly, social and technological factors relevant to the successful implementation of IT in manufacturing are now much better understood. Concerted attention to these factors will dramatically improve the prospects for using IT successfully in manufacturing in the future.

For example:

Page 23

Balancing Current Needs and the Development of Future Capabilities

At the same time that it acknowledges IT's promise, the committee also recognizes very clearly that for good and proper reasons, managers of manufacturing enterprises are much more concerned about turning out products today than about improving their operations tomorrow. In particular, they regard new technology as beneficial only insofar as it can help them to achieve their tactical

Page 24

goals of meeting schedules and generating revenue for next quarter's balance sheet. Thus, it is not surprising that many manufacturing managers, especially those who have a great deal of practical experience with the day-to-day problems in manufacturing, make an argument that goes something like this: "We have real and immediate problems in U.S. manufacturing today that need to be solved if we are to compete successfully in the future. These immediate problems are the ones that should be given the highest priority for the near future, and research on advanced information technology whose payoff lies decades in the future, if ever, is a poor use of today's limited resources." In the absence of an explication of the detailed benefits of new technology, in particular information technology, and the impact of those technologies on the bottom line, a plant manager may quite reasonable adopt a "show me first BEFORE you think of modifying MY factory" attitude.3

The committee recognizes the existence of substantial tension between the conservative nature of manufacturing as an activity and the radical change implied by IT in the long term. Clearly, the immediate problems of manufacturing warrant attention; if U.S. manufacturers do not survive because they are unable to solve today's immediate problems, all of our investments in technology research will go for naught. Much of what is needed today must be delivered today by today's suppliers; research (i.e., long-term work) holds promise only for the longer term.

Yet some degree of investment for the longer-term future is warranted, and investments made today in IT research for manufacturing may have high payoff in the future. Indeed, the committee believes that pursuit of a coherent research agenda in this area would exploit U.S. strengths in computing and communications

3Such concerns are buttressed by the history of the first wave of computer-integrated manufacturing (CIM). CIM, a concept advanced in the late 1960s for efficient factory production, encompassed a number of different visions of manufacturing under computer control and had both proponents and detractors in various segments of the manufacturing community. In its first decade or so, the strategic advantage of CIM was seen to be the considerable reduction of "blue-collar labor costs." Strategic planners hoped that U.S. products would then be made and assembled as cheaply as those in the newly industrialized countries where labor costs were relatively small. The capabilities and implications of the technology were overstated by the strategic planners of the time, and they did not recognize the sea change in the nature of business operations as corporations became more transnational and the idea of simply using cheap labor was already fading. In this context, both the benefits and failures of CIM were obscured. In concrete terms, the outcome of the first wave of CIM was mostly the installation of computerized machinery and robot arms on the factory floor, often in inappropriate applications or without the necessary expertise to use these systems. Much of the initial CIM investment provided a poor return, and today true computer-integrated manufacturing is far from commonplace in U.S. factories (as described in Box 1.3).For more discussion of CIM, see Merchant (1971), Bjorke (1979), and Harrington (1984). For reference to the increasing transnationality of manufacturing, see Bartlett and Ghoshal (1992).

Page 25

and enable the United States to regain a role in setting the world standard in manufacturing. Further, since technology is relatively easy to diffuse over national borders (and competitive advantages due to the possession of a given technology thus tend to be transient), a continuing role for research on IT related to manufacturing should be anticipated by policymakers and manufacturing managers.

Resolving the tension between taking care of immediate needs and investing in research with longer-term payoffs will require managers to understand the strategic importance of new technology even as they are enmeshed in their tactical environment, and technologists to develop new technologies with business goals in mind.

Looking Ahead

The committee fully recognizes that IT by itself is not a panacea. For example, a study of the auto industry by the Massachusetts Institute of Technology found that highly automated auto plants achieved only average productivity, and even though the automation was capable of substantial flexibility, these plants produced only two body styles of one product.4 But even in the research domain, research to fill the gaps in the non-IT-related scientific and engineering knowledge about products and processes to be supported by information technology is essential if the promise of IT is to be exploited fully. For example, a deeper basic understanding about materials and fluid behavior may be needed to support new fabrication processes or to improve old ones.5 A deeper basic understanding about fatigue and corrosion may be necessary to support product designers attempting to reduce maintenance requirements. A deeper basic understanding of relations between tolerances and function may be key to developing new assembly and shaping processes to be controlled by IT.

A critical concern for developers of technology should be identification of the decision maker who will decide whether or not to adopt a given information technology innovation. Such decision makers are found at nearly all levels in a company's hierarchy, but they have different concerns depending on where they sit, and they look to new technologies to answer different questions (see Chapter 7, Figure 7.1, which indicates the types of questions that may be asked at various levels of an organization's hierarchy, the relationships between the various levels

4See Dertouzos (1989). Similar examples in which the use of information technology does not correlate with marketplace success have been found in the service industries as well; see CSTB (1994a). The lesson is clear—companies (either manufacturing or service) that use information technology inappropriately are not likely to reap significant advantages from such technology.

5New processes may include laser processing, water-jet cutting, and deposition-based fabrication. Chemical vapor deposition is an example of a process that was introduced only recently but is now common.

Page 26

of the hierarchy, and the need for consistency and coherency throughout the enterprise). While the identification of these basic concerns is not IT research per se, technology researchers who wish their innovations to be adopted must craft and present their research in ways that address the needs and business aspirations of the decision makers in manufacturing. Without solving problems relevant to chief executive officers, technology researchers will seldom see their creations adopted.

Finally, in considering IT's potential for contributing to improvements in manufacturing, time line is a concern. The committee's vision is based on projections of how what is known today in IT and its applications to manufacturing might plausibly develop in the future. But forecasting time lines is difficult, and committee members' views on this issue reflect a spectrum of opinion. Even under the most optimistic view, it is not really plausible that most manufacturers, especially smaller ones, will have achieved this vision by 2010. However, it may be that major parts of it will have been achieved by a number of large manufacturing concerns, and that smaller firms will be learning how to use new information technology (or at least that they will be learning how to conduct their business in an IT-rich environment). Under a more pessimistic view, large-scale adoption may not happen until much later.

The committee emphasizes that its admittedly expansive vision for IT in future manufacturing should be recognized as just that—a vision—rather than as a definite prediction for the future. It is a vision driven both by threat (in that the vast power of information-driven manufacturing is increasingly recognized in other countries) and opportunity (in that the capabilities of information technology are growing at a rapid rate and that IT itself represents an area of U.S. comparative advantage), and describes the areas that are most likely to be essential to achieving a competitive (or indeed a leading) position in information-driven manufacturing.

The Potential Impact Of Information Technology
On The Manufacturing Enterprise

The Broad Vision

In a future manufacturing enterprise characterized by ubiquitous and integrated computing, IT will be important in every aspect of manufacturing. Computers will be everywhere—on factory floors, in products, in offices, in wholesale and retail outlets, in homes, and on the street. Computers will be embedded into products as invisibly as electric motors are today. Familiarity with personal computing and use of a national information infrastructure will be widespread, and the intimidation factor that currently often prevents the consideration of IT solutions will be greatly reduced. Manufacturing decision makers will use IT to make real-time determinations of, for example,

Page 27

In addition, the IT supporting diverse queries will be seamlessly integrated, so that information needed from one part of the enterprise by another part will be transported with minimal difficulty. For example, in the manufacturing enterprise envisioned by the committee, plans for new processes or products will be transferred electronically from development into production, significantly reducing the interval between the design and realization of a process or product. Simulation models at various levels of detail (from floor operations to strategic planning) will couple to each other, so that results of one simulation model can serve as input to another.

If this vision of IT-enabled manufacturing comes to pass, future manufacturing operations will realize many further and significant improvements in:

Improvements in the areas above through the use of IT would represent a fundamental paradigm shift from today's manufacturing enterprise, one that may already be under way (as suggested by Appendix B). The following sections elaborate the committee's vision of how various dimensions of manufacturing (product design, process design, production, and business processes) may be

Page 28

transformed by information technology; a further discussion of these dimensions themselves is found in Chapter 2.

Nearer-Term Prospects for Improvement

Product and Process Design

Product design has expanded in scope in recent years. It has been traditionally understood as the collection of geometrical, material, and system specifications that achieve functional performance as a finished product that meets customer perceptions of need; today it also includes attention to manufacturability, usability, and environmental concerns. Thus, a designer may well choose a product design with inferior performance characteristics in certain noncritical aspects that is much simpler to produce than the alternative. To a much greater extent than is true today, 21st-century design will address product design aspects that are not often associated with traditional design at all, such as designing so that servicing a product will be easy and error-free or so that a robust final product can be made from parts obtained from many different sources. Different dimensions of product design will thus need to be integrated to an unprecedented degree.

IT will continue to help to improve the quality of designs and reduce the cost and time needed to produce them. In particular, the 21st-century design environment envisioned by the committee would allow product designers to create a ''virtual" product and make extensive computer simulations of its behavior without supplying all of its details, and then "show" it to the customer for rapid feedback. The ability to undertake rapid electronic prototyping of designs would increase the ease of revising them, resulting in fewer design compromises and lower costs for making those changes. Product data would be represented in a uniform manner and would include information on all variants. Simulations would also enable the exploration of all of a product's behavior modes, including nominal behavior and major off-nominal variant behaviors that might affect performance, product quality, environmental quality, manufacturability, or user safety. The product design environment would provide computer assistance that would build on existing design knowledge and even existing designs but that would also be flexible enough to accommodate design innovation.

The designer would be able to specify a product in terms of function and performance rather than in terms associated directly with the production process (e.g., shape, tolerances, electrical inputs). For example, the designer of a motor would be able to use function-relevant parameters, such as the load capacity and life of a bearing or the allowed vibration frequencies and minimum fatigue resistance of a shaft. In addition, he or she would be able to express the desired functions of the product in a translatable and analyzable format that would permit

Page 29

decomposing a given design into functions and subfunctions and mapping these into engineering assemblies and subassemblies.

Whereas today's product designer builds a physical prototype or uses physical construction aids in the design process, tomorrow's product designer will certainly make extensive use of computer-based models. The construction of a physical prototype for use in testing a function or developing a tool necessarily freezes the design at the moment of construction and often leads to a loss of synchronization between model and product design when the design is altered after the prototype is built. An example of a more flexible approach to design is provided by the Boeing 777 airliner, whose computer-based design was undertaken by Boeing using a three-dimensional representation of a solid model. This application eliminated most paper documents and dramatically reduced design errors as measured by changes and corrections required to engineering drawings. Engineers were able to test small parts and whole sections of the fuselage to determine, long before these components were built, whether they would fit together properly on the factory floor. The capability for testing clearances between subassemblies to ensure human access was especially useful. Once conceptualized, changes to the virtual 777 could be realized immediately and communicated to all individuals interacting with the model, including distant contractors responsible for building components and subassemblies.

Another advantage is that these computer-based models often can enable different aspects of a product to be developed simultaneously. In the design of programmable electronic systems, for example, a linked product-and-process design environment would allow the supporting software to be tested on hardware simulators before the completion of hardware development. Since software development is often at least as time-intensive as hardware development, the use of simulators could cut overall development time by as much as a factor of two.

Extensive computational support would enable designers of manufacturing processes to explore, trace, and compare the production cost, quality, performance, safety, maintenance, and environmental implications of design decisions. The process designer would be able to specify a manufacturing process unambiguously and in ways that would yield information about, for example, its efficiency in advance of its actual deployment. Information technology would enable the process designer to identify the right manufacturing, assembly, and test processes for creating and verifying the elements of a product and matching them to specifications for functional behavior. He or she would also be able to develop models that could predict the expected results of variations implemented during design of a process. Such activities, which might be undertaken concurrently, could result in fewer errors, fewer repairs, shorter times to market, and lower cost. For example, flows of material through a production facility might be improved by evaluating various factory equipment layout concepts—to reduce the distance a product must travel before its completion or the number of hand-offs between process steps, or to allow people to perform more efficiently—or by

Page 30

evaluating different material-handling modes (e.g., whether or not to dedicate an automated guided vehicle to a particular process).

In computationally supported manufacturing, process design models would couple to production line implementations of those simulations. For example, code to control tools on the shop floor (e.g., numerically controlled machines, robots, and transfer systems) would be generated through simulations. If a product's specifications changed, the process model would be refined by the process designer, and such refinements would be reflected in the production line. When necessary, process designers would be able to modify manufacturing processes adaptively, taking advantage of the knowledge available at every step in a factory's entire manufacturing process in order to improve yield at a subsequent or preceding step. For example, detailed real-time knowledge of the status of each piece on the shop floor (both unfinished product and equipment) might enable managers to optimize equipment use and/or lot movement globally, rather than locally improving each step but possibly reducing overall factory performance.

With product design and process design coupled electronically, new processes or products would go directly from electronic blueprint or simulation to production; use of the process model as the blueprint for process transfer would significantly reduce process or product transfer times.

The common theme in all these design applications is that the manipulation of information (in the form of design simulations and the like) is likely to be much cheaper and faster than real experimentation within an operating manufacturing facility. As a result, the design space within which it is feasible to explore alternatives is much larger, giving designers more options for how to craft a product that will meet user needs.

Shop Floor Production

In the 21st century, order-driven production (Box 1.4) will increase compared to today's forecast-driven production (i.e., production volume sized in accordance with a forecast of what the demand for the product is likely to be), which runs a significant risk of overproduction (with the unsold products incurring carrying costs charged to the producer) or underproduction (in which case production must be ramped up at great expense on an emergency basis). Under a "build-to-order" strategy, production volume is much more closely correlated with orders, thus reducing the risk of production that does not meet demand and providing considerable flexibility in changing production priorities.6

6 Of course, when a customer needs an order filled in less time than it takes to produce a single item, "build to order" in a literal sense is not realistic. Thus, in practice, both forecasts and orders will influence production. Nevertheless, a production process can be configured to increase flexibility so that, for example, generic items can be produced in reasonable quantity and then customized on demand, thus reducing the time between customer order and fulfillment.

Page 31

Factories in the future will continue to use a variety of technologies and processes and varying levels of automation. However, with the help of IT, fabrication of unique items could be more convenient and less expensive, and products could be manufactured economically in smaller lot sizes than is characteristic today and with greater ease of production changeover (smaller setup times). Scheduling of people, machines, and plants would be based on all aspects of an enterprise's operations, taking into account sales and market projections as well as work in process and capital utilization to dynamically schedule operations for the best profit. Scheduling would optimize overall factory performance, rather than the performance of local area functions as it does today.

In the 21st-century production environment foreseen by the committee, the new devices and processes used would result in very little material waste, with shaping and assembling devices, for example, extending their reach to ever smaller spatial scales. Increasingly, the materials used will be synthetics, composites, and ceramics. If used, robot manipulators would operate on ever-smaller parts and assemblies. Information would be embedded in parts and products and read by material-handling, shaping, assembling, and processing equipment, further automating the flow of materials and work in process. Parts would be self-identifying not only in the production process but also throughout the life of the product. Controllers of material-handling equipment would be more tightly coupled with machine controllers and orchestrated by higher-level software in accordance with global plant goals. New shaping and assembly devices would be

Page 32

capable of high-precision fabrication at unprecedented levels of repeatability. Tool and equipment control strategies would involve all the control techniques—classical, modern, fuzzy logic, and neural network controllers—in an integrated approach to control.

Production engineering would rely heavily on a reusable base of acquired production knowledge. Manufacturing production lines would be designed using modular design and control components interconnected through well-defined interfaces. When necessary, production lines would be reconfigured to accommodate new generations or types of products, and certain changes to the production process would be implemented in a very short time (perhaps minutes or hours) as external circumstances changed.

To meet the need for replacing or upgrading equipment, the production line would accommodate the tooling and fixturing necessary for adding new parts. Programs for controlling equipment (e.g., for numerical control) and systems (e.g., for scheduling and execution) would be generated and modified to accommodate changes to product requirements, changes in the production process, and changes in management objectives (e.g., little or no inventory, rush orders).

The components that make up a production line would be structured in a manner that they could be "plugged and played" together in a modular fashion. "Plug and play" would characterize the physical connections for machines (power and communication), the functional connections for operations, and the tooling specific to products being produced. Equipment in a production line would be easily moved, replaced, modified (upgraded), and enhanced (adding machines), improving system flexibility and responsiveness. Functional control modules such as scheduling procedures and planning procedures would plug directly into the factory control system directing the operation of shop floor equipment. Similarly, the modules affecting a product (e.g., tools, fixtures, masks) would be extendable to accommodate new products. Changes to a production plan would be possible with minimal interruption.

The 21st-century computationally supported production line would be robust in the face of an uncertain mechanical environment. Detailed information on factory and equipment status, based on data from sensors and controllers located on the shop floor, would be visible to managers and shop floor personnel alike. Real-time control would enable dynamic rerouting and reconfiguration of work flows to bypass problem areas in the production facility with minimal disruption; in worst-case situations, a production facility would exhibit graceful rather than catastrophic degradation. The ability of automated shop floor tools to undertake self-diagnosis and self-correction of routine problems would improve reliability and reduce the need for people to tend such tools. The extensive simulation capabilities of detailed factory models would enable companies to design facilities that incurred minimal operational difficulties and to train new employees in the use, maintenance, and repair of complex equipment without risking the cessation of factory operations.

Page 33

Finally, the time needed to fabricate and deliver products would be significantly improved. Customized products would be delivered to consumers much more rapidly than today because of reductions in the time needed to make and assemble the various components of the product and in the time spent in waiting for processing by various components.7

Business Practices

Information technology could enable a high degree of integration between future manufacturing enterprises and their customers and suppliers. Working closely with a manufacturer via tie-ins to the manufacturer's information systems, for example, suppliers would know when to deliver supplies and would have blanket authority to deliver them as necessary. A precedent exists in certain car manufacturing operations today: suppliers query the manufacturer's production intentions and deliver their goods to the manufacturer's assembly plant in line sequence order, that is, in the order that they will be used. Information about needed supplies is thus passed automatically, without the manufacturer having to place an order explicitly. Suppliers may even be paid only when the final product is shipped, rather than when the parts are delivered, thus providing further incentive for just-in-time delivery. This practice, known as "pay on production," is increasingly common in the auto industry, because it lowers supply chain carrying costs that have created an artificial focus on inventory and it increases team play and product quality.

In computerized manufacturing, a supplier's own fabrication processes would easily accommodate changes required by the manufacturer. A customer's highly customized order for even a very complicated and sophisticated part (for which there may be many alternative choices) could be entered directly by the customer into a manufacturer's system: the order would be scheduled, and the customer would receive immediate acknowledgment and a product delivery date.8 A request for material replenishment could go directly and electronically from a manufacturer's factory floor to the supplier's system. With protections against

7 A good example is the manufacture of automobiles. The final assembly of a car takes a couple of days. Building the various components takes several days to a week. But the time from order to delivery for most American cars is usually at least several weeks. However, certain Japanese auto manufacturers now use sales forecasting techniques so sophisticated that most vehicles produced in a given production batch have a buyer by the time they come off the assembly line. In addition, the mix of options (a major variable in both throughput and production time) is roughly standardized to help match production batches with customer orders and is kept relatively small to simplify assembly. The result is that the time from order to delivery has been reduced to well under a week for these firms.

8 Some manufacturers have begun to implement or plan such systems, although with less customization of product than that envisioned in this report. See, for example, the discussion of Motorola's fusion program for producing pagers in Trobel and Johnson (1993).

Page 34

fraud and theft in place, the supplier would deliver the material directly to the factory floor, where receiving activities are performed. In short, the entire enterprise would be integrated all along the supply chain, from design shops to truck fleets that deliver the finished products.

Hints of this vision of integrated enterprises are available today, as businesses transfer information among related business organizations through electronic design interchange protocols. Some design information can be shared through common CAD systems or through information expressed through the International Graphics Exchange Standard. Suppliers to some manufacturers have been required to send progress reports of supplies in shipment and transit electronically for many years. A variety of industry-specific standards for transferring data electronically have been promulgated in the last 20 years. An advanced national information infrastructure such as that proposed by the Clinton administration would provide a widely accessible vehicle for linking manufacturers, suppliers, and customers to achieve the kind of integration and information sharing envisioned.

Factory decision makers' use of modeling would enable them to inquire about orders, market information, production status, product design, human resources, and financial information from a single unified source. Advanced modeling and analysis tools would make possible the collection and definition of data, processes, and associated knowledge at the depth and breadth necessary to support the simulation of business processes; models would take into account the form, meaning, and content of relevant data elements and processes. Dynamic simulations of an entire business would be possible, enabling decision makers to evaluate a wide range of "what-if" scenarios for the purpose of strategic and tactical planning. These simulations would operate much more rapidly than real time (simulating several months of factory operations in a few minutes), and so designers would be able to test a large number of alternatives for improving factory performance.

New Manufacturing Styles

If the nearer-term research challenges posed in this report are met fully and successfully and the anticipated nearer-term advances implemented so as to enable the capabilities outlined here, the committee believes that the result will be substantial improvement for manufacturing in cost, quality, asset utilization, productivity, environmental control, and time-to-market performance metrics. Implementation of these capabilities would also have profound implications for the workplace that would have to be addressed in a socially responsible manner. The longer-term concepts of a virtual factory and a programmable or reconfigurable factory as outlined below would embody IT to such an extent that the very character of manufacturing would be fundamentally altered. As presented,

Page 35

they pose targets and goals for research, whether or not they represent what future factories will actually be like.

The Virtual Factory

When a single factory may cost over a billion dollars (as is the case in the semiconductor industry), it is evident that manufacturing decision makers need tools that support good decision making about their design, deployment, and operation. However, in the case of manufacturing models, there is usually no testbed but the factory itself; development of models of manufacturing operations is very likely to disrupt factory operations while the models are being developed and tested. Today, decision makers have found that the use of good computer models of manufacturing facilities can provide valuable information that might otherwise have required time-consuming and expensive physical experimentation. More sophisticated versions of these simulations—what might be called virtual factories—call for a distributed, integrated, computer-based composite model of a total manufacturing environment, incorporating all the tasks and resources necessary to accomplish the operation of designing, producing, and delivering a product.

With virtual factories capable of accurately simulating factory operations over time scales of months, managers would be able to explore many potential production configurations and schedules or different control and organizational schemes at significant savings of cost and time in order to determine how best to improve performance.9 Since a factory model running in simulation mode would run thousands of times faster than real factory operations and would likely cost much less as well, managers would have a rapid, nondisruptive methodology for testing various manufacturing strategies. Improvements suggested by real operations could be tested without risk in the simulation. Simulations could also assist in training tool operators and floor managers, who would be able to use factory models in simulation mode much as pilots use simulators to gain experience in flying real airplanes, especially under stressful or unusual conditions.

Computer-based factory models might also be coupled to real factories in what could be called "control" mode. In control mode, the factory model would actually control and run the operation of the real factory through manipulation of the objects in the virtual factory. Operating procedures and scheduling protocols would be validated in the virtual factory and then applied in or transferred to the

9 Factory operation model development and testing are very different from process model development and testing in the sense that the disruption of an entire factory can be catastrophic for a firm's productivity. It would of course be possible to build a new factory that would carry the burden of experimentation. But factories are capital-intensive, and few companies are in a position to risk large amounts of money on time-consuming, expensive, or risky experimentation.

Page 36

real production facility. Control mode would enable the direct electronic transfer of modularized capabilities from computer simulation to production line.

Coupled to appropriate computer-based reasoning and decision-support tools, a virtual factory operating in control mode would be capable of a significant amount of self-diagnosis. Driven by data from the real factory, the virtual factory would be able to analyze the performance of the entire factory continuously to determine the potential for optimizing operations to reduce costs, reduce production time, improve quality, or reuse materials.10 For example, the virtual factory would be able to use the data collected by a factory monitoring system, analyze potential and actual failures, and identify the cause of a problem. Such a system assumes the availability of a knowledge base for every piece of equipment in the factory that, given certain monitored data, can be used in conjunction with a diagnostic system and reasoning and decision-support tools to identify the source of a problem.

Whether information is derived from a model run in simulation mode or control mode, results from good models can be examined from various user perspectives, including those of factory managers, product planners, and process equipment operators, in order to provide solutions to various types of problems that manufacturing personnel with different job tasks might encounter. Modeling and simulation are likely to become a basic tool used at all levels within the manufacturing environment, from senior management to equipment operator.

The Programmable or Reconfigurable Factory

A programmable or reconfigurable factory is one in which most or all of the information necessary for producing a product is embodied in a knowledge base and the associated programs, such that the factory can take in the information and produce the product with minimal human intervention. In such a factory, capital would substitute for labor both on the shop floor (with robots or other automated equipment undertaking many assembly and fabrication processes) and in real-time factory management (with computer-based decision-making functions operating on intellectual rather than physical inputs). Such a factory would be capable of a significant degree of "lights-out" operation, that is, operation with limited human involvement. Programmable or reconfigurable factories would also enhance managerial capability to cope with accidents and malfunctions on the work floor as work flow could be rerouted to bypass problem areas.

10 Simple versions of self-improving factories have been demonstrated in semiconductor and chemical plants, where neural networks and/or adaptive control have been used to monitor and adjust parameters to optimize operations. Although such plants differ in nature and character from those for assembling discrete products such as automobiles or airplanes, it may be possible to develop analogous improvements.

Page 37

Programmable factories expand the concept of the flexible manufacturing system, which is intended to be a set of machines and controllers that work cooperatively to manufacture similar parts in a product family in relatively small lot sizes but at unit costs and with qualities characteristic of mass production. But thus far, most such economically successful systems have produced parts with rather small variations in style and have exhibited limited software, little sensor feedback, and, still, relatively large lot sizes.

It is clear that a programmable factory would require software capable of operating on the entire range of product and process information. It would also require generic materials that could be used for different products made in the factory, generic tools and fixtures, programmable machine tools, and generic sensors. While not everything could be fabricated from generic materials, it might be possible to substantially reduce the number of different types of materials and tools needed to produce a diversity of parts. For example, it has been estimated that in a particular factory, 2,000 tools that were being used could be reduced to fewer than 50. In another case, 500 different types of steel could be replaced by 50 different types of steel. In a third case, 10 different types of 8-bit microprocessors could be replaced by a single 16-bit microprocessor. If more generic materials and tools could be used in production, high-volume purchases of these items could drive down costs.

In a programmable or reconfigurable factory, different products would be produced by changing software. Different software instructions would direct tool and machine controllers to perform different operations and to deliver different items to different work cells in different sequences. Although a single factory almost certainly could not produce computers and cars on different days, a highly programmable factory, in conjunction with new and more flexible fabrication processes, could produce cars in one week and trucks in the next week; the Toyota Motor Company does this in Taiwan today. Even without new fabrication processes, a highly programmable factory might be able to produce computers one day and defense electronics the next. Such a factory would depend on production facilities configured in such a way that a new production operation could be set up relatively rapidly. The ultimate goal is a paradigm known as "sell one, make one," as it is known in Japan.

In certain restricted domains, low-lead-time "reprogramming" of a factory to produce different products is possible today. The manufacturing facilities for books, very large scale integrated chips, and petroleum products produce many different products (different book titles, different chips, different types of fuel and lubrication) simply on the basis of a software change. Of course, the term "software" in this context refers to the particular text or masks that differentiate one book or chip from another, but the principle is the same—only intellectual inputs or changes are necessary to produce a different product, while the physical facilities of production remain much the same. But for other products, changeover times remain significant.

Page 38

A programmable factory is also necessary for the economic manufacture of highly customized products. An example from today is the ''on-demand" production of soft-cover textbooks in which chapters can be selected by a teacher based on his or her individual teaching needs. With increased product customization, customers would be able to obtain catalog items with the features, characteristics, and aesthetics they desired, at prices they could afford (probably comparable to the prices for mass-produced goods). The material and manufacturing cost per unit of producing a product likely would not be significantly different for a production run of 100 units or of 100,000 units in a single batch. Smaller lot sizes would also have major benefits with respect to quality control. When defects in a production process are caught early (as is the case when small lot sizes are produced), the amount of rework is minimized and fixes to the production process can be implemented more rapidly.

Note that high degrees of customization create additional stresses on scheduling. Since customization requires only small quantities of specialized materials, "just-in-time" scheduling either works properly or fails by idling the production machinery; the option of building inventory as a hedge against missed delivery times simply does not exist, since maintaining excess inventory is then a matter of purchasing unnecessary components rather than purchasing components that will ultimately be used.

The Networked Factory

In concept, a networked factory is one in which suppliers (both internal and external) and customers are connected electronically to a manufacturer (e.g., on the National Information Infrastructure). Manufacturers have been tied to suppliers and customers by telephone, mail, telex, and fax for years; the primary advantage of electronically networked connections would be the speed with which information could be exchanged and processed, sometimes automatically by intelligent agents that could respond to certain routine requests.

An electronically networked factory (hereafter a networked factory) would demonstrate significantly reduced transaction times as information technology reduced the delays of paper-based information transfer; information technology would facilitate instantaneous acknowledgement, scheduling of deliveries, and guaranteed service times. Many of the factors contributing to delays in the design and production processes would be significantly reduced within a networked factory. Reducing delay would contribute to reducing the time to market for new or improved products. A particularly important improvement would be a reduction of the time it takes a production facility to initiate the first step needed to respond to an order, since it is this time that often dominates the overall time required to fill an order.

Enabled through the National Information Infrastructure, networked factories would increase the options available to product and process designers.

Page 39

Today's designers are strongly constrained by the process capabilities of manufacturers—product designers do not design products that their factories cannot make, and process designers do not create processes that their factories cannot implement. Indeed, even approaches to design regarded today as sophisticated (e.g., design for manufacturability, design for assembly) are necessarily limited by preexisting production processes and facilities. When a single firm owns the means of production, such approaches make sense. But if constraints on ownership are relaxed (and process elements consequently can be linked on a regional, national, or even international basis), designers can be freed to focus primarily on the expressed needs of the customer without worrying about how best to use a single plant for which many costs have already been incurred. Designers using a networked factory would be able to "outsource" various production processes more easily and to coordinate their operation.

Of all the different modern concepts in manufacturing, the idea of a networked enterprise including a networked factory is perhaps the most widely accepted and adopted; in some circles, the term "agile manufacturing" is also used. Further, the evolving National Information Infrastructure is expected to facilitate networking of all sorts. Chapter 6 discusses this connection in greater detail.

Microfactories

A microfactory is a production facility whose output capacity can be scaled up by the replication of identical facilities. Since microfactories would not depend on economies of scale for economic viability, they would draw strongly on the technologies of programmable or reconfigurable factories as they attempted to produce small-scale output at unit costs comparable to or only slightly higher than those for mass-produced items. If microfactories prove to be feasible, a single, large, centralized manufacturing facility could be replaced by a large number of replicated, modular microfactories that could be geographically distributed and located close to customers.

For producing quantities of identical items, traditional factories oriented toward mass production will probably remain superior to microfactories, because anything that could be done to improve the production process in a microfactory could also be done in a traditional factory. On the other hand, today's mass-production factories are capital-intensive construction projects that are themselves custom-built. If a microfactory could itself be mass produced in quantities large enough to reduce the cost of an individual microfactory, it might be possible to amortize the cost of designing the microfactory over many such (identical) facilities. Even today, some steel micromills have drastically reduced the capital cost of steel production. In addition, microfactories might incur lower product transportation costs (as the result of placing microfactories near customers), lower

Page 40

inventory costs (as the result of production on demand), or lower labor costs (as the result of using locally available labor).

Irrespective of cost issues, however, microfactories might well provide advantages in other situations. For example, microfactories might provide a way for firms to insert local content into manufactured products, perhaps through local final assembly—a capability that could be desirable for political purposes (e.g., as a dimension of international trade relationships). A second example of a microfactory could be a mobile fabrication facility (e.g., a microfactory located on a large naval ship that produces replacement parts for the battle group with which it sails); in such a scenario, economic concerns might be secondary to the capability for a rapid response. A third example is that microfactories of a sort do exist today, although they make business sense for reasons other than lower unit production costs. Microbreweries for beer and street-corner copy shops are two examples of microfactories in which production costs are higher than those of larger facilities; nevertheless, such microfactories fill niches because they provide higher quality or greater convenience. The primary challenge remaining for microfactories is one of economics.

Getting From Here To There—
The Need For Balance And A Considered Approach

The various new manufacturing capabilities described above are tantalizing and appeal to many current notions of the progress possible in manufacturing. But for this vision to be realized, it will be necessary first to balance the responsibilities of factory managers and manufacturing decision makers to turn out quality products at low cost in a timely manner today against the desirability of planning to secure the potentially large improvements offered by judicious and innovative use of current and future information technology.

Even if these tensions are resolved, however, the full implications of successfully implementing IT are not known, and neither the committee nor the manufacturing community at large has thought through the many possible effects. To illustrate, success in some of the areas discussed above raises the following questions:

Page 41

These questions, and many others, are largely outside the scope of this report. But their identification, articulation, and eventual resolution are an integral part of moving toward a vision of IT-enhanced 21st-century manufacturing.

Whatever one's view, it is clear that a number of major technological and sociological barriers must be overcome before IT's potential to revolutionize manufacturing can be widely accepted and achieved. More rapid progress in

Page 42

overcoming these barriers will increase the likelihood of earlier acceptance and achievement. The nature of these challenges and the research needed to overcome them are the subjects of the next several chapters.

The Research Agenda

Technology Research

In formulating a research agenda, the committee was faced with the realization that manufacturing is fundamentally a complex activity, with many interactions among its various components. In the committee's view, this realization reflects the nature of manufacturing as "an indivisible, monolithic activity, incredibly diverse and complex in its fine detail … [whose] many parts are inextricably interdependent and interconnected, so that no part may be safely separated from the rest and treated in isolation, without an adverse impact on the remainder and thus on the whole" (Harrington, 1984). Thus, it is fruitless to seek the identification of specific "silver bullets" upon which all other progress in the field depends. That said, however, the committee has identified several general themes for technology research that would advance the capabilities of information technology to serve manufacturing needs; these themes include product and process design, shop floor control, modeling and simulation ("virtual factory") technology, and enterprise integration as it affects factory operations and business practices.

The sections below summarize a research agenda that is discussed in detail in Chapters 3 through 6. The largest part of the research recommended in this report is aimed at developing various IT-based tools to support advanced manufacturing.

Product and Process Design

Product design and process design depend heavily on human judgment. Research is needed both to develop information tools that can help human designers make good decisions in their design work and to increase understanding of the design process itself. Enabling the creation of better tools and facilitating the design process could, for example, make it easier to generate a requirements specification that meets customer needs or to design a product or a process "from scratch" and/or through the reuse of existing and validated designs. Although better tools and techniques are needed in all fields, product and process design for mechanical components and assemblies is especially important.

The committee believes that a research agenda for product design (especially for the design of mechanical products) should build to the extent possible on the lessons learned in the design of electronic products such as integrated circuit chips. For example, the design of electronics today is based on having, at each

Page 43

stage in the design process, design abstractions that contain only the detail relevant to that stage. Using these abstractions, the product designer can postpone decisions about details and focus on higher-order questions about function, leaving the detail to the subsequent stages in which lower-level issues can be resolved. Designers of electronic products also have an extensive set of predefined and prevalidated "parts" that can be used as building blocks in product design; altering the parameters of these predefined parts allows some customizing of the product being designed. Such capabilities also need to be made available to designers of mechanical products.

Although mechanical products differ qualitatively from electronic products (e.g., mechanical products are three-dimensional, and interactions among their components are more analytically intractable), making their mechanical design as easy as the design of electronic products today is a reasonable asymptotic goal to work toward. In addition, tools are needed that will enable the identification of trade-offs between cost and dimensions of performance such as reliability, power consumption, and speed; between cost and design choices; between alternate space allocations; or in functional decomposition, subassembly definition, three-dimensional geometric reasoning, and make-or-buy decisions.

In the domain of process design, tools for describing processes are critical for the design of individual products, the design and operation of factories, and the development of modeling and simulation technology. Formal descriptions are necessary if processes are to be represented in sufficient detail and with enough specificity to be adequately complete and unambiguous; such formalisms would allow designers to describe factory processes (involving both machines and people), design activities, and decision processes, among others. Languages for describing processes must facilitate checking for correctness and completeness and must be able to express variant as well as nominal process behavior.

New tools for describing and representing processes could also be used to enhance product design, so that by simulation and emulation the best process could be matched to the product design (and vice versa) for maximum economic advantage (or to satisfy whatever criteria—such as quality or time to delivery—are important for the particular case). Used in this way, simulation and emulation could facilitate "design for manufacturability" and "design for assembly," which should also encompass design for rapid testing and diagnosis, fast maintenance and repair, quality control, and material handling, as well as design for modeling more efficient factory processes and operations.

Shop Floor Control

By automating processes, extending the uses of sensors, and improving scheduling, information technology can play a vital role in improving the flow of material and the routine control functions of machine tools, robots, automated guided vehicles, and many other basic machines on the factory floor.

Page 44

Research is needed to advance the level of process automation, including greater ease of interconnection of factory equipment and more automated responses to problems. The current thrust today in systems development is toward open systems that allow equipment from different manufacturers to be "mixed and matched" as needed. Achievement of an open equipment controller architecture that would enable all factory and shop floor components to share the same programming environment, communication facilities, and other computer resources would contribute to the interconnection of factory equipment (as well as to enterprise integration). Advanced manufacturing languages that would be more flexible than existing languages for programming unit processes (such as APT and Compac for machining) would support the operation of accessory devices in conjunction with a particular process and be more closely coupled to product data generated by CAD/CAM systems (to facilitate direct transfer of products from blueprint to production).

Research is needed on advanced sensor systems as well. Sensors provide real-time feedback about the operation of a process during manufacturing (e.g., unpredicted part-tool interactions). Historically, sensors have served only as production monitors. Increasingly, they are becoming active components of production systems, integral to either a process or a finished product. Standardized sensor architectures must be developed so that sensors and actuators can be plugged into a common control system with only minor, automatic reconfiguration. Sensors connected through such architectures would be linked directly into databases for dynamic updates usable by machine controllers. Standardized sensor architectures will require a uniform method of characterizing sensors and actuators suitable for automation, applicable to a wide range of devices. Data fusion techniques for correlating inputs from multiple sensors would help overcome the difficulties of sensing in a relatively "dirty" environment. Intelligent sensors would be able to process shop floor data to higher levels of abstraction to determine their significance to manufacturing decisions.

Effective real-time, dynamic scheduling of factory operations on the shop floor remains a major problem but has great potential for improving factory performance. Dynamic scheduling is desirable because management priorities for production must be balanced moment to moment against circumstances prevailing in a plant and in the manufacturer's supply chain (e.g., sudden changes in conditions generated by drifts in machine capability, material shortages, unplanned equipment downtime, delays in arrival of necessary components). Dynamic scheduling would determine what should be done next by any particular piece of equipment at any particular moment based on current conditions throughout the factory. Research is needed to develop real-time scheduling tools that would provide capabilities for integrating scheduling and control reactively (e.g., tools that would make use of adaptive scheduling techniques based on the severity of the contingency at hand and the time available to adjust the schedule and/or

Page 45

scheduling techniques that could exploit windows of opportunity occurring fortuitously).

Also needed are information tools and techniques to ensure graceful degradation of plant operations in the event of local problems. Factory managers and operations teams will require effective means to support overall situation assessment, both within a factory (e.g., knowing when certain machines are inoperable, knowing the location of various parts) and outside it (e.g., knowing that a delivery will be delayed). They will also require tools that integrate multiple dimensions in which managers must make decisions, including decisions about product release, reordering, sequencing and batching, safety stock and safety lead-time, use of overtime, and order promising.

Autonomous agent-based architectures are a potential alternative to top-down scheduling. Autonomous agents (implemented as software objects or collections of objects, perhaps represented by physical robotic agents) could be attractive for manufacturing applications in the areas of planning, monitoring, and control. Important research problems connected with agent-based architectures include the level of autonomy that agents in various locations should have and how collections of agents would maintain stability when given potentially contradictory goals. The deployment of agents might well be risky until these issues are addressed in detail.

Modeling and Simulation

To realize a virtual factory that can faithfully reflect the operation of a real one in all relevant dimensions, it will be necessary to represent real manufacturing operations at different levels of abstraction. All objects in a real factory, whether they are pieces of equipment, product lots, human resources, process descriptions, data and information packets, or facilities, must have direct counterparts in the virtual factory; indeed, the actual production facility in which raw materials are transformed into physical products is itself one level of abstraction in a comprehensive virtual factory model. The boundaries of the virtual model must be flexible, capable of incorporating activities outside the factory or focusing only on entities within the factory structure as necessary for analytical purposes.

Central to simulation technology is research on modeling frameworks that can link the wide variety of models representing activities from all parts of manufacturing, from design through orders to multicountry manufacturing and distribution to customer delivery; a large number of models will be needed to simulate realistically even a modest factory. The models will be distributed in time and space; research will be required to understand how to link these essential pieces in a timely manner. Work is needed not only on general modeling techniques but also on fast methods of tailoring a specific model to local conditions. Hierarchical simulation models built from well-tested fundamental equipment

Page 46

blocks will be critical to the success of factory modeling. Similarly, procedures are needed that allow for the parallel processing of these hierarchical models so that very rapid (faster than real-time) simulation times can be achieved.

A second dimension of simulation technology is how to account for the stochastic nature of events on the factory floor. Given the multitude of unexpected events that can affect factory operations (e.g., tool breakdown, late supply shipments, personnel absences) as well as decisions influencing operations (e.g., which machine is to be changed, what personnel are to be used), no single simulation run will be definitive. Rather, tools are needed that can test a given configuration or plan in thousands of probabilistically determined runs.

A complete simulation that could be used for everything from analysis to control for even a modest factory is out of reach today. However, a first step would be the comprehensive simulation of an individual production line. For such a task, appropriately detailed models of individual tools are needed that can then be combined to provide overall realism. Equipment-level simulation models have been developed and used to analyze equipment-level characteristics, but the simulation of a production line would test the ability of such models to act in concert.

Validation of simulation models will be essential. Since a simulation is good only to the extent that it provides an accurate representation of reality, justified and well-grounded confidence in the model is critical for use and implementation. Simulation models can always be tweaked and otherwise forced to fit empirical data, but the purpose of simulation is to learn something reliable about a hypothetical factory operation for which no empirical data exist. Managers and decision makers will need high levels of assurance that a simulation's prediction of a new factory faithfully reflects what would actually occur, even taking into account the random events that affect today's manufacturing systems so adversely. Well-validated simulations would enable the creation of a demonstration platform that could compare results of a real factory system before the system ever operates. Tools to automate the process of sensitivity analysis for simulations would be particularly helpful in coping with the stochastic factory environment.

Ultimately, the modeling and simulation capabilities resulting from the research outlined here should be able to support configuring and constructing a real factory for high-level performance (on multiple dimensions), as well as planning how best to operate it once it has been constructed. A concrete demonstration of these capabilities would be the creation of a platform capable of comparing the results of real factory operations with the results of simulated factory operations using information technology applications such as those discussed in this report.

For modeling and simulation to serve manufacturing needs, two broad areas of research stand out for special attention: the development of information technology to handle simulation models in a useful and timely manner, and capture of the manufacturing knowledge that must be reflected in the models.

Page 47

Enterprise Integration and Business Practices

Research is needed to extend and enhance the information infrastructure supporting manufacturing enterprises, including both the internal infrastructure used within a factory (perhaps a dispersed one) and the external infrastructure that increasingly links an enterprise to its suppliers, partners, and customers. The use of networks by all kinds of personnel and of equipment to exchange all kinds of data (text, numeric, graphic, and video) calls for high bandwidth; greater dependability and security; greater support for real-time communication, monitoring, and control; and better interoperability (through architectures, standards, and interfaces) for component systems and networks of different types. Achieving ease of interconnection is essential; attaching equipment and subsystems to a factory information system should be as easy as plugging household appliances into outlets, at least in principle. Beyond better network-related facilities, there is a need for better technology for the exchange of information, information services to support integration of applications, and standard representations, protocols, libraries, and query languages.

In addition, enterprise integration requires research and development relating to the interconnection of applications. Indeed, much of today's manufacturing information technology can be characterized as islands of automation that are unable to communicate with each other due to incompatibilities in their representation of largely similar information. Enabling intercommunication will require the development of appropriate ways of explicitly representing information related to products, fabrication processes, and business processes, as well as how each element relates to itself and to other elements. These new representation schemes will themselves demand a deep understanding of the underlying information, an understanding that is sorely lacking in many of the domains that relate to manufacturing.

Research is also needed on organizing principles and architectures for connecting different network-based applications into a seamless environment. Such connection is necessary, for example, to link flexible manufacturing cells to the plant scheduling function and to link the scheduling function to the enterprise order, delivery, and financial systems. Enterprise integration also implies a need for research to enable the automatic interpretation of the type of transaction being executed, the routing of the message to the right location for processing, and the processing that must occur when the message for the transaction reaches the correct system. Integration of individual enterprises into the marketplace in the information age will require security and authentication features that guarantee the integrity of electronic transactions.

Non-Technology Issues

Expanding the scope of what is achievable by information technology is only one dimension of realizing a 21st-century vision of manufacturing. It is equally

Page 48

important to understand how manufacturing enterprises can actually make use of the technology. Even today, much useful technology remains unused. Innovators in manufacturing must ensure that human, institutional, and societal factors are aligned in such a way that information technology can be deployed meaningfully. This is a difficult but essential task, since even great technology that goes unused is not particularly beneficial to anyone.

Data, information, and decisions need to be communicated accurately across the breadth and depth of manufacturing organizations. Many mechanisms can contribute to enhancing communication, including sabbatical programs for industrialists and academics in each other's territory, teaching factories, and advanced technology demonstrations that illustrate how the use of information technology can benefit factory performance.

Considerable research in social science will be necessary to facilitate the large-scale introduction of information technology into manufacturing. In particular, fully exploiting new technologies generally requires new social structures. Innovators will have to confront issues such as the division of labor between human and computer actors, the extent and content of communications between those actors, and how best to organize teams of human and computer resources.

Matters related to education and training will be central to 21st-century manufacturing. Given an environment of increasingly rapid change, continual upgrading of skills and intellectual tools will be necessary at all levels of the corporate hierarchy. "Just-in-time learning," that is, learning things as it becomes necessary to know them, may assume added importance.

Finally, although businesses depend increasingly on their intellectual and information assets, generally accepted accounting principles that businesses use to audit their finances and operations are derived from a business philosophy in which capital expenditures (i.e., expenditures that relate to the long-term value of a company) are associated with buildings and pieces of equipment. Research is needed to develop valuation schemes that appropriately account for the contribution of knowledge and core competencies to manufacturing and enterprise values.