Information in the Design Process

Alice M. Agogino

Department of Mechanical Engineering

University of California at Berkeley

Intelligent Real Time Design (IRTD) is a framework that explicitly considers the cost of resources consumed during the design process (e.g., the value of the design team's time; the cost of further information gathering, analyses, or experimentation; and "time-to-market" opportunity costs) and trades these off against the optimum of the design solution found. Such a process dynamically takes into account the "real-time" aspect of the design process. IRTD helps designers focus their information-gathering efforts on reducing uncertainty in areas of the design that have the greatest impact on the design objectives (Bradley, 1993; Bradley and Agogino, 1991, 1992, 1993, 1994).

Uncertainty and ambiguity are greatest in the early conceptual stages of design, where the greatest potential for design improvement lies. Sources of uncertainty can be found in the design constraints, objectives, and parameters; evaluation models; and customer preferences. IRTD provides a prescriptive methodology for transforming the design space over the course of the design process. The results of an IRTD analysis are a set of expected costs associated with the uncertainty in each design decision. The designer must decide whether the impact of reducing uncertainty—by updating preference functions, design models, or design parameters—is worth the cost of information gathering (Wood, 1996; Wood and Agogino, 1996a).

The fundamental theory behind IRTD is the decision-analytic notion of the expected value of information (EVI). A design decision must be formulated as a nonlinear programming problem, with both objectives and constraints, as well as probability measures on the uncertain variables or parameters. EVI is a relatively simple concept, representing the expectation of the



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 13
--> Information in the Design Process Alice M. Agogino Department of Mechanical Engineering University of California at Berkeley Intelligent Real Time Design (IRTD) is a framework that explicitly considers the cost of resources consumed during the design process (e.g., the value of the design team's time; the cost of further information gathering, analyses, or experimentation; and "time-to-market" opportunity costs) and trades these off against the optimum of the design solution found. Such a process dynamically takes into account the "real-time" aspect of the design process. IRTD helps designers focus their information-gathering efforts on reducing uncertainty in areas of the design that have the greatest impact on the design objectives (Bradley, 1993; Bradley and Agogino, 1991, 1992, 1993, 1994). Uncertainty and ambiguity are greatest in the early conceptual stages of design, where the greatest potential for design improvement lies. Sources of uncertainty can be found in the design constraints, objectives, and parameters; evaluation models; and customer preferences. IRTD provides a prescriptive methodology for transforming the design space over the course of the design process. The results of an IRTD analysis are a set of expected costs associated with the uncertainty in each design decision. The designer must decide whether the impact of reducing uncertainty—by updating preference functions, design models, or design parameters—is worth the cost of information gathering (Wood, 1996; Wood and Agogino, 1996a). The fundamental theory behind IRTD is the decision-analytic notion of the expected value of information (EVI). A design decision must be formulated as a nonlinear programming problem, with both objectives and constraints, as well as probability measures on the uncertain variables or parameters. EVI is a relatively simple concept, representing the expectation of the

OCR for page 13
--> objective function over the uncertain variables with the new information, minus the expected value of the objective with the current state of information. Definition: Expected Value of Information (EVI) is the expected value of the objective function with the new information minus expected value of the objective given the present state of information In equation form (Wood, 1996): it is where vj is an uncertain design variable, obj is the value of the objective function, dec* is the current (best) decision, decj is one of the set of possible decisions, c is the constraint vector, and P(vj, c) is the probability of the value of the design variable and constraint set. Given the mathematical basis for the IRTD method, one might question its usefulness in conceptual design, where analytical models often are limited to those that can be written on the ''back of an envelope." Although the "envelope" in today's world of information technology is more likely to be an e-mail message, a document written on a word processor, a notation in a design database, a CAD drawing, or a sketch in a graphics program, the information in conceptual design is not likely to be in a form convenient for applying IRTD. Research on case-based reasoning, "data mining," and information retrieval (Dong and Agogino, 1996; Dong et al., 1995; Varma et al., 1996; Wood, 1996; Wood and Agogino, 1996a) is providing the tools needed for exploiting this structured and unstructured design information. These algorithms and techniques provide the theoretical engine for the UC Berkeley Concept Database Project (Bradley et al., 1994), aimed at supporting the conceptual design process by (1) identifying good prototype designs, (2) obtaining or deriving engineering models and uncertainty measures associated with the models, and (3) evaluating the value of reducing the uncertainty in the model (applying IRTD). The Concept Database methodology provides "smart navigation" through a hypermedia database of linked design concepts. Such a system expands the

OCR for page 13
--> amount of information available to the designer and other members of the product development team in a manner that is selective, using search techniques that are based on heuristic, deterministic, and decision-analytic methods. Prescriptive design methodologies are used to develop the structure of the search in order to guarantee that the full range of life-cycle issues is considered and to provide a framework for making design decisions that take into account multiple and often conflicting life-cycle objectives. Mechatronic design, the domain of our first prototype system, is an area of key strategic importance, playing an important role in the design of consumer products, computer peripherals, automation systems, and aerospace and defense systems. It is also an area where off-the-shelf component selection is essential and is one in which expertise crosses departmental boundaries: electronic design, mechanical design, interface design, supplier coordination, packaging, logistics, and support all are involved in the realization of a good mechatronic system design. Research from the Concept Database Project has been integrated also into the curricular reform efforts of the NSF-funded Synthesis Engineering Education Coalition, providing tools for developing educational case studies of engineering design and a database for archiving, searching, and retrieving engineering courseware (Agogino and Wood, 1994; Wood and Agogino, 1996b). Acknowledgments I wish to acknowledge the team of students I have worked with over the past 5 years, who have contributed to various aspects of this research, as listed in the References below. Steve Bradley was the first to suggest that my previous work in using information value theory in diagnostics and supervisory control might be modified effectively for use in real-time design environments, and he developed the first IRTD framework. Bill Wood brought in the notions of case-based reasoning, data mining, and information retrieval. Andy Dong continued this work by developing text analysis algorithms for constructing design representations for "smart drawing" documentation and concept retrieval. Anil Varma is extending this work to include neural network approaches to automated design classification, association, and retrieval. References Agogino, A. M., and W. H. Wood III. 1994. The Synthesis coalition: Information technologies enabling a paradigm shift in engineering education. (Keynote Address). Pp. 3-10 in Hyper-Media in Vaasa '94: Proceedings of the Conference on Computers and Hypermedia in Engineering Education, M. Linna and P. Ruotsala, eds. Vaasa, Finland: Vaasa Institute of Technology). Bradley, S. R. 1993. Design optimization under resource constraints. Ph.D. dissertation. University of California at Berkeley.

OCR for page 13
--> Bradley, S. R., and A. M. Agogino. 1991. Intelligent real time design: Application to prototype selection. Pp. 815-837 in Artificial Intelligence in Design '91, J. S. Gero, ed. Oxford, England: Butterworth-Heinemann Publishers. Bradley, S. R., and A. M. Agogino. 1992. Optimal design as a real time AI problem. Pp. 629-638 in System Modeling and Optimization, P. Kall, ed. Lecture Notes in Control and Information Sciences 180. New York: Springer-Verlag. Bradley, S. R., and A. M. Agogino. 1993. Computer-assisted catalog selection with multiple objectives. Pp. 139-147 in Proceedings of the ASME 1993 Design Theory and Methods Conference. Bradley, S. R., and A. M. Agogino. 1994. An intelligent real time design methodology for catalog selection. ASME Journal of Mechanical Design 116:980-988. Bradley, S. R., A. M. Agogino, and W. H. Wood III. 1994. Intelligent engineering component catalogs. Pp. 641-658 in AI in Design '94, J. S. Gero and F. Sudweeks, eds. Norwell, Mass.: Kluwer Academic Publishing, URL: http://hart.ME.Berkeley.EDU/~best/papers/AID94_paper/AID94.html Dong, A., and A. M. Agogino. 1996. Text analysis for constructing design representations. Pp. 21-38 in Artificial Intelligence in Design '96, J. S. Gero and F. Sudweeks, eds. Norwell, Mass.: Kluwer Academic Publishers. Dong, A., F. Moore, C. Woods, and A. M. Agogino. 1995. Managing design knowledge in enterprise-wide CAD. Pp. 329-347 in Advances in Formal Design Methods for CAD, J. S. Gero and F. Sudweeks, eds. Preprints of the IFIP WG 5.2 Workshop on Formal Design Methods for CAD. Sydney: Key Centre of Design Computing, University of Sydney. Varma, A., A. M. Agogino, and W. H. Wood III. 1996. A machine learning approach to automated design classification, association and retrieval. Pp. 429-445 in Artificial Intelligence in Design '96, J. S. Gero and F. Sudweeks, eds. Norwell, Mass.: Kluwer Academic Publishers. Wood III, W. H. 1996. Supplying concurrent engineering information to the designer: The conceptual design information server. Ph.D. dissertation. University of California at Berkeley. Wood III, W. H., and A. M. Agogino. 1996a. A case-based conceptual design information server for concurrent engineering. Computer-Aided Design (CAD) 28(5):361-369. URL: http://hart.ME.Berkeley.EDU/~best/papers/CAD_paper/CAD.html Wood III, W. H., and A. M. Agogino. 1996b. Engineering courseware content and delivery: The NEEDS infrastructure for distance-independent education. Journal of the American Society for Information Science 47(11):863-869.