Research and Advances
Architecture and Hardware

Thinking About Computer Systems to Support Design Synthesis

What should the CAD systems of the future be like?
  1. Introduction
  2. Guidelines for Design Support Systems
  3. References
  4. Authors
  5. Footnotes
  6. Figures

Designers currently have access to numerous computer systems that aid them in the different stages of the design process, such as the analysis-synthesis-evaluation, and throughout the phases a design goes through: conceptual design, embodiment design, and detailed design. But whether it is possible to develop a computer system for creative design that can automatically invent new solutions is still controversial. Some researchers believe the very nature of creativity is so unpredictable it cannot be explained in detail or envisaged by a computer system, although many believe it is feasible for future computer-aided design systems (CAD) to be expanded in order to support an interactive process with the designer [2, 3, 6]. Other researchers defend the potential capacity of computers to reach novel solutions through far different mechanisms from those humans use to be creative [1, 5].

In our opinion, many design process mysteries must be solved before a computer system can be developed that aids designers significantly more than current CAD systems. Furthermore, as far as we know, no autonomous design systems have been developed to date that offer the same degree of satisfaction as working with a human designer. Experimental research into how designers design can, therefore, provide us with valuable knowledge to help in developing new computer-aided design systems.

We attempted to uncover some of the mysteries of the design process by developing the MADIS project. The ultimate purpose of this project is to implement an architecture capable of supporting a design process that interacts with the designer, provides knowledge management capabilities, and stimulates creativity—while also simplifying collaborative work. Our first steps were to carry out experimental research involving the systematic observation of several groups of designers to determine how they generate solutions in the conceptual design phase, and how their effectiveness varies according to the means of expression used (words, drawings, or objects). This research work provided us with an initial idea of what functions future systems ought to include, and what the interrelation between user and computer should be like; it also enabled us to establish a series of guidelines with which to orient the design process. Although creative design is unpredictable, following certain guidelines can lead to more effective design.

Based on our review of the literature [3, 10, 12] and our research, we believe future design support systems must be capable of performing the following functions (see Figure 1):

  • Supporting and visualizing the design as it is carried out.
  • Stimulating creativity.
  • Knowledge management.
  • Allowing collaborative work.
  • Designing interactively with the user.

CAD systems were originally intended to serve as a platform on which to develop designs graphically. In recent years, however, these systems have advanced from 2D to 3D representation and now include different modules such as knowledge-based engineering (KBE) and computer-aided engineering (CAE), as well as tools that allow designers to work in collaboration with others.

Being able to visualize a design plays a fundamental role in the design process. Some studies [11] have suggested that in addition to acting as a means of communication, 2D and 3D visualizations are in fact an intrinsic part of the design process. They help to expand the designer’s short-term memory and make it easier to come up with solutions. But 3D visualization that simulates the actual handling of objects also makes the design process more effective.

Perhaps the most important point is that during the conceptual design phase, drawings can still be done by hand much faster than with computers. This limitation of computers has negative repercussions on the cognitive capacity to solve the design problem. To improve this situation, progress is being made in the definition of new interfaces and devices, such as pen-based input devices, expressive sketching techniques, and electronic whiteboards.

Another important capability needed in future CAD systems is the reconstruction of freehand sketches, which must allow for the application of operations such as partial erasure and the definition and modification of parameters, as if they had been introduced using the traditional software interface right from the outset.

Much of the theoretical and development work in the area of pen interfaces capable of tasks such as recognizing gestures and inferring 3D from 2D outline sketches has been accomplished by Mark Gross, Ellen Do, and others, under projects including Digital Clay, Digital Sandbox, Electronic Cocktail Napkin, and Gesture Modeling.

The first condition computer software for design support must meet is not being an obstacle to creativity. A good atmosphere for creative processes does not ensure creativity, but being creative does become far more difficult if the computer tools being used fail to promote a favorable environment because they are too complex, too rigid, or too slow.

A more significant requirement is that the design support system must also stimulate creativity. Although it is possible to find several computer programs aimed at stimulating creativity, they have been conceived and projected independently from other CAD tools. One of the research issues currently receiving much attention in engineering design is the selection of creative methods and techniques according to the stage of the design process, and the integration of both applications in the same system.

The design process also entails the acquisition and management of large amounts of knowledge. To obtain, understand, and adapt large amounts of data, and then manage data flow throughout the design process is extremely difficult without some kind of methodology. KBE is a methodology that allows us to organize the data flow and to develop an architecture under which design automation can be effectively implemented. KBE systems create, or in some way manipulate, detailed geometries and employ a method of representing knowledge that can be interpreted by the computer.

KBE packages such as ICAD and Intent have been applied to design automation in projects involving large amounts of knowledge and the handling of complex geometries. These turn out to be very sophisticated systems and are not directly linked to graphics software, which has led to their being basically (although not exclusively) limited to applications in the aeronautical or the automotive industries.

The latest generation of CAD systems is beginning to incorporate the tools usually found in a KBE system that expand the possibilities of parametric design by including “if-then” sentences, restraints, checks, and the use of spreadsheets to collect information. The possibilities of automation offered by these modules has aroused some curiosity within SMEs, although interfaces need some simplifying to make them more accessible to traditional CAD software users with no specific KBE training.

Another question related to the integration of the two systems, CAD and KBE, is whether CAD systems must acquire the functionality of a KBE system in order to support various aspects of the product life cycle or, conversely, whether CAD systems should be built with a completely open architecture to allow easy integration of a wide range of KBE software [7].

One can suppose that today’s CAD, CAM (computer-aided machining), and PIM (product information management) systems will become integrated product life cycle (PLC) systems, and indeed, a number of applications have already been developed that go in this direction. Additional PLC elements ought to be added to knit together the domains of CAD, CAM, CAE, and PIM. These will include, for example, information management systems for requirements and costing, as well as the use of warranties and financial information. The resulting architecture will allow the creation of an effective collaborative engineering environment that comprises the whole product management team. Kvan and collaborators [8] have noted the importance of using visual and textual communication and also virtual environments. Different capabilities will be included to enable the use of video conferencing, chat-line communication, electronic whiteboards, and other collaborative devices. We believe the best approach would be to develop systems that interact with the designer, instead of working independently, because the cognitive processes that take place during conceptual design are very unpredictable and without human involvement only the more obvious design solutions are reached.

Computational algorithms would be responsible for proposing solutions; the designer would process this information off-line in order to redefine the problem, and then feed it back into the computer and refine the design space. It would also be important for the design team to take part in the evaluation and the decision making concerning the concepts generated by the computer as a result of the computational procedures.

Back to Top

Guidelines for Design Support Systems

After a series of experimental sessions involving 12 groups of four designers, the design process protocols used by each group were analyzed to evaluate whether the ideas put forth by each participant were effective or not. At this point it should be noted that very little experimental research has been conducted on the design process using statistical techniques to check for the existence of repetitive patterns linked to a particular factor; such studies would enable researchers to draw up guidelines or recommendations that increase effectiveness.

The ideas were evaluated in terms of their quantity, and their fulfilment of the initial requirements. If the idea contradicted the initial requirements it was considered invalid, if it responded to the initial requirements without contradicting them it was considered valid, and if it did not meet these requirements it was cataloged as a non-related idea. Ideas were aggregated on a more general level (called an alternative). This aggregate included those ideas belonging or contributing to a potential design solution.

The representation of the FBS (function-behavior-structure) model of the experiment was also obtained to evaluate the progress of the functions and specifications of the design, and their relationship with the solutions generated. Taking the design protocol as our starting point, the different design entities of the FBS model can then be identified: the functions (F) describing the goals; behavior (B), which describes the changes in the state of the structures; and finally the structures or solutions (S) [4, 10].

To explain how the computer system would be able to design on an interactive basis with the designer, we made use of Boden’s classification of computational models of artificial intelligence for creativity in combinational methods and in exploratory-transformational methods [2]. Combinational methods generate novel ideas by making unusual combinations between ideas that share some similarities. The exploratory-transformational models include those based on heuristic searches and those based on evolutionary techniques. In the protocols of our experiment, creative computing methods were identified as being those in which the computer system generates ideas in the same fashion as designers.

We also set out with Takeda’s computational model of synthesis [9], the core of which consists of modelling the synthesis as an abduction process, according to which solutions stem from the requirements and functions that are selected from the problem and from the available knowledge. In turn, the analysis consists of using a process of deduction to derive new requirements and functions from the solutions and the knowledge obtained.

To illustrate how the design process works, we could use the following metaphor. Let’s suppose we have a large bag of balls that contain pieces of knowledge (we’re going to call them ideas). The design problem can be broken down into functions and requirements for which we have to provide a solution. We would take a ball from the bag (which corresponds to the exploratory model using Boden’s classification). If it fully satisfies our partial problem and does not conflict with the rest of the problem, then we would be finished (we would say it is an alternative). If the idea is of no use to us we would put it back in the bag, but if it helps to partially solve our problem, we would keep it out of the bag and go on removing balls, combining them with one another (combinational), or we would bring about changes in them (transformational) until we finally reach a valid alternative.

Our research shows that design is more effective when the number of alternatives is lower compared to the total number of ideas, and when these alternatives are more interrelated. These circumstances are favored when designers work with objects. When they only use concepts or static images, the creative stimulus to form alternatives is much lower. The effectiveness obtained in our experiment when conceptual design involves objects reaches 80%, yet it drops to 22% when conceptual design only makes use of verbal concepts [11].

This difference, which underlines the importance of objects in design systems, is produced for several reasons. First, there is a greater number of analysis-synthesis-evaluation cycles being carried out iteratively as new ideas are proposed; second, the partial solutions combine with one another from the outset, when solution principles are still being obtained without waiting for all the partial solutions to be produced; and third, the design solutions are more complete, that is, they provide an answer to a greater number of design functions.

Our research suggests computer systems should not only allow objects to be visualized and manipulated, but also offer the designer new alternatives for the different functions by search methods later completed by transformation procedures. These alternatives would be interrelated by combination procedures, and also with the proposals put forward by the designer himself.

Findings from our study suggest the analysis-synthesis-evaluation processes can be guided by selecting the right methods for obtaining ideas. Some of the tasks that could be carried out by the design support computer system in order to achieve a more effective design would include (see Figure 2):

  • Focusing the initial computer search for ideas on just a set of functions chosen by the designer as key functions the design must fulfil. It is important to avoid incorporating large numbers of additional functions at the beginning.
  • Combining partial solutions from the outset and throughout the entire design process by means of transformational methods.
  • Applying a higher proportion of transformational procedures than other types of procedure (exploratory and combinational).

Likewise, the computer system could perform several different tasks to guide the actions carried out by the designer, which would be achieved by:

  • Encouraging the designer to solve the initial functions and offer the partial solutions that have already been obtained so the designer can select the most suitable and combine them with one another from the outset.
  • Fostering designer associations by displaying on-screen the solutions that have already been obtained.
  • Keeping the designer working with a small number of alternatives and alerting the designer when it might be wise to work with another alternative or to seek a new one, after generating a reasonable number of ideas within the same alternative.

When the objective is originality, rather than effectiveness, a higher number of exploratory procedures must be applied and the designer will have to introduce a greater number of additional functions in order to increase the probability of generating novel ideas.

Back to Top

Back to Top

Back to Top

Back to Top


F1 Figure 1. Functions for design support systems.

F2 Figure 2. Guidelines for designing with a computer.

Back to top

    1. Bentley, P. and Wakefield, J. Conceptual evolutionary design by genetic algorithms. Engineering Design and Automation Journal 3, 2 (1997), 119–131.

    2. Boden, M. Computer models of creativity. Handbook of Creativity. R. Sternberg, Ed. Cambridge University Press (1999), 351–372.

    3. Candy, L. and Edmonds, E. Creativity enhancement with emerging technologies. Commun. ACM 43, 8 (Aug. 2000), 62–65.

    4. Gero, J. Design prototypes: A knowledge representation schema for design. AI Magazine 11, 4 (1990), 26–36.

    5. Indurkhya, B. Computational Modelling of Mechanisms of Creativity. Department of Computer Sciences, Tokyo University of Agriculture and Technology. Japan, Tokyo, 1998.

    6. Parmee, I. and Bonham, C. Towards the support of innovative conceptual design through interactive designer/evolutionary computing strategies. Artificial Intelligence for Engineering Design, Analysis and Manufacturing (AIEDAM) 14 (2000), 3–16.

    7. Penoyer, J., Burnett, G., Fawcett, D. and Liou, S. Knowledge based product life cycle systems: Principles of integration of KbE and C3P. Computer-Aided Design 32 (2000) 311–320

    8. Schabel, M.A., Kvan, T. Design, communication & collaboration in immersive virtual environments. International Journal of Design Computing 4 (2002).

    9. Takeda, H., Yoshioka, M. and Tomiyama, T. A general framework for modelling of synthesis—integration of Theories of Synthesis. In Proceedings of ICED `01, (Imeche, Glasgow, 2001) 307–314.

    10. Ullman, D. Toward the ideal mechanical engineering design support system. Research in Engineering Design 13 (2002) 55–64.

    11. Vidal, R., Mulet, E. and Gómez-Senent, E. Effectiveness of the means of expression in creative problem solving in design groups. Journal of Engineering Design 15, 3 (2004).

    12. Wiegers, T., Horváth, I., Vergeest, J., Opiyo, E., and Kuczogi, G. Requirements for highly interactive system interfaces to support conceptual design. Integration of Process Knowledge into Design Support Systems. H. Kals and F.V. Houten, Eds. Kluwer Academic Publishers, Enschede (1999), 69–78.

    This research is funded by Spanish Ministry of Science and Technology project DPI2002-04357-C03-01 and by Bancaja project P1 1B2003-36.

Join the Discussion (0)

Become a Member or Sign In to Post a Comment

The Latest from CACM

Shape the Future of Computing

ACM encourages its members to take a direct hand in shaping the future of the association. There are more ways than ever to get involved.

Get Involved

Communications of the ACM (CACM) is now a fully Open Access publication.

By opening CACM to the world, we hope to increase engagement among the broader computer science community and encourage non-members to discover the rich resources ACM has to offer.

Learn More