Keywords: design, software decision support, evolutionary computing, knowledge-based systems, case based reasoning.

Conceptual Design is generally accepted as being the part of the design cycle where the main decisions about what is to be designed and built are made. Current conceptual design practices are largely unaffected by the IT revolution, despite the fact that the research community has spent some two decades trying to develop techniques. The paper looks at the work that has been done during that period with especial emphasis on the Architecture, Engineering and Construction (AEC) industry and in particular the design of bridges and buildings.

The initial parts of the paper define exactly what is meant by conceptual design, placing it between client briefing and preliminary design with all stages having a degree of overlap with one another. Typically during conceptual design, there is a high degree of uncertainty and lack of definition. Humans have a need to use external memory aids and also need graphical and other forms of representation to communicate ideas to each other. The main form of representation currently used in conceptual design is sketching.

The difficulties of introducing new technology are discussed. If one develops computer aided tools to support conceptual design, should these just enhance current practices or should the best use of the technology be made, even though this will probably require a re-engineering of the process. The latter inevitably leads to resistance from practitioners and can severely hamper the introduction of new technology. However, current practices are restricted by human cognitive limitations and there is significant room for improvement.

The main body of the paper is a review of the main computing techniques which the research community have applied to conceptual design. Initial attempts made use of expert system / Knowledge Based System (KBS) technology using mostly <condition><action> production rules. These require a considerable amount of knowledge elicitation and knowledge engineering to produce the rule base. They also possesses a style of user interaction which is shown to be unsuitable for conceptual design. More recent research is reported which has led to the development of knowledge bases which can be built up and maintained in an easy and incremental fashion by users, thus avoiding the effort of knowledge elicitation. Also the style of interaction of this new style of KBS is specifically formulated to suit the needs of conceptual design.

The next section looks at Decision Support Systems (DSS). Two systems are discussed in this section, one which uses a visually based form of constraint handling and the other which has a high level of accuracy, thus effectively re- engineering the design process. There then follows a brief discussion of Case Based Reasoning.

Evolutionary computing based decision support is covered by comparing three systems, all of which are for the conceptual design of office buildings. The three systems are interesting because although they all support design in the same domain there are some significant differences between them. Each system has some features that are particularly innovative but also the differences show that there is no unique approach to the problem. All three systems show that evolutionary decision support is a good approach to conceptual design and the information requirements of such systems are considerably less than, for example, KBS and they are more successful than the other tools considered above.

There next follows a section on concurrent conceptual design. Little work has been done in this area to date although there are some research projects which are actively looking at its needs.

The final part of the paper discusses the use of evolutionary computing for shape discovery. A variety of primitives are looked at including voxels and shape grammars and an unusual technique where there is no genotype. It is concluded that the most promising approach to date has been one using genetic programming.

The overall conclusion is that for domains where the search space is relatively small, DSS are a good option. For more complex domains with large search spaces then evolutionary decision support techniques are far superior to anything else and overall offer the greatest promise. Some hints are given about areas which are currently neglected by the research community.

purchase the full-text of this chapter (price £20)

go to the previous chapter
go to the next chapter
return to the table of contents
return to the book description
purchase this book (price £92 +P&P)