1.0 INTRODUCTION
Preliminary design is defined to be the initial stage of understanding and formulating the behavior of a potential artifact. Little research and few computer tools are available for preliminary design, especially when compared to the sophistication and availability of numerical analysis tools available for the later stages of design, such as those simulating material flow or performing stress analysis.
The reason for so little research is that the early stages require a range of skills that one might call the "experience" of the engineer. The nature of computation is such that the broad experience that would be called for in preliminary design is difficult to capture and manipulate. In contrast, the numerical models available for the later stages of design are based on more strictly focus, formal theories, which can be easily interpreted for computational purposes.
The difference between the experiential nature of preliminary design and the formal theories available later in the design process point to a fundamental difference in the use of knowledge and the characterization of intelligence. Intelligence, as promulgated within the stringently formalist views of academic settings, depends on abstractions and categories to be manipulated by formal models. In formalist terms, theory has primacy over experience. Formal theories provide predictive capabilities, and so serve to provide an understanding of what is real. The basic paradigm of computing and of formal thinking is the abstraction of features from some background of data, the formation of rules and hypotheses, and the eventual distillation of a theory.
Given the chain from the background of experience, to observed data, to the construction of rules, and to the development of theory, an alternate view is to invert the status of experience to primacy. Experience is placed at the top. The justification of this is simple. The act of observation is the act of categorization and an abstraction away from the experiential background. In noting certain features and subsequently classifying, attention is shifted immediately away from the value to the thing known as the type, which is a model of that value. However, the categorization of data - the modeling of values - can abstract away information that may be valuable to other perspectives [Wegner 91]. This issue of robustness and the loss of information during modeling may seem obvious, but it is pivotal in the considerations of experiential and formal models of engineering design.
There is a loss of global perspective with formal models. This loss may be inconsequential during latter stages of design, when the components are largely fixed and it is accepted that potential changes would have only a localized effect. But in preliminary design, the artifact as a whole is not merely the sum of its features. There are inter-related concerns that are best resolved using the experience of many people, including the design and manufacturing engineers, as well as accountants and marketers.
The scope of general models of reality can be described as dictionary and encyclopedic [Wegner 91]. The dictionary models, like numeric simulation tools, are localized and limited in their focus and generally have proven computability. Encyclopedic models, of which preliminary design would have to avail, are global in nature and have a complexity which may forbid computation. There simply is no way to formalize and use the encyclopedic experience of one engineer, let alone the required and simultaneous experience of a whole engineering team.
Comprehensive engineering in the preliminary stages of design must make use of engineering experience distinct from theoretical models. The design environment must provide common ground upon which the background of experience and the abstractions of formal, theoretical models may interplay. In this environment, the initial results of preliminary design should assist the later stages of design.