Development of an Intelligent Genetic Design Tool (IGDT) for Use in Architectural Engineering

This paper describes the application of an Intelligent Genetic Design Tool (IGDT) in the design of architectural, structural elements. As a computer design aid an IGDT is innovative in its intelligent interaction with the designer. By always submitting multiple solutions for review by the designer, it is less likely to cause design fixation than other optimization techniques, and allows the user greater range in exploring hard-to-code design criteria such as aesthetics. As an example, the design of a cantilever truss is briefly explored. Using the coded optimization criterion of weight, and the designer's non-coded criteria of visual aesthetics and performance, a series of possible designs are explored. The ability of an IGDT to intelligently respond to the designer's preferences in a way that promotes creative thinking on the part of the designer is demonstrated. A final truss design is selected based on the use of the tool. It is concluded that an IGDT offers a significantly different approach to computer aided structural design which has the potential to enhance the user's own creativity in determining a good solution. 1 Creative Design Tool Criteria In engineering fields, accomplishing an objective with a minimum of effort, either in terms of material, time or other expense, is a basic activity. For this reason it is easy to understand the interest designers have in different optimization techniques. Mathematical, as well as, model based tools have traditionally been employed for such optimization. In recent times, mathematical methods executed on computers have become predominant. Unfortunately, computer derived solutions often obscure the range of possible solutions from the designer by only exhibiting a final, "best" solution. This not only deprives the designer of seeing alternatives, but also tends to promote designer fixation on the single, machine proffered solution (Purcell & Gero, 1996). Naturally, optimization methods can only respond to the objective parameters which are coded into the problem, and as a result, noncoded parameters, such as aesthetics, historic context, or meaning are left out of the optimization process, and ultimately left out of the final design solution. Lost too in most computer optimization, is the variety of information and creative stimulation that comes from manipulating physical models. Genetic Algorithms (GAs) are able to overcome some of the short comings which hinder traditional computer based optimization techniques in application as design tools. They can work to avoid designer fixation by presenting the designer with populations or successive generations of solutions rather than a single end result. By including interactive user selection, GAs can also incorporate non-coded design criteria into the search process. The idea of generating design forms or shapes using Genetic Algorithms (GAs) is not new. Many people are familiar with Richard Dawkins Biomorphs, described in his book The Blind Watchmaker (1986) or the graphic art of Karl Simms and William Latham (Kelly, 1995). In the area of engineering, form is closely coupled with performance. John Frazer is one who has used GAs in an engineering design sense. In an application dealing with sailing yacht design, the GA optimization considered both objective engineering parameters (stability, center of buoyancy, wetted surface area, prismatic coefficient, blocking), as well as designer criteria such as aesthetic appearance, historic tradition, allusion of form, and so on (Frazer, 1995). The Intelligent Genetic Design Tool (IGDT) described in this paper, develops many of these same ideas. 1.1 Analysis Tools vs. Design Tools In recent years a variety of computer based tools have been developed for the field of architectural engineering and design. Although originally applied to areas of computationally intensive analysis, with the ever increasing size and speed of processors, attempts have been made to develop design oriented applications as well. In the field of architecture engineering, the requirements of design tools are distinctly different from those of analysis tools. The analysis process can usually be executed in a regular, predetermined sequence of steps. The sequence may be iterative or have various paths based on the particulars of the problem, but there remains one "correct" solution for one stated problem. For example, for a given structural member, with a given load, a particular analysis will