A knowledge based expert system for the optimization of structural design at the conceptual stage

The conceptual design stage of a building is the stage where the geometry of the building is defined, and the major building systems are selected. At present there is no analytic method for conceptual design, and usually experience and intuition are combined to produce the conceptual design. The conceptual design is particularly suitable for an expert system as it possesses the following characteristics: The conceptual design problem is ill-structured, the objective of the design is usually hard to define and quantify, the number of feasible alternative solutions to a given design problem is very large, and experts use mainly heuristics in the design process. There are two main parts in this study. In the first, the process of conceptual design is investigated. In the second, CONCEPTUAL, a computer expert system to aid the conceptual design is created. In investigating the process of conceptual design, the objectives that a designer wants to accomplish are examined and the complexity of these objectives is shown. It is argued that in practice design is an approximate optimization process. This process draws heavily from the expertise of the designer. The CONCEPTUAL expert system consists of the following components: a knowledge base, an inference mechanism, an explanation module, and a user interface. The knowledge base includes several different types of knowledge: hard constraints, soft constraints, common building patterns, costs and loads of building components, heuristics, building codes, and user\u27s preferences. A number of tools is available for developing an expert system. The suitability of these tools for this study is explored. Prolog was the tool selected for this study. The unique features of Prolog, and the structure of the Prolog CONCEPTUAL program are explained. Advanced programming issues, which are of importance to the development of expert systems are discussed. The efficiency of CONCEPTUAL depends on the accuracy of the cost estimate the program performs in the search process. A way to increase the accuracy of the square-foot estimate is proposed, and is used in CONCEPTUAL. The proposed method is based on the results of a regression analysis performed on published building cost data. (Abstract shortened with permission of author.