Concept evaluation in structural design utilizing multiattribute utility theory and a knowledge-based approach.

摘要

The design process is considered a two-stage process. It includes a conceptual design stage, and a detailed or numerical design stage. The objective of the conceptual design stage is the development of a class of concepts; where a concept is defined as a solution in its most general form described by a set of qualitative and quantitative attributes.; In the past, the computerization of engineering design has been limited to the numerical stages of the design process. The present trend is to develop formal models of all stages of the design process, including the conceptual design stage, and to build computer programs based on these models as decision support tools.; A large number of concepts can be generated in the conceptual design stage. Therefore, there is a need for both systematic and logically defensible evaluation methodology and its computerized implementation. The evaluation method proposed in this study utilizes Multiattribute Utility Theory and a knowledge-based approach. A knowledge base is developed to provide designers with domain-specific data for the considered concepts while Multiattribute Utility Theory is used for the numerical assessment of the overall goodness of these concepts.; The feasibility of the proposed evaluation method and its computer implementation has been demonstrated and tested. It has been done in the area of structural engineering, and is related to the design and construction of steel roof trusses in industrial buildings. Twenty basic types of steel roof trusses were considered for evaluation. The evaluation was conducted considering various ranges of design variables which allow to study their influence on the overall worth and rank-ordering of the evaluated roof trusses. Based on the study and analysis conducted on the evaluation results, it has been observed that there is an 85% agreement among experts regarding the best five types of these roof trusses, and a 95% agreement among them concerning the five worst types. Moreover, when the results of application of the proposed method were compared with the results obtained using a single overall assessment of worth, a better agreement between the experts regarding both the most and least worth types was obtained.