Validating design-decision support models

Abstract

There is an ever growing body of work, in both academia and industry that approaches the engineering discipline under the assumption that at the heart of design is decision making. This design paradigm, known as Decision Based Design (DBD) has been embraced by many researchers in the last decade but like any research topic, there are many issues of debate. One issue that has become relevant in recent years within the DBD research community is that of validation. A large number of DBD models (tools, methods, processes) exist that are intended to help engineering designers make decisions during the design process. Validating these design-decision support (DDS) models is difficult as good outcomes of decisions do not necessarily indicate a valid process for decision making. Thus, the identification of criteria that validate the DDS models themselves is necessary. Under this motivation, the research of this dissertation looks at engineering design from the decision science perspective in an attempt to improve understanding of the concept of validation in decision making. Through a basic understanding of the components of the human decision making process and the cognitive difficulties that affect decision making, validation criteria for DDS models are proposed. The criteria represent a point from which designers can understand the limitations of DDS models used during the design process. As such, three DDS models known to industry and academia alike are investigated under the validation criteria. The DDS models are Pugh's Concept Selection Matrix, Suh's Axiomatic Design and the House of Quality. Through the investigation, the limitations and potential decision flaws that can result from these DDS models are discussed and suggestions for improvement are made in each case. The House of Quality (HoQ) represents an important DDS model of investigation in the dissertation. Beyond understanding the limitations under the DDS model validation criteria, study of the mechanics and assumptions behind the HoQ is conducted. Understanding of the HoQ mechanics through the study, both empirical and statistical, leads to the development of an improved HoQ method. The method utilizes "Design of Experiments" and a demand modeling technique known as "Conjoint Analysis". The result is a proposed methodology known as "Conjoint-HoQ". The research detailed within this dissertation relies on information from both the social and engineering sciences. The goal of exploring the overlap of these oft distinct fields through this research is two fold. First, an understanding of the decision based nature and cognitive aspects of engineering design are afforded through this work, resulting in criteria for validation of models that support design-decision making. Second, an improved understanding of design-decision support models in terms of their limitations and the impact of those limitations on the design process is detailed.