Using Parameterized Pareto Sets to Model Design Concepts

Abstract

The decisions designers make during conceptual design can have a large impact on the
success of a project. Although prior literature exists on making decisions during
conceptual design, the reported approaches each have limitations in certain contexts. A
gap exists between approaches that rely on qualitative models of a concept and those that
assume a detailed model already exists (or can be developed easily). In this article, we
investigate a novel approach to modeling design concepts quantitatively based on the
principles of decision making, data mining and predictive modeling. This approach
includes two main innovations. First, we combine a decision-based dominance analysis
step with traditional data mining and predictive modeling techniques. This yields models
that more accurately represent a design concept. Second, we introduce a generalization
of the classical Pareto dominance criterion, which we call parameterized Pareto
dominance. The new criterion addresses limitations of existing design concept
representations, allowing one to combine a design concept model with other models in a
decision problem. We demonstrate the application of dominance analysis using data for
commercially-available hydraulic cylinders and test whether a model fit to parameterized
Pareto set data is an adequate representation of a design concept using a gearbox
conceptual design problem.