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. Conceptual design decisions can be challenging because the imprecise nature of design concepts make them difficult to model. Prior literature exists on using Pareto sets to model design concepts abstractly in the space of decision attributes. However, this approach has limitations when the concept under consideration is a component of a larger system. The need to relate component-level decision attributes to system-level decision objectives can lead to a violation of the assumptions underlying classical Pareto dominance. The main contribution of this article is a new dominance criterion, called parameterized Pareto dominance, which is applicable in such situations. It is a generalization of the classical dominance rule and is found to be sound from a decision-theoretic perspective. A secondary contribution is the articulation of a generalized methodology for constructing concept models based on classical or parameterized Pareto sets using either observational or model-generated data. The modeling procedure, including the new dominance criterion, is demonstrated on observational data about hydraulic cylinders. The question of whether a parameterized Pareto set can be an adequate representation of a component-level design concept is evaluated on a gearbox conceptual design problem for which the component-level decision is subordinate to system-level vehicle design problem.