Composing Tradeoff Models for Multi-Attribute System-Level Decision Making Available to Purchase

In this paper, we study the prospects for modeling a system by composing tradeoff models of its components. A tradeoff model is an abstract representation of a system in terms of a predictive relationship between its top-level attributes. Designers can use this to predict the attributes they would achieve if they implemented the system. Prior approaches to generating tradeoff information are incompatible with model composition due to their reliance on classical Pareto dominance. We show that by using a generalization of this, called parameterized Pareto dominance, designers can produce tradeoff models that they can compose validly. The focus of this paper is on analyzing the modeling approach mathematically. The main result is proof that, under mild assumptions about how component-level attributes relate to system-level attributes, the approach is mathematically sound from a decision-theoretic perspective. The paper also includes a demonstration of the approach on the design of a hydraulic log splitter system using hydraulic component tradeoff models based on data about commercially-available components.