Analyzing the dynamic behavior of marine design tools using network theory

• A network metric has been developed to comprehend the behavior of design tools.
• The metric is the network equivalent of a first order Taylor series expansion.
• The metric is a lead indicator, providing the relative importance of interactions.
• Two formulations are compared against a full factorial design of experiments.
• Results demonstrate acceptable accuracy in predicting design tool behavior.

This paper introduces a new network metric, termed path influence, to analyze the dynamic behavior of marine design tools. Design tools are increasingly becoming opaque, if not outright black boxes, and engineers often do not have the resources to intuitively understand their behavior. Network representations of marine design tools have been created and analyzed in previous research, providing static insight into designer intent and tool formulation; this paper extends the research to include dynamic behavior. Two network weighting schemes were developed that enable path influence to approximate the impact of variable changes on an entire tool formulation. The first scheme utilizes partial derivatives and is demonstrated to be equivalent to a first order Taylor series expansion. The proposed approach was applied to the Sen Bulker problem, and then compared with a full factorial design of experiments. The benefits and drawbacks associated with path influence are discussed throughout the paper.