Modularity in design: formal modeling and automated analysis

• Authors: Kevin J. Sullivan;Yuanfang Cai
• Venue: Modularity in design: formal modeling and automated analysis
• Year: 2006

Reasoning about the evolvability properties and economic implications of design structures is critical to high-consequence decision-making, but it remains difficult. One key impediment is the lack of analyzable high-level design representations that both convey design architectures and enable designers to reason precisely about their modularity properties and economics. This dissertation contributes such a formal and analyzable representation. It supports formal design modeling and enables automation of a number of evolvability and economic-related analyses. We model both design decisions and relevant external conditions using an augmented form of constraint networks (ACNs). To support design impact analysis, we derive an intermediate, state-machine-based design space model from an ACN, which we call a design automaton (DA). To support traditional design coupling structure analysis, we derive a pair-wise dependence relation (PWDR) from a DA, based on which we can then derive design structure matrices (DSMs) to make use of a number of existing economic and engineering analysis techniques. We created a divide-and-conquer approach to addressing scalability issues in constraint solving and solution enumeration that a DA requires. We extend the ACN model into the complex augmented constraint network (CACN) to support the modeling and analysis of design decisions with structural impacts. This dissertation provides evidence to support the following thesis: this framework formally accounts for the key concepts of important but informal theories, enables automation of a range of formal architectural analysis methods related to evolution and economic value, and generalizes to provide an account of both object-oriented and newer aspect-oriented notions of modularity in a unified, declarative framework.