Methods for Analyzing Early Stage Naval Distributed Systems Designs, Employing Simplex, Multislice, and Multiplex Networks.

Abstract

Naval ships are some of the most complex systems ever engineered. The process by which they are designed is similarly complex. The complexity and disjointedness of this process leads to the creation of disparate and incomplete ship design information created by different systems of analysis, completed by different design groups, using different tools, at different levels of fidelity. Distributed system design decisions based off this disparate and incomplete information lead to unnecessary complexity when the design is transitioned from the early design stage to the detailed design stage. This dissertation presents novel network theory-based methods for better understanding and analyzing the implications of early stage distributed system design decisions. This new method introduces network theory concepts such as degree distribution, system interdependence, and community to the field of distributed systems design as metrics for determining system robustness, as well as develops new techniques for representing physical systems as networks. Additionally, a personnel movement modeling and analysis method, derived from the network concept of betweenness centrality, is developed. This dissertation documents the first use of multislice and multiplex structures in the analysis of physical systems. System design evolutions are analyzed using multislice network structures and the interactions between systems are investigated using multiplex network structures. These two structures are combined into a novel time-dependent multiplex network structure that is developed in this work. This new structure is used to track the evolution of systems interactions. A new network complexity metric based on the concepts of planarity and network communities is created for this research in a response to lack of methods for studying the planar and near planar networks that often arise in the study of real systems. The methods presented in this dissertation do not require complex 3D CAD models or simulations. Therefore, they can be used by a single naval architect to gain insight into the implications of design decisions in the early design stages. This will result in improved naval distributed systems designs that are easier to design, maintain, and upgrade.