Evaluating complex engineered systems using complex network representations

Graduation date: 2014This thesis is the combination of two research publications working toward a unified strategy in which to represent complex engineered systems as complex networks. Current engineered system modeling techniques segment large complex models into multiple groups to be simulated independently. These methods restrict the evaluations of such complex systems as their failure properties are typically unknown until they are experienced in operation.\ud In an effort to combat the computationally prohibitive simulations required for the analysis of complex engineered systems, complex networks are used to simplify the analysis and provide data during early design when costs for design changes and associated risk are lower. The first publication presents a methodology in which to model complex engineered systems as networks so that nodes are commensurate in ontological category under a common analysis goal. The second publication identifies a model scaling technique in which to evaluate network topology metrics for an evaluation of parameterized failure performance. Each publication utilized a drivetrain model to illustrate and simulate the methods and potential results. It was found that a bipartite behavioral network is capable of consistently identifying system failures within network topology. By analyzing complex engineered systems with complex network techniques, an evaluation of system robustness can be developed in an effort to eliminate variation in system performance