Mathematical Programming Approaches to System Partitioning

Effective engineering of large systems requires a careful delineation of major areas of responsibility for design and development. Some important considerations are critical interdependencies in performing the system mission, major data flows within the system, and similarities in technology between different parts of the system. This partitioning activity is viewed in two stages: synthesis-derivation of a reasonably small number of candidate partitions- and evaluation-more thorough investigation of detailed trade-offs. Partition synthesis is addressed by means of mathematical programming. A series of formulations are given, ranging in complexity from a simple linear transportation problem to quadratic integer programs requiring maximization of a convex function. Selection among alternative formulations depends on the actual partitioning criteria applied as well as the degree of preselection desired from the synthesis.