Decomposition for optimal synthesis in a parallel computing environment.

摘要

Many practical problems in science and engineering require extensive simplification and abstraction in order to be tractable in currently existing computing environments. Obvious examples are: computational fluid dynamics; nuclear and plasma physics; petroleum reservoir modelling; computer graphics and image processing; and structural synthesis. This research will be useful in these (as well as, other) application fields, but structural synthesis has been chosen as the area of application. The research has been inspired by a philosophy of design called multilevel decomposition, which enables solution of these problems in a reasonable time. In this approach a complex system (for example, a modern aircraft or automobile) is decomposed into several manageably smaller subsystems. These subsystems are solved independently without losing integrity with the main or parent system and ultimately achieving satisfactory results for the large scale system. In addition to making problem solution more tractable, the decomposition approach is compatible with a typical design office multidisciplinary organization and the parallel or distributed computing technology existing today. Several example problems (including classical problems in the field and practical applications from industry) have been used to check the validity of the approach.