Dekompositionstechniken zur optimalen Auslegung von Bauweisen unter Beruecksichtigung lokaler Effekte

With an increasing extent or complexity of a given problem an integral solution of structural analysis and optimization becomes more and more difficult due to both limitations in computer resources and a decreasing clarity of the modeling. By using decomposition techniques, the models will be partitioned and the determined partial solutions will then be coordinated. In the present thesis the structural behaviour is simulated by hierarchical detail analysis procedures; the optimization ist realized by means of an approximative decomposition strategy. A distributed computation, which is based on the message passing system PVM, is performed on a workstation cluster, using subtasks with a different degree of coupling for structural and optimization model. The above-named approaches form the basis of a superordinate decomposition strategy that is applied to the analysis and shape optimization of structures with local effects. The latter term describes a local, physical behaviour, e.g., in the form of stress increases. The distributed computations on the basis of decomposed models aim at both a reduction of the computation time and, in particular, the treatment of Large-Scale-Systems. The applicability and solution behaviour of the decomposition approaches for structural and optimization model are shown at two application examples. In particular, the optimal layout of a radial compressor wheel illustrates the practice-oriented use of the solution strategies. It is only the partitioned structural model that allows for the formulation of a robust model for the optimization. The use of distributed computations according to the Manager/Worker concept has also led to a substantial reduction of the computation time. Owing to its instable behaviour in the case of parameter variations, an integral modeling has proved to be insuitable for the optimization. (orig.)SIGLEAvailable from TIB Hannover: RR 2225(13-11) / FIZ - Fachinformationszzentrum Karlsruhe / TIB - Technische InformationsbibliothekDEGerman