Optimizing the Parallel Process Flow for the Individual Discipline Feasible Method
Hyung Wook Park, Min Soo Kim, Dong-Hoon Choi, Dimitri N. Mavris · 9th AIAA/ISSMO Symposium on Multidisciplinary Analysis and Optimization · 2002
For the parallel computation-based MDO methodologies such as IDF and CO, 1 a new decomposition method is presented. In this method, the design structure matrix (DSM), which is used to depict the coupling relations among disciplines, is divided into the sub-DSM groups whose number is directly equal to that of parallel computation procedures. Then, their analysis sequences were determined by a genetic algorithm in order to minimize the number of local feedback couplings in each sub-DSM group and the number of global feedback couplings among the sub-DSM groups simultaneously. Finally, two typical MDO problems are decomposed by the proposed method and their local and global feedback couplings are compared with those of the conventional method. These