The distribution of aircraft design oriented parallel algorithm
|School||Nanjing University of Aeronautics and Astronautics|
|Course||Applied Computer Technology|
|Keywords||Distributed parallel computing Substructure Modal synthesis method Matrix Storage Aircraft design|
Widely used in large-scale matrix computations aircraft design, spacecraft design , bridge design , biological sciences, computational physics and oil exploration and other fields. With the increase of the order of the matrix , the matrix solving and eigenvalue problems such as computation and storage capacity will rise sharply . Therefore, in aircraft design, the large matrix storage and computing has become an urgent problem. Thesis aircraft design large matrix generalized eigenvalue problem for generalized eigenvalues in the study of typical algorithms and distributed computing parallel algorithm based on the direct conversion method selected depth study , the method fewer communication for network-based connection distributed computing environments ; that direct transform matrix can be integrated in the process of parallelism , and the step of calculating the direct conversion method has been improved by using the first generation sub- modal stiffness ( mass ) matrix, and then assembled to form the overall modal stiffness ( mass ) matrix method can avoid the generation of large-scale matrix ; discusses the direct conversion method classification method for assembling , analyzing results : grading assembly is similar to the original problem solving process several times , the results obtained the error will be assembled with the increase of the number of conclusions ; proposes a method based on direct transformation matrix factorization algorithm can achieve load balancing among compute nodes ; matrix Storage Technology in the analysis , based on the given direct conversion method for high-level matrix storage methods ; MPI parallel program design in the study , based on the direct conversion method is designed based on aircraft design -oriented distribution of parallel algorithms , combined with mathematical packages CLAPACK implements the algorithm . Experimental results demonstrate that the proposed matrix decomposition algorithm, matrix storage method, based on an improved method for the distribution of the direct conversion parallel algorithm eigenvalue calculation precision aircraft structures for aircraft designed to meet the application requirements of large-scale projects , compared substructure space iteration method has greatly improved computing speed .