POD基底を用いた低次元化モデルによる解析の並列化

抄録

Although recent improvement in parallel computing enables us to perform large-scale analysis within practical computational time, it is necessary to accelerate simulations because many parametric studies are required in the product optimization design process. One of the promising approaches for reducing computational burden is reduced-order modeling (ROM). We have been working on proper orthogonal decomposition (POD)-based ROM in conjunction with empirical quadrature method, which is one of the hyper-reduction methods. The present study pursues further acceleration of simulations. To this end, we investigate efficient ROM under parallel computing environments.