Recent Progress on Numerical Wind Tunnel at the National Aerospace Laboratory, Japan

Abstract

Numerical Wind Tunnel (NWT) has been used in various computational fluid dynamics (CFD) computations at NAL. Most of applications use grid numbers of 0.1 to 2 millions. This order of problem size mainly comes from the past experiences of 3D NS computations. Therefore, one to 6–8 PEs are enough to run these codes. Since the parallelizing efficiency of NWT is linear up to the maximum node numbers and use of more PEs simply reduces computing time linearly as contrasted to other parallel machines. Because the problem size is not fine enough to obtain quantitatively accurate viscous flow field of high Reynolds number even for turbulent ideal gas flow past flat plate, more large size computation will be made in near future. Similar type of high performance computer with vector pipelines and large shared memory will be needed to suffice such demands. National Aerospace Laboratory (NAL) of Japan is the only one government research institution for the research of aerospace engineering and services of major R&D facilities such as lowspeed/transonic/supersonic/hypersonic wind tunnels and computer systems for national and civil aerospace programs of JAPAN. In the past, NAL has developed large scale scientific computer systems that are necessary for aerospace technology research but not available as a commercially available machine. 3-D Navier-Stokes computation became real thing on NS-I, although a typical computation required several to tens of hours of CPU Time. Based on this experience, a feasibility study to develop the next computer system, “NUMERICAL WIND TUNNEL (NWT)” started in 1989. The feasibility study concluded that machine with 100 times faster than VP400 can be built by using architecture of parallel vector processors.