Time accurate simulation of hypervelocity base flows on massively parallel computers.

摘要

The efficient design of thermal protection systems (TPS) for vehicles entering planetary atmospheres requires a detailed knowledge of the flowfield which envelops the vehicle during a hypervelocity entry. The size, shape, and thermodynamic state of the base flow generated behind such a body determines the size of the after body and the required heat shielding. For many years experimental evidence has indicated that the base flow was unsteady, but there was not sufficient data to understand or characterize this unsteady behavior.;In order to simulate numerically the time dependent behavior of these flow fields, it would be necessary to compute a three dimensional flow field on a mesh with a level of resolution adequate to resolve the complex features observed in hypervelocity base flows. These types of computations are prohibitively expensive in terms of total CPU hours and actual turn around times.;The focus of the current research was to investigate the time dependent nature of hypervelocity base flows while developing computational techniques to allow time accurate simulations to be more routinely performed in the future.;The latter goal was addressed by exploiting the emerging power of massively parallel supercomputers. In the course of this research a parallel extension to the traditional Gauss-Seidel line relaxation algorithm was developed and tested. This algorithm is referred to as Block Implicit Gauss-Seidel line relaxation and achieves high performance on parallel computers by limiting global communication. In general, large simulations can be run 4 to 8 times faster than on the Cray C-90 supercomputer.;The former goal was addressed by applying the previously developed computational tools to the time accurate simulation of hypervelocity base flows. The resulting computations verified that the unsteady character of the base flow was indeed computable. In fact, the computed results were in agreement with experimental observations. Furthermore, the computations strongly suggested that a mechanism of unsteadiness was the amplification of oscillations in the wake shear layers. Hence, shear layer instability may drive the time dependent nature of these flows.