Simulation of Hypervelocity Impact on Massively Parallel

Hypervelocity impact studies are important for debris shield and a.rmor/anti-armor research and development. Numerical simulations are frequently performed to complement experimental studies, and to evaluate code accuracy. Parametric computational studies involving material properties, geometry and impact velocity can be used to understand hypervelocity impact processes. These impact simulations normally need to address shock wave physics phenomena, material deformation and failure, and motion of debris particles. Detailed, three-dimensional calculations of such events have large memory and processing time requirements. At Sandia National Laboratories, many impact problems of interest require tens of millions of computational cells. Furthermore, even the inadequately resolved problems often require tens or hundred of Cray CPU hours to complete. Recent numerical studies done by Grady and Kipp [l] at Sandia using the Eulerian shock wave physics code CTH [2] demonstrated very good agreement with many featllres of a copper sphere-on-steel plate’ oblique pact experiment, fully utilizing the compute power and memory of Sandia’s Cray supercomputer. To satisfy requirements for more finelyresolved simulations in order to obtain a better understanding of the crater formation process and impact ejecta motion, the numerical work has been moved from the shared-memory Cray to a large, distributed-memory, massively parallel supercomputing system using PCTH [3-51, a parallel version of CTH. The current work is a continuation of the studies in [l], but done on Sandia’s Intel 1840processor Paragon X/PS parallel computer. With the great compute power and large memory provided by the Paragon, a highly detailed PCTH calculation has been completed for the copper sphere impacting steel plate experiment. Although the PCTH calculation used a mesh which is 4.5 times bigger than the original Cray setup, it finished in much less CPU time.