PARALLELIZATION OF A LARGE SCALE HYDROCARBON POOL FIRE IN THE UINTAH PSE

摘要

Realistic simulation of complicated systems such as large-scale pool fires requires the representation of relevant physical processes such as turbulent reacting flows, convective and radiative heat transfer, and fundamental gas-phase chemistry. Resolution of the length and time scales responsible for controlling the dynamic features of fire are also required to capture important fire physics. Resolving these length and time scales, however, requires massively parallel computations. To achieve coupling of these complicated processes in a massively parallel environment, software components that reuse physics-based, legacy fire codes (written in Fortran) are developed and integrated with Uintah, a component-based, visual Problem Solving Environment (PSE). Uintah provides the framework for large-scale parallelization for different applications. The integration of the new fire code in Uintah is built on three principles: 1) Develop different, reusable, physics-based components that can be used interchangeably and interact with other components, 2) reuse the legacy fire code as much as possible, and 3) use components developed by third parties, specifically non-linear and linear solvers designed for solving complex-flow problems. The simulation of a 10-m heptane pool fire illustrates the parallel scalability obtained with the integrated fire code. Linear scalability to 1000 processors is obtained on the SGI Origin 2000 at Los Alamos National Laboratory.