Description:Fire Simulation using the grid-enabled FDS system
Abstract:The Fire Dynamics Simulator (FDS) is a Computational Fluid Dynamics (CFD) model of fire-driven fluid flow, developed by NIST (National Institute for Standard and Technology, USA). It solves numerically a LES form of Navier-Stokes equations appropriate for a low-speed, thermally-driven flow, with an emphasis on the smoke and heat transport from fires. FDS has been designed to provide four programming models: sequential - for a single CPU, MPI parallel - for distributed memory systems, OpenMP multi-threading - for shared memory systems, and a combined model MPI and OpenMP - for e.g. multi-core clusters. FDS reads input parameters from a single text file which provides all the necessary information to describe the fire scenario. The numerical solution to the equations is written to a set of files of various types, depending on the input specification. All output quantities of interest must be explicitly declared at the start of the calculation. In the package a graphical post-processor tool SmokeView is included, which enables the animation and analysis of FDS results.
The Slovak research team has long-term experience with modelling and simulation of fires, starting with forest fire, going on with fires in automobiles, and nowadays we focus on simulations of fire in various closed and semi-closed human structures (tunnels, houses, etc.). FDS simulation with a real fire scenario represents a long-time, computational intensive and memory consuming job. To find the programming model exhibiting the best performance and producing authentic true results for a given scenario, a great number of experiments (i.e. simulation runs with various input parameters), need to be fulfilled. In order to automate this process of simulation runs with a minimum of human intervention, we have developed a set of scripts which made all of the FDS models running on both the local multi-core cluster and on the EGI infrastructure. The submission of jobs to the Grid was realized using the middleware EMI (gLite) and the tool MPI-Start. FDS computations on the local cluster run with real fire scenarios, and FDS computations carried out on the Grid were verified using simple fire variants taking short run-time.