Learning how to fight fires with HPC

Print Friendly, PDF & Email

insideHPC reader Sharan Kalwani sent us this pointer to a really cool application of the HPC resources at the EPCC, Edinburgh’s supercomputing center. Researchers  at the Building Research Establishment instrumented a mock apartment and set a fire so they could closely observe what happened as the fire burned and judge the effectiveness of an HPC-based numerical system called FireGrid that could eventually be used in real time by fire fighting teams:

A key element of FireGrid is its predictive capability, and this functionality was delivered using a fire/structure/egress code called K-CRISP, which ran faster than real time (or what its developers called “super-real time”) on a remote high performance computing (HPC) system, and was accessed over the Internet using grid protocols….K-CRISP was employed to predict both structural collapse and flash-over. To improve the reliability of model predictions, real-time data was assimilated into the computation at regular intervals. To ensure that the simulations were completed rapidly enough to be of use to firefighters in predicting what would happen next, K-CRISP was parallelized following the Task Farm paradigm, in which each slave ran a serial simulation of the fire, and where interprocess communication was handled via the file system.