Prediction of smoke filling in large volumes by means of data assimilation–based numerical simulations

Abstract

The concept of numerical simulations for real-time Numerical Fire Forecasting is illustrated for the case of natural smoke filling of a large-scale atrium in case of fire. The numerical simulations are performed within the Inverse Zone Modelling framework. The technique consists of assimilating collected data for a certain parameter, in casu the smoke layer height, into the zone model in order to estimate an unknown of the problem (‘model invariant’), mainly the fire heat release rate. A forecast in terms of evolution of smoke level and temperature can then be produced. Because zone model calculations are very fast, positive lead times of several minutes are obtained. The developed model produces reliable forecasts for the cases considered. Equally important, the robustness of the technique is illustrated: the sensitivity of the results to the ‘initial guess’ of the model invariants is small (i.e. the method converges easily); one model invariant is sufficient to obtain reliable predictions for smoke layer height evolution; the data assimilation window length does not affect the results significantly. The method automatically provides a different value for the plume entrainment constant, depending on the position of the fire (in the middle of the atrium or in a corner).