Early warning signals of flashover in compartment fires

Abstract

There have been efforts to predict the occurrence of flashover. Due to its sudden development and the difficulty in discerning warning signs during the induction phase, the approach to flashover prediction is still under investigation. This research tests a new approach for detecting flashover events within a compartment through single-signal processing of the heat release rate (HRR). A set of ordinary differential equations aligned with a two-zone model are formulated and transformed into stochastic differential equations, subsequently solved through a numerical method. Based on the noisy HRR readings, a dynamical marker is constructed as a product of two quantities: the smoothed HRR and the standard deviation of the noise component. The dynamical marker was found to increase prior to a rise in the HRR signal on its own, confirming its superiority as a sign of flashover detection. To assess the practical applicability of the dynamical marker, we computed it to detect flashover incidents using HRR obtained from an FDS simulation and the fire calorimetry database provided by NIST; the dynamical marker exhibited a significant rise before the transition to flashover, confirming its potential as an early warning signal.