Analysis, synthesis and simulation of fire signals as a tool for the test of automatic fire detection systems

A simulation method for the test of one input fire detection algorithm is presented. A signal model is described, capable of synthesising time variant random signal sequences with similar statistical properties, as observed for signals of smoke- and temperature-measuring devices in reality. The time varying model parameters are derived from recorded signals in realistic environmental conditions. The necessary analysis method is discussed.

The simulation method is equally well suited for synthesising signal sequences in the ‘fire-’ and ‘nonfire’ case, which is necessary for the detectivity test of a detection algorithm. A fast transputer implementation yields acceptable calculation times even for estimates of the false alarm rate, corresponding to years observation time in reality.