Genetic algorithms in automatic fire detection technology

The application of sophisticated signal processing and detection principles is one of the most important developments in automatic fire detection technology. In order to study the performance of a novel algorithm, simulation techniques based on a representative set of recorded fire signals are promising. These techniques require the adequate and efficient modelling of the recorded signals with the help of stochastic signal models. To this end, this paper presents a novel application of genetic algorithms in automatic fire detection technology. Specifically, the online identification of linear and bilinear stochastic signal models for measured fire signals is treated. The identification methodology is based on a prediction error filter, the parameters of which are adapted online with the help of the genetic algorithm, i.e. one generation corresponds to one time step of the stochastic signal. Experimental results are given that demonstrate the robustness and efficiency of the genetic algorithm in this digital signal processing application.