A Decision-Making Method for Fire Detection Data Fusion Based on Bayesian Approach

This paper introduces a method for fire alarm information fusion in feature level by means of Dynamic Bayesian Network (DBN), the method is effective to reduce the interference of environmental factors and to improve the accuracy of fire detection. With CO content, environment temperature and smoke concentration as input fire characteristic information of DBN, the DBN output for the probability of the naked flame, smoldering fire and no fire. A variety of fire scenarios were built for the training and validation of DBN by using the Fire Dynamics Simulator (FDS), the curve of the CO content, temperature and smoke concentration were fitted. The results show that the method can better identify fire characteristics between naked flame and smoldering fire.