The Research On Fire Information Fusion Based On Bayesian Network

The fire is a combustion process of lost human control. The basic element of fire is thefuel, the oxidizer and the source if ignition.The combustion process of physical and chemicalphenomena can be detected. The essential purpose of fire alarm is to get and dispose therelated information when a fire happens in order to make a timely and accurate alarm. Thetraditional detection method is acquisition a single fire characteristic parameter, like smokedensity, temperature and so on. Using threshold method to determine fire, will inevitably beaffected by environmental interference and limits its detecting performance, system highmisinformation rate is another problem.This paper introduces the fire detection technology development process and principle.And then introduces several types of fire detector, detection principle and traditional, theartificial intelligent fire information fusion algorithm, and analyzes their the advantages anddisadvantages. Fire information fusion algorithm is an important part of the fire detectionsystem, how to improve the detection system alarm accuracy, reduce the error rate is the keypoint. Later chapter introduces the multi-sensor information fusion technology basic principle,the information fusion structure forms and their advantages and disadvantages. Laid a goodtheoretical basis for the effective utilization of sensor information redundancy, diversity, andimproving the fire detection information extraction, and improving the reliability of firedetection system alarm accuracy. By using FDS simulation of several typical fire scene, getfire characteristic informations of parameters. Using BayesiaLab construct a bayesian networkmodel. Regard the temperature of the fire, smoke concentration, the concentration of CO asinput variables, and the smoldering, flame, no fire state probabilities as output variables.Using Microsoft Visual C++6.0edit interface, reading fire characteristic parametersinformations and constructing bayesian network model parameters, outputing the informationfusion results. By the example, validate that bayesian network model could intuitive show the probabilities of the fire states, and rapidly make fire alarm response.