Study On Fire Signal Processing Algorithms And Performance Evaluation Methods Of Algorithms

Automatic fire detection is a comprehensive multidisciplinary applied science newly formed and developed in the wake of rapid development of electronics, communication and information system, photo-electronics, fire safety science, sensor technology and automation technology. It reaches the goal of detecting fire by using sensor to detect the materials produced by fire. The real goal of fire detection is to give early alarm to a fire and reduce the fault alarm. In order to reach it, new developed technologies in signal processing field are continually applied to the fire signal processing and a lot of fire detection signal processing algorithms have been developed by using signal processing technology. Development of new fire signal processing algorithms, evaluation of algorithm performance and determination for engineering suitability of fire detectors are the hot research subjects in fire detection field. In this dissertation, fire signal characters, existing fire signal processing algorithms and the problems caused by applying the algorithms are discussed, two new developed algorithms are proposed, database of early fire and environmental information, methods of fire signal processing algorithm evaluation and determination for engineering suitability of fire detectors are also proposed in this dissertation.Although application of a lot of fire signal processing algorithms has brought benefit to the reduction of false alarm, improvement and development of fire signal processing algorithms is still a main research subject in fire detection field. In the practical application, smoke of cigarette and cook, water steam and so on are the main cause of many false alarms. In order to solve this problem, according to Mie theory, by using the fact that scattered lights from different angels reflect characters of different particles, a new algorithm based on signal character collecting and neural network is developed. Further, according to the fact that the energy released by fire or non-fire condition is different, a fire energy model is proposed and a new fire signal processing algorithm based on the model is developed.Evaluation of fire signal processing algorithms requires data samples of fire signals and environmental signals. Three signal information collecting systems and database built are introduced. The database included data of fire and environment for more than ten years, soft program for data comparison, processing and calling.Study on evaluation of fire signal processing algorithms is major subject in this dissertation. Three evaluation models, which are testing model, interface simulation model and soft program simulation model, are built to suit different situation. Method for single...