Key Technology Of Early Fire Monitor System In Library Based On Fuzzy Neural Network

The fire is an extremely complex combustion process. It not only deprives great estates and financial losses, but also threatens human's security. The university library is collecting the multitudinous precious materials, especially including many historical materials. In addition, the university libraries have many teachers and students every day .They all are our country precious wealth. Therefore, the library fire protection works is superiorly important.Our each librarian should pay attention to the preventing of fire and the reduction fire damage. The security guarantees of library are our responsibility and duty. The fire detection technology is the effective method of the fire preventing .How enhances the accuracy and the response time of the fire detection, how to reduce failure and misinformation, which are the following research direction and key. The paper studied early fire monitor system in the library and fire detection technology based on the fuzzy neural network. It has made the preliminary progress. The following work has been accomplished:This system completes the data acquisition, the pretreatment and the local monitor equipment's control by the monolithic integrated circuit. PC achieves each kind of complex data processing, controls the monolithic integrated circuit, and applies fire detection algorithm designed by author to draw and display the conclusion. The conclusion includes open-fire, smoldering fire and misfire.According to the library early monitor system design requests, the author selected the temperature, the smog, the CO density sensor as the detector of the system. The different detector combinations have been established in the different region by combining with characteristic of library including library construction, layout, humanities, and environment and so on. Three structures of fire detection system has been constructed by the artificial neural networks and the fuzzy logic technology, including signal processing level, signal judgment level and information decision-making level. The signal pretreated is sent separately into the fuzzy controller and the neural network. Then 4 conclusions (fuzzy controller fire probability, neural network open fire probability, lights a fire probability cloudy and misfires probability) obtained is gave the information decision-making strata to draws the final conclusion.Fuzzy logic has the characteristics of strong comprehensive determination ability and identification ability. Making use of these two characteristics, fuzzy control is used in fire alarm system. The design of fuzzy controller in the fire detecting system is fulfilled with the use of MATLAB.The neural network fire detector model has been established with BP and RBF by using self-study and self-adaptation of neural network. At the time, the author edited algorithm design of neural network in fire detecting system and practiced and tested the network. Through comparing the restraining effect and the erroneous size of each kind of network above-mentioned, three feedforward BP neural network has been designated, the unit number of hidden layer was 12, the training function was traingda.Simulation experiments about national standard fire (including SH1, SH2, SH4) have been carried out. This system can accurately differentiate open-fire and smoldering fire and have high sensitivity to the different type fire. It has verified feasibility and availability of the system designed in the paper.The system can effectively reduce false alarm rate and missing report rate, and have higher sensitively, shorter response time and high anti-interference. It has more reference and inspiration at research and design about early monitor system and fire detection technology in libraries, archives, stacks and reference rooms.