Research Of Intelligent Fire Warning System Based On Neural Network

Fire is one of the most frequent and most frequent natural disasters in today's society,which seriously endangers people's lives and property.In order to reduce the harm caused by fire,on the one hand,it is necessary to strengthen the safety awareness of the fire;on the other hand,it is necessary to increase the research and application of fire warning systems so that fires can be detected and controlled in time.However,for the traditional fire warning system,it is impossible to effectively detect fire conditions,and false alarms and missed reports often occur.Based on this,this dissertation deeply researches and analyzes the intelligent fire warning system under neural network,proposes an intelligent algorithm that is cascaded based on improved neural network and softmax classifier,and applies it to fire classification and recognition.Firstly,the mechanism of fire generation and related technology of fire warning were studied and analyzed.The platform of fire warning system based on ARM was designed,including the main controller of fire warning,data acquisition device,emergency handling and alarm device.The use of wired transmission to achieve communication between the main controller and data acquisition devices,emergency handling and alarm devices.The main controller receives the data sent by the data acquisition device,and realizes real-time monitoring of various emergency events in the room.At the same time,the data is uploaded to the upper computer monitoring interface,and the user can view it directly.Build a fire data acquisition experimental platform,completed the open flame and smoldering fire experiment,obtained the fire data corresponding to the data support for the study of the following algorithm.In order to solve the problems of slow convergence speed and easy to fall into local minimum in BP algorithm,this paper put forward a cascaded fire classifier by neural network based on L-BFGS algorithm and softmax.Take the fire signal collected in the early stage as the target,and conduct real-time monitoring of the main characteristic signals generated during the fire.A cascaded monitoring system based on L-BFGS algorithm for neural network and softmax is designed.The cascaded structure model is established to realize the classification and identification of fire.In the early stage,the simulation of the cascaded model was completed using MATLAB simulation software using fire data.The results show that the algorithm greatly improves the system's convergence speed and prediction accuracy.In the later stage,a fire experiment platform was set up based on the designed fire warning system to perform functional verification experiments,response time experiments,and reliability tests on the system.The experimental results show that this system can well monitor and prewar the fire,improve the stability and accuracy of fire prediction,and effectively reduce the false positives and false negatives.