Study On The Smart Household Fire Prevention System Based On Fuzzy-neural Network

With the development in science and technology, and the improvement of livingconditions, people pay more and more attention to the comfort, safety and intelligence ofhousehold life. Smart home has become the development trend of the future furniture. Fireprevention is an important part of intelligent home security system. Especially in the recentyears, since household appliances are always in working mode for a long time, household fireaccidents occasionally occur, so the study of household fire prevention system has a lot ofpractical significance.The household fire prevention system is designed to accurately forecast a fire in theearly stages, thereby reducing the loss of life and property. Currently most of the householdfire prevention systems still make inaccurate reports of fires or miss report of fire accidentssince they are using a single detection method, such as, the detection of smog information.This paper combines different methods in fire detection in order to improve timeliness andreliability of fire detection, which is to detect physical parameters ofthe fire temperature,smoke, carbon monoxide d, and get a final forcast result through Fuzzy neural networkalgorithm. Hardware with HY-2440A Development board serves as main controller of thehardware, and detectors are monitored through the built-in micro processing in wirelessmodule CC2530F256on sensor management, will be collected to abnormal signal is sent tothe master controller algorithm fusion. The main research contents of the thesis are asfollows:In the beginning of the paper, the author analyzes the current situation of fire protectionsystem at home and abroad, analysis the technical problems and research significance ofhousehold fire prevention systems,and lists the characteristics of the phenomenon at allstages of an fire outbreak, By studying the conditions adaptable to various detectors, and theadvantages and disadvantages of various detectors, the author decides a composite sensorshould be adopted in fire detection and forecast.The second part of the paper discusses the overall design of fire protection systemincluding the choice and on-site installation field of detectors. Combined with the features ofvarious information fusion algorithms, the author selects the neural-fuzzy inference systemfor data fusion algorithm, and conducts fire prevention system training for BP algorithm andits improved algorithmIn the end the RLS-BP algorithm is chosen to carry on the samplestudy. Establishing the neural-fuzzy inference system for fire prevention system isestablished and fire accident samling data is studied to adjust The inference system membership function parameter.The simulation test has been conducted for the fireprotection system. To further improve the timeliness and reliability of the system, smoketime of the fires is added into the fire detection system in the end for a fuzzy inference andcombining all elements to make final detection of fire forecastThe hardware for the household fire detection system has been designed, including maincontrol module, data acquisition module, wireless transfer module, power supply module andalarm module. Diagrams for partial layout and each module's process are given.The paper draws an conclusion in the end and raises further research direction for thistopic.