The Research Of Fire Data Processing Methods Based On Fuzzy Neural Network

With the rapid growth of population today, people create a large number ofwealth to speed up the process of urbanization and result in a high concentration ofpopulation and wealth.But the modern fire is frequent and the scale is expanding.Sothe loss is increasing. In order to avoid such disastrous consequences, it should raisethe accuracy of fire detection in demand to reduce false positives, false negative andextended reporting as far as possible ahead of forecast accuracy.According to the complexity of the fire signal, nonlinear structure and othercharacteristics, we design a signal processing system of fire based on fuzzy logic andneural network with the fuzzy logic theory and neural network theory principle andwe simulate it with MATLAB. From the simulation, we can see that it can not achievegood recognition results if the fire signal processing only relies on neural networks orfuzzy logic.Based on the above analysis, we propose that we can use the fuzzy neuralnetwork in fire signal processing. And the fuzzy neural network is six-story. It is usedto connect the fuzzy logic inference system. Fuzzy rules and membership functionsare shown by neural network in the six-story of the fuzzy neural network. With theweight of the membership function given to neural networks, it is used to implementfuzzy inference. We use the training data through the error back propagation methodto train the neural network, to modify the weights of neural networks for on-siteenvironment obtained precise fuzzy rules. Extracting from the modified neuralnetwork membership functions and fuzzy rules, we save it and regard it as a fuzzyreasoning in field environment. When the real external signals input, we calculate it inaccordance with the trained fuzzy neural network structure and get the probability ofcorrect fire. Conecting output of the fuzzy neural network with the fuzzy logicinference system, we introduct the smoke duration function T (n). Then we can get anaccurate fire judgment even in difficult situations around0.5for further processing todetermine.From the simulation, we can see that the design achieves the desired results.Andit proves that the fuzzy neural network signal processing applied to the idea of fire isreasonable. So, it can achieve accurate prediction ahead of schedule to reduce thefalse positive and false negative fire detection.And it coule get the purpose of improving reliability and capacity of the environment in the fire detection system.