基于可拓神经网络的火灾探测算法

Fire detection algorithm based on extension neural network

为解决传统单一传感器式的火灾探测器容易造成火灾报警的漏报和误报的问题,采用多传感器信息融合技术,将温度、烟雾浓度和CO浓度等多个参数相结合,进行综合分析,对火灾进行早期预测。采用可拓神经网络作为数据融合算法,以温度、烟雾浓度、CO气体浓度三个物理参量作为输入,以三种火灾预警等级作为输出。通过仿真分析结果表明:火灾正确识别率很高,达到93.9%以上。同时通过与传统BP神经网络的对比,表明可拓神经网络在数据融合的速度和可靠性上有突出的优势,从而使可拓神经网络实际应用于火灾早期预测成为可能。

In order to solve problem of failing or false alarm in traditional single sensor type fire detector, multi- sensor information fusion technology is applied to fire early prediction, it combines temperature, smog concentration and CO concentration together and analyze comprehensively. Extension neural network is used as data fusion algorithm, input values are temperature, smog concentration and CO concentration, and output values are three kinds of fire warning level. Simulation analysis result shows that correct identification rate of fire is in a very high level, which reaches above 93.9 %. At the same time, compared with traditional BP neural network, it shows that extension neural network has a prominent advantage in speed of data fusion and reliability, so it is possible to apply extension neural network to fire early prediction.