摘要
近年来,WSN应用趋向于网内节点数量增多、模块功能多样、应用环境复杂,由此基于WSN的火灾监测预警系统容易因节点故障出现数据融合异常的现象。为提高火灾数据融合精度,引入高斯模型,通过对不同节点间同类信息融合形成的熵值,表示融合结果的不确定性,以鉴定融合效果。由此推理出一种正态分布的贝叶斯网络算法。在仿真实验中,将三种常用火灾传感器探测信息融合,分析改进后的静态、动态贝叶斯网络特点。用FDS平台模拟火灾场景,实验得到探测信息离散区间与发生率,再以Bayesia Lab计算输出节点的条件概率。最后通过Visual C++离散化选取探测阈值下限的判定依据,实现全网信息融合,作出正确、快速的报警反应。
In recent years,application of WSN tends to increasing node number,diversing module functions and complex application environment. Fusion data of fire monitoring and warning system are easy to be abnormal for node failure. In order to improve the precision of fire fusion data,this paper introduced the Gauss model. According to the entropy from fusing similar information between nodes,it ultilized the fusion outcome to express uncertainty. And it identified the fusion effect. The data fusion of Bayesian network based on the normal distribution was reasoning. In the simulation experiment,3 kinds of commonly used fire detection information were fusion. The improved static and dynamic Bayesian networks were used to be analysis. Simulating fire scene with FDS,detection information discrete interval and incidence could be obtained by experiment. The conditional probability of the output nodes was calculated with Bayesia Lab. Finally,the conclusion was obtained with Visual C ＋ ＋ and the discretization step. It was the basis of selecting the threshold limit of detection information. And the early fire alarm response could be accurate fast recation.
出处
《计算机应用研究》
CSCD
北大核心
2016年第5期1473-1476,1485,共5页
Application Research of Computers
基金
国家自然科学基金资助项目(61201179)
关键词
正态分布
贝叶斯网络
火灾数据融合
预警
normal distribution
Bayesian network
fire data fusion
early warning
