基于贝叶斯网络的火灾信息融合方法研究

发布时间：2018-07-25 09:03

【摘要】：火灾是一种失去人为控制的燃烧过程，产生火灾的基本要素是可燃物、助燃物和点火源，燃烧过程中的物理和化学现象是可以被探测到的。灾报警的根本目的是获取火灾发生时的相关信息，并进行处理，达到及时准确报警的目的。传统的传感方法仅仅通过采集烟、温、光、气等单一的火灾特征参数信息，采用阈值法来判定火灾，不可避免地会受到环境的干扰，限制了其传感性能，系统高误报率的问题比较突出。本文介绍了火灾探测技术发展过程及原理。介绍了几种火灾探测器的类型、探测原理及各类型探测器的优缺点。详细介绍了传统及人工智能火灾信息融合算法，并分析各自的优缺点。火灾信息融合算法是火灾探测系统的重要部分，如何提高探测系统报警的准确率、降低误报率是研究的重点。之后章节介绍了多传感器信息融合技术的基本原理，通过对原理的分析，了解各信息融合结构的形式和优缺点。为有效的利用传感器信息的冗余、多样性，提高火灾探测信息提取的及时性和可靠性，提高火灾探测系统报警的准确性、减少误报打下了良好的理论基础。运用FDS软件模拟几种典型的火灾场景，得到火灾特征参数的信息。分析连续属性离散化方法，运用BayesiaLab构建贝叶斯网络模型，以火场的温度、烟气浓度和CO浓度作为输入变量，以阴燃、明火和无火状态的概率作为输出量。运用Microsoft Visual C++6.0编辑界面，读取火灾特征参数信息及贝叶斯网络模型参数，输出最后信息融合的结果。通过实例，验证可以明确直观的表示出火灾状态的概率，并能较快的做出火灾报警响应，可以很好的识别阴燃火的状态。
[Abstract]:Fire is a kind of combustion process without artificial control. The basic elements of the fire are combustible, combustible and ignition sources. The physical and chemical phenomena in the combustion process can be detected. The basic purpose of the disaster alarm is to obtain the relevant information when the fire occurs and deal with it to achieve the purpose of timely and accurate alarm. Traditional sensing methods only collect smoke, temperature, light, gas and other single characteristic parameters of fire, and adopt threshold method to judge fire, which will inevitably be disturbed by the environment, which limits its sensing performance. System high false alarm rate of the problem is more prominent. This paper introduces the development process and principle of fire detection technology. The types, principles, advantages and disadvantages of several kinds of fire detectors are introduced. The traditional and artificial intelligence fire information fusion algorithms are introduced in detail, and their advantages and disadvantages are analyzed. The fire information fusion algorithm is an important part of the fire detection system. How to improve the accuracy of alarm and reduce the false alarm rate is the focus of the research. After the chapter introduces the basic principle of multi-sensor information fusion technology, through the analysis of the principle, understand the form of each information fusion structure and its advantages and disadvantages. It lays a good theoretical foundation for effectively utilizing the redundancy and diversity of sensor information, improving the timeliness and reliability of fire detection information extraction, improving the accuracy of fire detection system alarm and reducing false alarm. Several typical fire scenes are simulated by FDS software, and the information of fire characteristic parameters is obtained. The method of continuous attribute discretization is analyzed, and the Bayesian network model is constructed by using BayesiaLab. The temperature, smoke concentration and CO concentration of the fire field are taken as input variables, and the probability of smoldering, open fire and non-fire state is taken as the output quantity. The fire characteristic parameter information and Bayesian network model parameters are read by using Microsoft Visual C 6.0 editing interface, and the final information fusion results are outputted. Through the example, it can show the probability of fire state clearly and intuitively, and can make fire alarm response quickly, and can recognize the state of smoldering fire very well.
【学位授予单位】：沈阳航空航天大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TU892

【参考文献】