火灾早期过程特征提取与火灾源识别方法  

Characteristic extraction from early fire-initiating process and fire sources identification

作　　者：张认成[1] 杜建华[1] 杨建红[1] 

机构地区：[1]华侨大学机电及自动化学院,福建厦门361021

出　　处：《安全与环境学报》2010年第2期152-156,共5页Journal of Safety and Environment

基　　金：福建省自然科学基金项目(2009J01290;2008J0134)

摘　　要：针对传统火灾探测方法的准确性与及时性、误报率与漏报率相互制约的技术问题,利用傅里叶变换红外光谱分析技术,测量火灾初期的特征气体CO和CO_2体积分数。以CO与CO_2的体积分数比值为观察变量,运用移动平滑滤波器提取基于观察变量的大小、变化速度以及加速度的火灾过程特征向量。视虚假火灾源为具有有限扰动的固定目标,真实火灾源为具有相应位置、速度以及加速度的运动目标。采用数据融合技术,经过目标观察、特征提取以及身份识别,对火灾源的类型进行识别。研究表明,利用火灾过程特征信息与数据融合技术可以对真假火灾源进行有效识别,提高了火灾预警的准确性与及时性。This paper is intended to present its authors＇ extraction of the characteristic features of the early fire-starting and spreading process and its source identification. As is known, traditionally speaking, fire-starting and spreading process tracking or finding technologies used to measure the density of particle density in the smoke or monitor the rising temperature resulting from the fire. But, in fact, in the early stages of a fire, especially in the smoldering fire, only a small quantity of particles and heat are likely to come out. On the other hand, the later fire detection algorithms tends to make a fire alarm by comparing the fire variables sampled instantaneously with a given absolute threshold value, and as a result, the transient sampled values of the fire variables are usually regarded as the effect produced by some stochastic disturbances. Therefore, the above mentioned techniques can hardly help us to alarm the early fire fleetly and reli- ably. To reduce such wrong alarms of false positives and false negatives, it seems necessary to bring forward a new fire source identification method or theory of the early fire process characteristic features extraction and fire sources identification. Therefore, based on our research, it is found that the concentrations of carbon monoxide and carbon dioxide in the environment under discussion can be measured continuously by using NEXUS Fourier transformation infrared spectrometer with a 10-meter gas cell. For research convenience, fire variables can be defined as the ratio of CO concentration to CO2 concentration, whereas the process features characteristic of early fire occurring and spreading can be expressed by a 3-dimensional eigenvector consisting of the ratio as well as its increasing velocity and acceleration, which can be extracted from the experimental data by applying a movable smoothing filter. The process eigenvectors of nonfire nuisance sources and the authentic fire sources could be divided into two clusters obviously in a 3D phase space. Since t

关 键 词：安全工程 火灾早期探测 过程特征 身份识别 

分 类 号：X924.2[环境科学与工程—安全科学] TN911.23[电子电信—通信与信息系统]