Download as pdf or txt
Download as pdf or txt
You are on page 1of 2

A Framework of Fire Monitoring System

Based on Sensor Networks

Longjiang Guo1,2 , Yihui Sun1 , Jinbao Li1,2 ,


Qianqian Ren1,2, , and Meirui Ren1,2
1
School of Computer Science and Technology, Heilongjiang University, China
2
Key Laboratory of Database and Parallel Computing, Heilongjiang, China
{longjiangguo,sunyihui1,wsn.qqren}@gmail.com

Abstract. Sensor networks have been widely applied in harsh environ-


ment monitoring. Fire monitoring is one of the extensive applications.
But existing fire monitoring systems based on sensor networks fall into
two problems. First, since sensing ability of sensor nodes is limited, the
fire alarm may be delay or even fail to report. Next, because of the fire’s
uncertainty, it is difficult to accurately determine whether the fire break
out or not. This paper proposes a new framework of fire monitoring sys-
tem based on sensor networks to conquer the above two problems. The
system consists of data collection mechanism adopting improved time
series prediction algorithm (for short TSDC) and fire detection mech-
anism using neural network model. Experiment results show that our
fire monitoring system can recognize the flaming fire nearly 100%, and
fire warning delay can be controlled within 30s. The slow smoldering
fire recognition rate can be controlled within 80%, alarm delay can be
controlled within 1 minute.

1 Introduction
Traditional fire monitoring system adopts wire and the alarm information is
transmitted to center with cable. In particular scenario such as old protection
building, the deployment of wired fire monitoring system will influence aesthetics

This work is supported by Program for New Century Excellent Talents in University
under grant No.NCET-11-0955, Programs Foundation of Heilongjiang Educational
Committee for New Century Excellent Talents in University under grant No.1252-
NCET-011, Program for Group of Science and Technology Innovation of Heilongjiang
Educational Committee under grant No.2011PYTD002, the Science and Technol-
ogy Research of Heilongjiang Educational Committee under grant No.12511395, the
Science and Technology Innovation Research Project of Harbin for Young Scholar
under grant No.2008RFQXG107, 2009RFQX080 and No.2011RFXXG014, the Na-
tional Natural Science Foundation of China under grant No.61070193, 60803015,
Heilongjiang Province Founds for Distinguished Young Scientists under Grant
No.JC201104, Heilongjiang Province Science and Technique Foundation under Grant
No.GC09A109, The Natural Science Foundation of Heilongjiang Province under
Grant No. F201038.

Corresponding author.

X. Wang et al. (Eds.): WASA 2012, LNCS 7405, pp. 398–410, 2012.

c Springer-Verlag Berlin Heidelberg 2012
A Framework of Fire Monitoring System Based on Sensor Networks 399

and function in a sense. There is no wiring requirement of fire monitoring system


in WSN. The convenient installation can reduce building destruction greatly.
That is a good way for fire protection of old buildings. Due to self-organization
of WSN, detection system can construct network in a short time. Therefore,
WSN can be applied in fire detection.
Due to sensing limitation of low-power and low-cost wireless sensor node,
sensing delay occurs when get environment information such as temperature. The
temperature sensing component adopted in this paper is Sensirion SHT11 digital
temperature/humidity sensor. We measure heat source of 70 degree centigrade in
experiment. It costs 1 minute for the ascending of temperature sensing from 25 to 70
degree centigrade. The sensing delay is the main reason for fire alarm delay. Fire is
indeterminate so that fire parameter is different when fire occurs in different place,
different type of fire and at different time. It is hard to judge whether fire occurs in
determined algorithm. Traditional methods based on threshold produce high false
positive and false negative. We mainly focus on these two problem in system.
Time series prediction algorithm is a sort of data statistical algorithm. The
algorithm summarizes a series of discrete data, extracts the trend function of
these data, and takes proper variable into the function to estimate correspond-
ing value. We adopt time series prediction algorithm to process data which is
collected in a certain time in sensor node, and obtain the trend function of envi-
ronment parameter. Environment parameter in future time t will be forecasted
when time variable is taken into the trend function. Therefore, the sensing delay
of sensor node can be reduced.
Neutral network algorithm is a multiple data fusion learning algorithm. The
neutral network model adapted for certain environment can be computed by
training environment data samples. System with multiple data fusion neutral
network algorithm detects whether fire occurs.The current fire occurring prob-
ability can be obtained when input parameters are taken into trained neutral
network. Compared with system initialized threshold, the accuracy of fire alarms
by using neutral network algorithm is greater than the threshold.
The rest of this paper is organized as follows: Section 2 introduces related
work. Section 3 presents architecture of fire monitoring system and the design
of main module in system. Section 4 describes TSDC algorithm and neutral
network algorithm. Experiment and analysis are introduced in Section 5. We
conclude the whole paper in Section 6.

2 Related Work
Traditional fire detection system adopts wire. Alarm information is transmitted
to center with cable. Usually the components need to be connected by wiring.
The installation always takes long time, even wrecks the building structure. The
hardware circuitries can easily aging, and the false positive will be arise. The
number of circuitries increases as the number of sensors increases. The imple-
mentation of fire protection in WSN can effectively prevent from the problems.
Due to the low cost in price and the ability of adapting harsh environment, fire
protection in WSN attracts widespread research.

You might also like