To recognize a fire, a person usually considers all information he can obtain such as the size of flame, the amount of smoke generated by the source, the location of flammable materials and other related conditions. The criteria for recognizing a fire are unclear and vary from person to person but human judgment is generally correct. This paper describes a fire detection system which uses a multi-layered neural network, which has been considered effective in recognizing and judging a situation with obscure factors such as a fire. This system uses the output of three different sensors: temperature, smoke and gas, and processes their output data to obtain information about the fire source, such as the heat release rate and the generation rate of smoke and gas. The real time data values and the previously collected data are then applied to a multi-layer neural network to obtain judgments about the state of the fire. After intensive studies, a new type of fire detection system has been achieved, which can not only form a proper analysis of gradually spreading fires but can also resolve one of the existing problems, false alarms caused by transient inputs, by identifying then suppressing them.
[1]
William D. Walton.
ASET-B: A Room Fire Program for Personal Computers. | NIST
,
1985
.
[2]
Hiromitsu Ishii,et al.
Multi-Dimensional Measurement of Early Fire Stage
,
1988
.
[3]
Hiromitsu Ishii,et al.
The Frequency Response Required for an Intelligent Fire Alarm System
,
1986
.
[4]
Leonard Y. Cooper,et al.
A mathematical model for estimating available safe egress time in fires
,
1982
.
[5]
Takashi Ono,et al.
An algorithm for improving the reliability of detection with processing of multiple sensors'signal
,
1991
.
[6]
William D. Walton.
ASET-B: A room fire program for personal computers
,
1985
.
[7]
L. Y. Cooper.
Estimating safe available egress time from fires
,
1981
.
[8]
Geoffrey E. Hinton,et al.
Learning representations by back-propagating errors
,
1986,
Nature.