Decision Algorithms in Fire Detection Systems
SERBIAN JOURNAL OF ELECTRICAL ENGINEERING
Vol. 8, No. 2, May 2011, 155-161
UDK: 654.924.5:004.21; 614.842.4:004.021
Decision Algorithms in Fire Detection Systems
Jovan D. Ristić1, Dragana B. Radosavljević2
Abstract: Analogue (and addressable) fire detection systems enables a new
quality in improving sensitivity to real fires and reducing susceptibility to
nuisance alarm sources. Different decision algorithms types were developed with
intention to improve sensitivity and reduce false alarm occurrence. At the
beginning, it was free alarm level adjustment based on preset level. Majority of
multi-criteria decision work was based on multi-sensor (multi-signature)
decision algorithms – using different type of sensors on the same location or,
rather, using different aspects (level and rise) of one sensor measured value.
Our idea is to improve sensitivity and reduce false alarm occurrence by forming
groups of sensors that work in similar conditions (same world side in the
building, same or similar technology or working time). Original multi-criteria
decision algorithms based on level, rise and difference of level and rise from
group average are discussed in this paper.
Keywords: Decision algorithms, System architecture, Fire detection systems.
1
Introduction
The application of multi-criteria fire-detection technology primarily started
with the introduction of addressable analog detectors. Advances in
microprocessor electronics first allowed detectors to be more intelligently
monitored and controlled by a supervisory control panel. In more recent years,
further advancements in microprocessor electronics have allowed the
development of intelligent detectors. In this case, data processing can occur in
the detector itself, independent of the control panel.
The use of multi-criteria-based detection technology continues to offer the
most promising means to achieve both improved sensitivity to real fires and
reduced susceptibility to nuisance alarm sources [1, 6]. A multi-criteria detection
system can be developed by properly processing the output from sensors that
measure multiple signatures of a developing fire or by analyzing multiple
aspects of a given sensor output (e.g., absolute value, rate of rise or fluctuation).
1
Ph. D. Jovan D. Ristic, Associate professor, Faculty of Technical Sciences, Kneza Milosa 7, Kosovska
Mitrovica, Serbia, e-mail: [email protected]
2
M.S.C. Dragana B. Radosavljevic, Assisstant, Faculty of Technical Sciences, Kneza Milosa 7,
Kosovska Mitrovica, Serbia
155
J.D. Ristić, D.B. Radosavljević
All of the work in the area of multi-criteria fire detection has focused on
processing data from different types of detectors from the same point on the
object, so cold multiple sensors (i.e., multi-signature detection) [2-4, 6, 8, 9].
All of those sensors operate with adjustable, but still predefined alarm levels.
Much of this research has focused on the development of alarm algorithms
using fuzzy logic and neural networks for event classification and
discrimination between fire and nuisance sources [7, 10].
Our idea is to consider same type (or types) of sensors on the different
locations on objects as a group.
Grouping sensors according some criteria (topological, working time or
technology), calculating group average level and using it in alarm level
calculating could give the new quality in fire detection.
2
Alarm Level Settings
Independently of the place of execution of algorithms (in the detector or in
the control panel), free alarm setting level is enabled.
Changing of alarm level could be automated on the different bases such as:
winter, summer, presence or absence of heating, part of a day and similar. Fig. 1
shows natural course of daily temperature changes in a non air-conditioned
working room during winter and during summer period. By changing of alarm
level for daily and annual level, in advance settled curve was followed. Thus, it
is possible to alarm fire danger already at 10°C (in a cold winter night) but only
at 50° C (in a hot summer afternoon), in the same room. It is clear that similar
curves of changes may be settled for smoke concentration.
Decision about alarm condition is made in the detector itself or in the
control panel in the same way as with classic detectors by comparing the
measured value with the alarm level. By such an approach, sensitivity of the
detector is increased, and increase of false alarms is avoided [5].
Further increase of the quality of the announced alarms can be reached by
uniting of detectors on the eastern and western side of a building, by detaching
of their curves for temperature changing and by comparison of the measured
temperature with the temperatures of detectors from the group. Fig. 2 shows the
decision algorithm about the fire alarm comparing the measured level of fire
parameter with the average value of the reached levels of other detectors from
the group.
156
Decision Algorithms in Fire Detection Systems
Time (h)
Fig. 1 – Daily changes in temperature during summer and winter period in a not
acclimatized and unheated room.
if Ei ≥ Eial
then if Ei ≥ Eg + Δ
then ALi = 1
endif
endif
(a)
if Ei ≥ Eial
then ALi = 1
else if Ei ≥ Eg + Δ
then ALi = 1
endif
endif
(b)
Fig. 2 – Decision algorithm based on the reached level and average value of the group:
(a) increased reliability; (b) increased sensitivity.
Ei – measured value on detector i, Eial – alarm value for detector i,
Eg – average level of the group,
Δ – allowed difference from average value of the group.
3
Gradient Introduction
Observation of analogue values of detectors enables us to follow the speed
of increase of the observed fire parameters. With the classic detectors, there is a
well-known type of thermo differential fire detectors. Now, it is possible, with
the help of corresponding software, to treat each temperature sensor as both top thermal and thermo differential. Besides, each smoke detector could be
observed as a top and a differential detector. Now we can talk about the gradient
(speed of changing) of the observed fire parameter. Fig.3 shows the decision
algorithm on the base of the reached level and the gradient of the observed fire
parameter. This type of algorithm could be executed in the detector itself as well
as in the control panel and it gives more earlier information about the fire
157
J.D. Ristić, D.B. Radosavljević
danger. If we make correct choice of alarm gradients, there will not appear an
increased number of false alarms.
if Ei ≥ Eial
then if Gi ≥ Gial
then ALi = 1
endif
endif
if Ei ≥ Eial
then ALi = 1
else if Gi ≥ Gial
then ALi = 1
endif
endif
(a)
(b)
Fig. 3 – Decision algorithm based on the reached level and gradient:
(a) increased reliability; (b) increased sensitivity.
Ei – measured value on detector i, Eial – alarm value for detector i, Gi –gradient on
detector i, Gial - alarm level for gradient for detector i.
4
Multicriteria Decision
Measuring of one fire parameter (example temperature) it is possible to
recognize four different alarm conditions caused by:
Reaching of preset alarm level – ALL;
Reaching of preset gradient level – ALG;
Greater than allowed levels differ from average level of a group – ALLΔ;
Greater than allowed gradients differ from average gradient of a group –
ALGΔ.
Maximal adaptability of alarm level to the conditions in observed room was
reached by adapting alarm level during the year (because of yearly oscillations
of the temperature) and during the day (because of daily oscillations of the
temperature, people presence and technology).
Maximal adaptability of gradient alarm level to the natural gradient
oscillations (caused by heating, cooling, people presence or technology) was
reached by adapting gradient alarm level during the year and during the day.
Further decreasing of false alarm occurrence and increasing of sensitivity
could be reached by appropriate grouping of detector on some reasonable base
and by defining of allowed deviation from average level and gradient of the
group.
Decision about fire alarm level could be made on the base of four
parameters and choice of one of many decision algorithms. Four Boolean
variables enable (theoretically) sixteen decision algorithms, but this number will
158
Decision Algorithms in Fire Detection Systems
be reduced on a few, in practice. Table 1 shows all possible combinations of
alarm variables on one detector and three of sixteen decision algorithms.
Analogue values of level and gradient compared with pre defined alarm levels
and average group levels and allowed differs gives values for four Boolean
alarm variables (ALL, ALG, ALLΔ and ALGΔ), as shown on Figs. 2 and 3.
Based on the values of four Boolean variables and using one of 16 possible
algorithms we can make a decision about fire hazard as L – low, PA – pre alarm
conditions, A – alarm. Some decimal value corresponds to any combination of
Boolean variables ALL, ALG, ALLΔ and ALGΔ. Selecting one of possible
algorithms (working regimes), in adequate row is fire hazard level.
No
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
ALL
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
Table 1
Alarm values and fire hazard.
Alarm values
Algorithms
ALG ALLΔ ALGΔ 1
2
3 4
0
0
0
L
L
L
0
0
1
PA L
L
0
1
0
PA L
L
0
1
1
A PA PA
1
0
0
PA L
L
1
0
1
A
A PA
1
1
0
A PA PA
1
1
1
A
A
A
0
0
0
A
L
L
0
0
1
A PA PA
0
1
0
A
A PA
0
1
1
A
A
A
1
0
0
A PA PA
1
0
1
A
A
A
1
1
0
A
A
A
1
1
1
A
A
A
It is possible to make decision based on a couple of algorithms based on
majority.
Systematically monitoring of false alarm and pre alarm and adjustment of
alarm levels and allowed differs from group average of level and gradient
requests organization of database about events in fire detection system and
adequate programs for analyse and parameter correction.
Such algorithms are possible only in control panel, not in detector.
159
J.D. Ristić, D.B. Radosavljević
5
Supervision of Sensor Dirtiness
It is quite an expense for maintenance of the fire detection system. Also, for
periodical checking of sensor dirtiness together with the additional
unpleasantness that is obviously in connection with switching off of the certain
informing sectors. Automatization of this process increases reliability of the
whole system significantly by alarming of sensor dirtiness. Dirtiness level
decision algorithm is shown in Fig. 4.
6
Conclusion
Number of variables used in decisions and the freedom of their assignment,
undoubtedly brings new quality to the fire detection systems. Usage and
maintenance by such a number of parameters represents also a potential source
of errors. Further development of fire detection and information system will
demand more attention to be paid to the so-called user’s program-set of tools,
which will enable the user more simple way of resourcefulness in abundance of
possibilities.
Fuzzification of the observed fire parameters, by applying fuzzy logics and
by forming of knowledge base, performances of the security system can be
improved.
Addressable and analogue fire detection systems have opened many
possibilities. Main directions and the way of using these possibilities were
emphasized in this paper. Authors do not know fire alarm algorithms that use
grouping sensors according some criteria (topological, working time or
technology), calculating group average level and using it in alarm level
calculating.
if Δiд ≤ Δiw – C1;......................daily difference min. and max. is different from
weekly difference for С1
then if Δiд ≤ Δim – C1; .........daily difference is different from monthly
difference for С1
then if Δiд ≤ Δim – C2; ......daily difference is different from
monthly difference for С2
then SZi = 2;.................dirtyness level = 2
else SZi = 1;..................dirtiness level = 1
endif
endif
else SZi = 0;...............................................dirtiness level = 0
endif
Fig. 4 – Dirtiness level decision algorithm:
Δiд – daily difference on detector i; Δiw – weekly difference on detector i;
Δim – monthly difference on detector i; C1, C2 – constants.
160
Decision Algorithms in Fire Detection Systems
Original algorithms based on original idea were expressed in that paper.
If we succeeded to induce researches and planning engineers of some new
thinking, and users to ask for explanations from their suppliers, we shall
consider that the aim of this paper is reached.
7
Acknowledgement
This paper is supported by Project: Interdisciplinary Research of Cultural
and Linguistic Heritage of Serbia, Development of Multimedia Internet Portal:
Vocabulary of Serbian Culture, No. III 47016, Ministry of Science and
Technological Development of Republic Serbia.
8
References
[1]
C. Todd, Fire Detection - Overcoming the False Alarm Problem, International Fire and
Security Product News, Vol. 23, No. 3, 1998, pp. 39 – 40.
[2] D.T. Gottuk, M.J. Peatross, R.J. Roby, C.L. Beyler: Advanced Fire Detection using Multisignature Alarm Algorithms, Fire Safety Journal, Vol. 37, No. 4, June 2002, pp. 381 – 394.
[3] J.A. Milke, M.E. Hulcher, C.L. Worrell, D.T. Gottuk, F.W. Williams: Investigation of
Multi-sensor Algorithms for Fire Detection, Fire Technology, Vol. 39, No. 4, Oct. 2003,
pp. 363 – 382.
[4] J.B. Bowles, C.E. Pelaez: Application of Fuzzy Logic to Reliability Engineering,
Proceeding of the IEEE, Vol. 83, No. 3, March 1995, pp. 435 – 449.
[5] J. Reed: Control and Indicating Equipment for Fire Alarm Systems, International Fire and
Security Product News, Vol. 20, No. 5, 1995, pp. 41 – 42.
[6] L.A. Cestari, C. Worrell, J.A. Milke: Advanced Fire Detection Algorithms using Data from
the Home Smoke Detector Project, Fire Safety Journal, Vol. 40, No. 1, Feb. 2005, pp. 1 – 28.
[7] P. Radonja, S. Stankovic: Generalized Profile Function Model Based on Neural Networks,
Serbian Journal of Electrical Engineering, Vol. 6, No. 2, Nov. 2009, pp. 285 – 298.
[8] R. Siebel: Test of Fire Detection Algorithms using Artificially Generated Events, Fire Safety
Journal, Vol. 41, No. 4, June 2006, pp. 258 – 265.
[9] V. Gurevich: Microprocessor Protection Devices: The Present and the Future, Serbian
Journal of Electrical Engineering, Vol. 5, No. 2, Nov. 2008, pp. 325 – 339.
[10] Y.C. Tsai: The Design and Implementation of Early Fire Detection and Hierarchical
Evacuation Alarm System, Master Thesis, Graduate Institute of Networking and
Communication Engineering, Taiwan, 2007.
161
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement