World Journal of Engineering and Technology, 2014, 2, 36-41
Published Online September 2014 in SciRes. http://www.scirp.org/journal/wjet
http://dx.doi.org/10.4236/wjet.2014.23B006
How to cite this paper: Lebey, M., Bouaoud, A. and Lambert, E. (2014) Danger Detection during Fight against Compart-
ment-Fire Using Moving Averages in Temperature Recordings. World Journal of Engineering and Technology, 2, 36-41.
http://dx.doi.org/10.4236/wjet.2014.23B006
Danger Detection during Fight against
Compartment-Fire Using Moving Averages
in Temperature Recordings
Michel Lebey, Amal Bouaoud, Eloi Lambert
Laboratoire Ondes et Milieux Complexes (LOMC, UMR CNRS 6294), Uni ve rsit y of Le Havre, Le Havre, France
Received June 2014
Abstract
In compartment fires (houses, buildings, underground, warehouse, etc.), smokes are a major dan-
ger during firemen intervention. Most of the time, they are at high temperature (>800˚C) and they
flow everywhere through many kinds of ducts, which leads to the propagation of the combustion
by the creation other fires in places which may be far away from the initial fire. In this paper, we
present a new approach of the problem, which allows to better follow the fire behavior and espe-
cially to detect the dangers that may appear and endanger firefighters. This approach consists in a
mathematical analysis based on the comparison of moving averages centered in the past, calcu-
lated on the temperature recordings of the smokes. As a consequence, this method may allow to
improve decision support in real time and therefore to improve the security and the efficiency of
firefighters in their operations against that kind of fires.
Keywords
Compartment Fire, Decision Support in Real Time, Moving Average, Danger Detection
1. Introduction
In fire fi ghti ng, kno wled ge o f fir e d ynami cs a nd haz ard s wo uld allo w b uildi ng b ett er d ecis io n sup por ts. T his can
be done by developing the ab ility to obtain p hysical measu rements which are specific to the fire and to be able
to transmit them outside to a control centre and analyse them in order to have a more precise assessment of the
situa tio n and co nseq uent ly to i mprove dec ision supports. Moreover, this may help to improve the global organi-
zation of firefighting depart ments.
From a physical point of view, the spread of a fire can occur in three ways: conduction, convection and radia-
tion. The tr ansition fro m a growing fire to a widespread fire can take many forms. But one of the most important
in compartment fires is the propagation of the fire by the smokes, because of their mobility associated to their
high temperature (often higher than 800˚C), e.g. they can move everywhere through any type of duct such as
corridors, staircases, rows of rooms, but also through technical ducts such as elevator shafts, ventilation and
conditioning-air ducts, etc. When they are drawn into a non-burning space, even far away from the start of the

M. Lebey et al.
37
fire, an auto-i gnitio n ma y oc cur i f they ar e mixed with t he ox ygen in a ir, or they ma y ignit e anot her fue l of lo w
auto-ignition such as for instance paper, cardboard, or wood. This situation leads to the extension of the fire
thro ughout the who le bui ld i ng . I n op er a tio n, vent ila ti o n in fi r e usin g p o rt ab le fans c an he l p to d r aw the m out s id e
and so to protect firefighters. Many studies have been carried out on this subject in different situations, such as
Fire Ventilation [1], ventilation in compartment fire [2], in tunnel s [3], ventilation in rooms on fire [4] or in an
underground s tat ion [ 5].
On the other hand, if smokes are a major hazard in compartment firefighting and thus remain to be studied,
they are, however, of major interest because they contain and convey a certain amount of information (tempera-
ture, pressure, sp eed, physical compositio n and che mical) correspond ing to the fire, its nature, its evolution and
its risks.
In the first part o f this stud y, we show how it is possible to get precise infor mation abo ut the fire and its evo-
lution, and about the efficie ncy of the strategies e mplo yed to fight it. In the second part, we show how it is pos-
sible to detect the dangers that may occur by the use of the co mparison of two moving averages in the past cal-
culated in a temperature recording. At the end, we show how this comparison of averages in the past may allow
to improve decision support in real time and therefore may allow to improve the security and the efficiency of
firefighters i n their operations against compartment fires. Moving averages are often used in other fields such as
Energy [6], Knowledge-Based Systems [7], and Procedia Technology [8].
2. Follow-Up of Temperatures in the Development of Fire and of the Effectiveness
of Strategies Put in Place to Fight against Fire
The demonstration of the use of the temperature recording analysis in order to follow the evolution of the fire
and the effects of the means used to fight it, has been made with an experiment of a simulation of a fire in real
conditions. This experiment has been made in hospital room, 6 m long, 5 m wide and 3 m hig h, in wh ich there
was about 300 kg of combustibles: a table and a chair made of wood, an armchair made of imitation leather, a
wardrobe full of clothes, another table with a computer and a TV, and two beds with blankets. Temperatures
were recorded at different locations as shown on Figure 1. Temperatures were measured at different locations as
sho wn on Figure 1, using K thermocouples and recorded and analysed with the LabView software.
In this experiment, the fire intervention began about 5 minutes after the fire start. This lapse of time (5 min-
utes) represents the mean usual time between the alert and the arrival of the firefighters on the spot. The fire at-
tack was made using two portable fans only, without water hose at the beginning, in order to show the effects of
directed air flows on the smoke movement control. Obviously, in a real situation, the attack would have been
made using a water hose at the sa me time.
The temperature curves shown on Figure 2 are significant of the obtained results. In the first stage without
any action, a temperature plateau appears at about 550˚C. This is the temperature of auto-ignition of carbon
monoxide, a gas contained in great amount in the smoke of compartment fires. Just after this plateau a sudden
global temperature rise occurs to a level higher than 800˚C. This phenomenon is named the flashover, well-
known by firefighters because it is very dangerous, even fatal. On Fi gure B , t he ver tical V 1 ind icates the star t o f
the first fan, which produces a global temperature decrease. This is due to two reasons. Firstly, the glass windo w
had exploded under the effect of the heat. Secondly, the fan located in front of the front door of the buildin g cre-
ated an inflow of air, which forced the smokes to be drawn outwards through the broken glass window. Never-
theless, this temperature decrease stopped after a few minutes and was followed b y another sudden violent tem-
perature rise because the combustion was reactivated by the oxygen drawn inwards by the air flow. This shows
that the fan used is not powerful enough to draw outwards all the smoke produced by the fire. The vertical V2
shows the start of the second fan and from that moment the power of the two fans has become sufficient to force
all the smoke to flo w outwards and to carry the heat outside.
Two main remarks can be made from this test:
Remark 1: The temperature curve under the ceiling inside the corridor shows the fire’s evolut ion a nd the e f-
fect of the action that was under taken and the lack of power for the first fan alone, particularly. Finally, it clearly
shows the efficiency of the ventilation after the second fan starts when the ventilation has beco me strong enough.
So, a smart se nsor set a t this loca tion would communicate clea r information abo ut the fire’s evolution.
Remark 2: Curves B and E clearly show the temperature plateau which occurs just before a flasho ver, a well-
known dangerous phenome non whic h is fo llowed by a drastic temperature rise higher than 800˚C.

M. Lebey et al.
38
Figure 1. The smoke movement control by ventilation with portable fans; left: details of the experimental situation;
right: photo showing the effectiveness of the ventilation in the fire, the flow of combustion gas is discharged to the
outside, so the dangers of flashover and spread are much better controlled, the intervention is secured.
Figure 2. Experiment for ventilation efficiency in firefighting and danger detection: time temperature
evolut ions all alon g the t est and port abl e fan star t effects. Te mper at ur e cur ves are as foll ows: 1 ) A: 20 cm
under the ceiling in the corridor, 2) inside the room on fire just close to the door: B, 20 cm under the
ceiling, C at a height corresponding knees-to-head for a firefighter (which is the progression position into
a fire), D, 20 cm above the floor, and 3) inside the burning room just in front of the glass window: E, 20
cm under the ceiling and F at mid height. V1 shows the first fan start and V2 the second fan start.
3. Risk Anticipation in Firefighting Using Moving Averages in Temperature
Recordings
3.1. Time Series of Temperatures and Moving Averages Centered in the Past
In that kind of situation, the temperature recordings can be analysed as temporal series on which it possible to
apply laws which allow to detect some events. In severe situations, such as blaze, identifying dangerous events
become a way of security to protect firefighters from the d angers of hot smokes which ma y explode or against
the fla shove r. I n this s tudy, on l y the t e mp er at ure s a re co nsi d e r ed kno win g t hat o t her p hys i ca l q uant iti e s c o uld b e
use and give more information about the fire and its dangers.
In this analysis we use moving averages centered in the past, calculated as follow:
(t
i
−d+a)(t
i
−d)
(t
i
−d−a)t
i
t
I
c
( )
( )
( )
,
21
i
i
tt da
tt da
i
da
Tt
Tt a
=−+
−
=−−
=+
∑

M. Lebey et al.
39
With: Ic, the interval in which the moving average is calculated, d: the temporal distance in the past, a: the
half a mplitud e of the c alculat ing inte rval, ti: the date at which the moving average is calculated in the past, and
with the fo l lowin g condition: ti – d + a < ti.
Through this way it possible to define moving averages centered in the past at a temporal distance d. In a
simple step, thr e e moving aver ages could be defined:
- a first one at a short te rm, such that ti – d is close to ti, i.e. d small,
- another at lon g term, such that ti – d is far from ti, i.e. d great,
- at least, an intermediate moving average at a distance d chosen between the two previous.
In this calculatio n, a and d are parameters that may be chose n in order to obtain the best possible results, as it
will show in the analysis of the experiment detailed in the previous section.
3.2. Prediction of the Eve nts in a Compartment Fire Using Two Moving Averages Centered
in the Past
In this part, the study is carried out with the recording of the temperature under the ceiling close to the door in-
side the room on fire in the experiment described in the previous section. Two movin g aver ages c enter ed in t he
past are compared thus two main characteristics can be highlighted:
1) W hen the evo lut ion o f t he movi ng ave ra ge ce ntr ed in t he pa st at a lon g ter m (d 2) cro sses t he mo vin g ave r-
age at short term and beco mes higher, this means that the te mperature is going to decrease, and this mea ns that
the power of the fire is going to decrease, and therefore this means that the situation is going to be made safer.
2) W hen the evo lut ion o f t he movi ng ave ra ge ce ntr ed in t he pa st at a long term (d2) crosses the moving aver-
age at short term and becomes lower, this means that the temperature is going to increase, and this means that
the power of the fire is going to increase, and therefore this means that dangers may occur, such as smoke ex-
plosion or a flashover.
Figure 3 shows the temperature evolution recorded and two moving averages, one centered at −15 s a nd the
other centered at −30 s. T he inter sectio ns of the cur ves o f these t wo mo ving a vera ges hi ghlight dif ferent p hase s
of the fire with specific events:
- From the beginning to A: the moving average at −15 s is above the one at −30 s, this means that the tem-
peratures are increasing, there fore the fire power is increasing,
- at A point, the average at −15 s becomes higher than the average at −30 s, this means that the fire power
does not increase any more, which is confirmed by the temperature plateau appearing on the whole tempera-
ture recordings,
- between A and B, the temperature plateau, there are oscillations i n t he curves o f b ot h a ve r a ges, whi c h s ho ws
the possible appearing of a flashover because the temperatures are higher than 550˚C, the self-ignitio n tem-
perature of the CO gas,
- between B and C once again the moving average at −15 s becomes lower than the movi ng aver age at −30 s
with a sudden drastic temperature increase, this is the flashover, the well-known d ange rous event,
- between C and D point, there are once again oscillations in bo th averages and temperatures remain quite sta-
ble but at a high level , thus the situa tion remains dangerous,
- at D point, the average at −15 s becomes o nce again under the average at −30 s, this means that a ll tempera -
tures are going to decrease, therefore the situation may become safer, which shows that the ventilation ap-
plied seems to be efficient,
- at E point, there are once again a short period of oscillations and then the average at −15 s becomes higher
than the average at −30 s, clearl y; this event cle arly sho ws that if t he first ventilation was efficien t at the be-
ginning it beco mes insufficie nt because it br ings in fresh air that increase s the combustio n dynamics withou t
being powerful enoug h to fo rc e all the gas in comb ustion to flow outward,
- at F point to E point, the a verage at −15 s becomes o nce again lower than the average at −30, which clearly
shows the positive effect of the start of the second fan which increase the ventilation and can force all the gas
in combustion to flow outside. All the temperatures inside the room decrease to under 220˚C, therefore the
situation is u nder co ntrol and completely secured even if there is a small temperature increase from E point.
The comparison of the two moving averages centered in the past clearly shows that it is possible to detect
events which may be dangero us such as flashover, or to know that the firs t ventilat ion is not po werful enough.
This informatio n may become ver y importa nt for the firefighter commander in orde r to impro ve its decision ca -
pacities in real time. This is possible because all the calculations could have been made in a very short time due

M. Lebey et al.
40
to the fact that the time distance in the past of each moving average was very short, lower than 30 s.
3.3. Analysis of the Distance in the Past of the Moving Averages
In the analysis carried out in the previous section the temporal distances in the past d1 an d d2 have been chosen
in order to obtain significant results. However the question is how to choose the value of these temporal dis-
tances in the past and how long time it is necessary to obtain these values. As it was shown in the previous sec-
tion the delay to obtain one value is very short. In order to analyse this problem, other calculation were also
made in the same time with in addition three te mporal distan c e s in the past, d1 = −15 s, d2 = −30 s, d3 = −1 min,
d4 = −5 min, and d5 = −10 min. Figure 4 shows the results of these comparisons with the two used in the pre-
vious section. It is clear that these latter two moving averages give the best results. Moreover they are calculated
with the two shortest times in the past, thus they are useful and completely compatible with the characteristic
time of a compart me nt fire, about 10 minutes to 1 hour.
The moving average a t −1 min gives also useful information when it is compared to that at −30 s, howe ver the
temperature plateaus at 550˚C and 750˚C are not any more clearly visible. This moving average could be used
but for blazes of characteristic time more long, such in warehouse fire . At last, it is cle ar that the moving aver-
ages at −5 mi n a nd −10 min ha ve no sen se for the fir e d ur atio n of t he p re se nte d exp er i me nt in t hi s s t ud y, e ve n if
they may be useful for longe r duration of fire s.
Figure 3. Temperature vs time inside the room on fire close to the door under the ceiling and the two
moving averages, at −15 s and at −30 s.
Figure 4 . Comparison of different moving averages.

M. Lebey et al.
41
4. Conclusion and Applic ation in Decision Support for Fig h t against Comportment
Fire
As it was reminded in the beginning of this study, fighting against compartment fires may causes many dangers
for firefighters, especially backdraft and flashover, which can involve fatal consequences on firefighters. The
possibility to detect these dangers as soon as possible can ensure a better security and a better efficiency for the
firef ight ers. Thi s stud y ha s al lo wed to sho w on o ne ha nd that the a nal ysis o f the te mpera tur e re cor dings gi ve in -
formation ab out the evolution of the fire and ab out the efficiency of the applied strategy. And on the other hand
it has allowed to show that the use of the comparison of two moving averages centred in the past is useful and
may be efficient to detect dangers that may occur, and therefore in order to protect themselves. The comparison
between two moving averages centred in the past can be made by the records of the temperatures given by sen-
sors located at strategic locations. Many solutions are available to place these sensors. They can be divided in
two categories, either before the intervention on the fire, i.e. during the constructio n, or during the intervention
on the fire.
In the first case when sensors have been placed durin g the construction, the information can be collected and
sent in real time to a control centre and/or to the fire department centre, thus the firefighter commander can have
information about the fire b efo re go ing on s ite.
During the intervention on the fire, when there are no sensors placed during the construction, other kinds of
sensors can be used. Either, sensors may be brought by firemen on the fire suit, or they can place sensors on
strategic locations, these sensors being able to give their position and to send information outside to a control
centre, especially temperatures.
It is also po ssi ble to get information about temperatures using an infrared camera outside. This camera may be
placed on the floor or brought by a drone. This last case gives a new possibility to have a look around the fire in
order to have a global view of the situation. The whole of these means allows to consider important improve-
ments for the securit y and for the ef ficiency of the firefighte rs in their fight against compa rtment fires, and con-
sequently a decrease of goods losses and above all human losses.
Acknowledge ment s
The authors thank the Paris Fire Brigade (Brigade des Sapeurs-Pompiers de Paris, FRANCE) for its help to
carry out the experiments of real fires used for this study.
References
[1] Svensson, S. (2000) Fire V entilation. In: Göransson, A.-L., Ed.
[2] Hume, B. (2004) Fire Fighting in Under-Ventilated Compartments: Literature Review. Fire Stat isti cs and Resear ch Di-
vision, Office of the Deputy Prime Minister, London.
[3] Ura, F., Kawabata, N. and Tanaka, F. (2014) An Experimental Study on Critical Velocity in Sloping Tunnel with Lon-
gitudinal Ventilation under Fire. Fire Safety Journal, 67, 96-106.
[4] Lebey, M., Lambert , E. and Cité, O. (2004) Essai de démonstration de l’efficacité de la ventilat ion dans les in cendies
dans un feu d’hôpital, Rapport interne Université du Havre & Brigade d e Sapeurs-Pompiers de Paris.
[5] Lebey, M., Massouh, F., Vidor, R. , Le Corre, F ., Calas, C. and Ziliox, Cb. (2001) Contrôle des mouvements de fumées
dans une station de métro lors d’un feu de rame d e rame dans l’un des tunnels, Rapport interne, Université du Havre,
ENSAM Paris-Tech, RATP & Brigade de Sapeur s -Pompiers de Paris.
[6] Jeong, K., Koo, C., and Hong, T. (2014) An Estimation Model for Determining the Annual Energy Cost Budget in
Educational Facilities Using SARIMA (Seasonal Autoregressive Integrated Moving Average) and ANN (Artificial
Neural Network). Department of Architectural Engineering, Yonsei University, Seoul, Republic of Kor ea, Energy, 71.
[7] Lee, Y.-S. and Tong, L.-I. (2011) Forecasting Time Series Using a Methodology Based on Autoregressive Integrated
Moving Average and Genetic Programming. Department of Industrial Engineering and Management, National Chiao
Tung University, 1001, Ta-Hsuch Rd., Hsinchu 300, Taiwan, Knowledge-Based Systems, 24, 66-72.
[8] Pal, A., Singh, J.P. an d Dutta, P. (2013) The Path Length Prediction of MANET Using Mo vin g Aver age Model. Inter-
national Conference on Computational Intelligence: Modeling Techniques and Applications (CIMTA) 2013, Procedia
Technology, 10, 882-889.