Microsoft Word - 025.docx
CHEMICAL ENGINEERING TRANSACTIONS
VOL. 48, 2016
A publication of
The Italian Association
of Chemical Engineering
Online at www.aidic.it/cet
Guest Editors: Eddy de Rademaeker, Peter Schmelzer
Copyright © 2016, AIDIC Servizi S.r.l.,
ISBN 978-88-95608-39-6; ISSN 2283-9216
Detection of Smouldering Fires by Carbon Monoxide Gas
Concentration Measurement
Jim Vingerhoets*, Jef Snoeys, Stijn Minten
Fike Europe, Toekomstlaan 52, BE-2200 Herentals, Belgium
jim.vingerhoets@fike.com
Smouldering fires pose a serious threat in processes that handle organic dusts. They are difficult to detect but
can have drastic consequences: besides deteriorating valuable product, they may evolve to open fires and
even to dust or gas explosions with potential massive destructive effects. A large number of incidents have
occurred in agricultural, food and chemical processing plants, in large storage silo’s, drying equipment and
mechanical processing apparatuses, such as mills, cutters or grinders. Smouldering is difficult to detect,
because combustion happens at lower temperature, without flames, often in the internal of a porous heap of
organic material where the oxygen level is reduced but maintained by diffusion through the porous medium.
Detection of smouldering based on carbon monoxide concentration measurement is regularly used in industry.
This work presents experimental results of smouldering of several organic products, where the size and
temperature of the smouldering nest and the available oxygen is correlated with the emission level of carbon
monoxide. This information is essential for the design of reliable smouldering detection systems based on CO
gas sensors, especially when process ventilation dilutes combustion gases and low detection levels are
required for an early response. Based on new and earlier reported experiments, required detection levels for
use in milk spray driers are given.
1. Introduction
Several investigations (BIA, 1997, CSB, 2005, Febo, 2013) indicate that smouldering is by far the most
occurring ignition source for fires and dust explosions in silos and dryers processing organic products. Also in
milling systems, smouldering often occurs as a result of mechanical malfunction, before a larger fire or dust
explosion is ignited.
Short residence time, thorough cleaning practices and low process temperatures lower the risk for product to
start smouldering. Inerting may be a solution but impractical for large throughput systems. A more cost-
effective solution is detection and control of smouldering in an early phase. Possible detection techniques
include temperature sensors, infrared light sensors and combustion gas sensors. The combustion gas sensors
are typically installed on the silo roof or in the exhaust air of continuous air flow systems, such as driers or
milling systems. Industry standards exist which specify procedures and evaluation criteria for the smouldering
ignition behavior of dust accumulations. The hot plate test (ASTM E2021, IEC 61241-2-1, EN 50281-2-1, VDI
2263-1) determines the minimum ignition temperature of dust layers on a hot plate. The hot storage test (UN
N.4, EN15188, VDI 2263-1) determines the self-ignition temperature of dust samples when stored in ovens at
constant temperature. The Grewer test (VDI 2263-1) determines the self-ignition temperature of dust samples
in a stream of air with temperature increasing at 1°C/min.
Frank-Kamenetskii (1969) and Bowes (1976) showed that the process of smouldering initiation can be
modeled as a heat balance process where the temperature rise of the product sample is governed by heat
transfer between product sample and environment (hot plate, ambient air, oven air, ..), conductive heat
transfer through the product sample and heat generation inside the product sample due to combustion. If the
product sample is excited such that it reaches a temperature where the heat generation due to combustion
exceeds the heat conducted to the environment, the combustion reaction runs away and auto-ignition occurs.
Once smouldering is initiated, it progresses as a slow, flameless form of combustion that emits heat, carbon
monoxide gas (CO), carbon dioxide gas (CO2), water vapor (H2O) and some product specific volatile organic
DOI: 10.3303/CET1648077
Please cite this article as: Vingerhoets J., Snoeys J., Minten S., 2016, Detection of smouldering fires by carbon monoxide concentration
measurement, Chemical Engineering Transactions, 48, 457-462 DOI:10.3303/CET1648077
457
compounds generated during decomposition and volatilization of the heated solid smouldering product.
Because CO2 and H2O are more abundantly present in normal air, CO gas is the most distinctive reaction
product.
The correct and safe application of smouldering detection systems based on CO measurement requires
knowledge on how much carbon monoxide is emitted under which circumstances. Especially in processes that
operate under continuous air flow, such as driers and mills, combustion gases are diluted and detection levels
should be set accordingly to prevent that the smouldering stays undetected.
This paper focuses on methods to determine CO emission amounts associated with smouldering, which is
fundamental knowledge to the correct application of explosion prevention by means of CO detection at early
stages.
2. Literature review
Most of the documented investigations were related to the use of early smouldering detection on dryers in the
dairy and animal feed production industry. Relatively little is reported in literature on CO emission amounts
during smouldering of organic powders.
Verdurmen et al. (2006) installed samples of skim milk powder in an oven set at 180°C and measured sample
temperature, oven temperature and CO emission rate. The oven was poorly ventilated. They report peak CO
emission rates for skim milk powder of 310 mL/min for a sample of 200 g. A sample of 950 g emitted up to
970 mL/min under the same conditions.
Steenbergen et al. (1989) investigated CO emission for different types of dairy powders, such as skim milk
powder, fat milk powder and whey powders. Samples of these products were heated in poorly ventilated
ovens at constant temperature (200 – 250 °C) and CO emission was measured. The influence of powder type
and powder size was investigated. For skim milk powder peak CO emission rates of 290 mL/min, 472 mL/min
and 624 mL/min were measured for sample sizes of 400 g, 900 g and 1900 g respectively. Fat milk powder
produced less CO. Whey powders produced similar CO amounts as milk powders.
In a next experiment, a lump of 500 g of fat milk powder was installed and brought to smouldering at the
ceiling of an industrial spray dryer, so close to the hot air inlet. Now the CO emission rate peaked to 2480
mL/min.
Zockol (1985, 1992, 1996, 2011) investigated CO emission for cacao powder, for cereals such as wheat,
starch and rapeseed and for dairy products such as milk powder. For the milk powder, Zockol (1992)
preheated 1 kg of skim milk powder to 200°C, installed the heated sample at different heights above an
industrial fluid bed dryer, and measured the air velocity at the sample installation position and CO emission
rate. He showed a strong dependency of sample installation position and CO emission rate. Close to the fluid
bed where air velocity was around 5 - 6 m/s, CO emission rate was 5600 mL/min, further away from the fluid
bed where the air velocity was around 3 m/s, CO emission rate was only 2415 mL/min.
Chong et al. (2006) investigated CO emission for small skim and fat milk powder samples in a 300°C hot air
stream. It was found that small samples (< 6 g) of skim milk powder were difficult to ignite and produced very
little CO whereas small samples of fat milk powder reached smouldering temperatures up to 965 °C and
produced much more CO (235 mL/min peak emission for 4,2 g of fat milk powder). They conclude that
smouldering is much more difficult to detect in skim milk powder production plants than in fat milk powder.
This is in contradiction with the findings of Steenbergen et al.
From these few studies it may be concluded that the amount of carbon monoxide emission during smouldering
of organic products is highly dependent on product composition, amount of product smouldering and the level
of aeration.
3. Experiments
3.1 Hot storage tests
In a first set of experiments samples of organic powders were put in cylindrical wire mesh baskets of 400 mL,
8 cm in diameter and stored for several hours in an oven. See Figure 1. The oven was set at a constant
temperature, above the self-ignition temperature of the product sample, so to initiate smouldering. The oven
was ventilated with a small amount of air (2 L/min) to prevent oxygen depletion during smouldering. The
exhaust air of the oven became diluted in a secondary air stream into which a CO concentration sensor was
installed. The dilution air stream was regulated such that CO concentration stayed within the CO sensor’s
measurement range. Oven and sample temperature (centre), as well as air flow and CO concentration were
continuously recorded. The experiment was stopped when the smouldering had finished, when the sample
temperature and CO concentration had dropped back to nominal values.
458
Figure 1: Hot storage test setup
Figure 2 shows hot storage test results for a sample of 400 mL (190 g) of skim milk powder (52% lactose, 26%
protein, 0.5%fat, self-ignition at 156°C according Grewer) and for a sample of 400 mL (76 g) of
methylcellulose and oven temperatures around 200°C. The CO emission rate was determined as the product
of CO concentration and dilution air flow.
Figure 2: Hot storage test results; left 190 g of skim milk powder; right 76 g methyl cellulose
According to the evaluation criteria of EN15188, self-ignition of the milk powder occurs after 2 h 7 min when
the milk powder temperature evolution over time shows an inflection point. The milk powder temperature at
ignition is 198 °C, somewhat higher than what the Grewer test had indicated (156 °C). First CO emission was
measured after 2 h 44 min when the milk powder temperature had reached 303°C. The delay between ignition
and detection may have to do with the lower detection limit of the CO sensor, not measuring very small
amounts of CO. CO emission peaked to 188 mL/s after 4 h 55 min. Milk powder temperature peaked to 769°C
after 5 hours and 17 min. CO emission peak falls somewhat earlier than milk powder temperature peak, but
generally speaking it can be said that CO emission is more or less proportional to sample temperature. The
CO emission rate seems to be a good measure of combustion reaction intensity. The total amount of CO
emitted during the experiment is 33,2 L or 38 g. Assuming lactose has a chemical composition similar to
C12H22O11 and protein similar to C47H48N3O7S2Na and assuming all carbon in the CO emission originates from
the milk powder, it can be said that around 20 % of the milk powders carbon has reacted to CO, the other 80%
must have been reacted to CO2 or has been volatilized as organic compound.
The bulk density of methylcellulose (76 g / 400 mL) is much lower than skim milk powder (190 g / 400 mL).
The porosity of the methylcellulose sample is therefore much higher. A drying phase is recognized as a
‘plateau’ in the methylcellulose temperature over time evolution somewhere halfway the first hour of the
experiment. During drying, the sample temperature rises slower due to latent heat consumption. Start and end
of drying phase can be identified by inflection points in the methyl cellulose temperature over time evolution.
Using the inflection point criterion, drying occurs between 18 and 37 min, respectively between 48 °C and 68
°C methylcellulose temperature. Duration of combustion based on methylcellulose temperature is much longer
than duration of combustion based on carbon monoxide emission. A possible explanation is that the methyl
cellulose sample may have fallen apart shortly after auto-ignition and that remaining combustion happened in
an oxygen rich environment favouring CO2 emission at the expense of CO emission.
The peak CO emission rate (506 mL/min) for methylcellulose is much higher than for skim milk powder. Since
the peak emission rate occurs before sample temperature peaked, the increased peak emission is probably
due to the increased initial availability of oxygen in the more porous sample.
Oven Temp. Sample Temp.
CO concentration
Air flow ~
2L/min
Oven – 115 L
Air flow ~20 m3/min
0
100
200
300
400
500
0
200
400
600
800
0 1 2 3 4 5 6 7 8
C
O
e
m
is
si
on
ra
te
(m
L/
m
in
)
Te
m
pe
ra
tu
re
(°
C
)
Time (h)
Oven Temp.
Sample Temp.
CO emission rate
0
200
400
600
800
1000
0
200
400
600
800
0 1 2 3 4 5 6
C
O
e
m
is
si
on
ra
te
(m
L/
m
in
)
Te
m
pe
ra
tu
re
(°
C
)
Time (h)
Oven Temp.
Sample Temp.
CO emission rate
459
3.2 Spray dryer experiments
In a next experiment, an almost equal sample of skim milk powder (400 mL // 220 g) was installed at the
ceiling of an operational spray drying, close to the hot air inlet. The hot air inlet flow velocity was around 20
m/s. Air stream temperature, product sample temperature (centre) and emitted carbon monoxide were
measured. See results in Figure 3.
Figure 3: Spray drier test results 220 g of skim milk powder
Start of self-ignition based on the inflection point criterion is 156°C, a value lower than for the hot storage test,
and the exact same value as determined by the Grewer test, which was also performed under aerated
conditions.
The peak sample temperature (900°C) and peak CO emission rate (2110 mL/min) are much larger than for the
hot storage test. However, the total combustion duration (1 h 38 min) is much lower than for the hot storage
test. Aeration intensifies the combustion reaction. Remarkably, the total amount of CO (50.0 L) is also
considerably larger than for the hot storage test. This seems in contradiction with what may be expected
knowing that the availability of oxygen should favour the reaction to carbon dioxide at the expense of carbon
monoxide. It may be that due to the higher sample temperature, volatilized organic compounds that stayed
unburnt in the hot storage test, now do burn, eventually in gas phase, and contribute to both the CO and CO2
emission level.
3.3 Hot air stream tests
In a third set of experiments different sizes of small skim milk powder samples were inserted in a +/- 300°C
hot air stream and CO emission rate was measured. Refer to test setup in Figure 4 and test results in Table 1.
All samples used a cylindrical basket of 100 mL and 5 cm in diameter which was filled to different levels. The
air and sample temperatures were not measured.
Figure 4: Hot air stream test setup
Table 1: Hot air stream test results
Sample size CO emission duration Peak CO emission Total CO emission
5 g 5.8 min 95 mL/min 0.32 L
10 g 6.8 min 116 mL/min 0.69 L
15 g 12.8 min 163 mL/min 1.42 L
20 g 21.0 min 184 mL/min 2.13 L
0
500
1000
1500
2000
2500
3000
0
200
400
600
800
1000
0 1 2 3 4
C
O
e
m
is
si
on
ra
te
(m
L/
m
in
)
Te
m
pe
ra
tu
re
(°
C
)
Time (h)
Air stream Temp.
Sample Temp.
CO emission rate
Air flow ~ 2L/min
Air flow ~20 m3/min CO concentration
460
Progressive browning of the milk powder was observed indicating that the combustion front propagated both
from sample bottom upwards and from sample side inwards.
CO emission duration, peak CO emission rate and total CO emission clearly scale with sample size, but not
linear. The CO emission duration is just a little less than linear proportional to the sample size. If the
combustion front would only propagate from the sample side inwards, one would expect constant combustion
duration, independent of sample size, since all samples had identical diameter. The almost linear correlation
shows that bottom up combustion propagation is most determining for the total CO emission duration. This is
in its turn due to the fact that the air temperature is higher at the sample bottom than at the sample side, since
the hot air tube was not insulated.
The peak CO emission is far less than linear proportional to the sample size. This is due to the fact that not
the entire sample burns simultaneously. The empirical correlations of this experiment are expressed as:
PeakCORate ~ TotalCO / CODuration ~ SampleSize1.4 / SampleSize0.9 ~ SampleSize0.5
4. Design of early smouldering warning systems
Experimental investigations clearly indicate that the CO emission amount during smouldering of organic
products is highly dependent on product composition, size of smouldering, intensity of the smouldering and
the level of aeration. The largest experimental data set is available for skim milk powder. The chart in Figure 5
plots all available results for skim milk powder and illustrates the importance of smouldering size and level of
aeration.
Figure 5: Summary of experimental data for skim milk powder
Early warning smouldering detection systems typically use sensors at the ceiling of silos or at the exhaust air
of dryers and mills. These sensors measure the concentration of carbon monoxide in the process air, in
percentages or in parts per million.
For silos, without continuous air flow, smouldering - even at a low intensity level - causes a continuous rise of
carbon monoxide concentration. Therefore, the CO concentration in silos is a measure of how much material
has smouldered after the silo was last filled with a new batch of product:
CCO = QCO, smouldering .dtVprocess
A smouldering detection system in silos is unable to discriminate between intense, hazardous smouldering -
with an immanent risk for fire or dust explosion - and harmless, less intense smouldering continuing over a
longer period of time. A minimum requirement is that when the CO concentration in the silo approaches its
lower explosion limit (12%) a high priority alarm should be initialised.
In dryers and mills that operate under continuous air flow however, the CO concentration at the exhaust air is
a measure of the actual size and intensity of smouldering. The CO concentration is now the ratio of CO
emission rate and process air flow
CCO = QCO, smoulderingQprocess air
In large processes that operate at high air flows, the CO emission can become diluted to such an extent that
only low CO concentration levels appear at the exhaust air – in the order of parts per million or even lower and
the smouldering must evolve to a larger size or intensity before it becomes detected. Table 2 specifies the
minimum required detection level for early warning systems on skim milk spray dryers. It is based on
experimentally measured CO emission rates in aerated conditions when the smouldering temperature was
0
1000
2000
3000
4000
5000
6000
0 500 1000 1500 2000
P
ea
k
C
O
e
m
is
si
on
ra
te
(m
L/
m
in
)
Sample size (g)
Verdurmen et al. (2006)- unaerated
Steenbergen et al. (1989) - unaerated
Steenbergen et al. (1989) - aerated
Zockol (1992) - aerated
Vingerhoets et al. - unaerated
Vingerhoets et al. - aerated
aerated
unaerated
461
600°C and 900°C and assumes a linear scaling of CO emission rate with smouldering size. The smouldering
size should then be interpreted as “amount of simultaneously smouldering product at equal intensity”.
It seems more appropriate to apply CO emission values under aerated conditions instead of unaerated values,
because the risk for smouldering to start and to evolve to a fire or initiate a dust explosion is much higher in
aerated conditions and locations of the drying plant. Experiments indeed do indicate a lower self-ignition
temperature and higher peak temperatures in these cases.
Table 2: Required detection level for early smouldering warning systems
Process air flow
200 g 750 g 2000 g
600°C 900°C 600°C 900°C 600°C 900°C
20,000 kg/h 2.3 ppm 6.9 ppm 8.6 ppm 25.9 ppm 23.0 ppm 69.0 ppm
100,000 kg/h 0.5 ppm 1.4 ppm 1.7 ppm 5.2 ppm 4.6 ppm 13.8 ppm
200,000 kg/h 0.2 ppm 0.7 ppm 0.9 ppm 2.6 ppm 2.3 ppm 6.9 ppm
5. Conclusions
Smouldering can be initiated in laboratory experiments and CO emission levels can be measured for organic
powders. A significant increase in CO emission is observed when the product sample is aerated. When
product is brought to smouldering in industrial process equipment, similar CO emission levels than in lab test
conditions are found. The level of aeration in the lab test should however match the condition of the industrial
apparatus. Scaling for larger smouldering lump sizes is linear or less than linear. Set detection levels for early
warning smouldering detection systems should take into account the anticipated size of the smouldering, the
intensity of smouldering and the dilution effect of process air flow.
References
BIA, 1997, Dokumentation Staubexplosionen - Analyse und Einzelfalldarstellung <http://www.dguv.de>.
Bowes P.C.,1976, Process industry hazards. Institution of chemical engineers, Rugby, 93-107.
Chong L.V., Dong Chen X.D, Mackereth A.R., 2006, Experimental results contributed to early detection of
smouldering milk powder as integrated part of maintaining spray drying plant safety. Drying technology,
24, 783-789.
CSB (Chemical Safety Board), 2006, Combustible dust hazard study <http://www.csb.gov>.
Febo H., 2013, Processes for drying powders – hazards and solutions. Chemical Engineering Transactions,
31, 709-714
Frank-Kamenetskii D.A., 1969, Diffusion and heat transfer in chemical kinetics. Plenum Press, London,
Second edition (translated by Appleton), 374 – 421.
Steenbergen A.E., van Houwelingen G., Straatsma J., 1989, Signaleren van smeulend poeder in een
droogtoren door CO-meting. NIZO nieuws 9, 21-23.
Verdurmen R.E.M, Huijs G., van Houwelingen G., Blommestein B., 2006. CO Detectie voorkomt brand in
sproeidrogers, Bulk, Februari 2006, 20-23.
Zockol C., 1985, Untersuchungen zur Bestimmung des CO gehaltes in Kakaoschüttungen und die Bedeutung
der Ergebnisse für den Aufbau eines Brandfrüherkennungsystem. Süsswaren, 83-86.
Zockol C., 1992, Brandfrüherkennung durch CO Detection am Biespiel von Sprühtrocknern in der
Milchindustrie. VDI Berichte 975.
Zockol C., 1996, Früherkennung von Bränden durch CO-Detection. VDI Berichte 1272.
Zockol C., 2011, Brandfrüherkennung bei der Getreidetrocknung durch CO Detection. FSA research project
F05-0803.
462