Content uploaded by Jian Wang
Author content
All content in this area was uploaded by Jian Wang on Mar 10, 2016
Content may be subject to copyright.
Experimental Study on the Combustion
Characteristics of Primary Lithium
Batteries Fire
Mingyi Chen, Chuang DeZhou and Jian Wang, State Key Laboratory of Fire
Science, University of Science and Technology of China, Hefei 230026
Anhui, People’s Republic of China
Yaping He, School of Engineering, University of Western Sydney, South
Penrith, NSW, Australia
Mingyi Chen and Yuen Richard, Department of Civil and Architectural
Engineering, City University of Hong Kong, Hong Kong 999077, People’s
Republic of China
Received: 12 May 2014/Accepted: 21 November 2014
Abstract. The use of lithium batteries requires understanding their fire and explo-
sion hazards. In this paper, a report is given on an experimental study of the com-
bustion characteristics of primary lithium batteries. Burning tests of single and
bundles of primary lithium batteries were conducted in a calorimeter to measure their
heat release rates when exposed to an irradiance of 20 kW m
-2
. Several variables
including time to ignition, mass loss, heat release rate and plume temperature were
measured to evaluate the ignition and combustion characteristics. The burning batter-
ies were observed to have flame temperatures in excess of 1,200C and to release cor-
rosive compounds. The experimental results show that the combustion efficiency,
carbon dioxide yield and mass loss are proportional to the number of batteries in the
bundle. The total heat released by battery bundles was deduced empirically to be pro-
portional to the number of batteries with a power of 1.26. The results provide experi-
mental basis for the development of fire protection measures during the use, storage
and distribution of primary lithium batteries.
Keywords:Lithium batteries,Calorimeter,Fire hazard,Heat release rate,Mass loss,Temperature,
Thermal runaway
List of symbols
APower factor
mMolecularity of CO
2
nNumber of batteries in a bundle
_
QHeat release rate (kW)
_
QnHeat release rate of nnumber of batteries (kW)
Q
n
Energy released from nnumber of batteries (kJ)
_
Qpk Peak heat release rate (kW)
RH Relative humidity (%)
* Correspondence should be addressed to: Jian Wang, E-mail: wangj@ustc.edu.cn
Fire Technology, 52, 365–385, 2016
2014 Springer Science+Business Media New York. Manufactured in The United States
DOI: 10.1007/s10694-014-0450-1
12
SOC State of charge (%)
TTemperature (C)
T
¥
Ambient temperature (C)
tTime (s)
t
ig
Time to ignition (s)
t
ex
Time to extinction (s)
WFuel mass (g)
W
a
Ash mass (g)
W
t
Initial fuel mass (g)
W
*
Dimensionless mass loss
DH
e
Effective heat of combustion (kJ g
-1
)
DWMass loss (g)
uCO2=CO Ratio of mole fractions or volumetric concentrations of CO
2
to CO
r
pk
Standard deviation of peak
_
Qpk (kW)
r
Q
Standard deviation of energy released (kJ)
Subscripts
max maximum or peak value
1. Introduction
Lithium batteries are nowadays widely used in portable electronic appliances.
Meanwhile, a number of lithium battery safety problems, especially fire safety in
usage and transportation, have aroused public’s concern. These problems are
mainly caused by lithium battery thermal runaway, which can lead to fire and
explosion [1]. However, knowledge of appropriate fire protection agents has not
kept pace with the industrial use of lithium batteries that are growing substan-
tially over the past few years.
There are two types of lithium batteries. One type is the lithium-ion battery
which is a member of a family of rechargeable batteries. The other is non-
rechargeable, or primary type. The mostly popular lithium-ion batteries are based
on lithium cobalt oxide (LCO), lithium-iron phosphate (LFP), lithium manganese
oxide (LMO) and lithium nickel manganese cobalt oxide (NMC). The most widely
used primary lithium battery is based on the lithium manganese dioxide (MnO
2
)
couple. Past studies on battery safety focused on chemical, electrical and environ-
mental aspects in relation to the materials, components (such as electrode materi-
als, the separator, the electrolyte and its additives), structures and the
manufacturing process [2–4]. Studies on the abusive tests of lithium batteries were
also conducted such as the oven heating, overcharge, nail penetration, crush and
internal short circuit [5,6]. Tobishima and Yamaki determined the thermal stabil-
ity limit of primary lithium cells to be 150C[7], so the temperature of most oven
tests were limited to 150C. The problems of thermal runaway caused by battery
abuses were investigated recently by Wang and Zhao [1]. The heat generation pro-
cess in batteries, which is considered to be the major threat to battery safety,
includes the decomposition of the solid electrolyte interface (SEI) and electrolytes,
the reactions of anode with electrolyte [8,9]. Thermal stability of lithium-ion
batteries with different materials has also been studied [10]. The aforementioned
366 Fire Technology 2016
studies mostly focused on the internal heating reactions, but the study of combus-
tion after heating is also very important for developing fire protection strategies
for usage, storage and transportation of lithium batteries. International Air Trans-
port Association Dangerous Goods Regulations has put forward special regula-
tions for lithium battery transport [11]. A series of tests was conducted by US
Federal Aviation Administration to determine the flammability characteristics of
primary lithium batteries and the dangers associated with shipping them in bulk
form on commercial transport category aircraft [12].
Heat release rate (HRR) as a function of time is a critical parameter for the
evaluation of fire hazards [13,14]. It has been an essential input variable for many
analytical and empirical correlations, and numerical simulation models to deter-
mine many other quantifiable consequences of fires [15]. Batteries are used and
transported in many forms and ranges of quantity. However, the investigations of
the HRR of lithium battery fires reported in the literature tend to fall into two
extremes. At one end, the single cell burning data was found in a recent paper by
Perrine et al. [16]. In their study the 2.9 Ah pouch type batteries were tested at
100%, 50% and 0% state of charge (SOC), and the relationship between HRR
and SOC was investigated by means of the Fire Propagation Apparatus. The stor-
age and transport of goods is hardly in the form of a single product and it is well
known that the burning behavior of a pile of multiple products can be drastically
different from that of a single one [17]. Therefore, FM Global, at the other end,
conducted tests on the flammability characterization of lithium-ion batteries in
bulk storage with the aim to develop protection system guidance [18]. However,
the HRR data of multi-cell bundles between the two extremes is not readily avail-
able, and furthermore a question still remains as to how the bulk burning behav-
ior is correlated to that of single cells.
In order to fill in the gap and obtain the HRR and other burning characteristics
of multiple primary battery cells, more experiments involving multiple primary
lithium batteries were conducted in current study. The attention was given to the
investigation of the combustion characteristics of lithium batteries as a conse-
quence of thermal runaway. The heat release rate, mass loss and temperature were
measured and analyzed for battery piles of different numbers.
2. Thermal Runaway, Ignition and Burning of Lithium
Battery Fires
The ignition and combustion associated with batteries have quite different mecha-
nisms to the ignition and combustion of ordinary or exposed hydrocarbon fuels in
the following two aspects: (1) the batteries do not have exposed fuel; (2) oxidant
for combustion can be generated within the batteries. The usage of batteries
involves complex chemical reactions and some of them are exothermic. The heat
generated by these exothermic reactions usually disperses to the surrounding envi-
ronment and no hazard is created. However, if the heat is not dispersed or when
external heat is applied to the batteries, the resultant temperature increase will
accelerate some exothermic reactions which will lead to a phenomenon called
Combustion Characteristics of Primary Lithium Batteries Fire 367
thermal runaway [1]. In particular, descriptions of the chemical reactions and ther-
mal runaway associated with the batteries of the lithium manganese dioxide
(MnO
2
) couple type are given below.
The electrolyte of this type of battery is lithium perchlorate (LiClO
4
), propylene
carbonate (PC) and dimethoxyethane (DME). When the battery is heated some
reactions between electrolytes, positive active material and negative active material
may take place at high temperatures. Equations (1) and (2) describe the reactions
of intercalated lithium with the organic solvents in the electrolyte.
2Li þC4H6O3PCðÞ!Li2CO3þC3H6ð1Þ
4Li þ3C
4H10O2DMEðÞ!2Li
2CO3þ5C
2H6ð2Þ
Early research shows that these reactions typically start at 100C inside the bat-
tery [1]. These reactions create flammable gases, which result in very high pressure
inside the battery shell. However, no oxygen is released to facilitate burning. Fur-
ther heating and temperature rising to 430C will cause the decomposition of
LiClO
4
into lithium chloride and release of oxygen as shown in Eq. (3)
LiClO4!LiCl þ2O2ð3Þ
The presence of oxygen may lead to the combustion of the electrolyte and the
hydrocarbon gases released in reactions (1) and (2).
Once the hydrocarbon gases, as the products of the reactions described in
Eqs. (1) and (2), are ignited, a series of combustion reactions will take place.
Equations (4–7) show some of these combustion reactions:
C4H6O3þm1þ4
2O2!m1CO2þ4m1
ðÞCO þ3H2Oð4Þ
C4H10O2þm2þ7
2O2!m2CO2þ4m2
ðÞCO þ5H2Oð5Þ
C3H6þm3þ6
2O2!m3CO2þ3m3
ðÞ
CO þ3H2Oð6Þ
C2H6þm4þ5
2O2!m4CO2þ2m4
ðÞCO þ3H2Oð7Þ
where m
i
(i= 1, 2, 3, 4) is the molecularity of CO
2
in the corresponding reaction
and is an indication of the completeness of the hydrocarbon fuel combustion. For
example in reaction Eq. (4), the ratio of mole fractions of the product gases CO
2
and CO is expressed as
368 Fire Technology 2016
uCO2=CO;1¼m1
4m1
ð8Þ
As m
1
approaches 4, less CO and more CO
2
are produced, the value of uCO2=CO;1
increases and the reaction is more complete, or the combustion efficiency increa-
ses.
Since these reactions may not be completed, the product gas will contain CO,
CO
2
,H
2
O, C
2
H
6
,C
4
H
10
O
2
. Other toxic and/or corrosive compounds such as HF,
HCl, and SO
2
may also exist in the product gas mixture [16].
3. Experimental
3.1. Battery Specimen
The most widely used primary lithium cell is the lithium manganese dioxide
(MnO
2
) couple. Batteries of a commercially available type in 100% SOC were
chosen as the experimental specimens. The normal voltage of these batteries is
3 V, with energy density ranging from 280 Wh kg
-1
to 500 Wh kg
-1
. Various bat-
tery configurations or bundles were tested in the experiment. The physical dimen-
sions and some of the configurations of the battery bundles are shown in
Figure 1. The burning test of each configuration was repeated five times under the
same condition. Table 1gives a summary of all tests conducted.
In addition, some 6 96 and 10 910 tests (repeated three times) were con-
ducted and the results are shown in Table 3.
3.2. Experimental Facility
The experimental facility consists of a heating equipment, a mesh specimen
holder, a calorimeter and a range of instruments to measure temperature and fuel
mass as revealed in Figure 2. Details of the components of the facility are descri-
bed in the following subsections.
3.2.1. Heating Equipment and the Specimen Holder. In the current study a 2 kW
electrical heater was chosen as the heating source to ignite the battery specimen.
The heater was made from a coiled electrical element of 10 mm in diameter form-
ing a heating spiral of 10 cm in diameter. The battery specimen was rested on a
Figure 1. Batteries size and configurations.
Combustion Characteristics of Primary Lithium Batteries Fire 369
steel mesh made from 1 mm steel wires. The mesh size was 10 910 mm. The gap
between the mesh and the surface of the electrical heater spiral plane was 2 cm.
The specimen batteries were bundled together with a 5 95 mm mesh. The close
views of the heater, the specimen holder and bundled batteries are shown in
Figure 3.
3.2.2. Calorimeter. The commonly used heat release rate tests of materials and
goods are full-scale room test (ISO 9705) [19] and the cone calorimeter test (ISO
5660) [20]. The former is intended to evaluate the contribution to fire growth pro-
vided by a surface product using a specified ignition source. The latter is a
method for assessing the heat release rate of a specimen exposed in the horizontal
Figure 2. Schematic diagram of the test rig and instrumentation.
Figure 3. Close view of the heater and the specimen holder.
370 Fire Technology 2016
orientation to controlled levels of irradiance with an external igniter. Lithium bat-
teries are a quite unique fuel in that their exposed surface is not combustible. As
discussed in Sect. 2, the ignition and exothermic reactions may take place within
the body of the objects and the ignition is most likely non-piloted. Since the bun-
dled lithium batteries are of bench-scale, the suitable facility to measure their
HRR seemed to be the cone calorimeter. However, previous and preliminary tests
revealed that primary lithium battery fires can be a ferocious combustion process
coupled with the discharge of corrosive substances and high flames that extend far
beyond the dimension of a cone calorimeter. On the other hand, the size the bat-
tery specimen were too small for the ISO 9705 test room. To cater for the current
need, an in situ calorimeter was constructed. This calorimeter is a scaled-down
version of the ISO 9705 test room by a factor of 0.4 as shown in Figure 2. It con-
sisted of a burning chamber, an exhaust system, and a catharometer system. The
dimensions of the burning chamber were 1.2 91.2 91.2 m. The chamber had a
0.15 m gap at the bottom to allow the intake of fresh air. The product gases were
collected by a hood connecting to an exhaust duct of 160 mm in diameter. Baffles
were mounted at the top of the hood and two guide vanes at both ends of the
exhaust duct to make the gases uniformly mixed. A volumetric air flow rate of
0.18 m
3
s
-1
was set to remove gases by a fan through the duct. The corresponding
oxygen mass flow rate is 5.0 910
-2
kg s
-1
, which ensured the over-ventilation
condition for the test fires. A further verification of the ventilation condition is
given in Sect. 4. There was a 60 940 cm glass door at the front of the burning
chamber for video recording.
Same as ISO9705 and ISO 5660, the heat release rate measurement of the
in situ calorimeter is based on the oxygen consumption principle [21,22]. The fire
products and air mixture captured by the exhaust duct was sampled continuously
using the Servomex 4100 analyzer as shown in Figure 2. Gas composition through
the pump was continuously monitored for oxygen, carbon monoxide and carbon
dioxide concentrations. The ranges for measuring oxygen, carbon monoxide and
carbon dioxide are 0% to 25%, 0% to 1% and 0% to 10% respectively, all with
0.001% resolution. The operating condition of the in situ calorimeter was the
same as the ISO 9705 test room calorimeter to ensure that all product gases were
captured. In other words, the scaling down of the ISO 9705 test room calorimeter
was only in the physical dimension and had virtually no impact on the measure-
ment.
3.2.3. Other Instruments. The mass loss of specimen was measured with a Mettler
Toledo XP10002S electrical scale. The resolution of the scale was 0.01 g and the
range was 9 kg. An array of ten K-type thermocouples (numbered T0 to T9,
0.5 cm in diameter, and 0.1C accuracy) was placed vertically along the center
axis of the specimen to measure the temperature of the flame. The lowest thermo-
couple (T9) was 3.5 cm above the top of the specimen, and the spacing between
two consecutive thermocouples was 5 cm.
The experimental process was recorded using a high-speed video camera with
500 frames per second.
Combustion Characteristics of Primary Lithium Batteries Fire 371
3.2.4. Experimental Procedure. In each test, the specimen was placed on the
holder first before the measuring instruments, the exhaust fan and the heater were
turned on. The electrical heater was switched on with full power and the heating
process was not moderated. It usually took 2 min for the measured irradiance to
reach the steady state reading. The maximum steady state irradiance measured at
2 cm above the centerline of the electric heater spiral plane where the specimen
was placed was approximately 20 ±0.3 kW m
-2
. Note that the heater was cooled
to room temperature between tests.
4. Results and Discussion
This section presents general observations, measured and derived quantities in
various fire tests. Unless mentioned otherwise, all quantities presented are the
average of the test group.
4.1. General Observations
The combustion processes of single and bundles of primary lithium batteries were
similar, and the ignition time was quite random. More batteries or larger size bat-
tery bundles tended to have longer ignition times as shown in Table 1. Some
snapshots of the burning of one of the 3 93 battery bundle fire tests are shown
in Figure 4. After about 6 min of heating by the electric heater disc, the top cov-
ers of the batteries were ruptured by the high pressure product gases as the result
of the initial reactions expressed in Eqs. (1), (2) and (3). A small flame was issued
from the top of the specimen (a), and then combustible solids and liquid droplets
were ejected from the batteries with high speed, forming white sparks (b). The
sparks nearly reached the top of the calorimeter chamber. The sparks were then
accompanied by a jet flame (c), indicating that gaseous combustibles were ejected,
which gradually replaced the sparks (d). Finally the flame disappeared, leaving the
hot red remains of the batteries (e). It is noted that in some tests of the same con-
figuration, white sparks were ejected suddenly without the initial flame. The igni-
tions of multiple batteries were usually sporadic or non-simultaneous. For
example, the two batteries in the 2 91 bundle might burn either quite closely or
quite separately in time. For larger battery bundles, the ignition of one battery
could trigger the ignition of the neighboring batteries due to heat transfer among
batteries as well as from the external heat source. As a consequence, the consecu-
tive ignitions were quite close in time for batteries in the larger size bundles. More
discussion on the ignition time is given in Sect. 4.4.
It was also observed that the orientation of the flame was intermittent and ran-
dom at the ignition and during the burning process. See frames (a), (b), (c) and
(d) of Figure 4.
4.2. Concentration of Product Species
Chemical reactions as described in Eqs. (4–7) are considered the predominant
reactions for heat release and species O
2
,COandCO
2
are the predominant
372 Fire Technology 2016
species for heat release rate estimation using the oxygen consumption principle
[19,21,22]. The measured concentrations of these three species are presented in
Figure 5.
Figure 4. The ignition and burning process of a 3 33 battery bundle.
Figure 5. Measured O
2
, CO and CO
2
concentrations as functions of
time. (a) 1 31 battery (b) 2 31 batteries (c) 2 32 battery (d) 3 33
batteries.
Combustion Characteristics of Primary Lithium Batteries Fire 373
The change in O
2
concentration was no more than 0.3% as shown in Figure 5d.
This result presented the strongest evidence that the test condition is well-venti-
lated. The measured maximum net increments in concentrations, which are the
differences between the measured maximum concentrations and their initial values,
are plotted in Figure 6a versus the battery numbers in test bundles. The measured
product species concentration increments are proportional to the number of bat-
teries in the specimen. The amount of CO
2
increased more distinctly than that of
CO. The ratio of the measured CO
2
concentration increment to CO concentration
increment vs the bundle battery number is plotted in Figure 6b. As discussed ear-
lier in Sect. 2, an increase in CO
2
/CO ratio indicates an increase in combustion
efficiency.
4.3. Mass Loss
The mass loss of the specimen is a result of the ejections of solid particles, liquid
droplets and gases from the heated batteries. It is directly related to the heat
release rate of the fires too. Since little mass loss occurred prior to the rupture of
the battery tops, the plot of mass loss data in Figure 7starts at 200 s from the
commencement of the tests.
Dimensionless mass loss is defined as the fuel mass loss in terms of the fraction
of the total consumable fuel mass [17].
W¼WtW
WtWa
ð9Þ
where Wis fuel mass; W
t
the initial fuel mass; W
a
the remain ash mass. Quanti-
ties W
t
and W
a
are given in Table 1.
Figure 7is a plot of W* versus time of typical four groups of battery fire tests.
The ignition of the fires is marked by the rapid increase in non-dimensional mass
loss. The extinction, or the burn out of the fires is marked by the attainment of
the value of 1. The relatively small fluctuations in the W* readings during the ini-
tial incipient phase and the final burn out phase were likely the result of the dis-
turbance from the air movement surrounding the scale. The fluctuation during the
Figure 6. Gas concentration increments with different number of
batteries: (a) The maximum concentration increments; (b) Overall
ratio of CO
2
/CO concentration increments.
374 Fire Technology 2016
burning period was the combined effect of fuel mass loss and the reaction to fuel
ejection motion.
The mass loss rates, or the burning rates (dW/dt) for various test groups are
shown in Figure 8. It appears that the time to ignition and the burning duration
are generally proportional to the size of the bundle. Obviously, the maximum
mass loss rate increases with the battery number. The multiple humps in this dia-
gram indicate the sequential ignition in the multi-cell bundles. The delay in igni-
tion for large bundles can be explained by the reduced radiant heat exposure per
unit volume (or per unit mass) of battery material. For a single cell, the almost
the entire battery surface (except the top cover) was exposed to the thermal radia-
tion field during the test and the exposed surface to volume ratio was high. For
larger bundles, this ratio is much lower and the heating process, or the incipient
phase, was much longer. In addition, the heating process was uneven. Some cells
in the large bundle might be heated up earlier and, therefore ignited earlier whilst
others ignited later. The desynchronized ignition may also be attributed to the
deviation in the manufacturing quality of the batteries. The desynchronized igni-
tion resulted in the apparent longer burning duration for larger bundles.
The net mass loss DWð¼ WtWa) of batteries with different numbers is fitted as
shown in Figure 9a. The data shows a good liner behavior that the mass loss
Figure 7. The dimensionless mass loss.
Figure 8. The mass loss rate.
Combustion Characteristics of Primary Lithium Batteries Fire 375
increases with the numbers. The fitted curve is y= 4.77n+ 1.03 with fitting coef-
ficient of 0.997. The linear fitting is based on the assumption that the mass loss of
each battery is approximately constant and independent of whether it is burnt as
an individual or in a bundle. However, when DWis divided by the number of bat-
teries in the bundle, the result DW
c
(=DW/n) reveals a variable as in Figure 9b.
The mass loss per unit battery appears to decrease with the bundle size. The rea-
son is not yet known. It might be possible that the multiple jets associated with
the large bundles helped reducing the amount of solid residual falling outside of
the electrical scale catchment area.
4.4. Heat Release Rate (HRR)
Figure 10 shows the HRR data as functions of time for different numbers of bat-
teries. The fluctuation or non-zero readings before and after the fire indicate that
the measured HRR data contained system drifting error. This error can be as high
as 10% of the measured maximum HRR for the 1-battery and 2-battery fires, less
than 4% for the 4-battery fire and approximately 2% for the 9-battery fire.
Comparing Figure 10 with the mass loss rate in Figure 8, the curves in the two
figures are largely consistent with each other. Instead of exhibiting multiple humps
as in the case of dW/dt in Figure 8, the HRR curves in Figure 10 appear smooth,
indicating that the response time of the oxygen consumption based calorimeter is
much slower than that of the electronic scale used to measure the mass loss.
The time to ignition t
ig
is indicated by the rapid rise in HRR. Although t
ig
is
largely proportional to the size of the battery bundle, there exists considerable
variation in ignition time (see also Table 1). The variation is attributed to devia-
tions in battery manufacturing quality as well as in the heating process. In the two
cases of bundle of one and bundle of two batteries, the exposures to the external
heat source were similar, or the difference in volume to exposure surface ratio was
relatively small in comparison with other bundles. Therefore, it is possible that
some two-battery bundles ignite faster than a single battery. See Figure 10. The
two peaks of the 2-battery fire indicate that the two batteries were ignited sequen-
tially rather than simultaneously. As a consequence, the values of the correspond-
ing two peaks were approximately equal to that of the 1-battery fire.
Figure 9. The relationship between the net mass loss and cell
numbers.
376 Fire Technology 2016
Sequential ignitions also occurred in 4-battery, 9-battery and even larger battery
bundle fires. This is reflected in the mass loss rate curves in Figure 8. However,
larger bundles exhibit group behavior in that the differences between individual
group members are relatively smaller, or the coefficient of variation for larger
groups is smaller than that for smaller bundles. Therefore, the multiple peaks in
HRR were less pronounced for larger bundles than for smaller bundles.
The total energy released in a test is obtained by integrating HRR
_
Q(t) over the
duration of the test which is symbolically denoted by ¥.
Q¼Z
1
0
_
QðtÞdt ð10Þ
The results are listed in Table 1together with the maximum, or the peak, HRR
results for various tests. It is seen from Table 1that significant deviations exist for
the same bundle configurations. The ratio of combustion energy to the net mass
loss is defined as the effective heat of combustion, DH
e
=Q/DW. This parameter
vs cell numbers was plot in Figure 11. An increasing effective heat of combustion
with the battery number is observed. This result implies the increasing combustion
efficiency and the effective heat of combustion with the battery number. This is
consistent with the result derived from the obtained ratio of CO
2
/CO concentra-
tion increments (see discussion in Sect. 4.2).
The mean values and the variation coefficients, or the relative error, of the peak
HRR and the total heat released are tabulated in Table 2.
The large relative errors associated with the 2 91 battery fires were likely due
to the significant deviation in battery quality.
Dividing the mean total heat release by the number of battery in a bundle, the
resulting parameter is the heat release per cell Q
c
(=Q/n). The values of this
parameter for various configurations are also included in Table 2. It can be seen
that the energy released per battery is not a constant value, but related to the
bundle size, or the number of batteries per bundle. The fire behavior of bundle
batteries was not a simple relationship of multiplication of one battery.
Figure 10. Typical HRR data of four test cases with different number
of battery cells.
Combustion Characteristics of Primary Lithium Batteries Fire 377
Figure 12 is a log–log plot of heat vs cell numbers (n).
From the log–log plot in this figure, it is not difficult to assume that a power
law of the form
Qn¼Q1nað11Þ
may exist, where Q
1
can be regarded as the heat release from a single battery, Q
n
is the heat release of a bundle of nbatteries and ais a constant. The regression
using Eq. (11) yields that Q
1
= 31.8 kJ and a= 1.26. The correlation coefficient
is 0.997.
Equation (11) can be used to estimate the total heat release of a given number
of batteries. Indeed, the results of 6 96 and 10 910 configuration tests are
shown in Table 3to compare very well with the estimates using Eq. (11).
It needs to be pointed out that Eq. (11) does have its limitation. By dividing
Eq. (11)bynand the effective heat of combustion for lithium battery DH
e
, one
obtains
DWc¼Qc=DHe¼Qn
nDHe
¼Q1na1=DHeð12Þ
Table 1
Summary of Test Conditions and Results
Test configuration
and label W
t
(g) W
a
(g) RH (%) T
¥
(C) T
ig
(s)
_
Qpk (kW) Q(kJ) T
ex
(s)
191
No. 1 16.1 10.3 44 20 345 2.0 31.1 384
No. 2 16.0 10.5 52 20 330 1.5 34.0 362
No. 3 16.0 10.2 52 20 357 2.0 34.9 397
No. 4 16.0 10.5 50 20 340 2.2 31.7 370
No. 5 16.1 10.0 54 18 353 1.8 36.3 395
291
No. 1 32.3 20.0 40 20 323 3.3 56.8 384
No. 2 32.3 21.8 46 18 345 2.8 68.1 387
No. 3 32.3 19.8 50 19 385 4.0 84.0 420
No. 4 32.1 21.1 46 20 374 2.5 81.4 403
No. 5 32.4 21.4 46 20 335 2.5 64.8 370
292
No. 1 63.8 43.0 42 21 385 5.8 175.1 459
No. 2 64.4 47.5 42 22 368 8.2 172.0 444
No. 3 64.1 43.2 42 20 415 4.7 182.7 469
No. 4 64.6 46.6 40 24 377 7.7 189.4 437
No. 5 64.7 44.6 36 24 374 6.5 175.1 428
393
No. 1 144.8 98.5 38 23 415 13.5 520.0 493
No. 2 145.7 97.7 38 23 388 16.0 440.9 463
No. 3 145.5 107.1 42 21 400 13.4 534.0 478
No. 4 145.9 101.6 30 18 376 13.5 522.7 457
No. 5 146.0 98.6 34 15 402 14.4 525.4 482
378 Fire Technology 2016
where DW
c
is the net mass loss per battery. Since Q
1
and DH
e
are approximately
constant and a>1, a sufficiently large number nmay result in a net mass loss
per battery that is greater than the weight the single battery. Caution should be
taken when using Eq. (11).
The peak HRR of the 36 and 100 battery fires are also presented in Table 3.It
is discerned from this table that the maximum measured HRR is approximately
70 kW for the 10 910 test. Based on Huggett’s [21] oxygen consumption constant
Figure 11. The relationship between Q/DWwith cell numbers.
Table 2
The Means and the Variation Coefficients of the Peak HRR and the
Total Heat Released
Configuration
_
Qpk (kW) rpk=
_
Qpk (%) Q(kJ) r
Q
/Q (%) Q
c
(kJ) r
Qc
/Q
c
(%)
191 1.9 13.7 33.6 6.5 33.6 6.5
291 3.0 21.3 71.0 16.1 35.5 16.1
292 6.6 21.3 178.9 4.0 44.7 4.0
393 14.3 10.3 527.0 9.5 56.5 9.5
Figure 12. The relationship between heats released with cell
numbers.
Combustion Characteristics of Primary Lithium Batteries Fire 379
(1.31 910
4
kJ kg
-1
) for hydrocarbon fuels, the 70 kW HRR corresponds to an
oxygen mass consumption rate of 5.3 910
-3
kg s
-1
which is almost an order of
magnitude less than the oxygen supply rate of 5.0 910
-2
kg s
-1
(see Sect. 3.2.2).
Therefore, the test fires were over-ventilated, or fuel controlled.
4.5. Temperature
Flame temperature is an important parameter in the combustion and fire engi-
neering study. However, little record could be found in the literature on measure-
ment of lithium battery fire flame temperatures. The bright colored flame as
presented in Sect. 4.1 indicates high flame temperature. Figure 13 shows the tem-
perature rise along the vertical centerline of the flame from 200 s to 600 s. At the
initial stage of the test, the temperature increases were a result of the radiation
and the formation of thermal plume from the heater. The temperature curves
remained relatively steady before the ignition of the battery fire. Rapid increases
in the measured temperatures were observed after the ignition. The time to reach
the peak increases when the number of batteries increases. The highest tempera-
ture of one battery fire is nearly 700C. For nine battery fires, this temperature is
over 1,200C. The thermocouple which was nearest to the batteries was damaged
by exceedingly high temperature jet of gases and molten metal compounds
released from the batteries. The high temperature reading in the large battery bun-
dle fires agrees well with the past studies [23–25] which showed that the tempera-
tures of lithium fires can be above 1,000C. This result indicates that not only the
hydrocabon compounds, but also the lithium metal inside the primary battery
were involved in the burning.
The measured maximum temperatures correlate well with the maximum (or
peak) heat release rates shown in Figure 10. Away from the fire base
(d>8.5 cm), temperature readings quickly returned to their initial values after
the completion of flaming combustion. However, near the fire base (d<8.5 cm),
this return was slow because of the hot residual of the burns.
At a given time, the temperature distribution along the axis of fire flame and
plume is dependent on the HRR of the fire [26]. Figure 14 shows the distribution
of the average measured maximum temperature along the axis for various battery
bundle fires correspondent to the moment of the maximum HRR. Significant
Table 3
The Results of 6 36 and 10 310 Configuration Tests
Configuration Equation (11) (kJ) Measured Q
n
(kJ) Relative error (%)
_
Qpk (kW)
696 2,802 No#1 2,925 4.5 37.3
No#2 2,800 0.1 38.6
No#3 2,834 1.1 35
10 910 10,388 No#1 10,080 3.0 70
No#2 9,985 3.9 68.2
No#3 10,290 0.9 65.4
380 Fire Technology 2016
differences are seen near the base of the fires (d<8.5 cm). These differences
quickly diminish with the increasing vertical distance d.
As discussed in Sect. 4.1, flame orientation exhibited intermittent behavior.
Form fires of smaller bundles, the flame diameter was small and the fluctuation in
flame orientation could result in the thermocouple array not being aligned with
the true axis of the flame. Particularly for the 2 91 battery fires, the fuel center-
line was not aligned with the jet flames of the two batteries which were ignited at
different times. As such, the measured ‘‘centerline’’ temperature was not the flame
centerline temperature. The misalignment of the thermocouple array and fire axis
and non-simultaneous ignition of batteries resulted in why the small difference
between the thermocouple readings of the 2 91and191 battery fires, and even
occasionally the former being lower than the latter.
4.6. Further Discussion
The heat release rate measurement error has been discussed in Sect. 4.4. Another
uncertainty in the measured HRR came from the O
2
generated in the battery
itself. The oxygen consumption based calorimeter assumes that all consumed oxy-
gen comes from the atmosphere with known oxygen concentration. However, the
batteries tested in the current study generate oxygen within themselves. The
hydrocarbon fuel has been premixed with oxygen before ejecting out of the bat-
tery shell. Therefore, the combustion is a kind of premixed (though not stoichio-
Figure 13. The temperature rise along the vertical centerline of the
flame.
Combustion Characteristics of Primary Lithium Batteries Fire 381
metrically). This offers an explanation why the combustion was fierce. The HRR
_
Qtcan be expressed as
_
Qt¼
_
Qþ
_
Qsð13Þ
where
_
Qis the HRR measured from the calorimeter,
_
Qsis the HRR from the
combustion with self-generated oxygen. This issue was not discussed in the previ-
ous studies. The measurement of the second term is difficult to achieve with the
oxygen consumption based calorimeters. The estimate of it is possible so long as
the chemical reaction of Eq. (3) can be quantified. This issue was not discussed in
the previous studies, and further studies should be considered to revise the HRR.
The flame temperature is strongly influenced but not entirely dictated by the
chemical reactions. Air entrainment to the flame and plume regions also has
strong influence. On the other hand, the temperature of the flame also influences
the combustion dynamics [27]. This is why the difference in combustion efficiency
was observed for different flame temperatures. The higher the flame temperature
is, the higher is the combustion efficiency as in the case of larger bundle fires.
Thermocouple readings are results of heat balance at the thermocouple junc-
tions. The measurement of local flame temperature is plagued by thermal radia-
tion [28]. Uncertainties or measurement errors up to ±15% may exist in the
presented temperature results. Owing to the uncertainty in the alignment of flame
axis or in the location of thermocouples within the flame, the significance of accu-
rate reading is diminished. Therefore, the detailed error analysis was not con-
ducted and the results presented in Sect. 4.5 only give indicative temperature
distributions.
5. Conclusions
This study has examined the ignition and burning behavior of single and multiple
primary lithium battery bundles. The ignition time, mass loss, the heat release rate
Figure 14. The T
max
distributions along the flame centerline above
the battery bundles of different configurations.
382 Fire Technology 2016
and the flame temperature distribution were measured in the combustion tests.
The total heat released was also deduced from the measured heat release rate. The
following conclusion may be drawn:
Unlike most other natural fires, the primary lithium battery fires can be clas-
sified as partially premixed jet fires. The chemical reactions during the pre-heat-
ing phase cause a pressure build-up inside the battery shell followed by a
violent rupture of the battery top cover. The combustion can take place in
solid phase in the form of sparks as well as in the gas phase in the form of a
flame. It was observed that the time to the top rupture or ignition was quite
random or varied significantly under the same external heating condition. The
primary lithium battery fires are specifically harmful, which not only result in
high flame temperature in excess of 1,200C, but also release hot and corrosive
solid compounds.
The experiments revealed that the ignition and burning characteristics of
multi-cell bundles are different from that of the single cell when subjected to
the same thermal radiation environment. The combustion efficiency, the
amounts of carbon dioxide and thermal energy released per battery are propor-
tional to the number of batteries in the bundle. The amount of carbon monox-
ide released per battery, however, decreases with the increasing bundle size.
Because of the sporadic ignition of individual batteries in a multi-cell bundle,
the maximum heat release rate of the bundle is not equal to the multiplication
of the single battery maximum heat release rate by the number of batteries in
the bundle. An empirical correlation in the form of a power function was
derived from the experimental data.
The oxygen consumption principle based calorimeter produced under-measure-
ments of heat release rate for lithium batteries which generate oxidant during pre-
heating and combustion process. In order to achieve more accurate measurements,
the oxygen content inside the battery needs to be measured, in further studies.
The ignition temperature is another fundamental quantity in understanding the
ignition and burning of battery fires. As discussed in the Sect. 2, the ignition and
initial combustion were likely to take place inside the heated batteries. It is very
difficult to gauge the battery interior temperature without tempering with the
structure of the batteries. Considerations should be given in future studies to the
employment of appropriate instruments and/or techniques to measure the interior
temperature and pressure inside the heated batteries prior to the rupture and igni-
tion.
The attempt to measure the flame axis temperature distribution of the primary
lithium battery fires was probably the first in the literature. However, the intermit-
tent behavior and unpredictable flame orientation of lithium battery fires pose
challenges to the measurement using stationary thermocouple device and to error
analysis. Advanced or non-intrusive temperature measurement devices or tech-
niques and thorough error analysis are desired to tackle these challenges.
From fire safety point of view, quantification of other toxic gas species, such as
HF, HCl, and SO
2
, is also very important. Future studies should also include the
measurement of these species.
Combustion Characteristics of Primary Lithium Batteries Fire 383
Acknowledgments
This research was supported by the National Natural Science Foundation of
China (No. 51376172).
References
1. Wang Q, Ping P, Zhao X (2012) Thermal runaway caused fire and explosion of lithium
ion battery. J Power Sources 208:210–224
2. Farrington MD (2001) Safety of lithium batteries in transportation. J Power Sources
96(1):260–265
3. Lisbona D, Snee T (2011) A review of hazards associated with primary lithium and lith-
ium-ion batteries. Process. Saf. Environ 89:434–442
4. Robert Selman J, Hallaj Said Al (2001) Cooperative research on safety fundamentals of
lithium batteries. J Power Sources 97–98:726–732
5. Fan Jiang (2004) Studies of 18650 cylindrical cells made with doped LiNiO
2
positive
electrodes for military applications. J Power Sources 138:288–293
6. Underwriters Laboratories Inc (2005) UL Safety Issues for Lithium-Ion Batteries.
Underwriters Laboratories, Northbrook
7. Tobishima Shin-ichi, Yamaki Jun-ichi (1999) A consideration of lithium cell safety. J
Power Sources 81–82:882–886
8. Wang Q, Sun J, Yao X, Chen C (2006) Micro calorimeter study on the thermal stabil-
ity of lithium-ion battery electrolytes. J Loss Prevent Proc 19:561–569
9. Mandal B et al (2006) Thermal runaway inhibitors for lithium battery electrolytes. J
Power Sources 161:1341–1345
10. Roth EP, Doughty DH (2004) Thermal abuse performance of high-power 18650 Li-ion
cells. J Power Sources 128:308–318
11. International Air Transport Association (2013) Lithium Battery Guidance Document,
Revised for the 2014 Regulations
12. Webster H (2004) Flammability Assessment of Bulk-Packed, Report No. DOT/FAA/
AR-04/26, Federal Aviation Administration, US Department of Transportation
13. Yao W, Hu X, Rong J, Wang J, Zhang H (2013) Experimental study of large-scale fire
behavior under low pressure at high altitude. J Fire Sciences 31(6):481–494
14. Ko YJ, Michels R, Hadjisophocleous GV (2011) Instrumentation design for HRR mea-
surements in a large-scale fire facility. Fire Technol 47(4):1047–1061. doi:10.1007/
s10694-009-0115-7
15. Overholt KJ, Ezekoye OA (2012) Characterizing heat release rates using an inverse fire
modeling technique. Fire Technol 48(4):893–909. doi:10.1007/s10694-011-0250-9
16. Perrine R et al (2012) Investigation on the fire-induced hazards of Li-ion battery cells
by fire calorimetry. Energy & Environ. Sci 5(1):5271–5280
17. Niu Y, He Y, Hu X, Zhou D, Lin C, Yin J, Yao W, Wang J (2013) Experimental study
of burning rates of cardboard box fires near sea level and at high altitude. Proc Com-
bust Inst 34(2):2565–2573
18. Ditch B, Vries J (2013) Flammability characterization of lithium-ion batteries in bulk
storage. FM Global, Norwood
19. ISO 9705-2003 Fire tests—Full-scale room test for surface products
20. ISO 5660:1-2002 Reaction-to-fire tests—Heat release, smoke production and mass loss
rate—Part 1: Heat release rate (cone calorimeter method)
384 Fire Technology 2016
21. Huggett C (1980) Estimation of rate of heat release by means of oxygen consumption
measurements. Fire Mater 4(2):61–65
22. Janssens ML (1991) Measuring rate of heat release by oxygen consumption. Fire Tech-
nol 27(3):234–249. doi:10.1007/BF01038449
23. Rhein RA (1992) An experimental study of the use of liquid argon and argon-filled
aqueous foams for the extinction of lithium fires. Fire Technol 28(4):290–316.
doi:10.1007/BF01873399
24. Rhein RA, Carlton CM (1993) Extinction of lithium fires: thermodynamic computa-
tions and experimental data from literature. Fire Technol 29(2):100–130. doi:10.1007/
BF01038535
25. Rodgers SJ, Everson WA (1965) Extinguishment of alkali metal fires. Fire Technol
1(2):103–111. doi:10.1007/BF02588480
26. Heskestad G (2008) Fire Plumes, Flame Height and Air Entrainment. In: Dinenno PJ
(ed) SFPE Handbook of Fire Protection Engineering, 4th edn. Society of Fire Protec-
tion Engineers, Boston, Massachusetts, USA
27. Glassman I (2008) Combustion. Boston, Academic Press, Amsterdam
28. Yang M, He Y, Li H, Zhang Z, Li Z, Wang J (2012) Combustion of laminar non-pre-
mixed acetylene jet at two different altitudes. Combust Sci Technol 184(12):1950–1969
Combustion Characteristics of Primary Lithium Batteries Fire 385
