Content uploaded by Markus Spielbauer
Author content
All content in this area was uploaded by Markus Spielbauer on Sep 25, 2021
Content may be subject to copyright.
Journal of Energy Storage 43 (2021) 103213
2352-152X/© 2021 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Contents lists available at ScienceDirect
Journal of Energy Storage
journal homepage: www.elsevier.com/locate/est
Experimental investigation of the failure mechanism of 18650 lithium-ion
batteries due to shock and drop
Markus Spielbauer a,b,∗, Philipp Berg b, Jonas Soellner b, Julia Peters c, Florian Schaeufl a,
Christian Rosenmüller a, Oliver Bohlen a, Andreas Jossen b
aDepartment of Electrical Engineering and Information Technology, Munich University of Applied Sciences, Lothstr. 64, 80335 Munich, Germany
bInstitute for Electrical Energy Storage Technology, Technical University of Munich, Arcisstr. 21, 80333 Munich, Germany
cTÜV SÜD Battery Testing GmbH, Daimlerstr. 15, 85748 Garching, Germany
ARTICLE INFO
Keywords:
18650 lithium-ion battery safety
Mechanical abuse test
Vibration
Shock and drop
Failure mechanism
Current interruptive device
ABSTRACT
This work presents an experimental investigation of the failure mechanism of 18650 lithium-ion batteries
subject to dynamic mechanical loads and the implications of severe damages on the safety function of the
current interruptive device (CID), as current literature offers no insight in this topic. First, a conducted shock
test series with loads beyond automotive standards showed no distinct impact on various modern cell types in
impedance and computed tomography (CT) analysis, while older cell types exhibited signs of damage such
as mandrel displacement and increase of ohmic resistance, as had already been reported in literature. A
following investigation with acceleration measurements of drops of power tool battery modules revealed that
accelerations in some applications can exceed even high load-level standards significantly. In a subsequent test
series with axial drop tests in both orientations with various cell types, impact surfaces and states of charge
(SOC), multiple cell types exhibited high ohmic failure without a thermal event. Computed Tomography (CT)
and Post Mortem analysis revealed that, among various observable damage mechanisms, the predominant
failure mechanism is contact loss in the CID region. Even severe mechanical damages, although influencing
electrical and thermal behavior, showed no impact on the functionality of the CID in overcharge tests
1. Introduction
In many applications such as industrial use, power tools or mobility,
cylindric 18650 lithium-ion battery cells are used as they offer high
energy densities, are inexpensive, widely available and allow flexible
module design. In these and other operational areas, the cells can be
subject to severe mechanical loads from shocks, drops or vibration.
As such load types are believed to be potential causes of catastrophic
failure in battery packs and pose a risk to the safety of users and
the environment, great effort is put into the testing of battery cells
according to various standards and regulations [1] to ensure their
safety during operation and into robust battery pack design to mitigate
external loads on the cells [2].
Automotive test standards for lithium-ion batteries such as the SAE
J2464 or SAE J2929 test small cells below 0.5 kg with repeated shocks
with 150 g (not an SI unit but widely used in relevant literature,
1 g equates to 9.81 m∕s2) peak acceleration and a pulse duration
of 6 ms in multiple directions, while larger cells generally have to
withstand smaller loads [8,9]. Other automotive standards apply even
∗Corresponding author at: Department of Electrical Engineering and Information Technology, Munich University of Applied Sciences, Lothstr. 64, 80335
Munich, Germany.
E-mail address: markus.spielbauer@hm.edu (M. Spielbauer).
smaller loads. The commonly used standard UN 38.3 [10] for the
transport of dangerous goods recommends shock testing similar to the
SAE J2464. [1,17,18].
For space applications, NASA/TM-2009-215751 recommends pyro
shock testing of lithium-ion batteries from 20 g at 100 Hz up to 2000 g
peak acceleration above 1.6 kHz [12]. The military standard MIL-STD-
202G, not for lithium-ion batteries but for electronic and electronic and
electrical component parts in general, defines different test levels up to
1500 g peak acceleration with 0.5 ms half-sine pulses. [11]
More general standards for shock testing such as the IEC 60068-2-27
define a range of severity levels for shock tests with peak accelerations
up to 3000 g [13]. The same standard presents an overview of typical
shock forms and loads from 15 g for stationary or shock-protected de-
vices during transport up to 1500 g for semiconductors or microcircuits.
A summary of commonly applied standards for shock tests is depicted
in Table 1.
Drop tests, for example in the UL 2580 [6] or in the Freedom-
CAR [15], are usually free fall tests from a wide range of heights up to
https://doi.org/10.1016/j.est.2021.103213
Received 15 March 2021; Received in revised form 26 July 2021; Accepted 1 September 2021
Journal of Energy Storage 43 (2021) 103213
2
M. Spielbauer et al.
Table 1
Overview of shock test standards and regulations for small format battery cells and other components [1].
Standard Application Orientation Number of Shocks Peak Acceleration Pulse width
UN/ECE-R100.02 [3] Automotive longitudinal, lateral Either positive or negative direction or both 20–28 g, 8–15 g 80–120 ms
IEC 62660-2(3) [4] Automotive Analog to vehicle, all if unknown 10 per direction 51 g 6 ms
ISO 16750-3 [5] Automotive Analog to vehicle, all if unknown 10 per direction 51 g 6 ms
UL 2580 [6] Automotive 3 axes 3 shocks in each direction, total 18 25 g 15 ms
UL 1642 [7] Lithium batteries 3 axes, 2 for symmetry 3 shocks in each direction 125–175 g 75 g average in first 3 ms
SAE J2464 [8] Automotive 3 axes 3 shocks in each direction, total 18 150 g 6 ms
SAE J2929 [9] Automotive longitudinal, lateral One repetition each in positive and negative direction 150 g 6 ms
UN 38.3 [10] Transportation 3 axes 3 shocks in each direction, total 18 150 g 6 ms
MIL-STD-202G [11] Military, electrical component parts 3 axes 3 shocks in each direction, total 18 up to 1500 g at 0.5 ms
NASA/TM-2009-215751 [12] Space applications 3 axes 2 shocks in each direction 20–2000 g Pyro-Shock, 0.1–10 kHz
IEC 60068-2-27 [13] General 3 axes 3 shocks in each direction, total 18 up to 3000 g At 0.2 ms
Table 2
Overview of drop test standards and regulations for battery cells and other components [1].
Standard Application Drop height Impact surface
UL 2580 [6] Automotive 1 m Flat concrete surface
ISO 16750-3 [5] Automotive 1 m Concrete ground or steel plate
QC/T 743 [14] Automotive 1.5 m Hardwood floor
FreedomCAR [15] Automotive 10 m Steel object
MIL-STD-810H [16] Military, Logistic transit 1.22 m Steel backed by concrete
10 m with various impact surfaces and number of repetitions as shown
in Table 2. [1]
While testing according to standards is necessary for quality control,
it does not promote a better understanding of the underlying failure
mechanisms, as they define simple evaluation criteria such as ‘‘no
leakage, no venting, no disassembly, no rupture and no fire" or no drop
in open-circuit voltage (OCV) larger than 10% [10]. With a wide range
of peak accelerations defined in different standards for shock tests, for
many applications, it is also unclear what the actual loads are that
lithium-ion batteries are subjected to. Apart from this, there is also no
information available, in which magnitude the loads occurring in drop
tests are in comparison to shock tests.
Scientific literature offers more profound investigations regarding
shocks and drops, but the number of available publications is relatively
small. Brand et al. [19] used the pulse shape as defined in the UN38.3
to shock 18650 and pouch cells, but rather than applying 3 shocks
for each axis, they conducted 300 in the axial direction. Despite this,
they reported no significant changes in capacity or ohmic resistance
or damage to the electrodes. For the 18650 cells, deformation of the
current interruption device (CID) was observed, but it was not tested
whether this deformation has negative effects on the functionality of
the CID. Tsujikawa et al. [20] performed half-sine shock tests with 50 g
peak acceleration and 11 ms pulse width in multiple orientations, as
well as free-fall tests from up to 3 m height on large format cells and
reported no leakage or voltage drop for either. Ebert et al. [21] applied
automotive crash profiles with peak accelerations up to 45 g to 18650
and pouch cells and analyzed long-term cycling up to 80% state of
health (SOH), for which they reported slightly accelerated aging for
18650 cells, while no significant changes were observed for the pouch
cells.
More publications applying vibration tests than shock tests are
available, whose results, even though vibrations are not investigated
in this publication, are also concluded shortly here, as there is reason
to assume that the potential failure mode is the same as for dynamic
loads with higher accelerations.
Brand et al. [19] investigated vibrations with sine vibrations ac-
cording to UN 38.3 and long-term sine sweep vibration in different
orientations for 18650 and pouch cells. The pouch cells proved to be
very resilient, and the tests of the 18650 cells showed no changes
apart from a slight mandrel displacement in CT analysis either. Long-
term vibration tests caused no cell failure but did cause increases in
internal resistance, mandrel displacement, a partly melted separator
and even a hole punched out in the negative pole by the mandrel for the
18650 cells. The authors suggested that this might cause the resistance
increases but did not conclude a hypothesis about the ultimate failure
mode.
Hooper and Bruen et al. [22,23] applied various load profiles to
18650 cells in different orientations and at various SOCs. No cell
failures were observed, but for some of the tested cells they reported
an increase in ohmic resistance, which they attributed to increasing
contact resistance or delamination, but without further investigating
the actual reason for this behavior or their implications on cell safety.
Another study by Hooper et al. [24] on a different cell type with
multi-axis vibration type presents similar results but with significantly
smaller increases of ohmic resistance. Somerville [25] from the same
research group attributed power and capacity loss from vibration to
the damaging of boundary layers by analyzing electrolyte-deposited
products after vibration testing. However, this publication does not
conclude which mechanism will ultimately lead to cell failure either.
Berg et al. [26] tested the durability of 18 types of 18650 cells with
different pre-cycling conditions various random vibration profiles and
reported no degradation in resistance or capacity for any of the cells.
For some cells with mandrel, CT scans revealed imprints of the mandrel
in the anode current collectors, as had already been reported by Brand
et al. [19]. As potential failure mechanisms, high ohmic failure due to
contact loss as well as cut-out particles of the current collectors causing
internal short circuits (ISCs) were assumed, but without being able to
confirm or evaluate the probability of such events.
To conclude, neither standards nor literature offers a clear insight
into the failure mode of lithium-ion batteries when undergoing critical
mechanical load from shocks, drops or vibrations. On top of this, it is
unclear if severe damages which do not cause an instantaneous event
might harm the safety functions of the cells. However, such knowledge
could be valuable, for example, for researchers investigating dynamic
loads on battery cells or for cell manufacturers, who are still unaware
of the predominant failure mode and who could thereby improve the
cell design of upcoming cell generations. Also, companies using 18650
batteries in rough environments could benefit from such knowledge,
as it makes a significant difference to safety requirements if cells drop
out high ohmic or if they pose the threat of causing thermal runaways
after undergoing severe shocks or vibrations. Moreover, information on
the dominating failure mechanism is important for modeling of battery
deterioration and research on possible fault-detection and prognosis.
For this reason, this paper investigates the failure mode of 18650
commercial lithium-ion batteries in dynamic load scenarios. Therefore,
an experimental approach consisting of four successive steps, as de-
picted in Fig. 1, which also represents the structure of this paper,
was chosen. As the steps are consecutive and to improve readability,
each conducted test series is presented and discussed within the same
section.
Journal of Energy Storage 43 (2021) 103213
3
M. Spielbauer et al.
Fig. 1. Overview of the methodology applied in this paper to investigate the failure mode of 18650 lithium-ion battery cells.
2. Test of the resilience of lithium-ion batteries against shocks
beyond transport standards
A first test series to investigate the resilience of lithium-ion batteries
against shock and the relevant failure mode was performed by TÜV
SÜD Battery Testing GmbH on a shock test machine.
2.1. Test setup
For this test series, a pneumatic shock test machine AVEX SM-110
by Benchmark Electronics Inc, as depicted in Fig. 2(a), was employed.
The test machine is suitable for test specimens up to 90 kg and can be
used to generate half-sine, sawtooth and square wave shocks using dif-
ferent pulse generator pads. For rigid fixation, the cells were mounted
with M10 steel screws and 3D-printed cell holders.
To analyze the electrical behavior of the cells when undergoing
shocks, electrochemical impedance spectroscopy (EIS) measurements
were conducted before and after the tests. For this purpose, multiple
Gamry Reference 5000P were used in galvanostatic mode with 100 mA
AC RMS and a variable number of measurement points per decade with
frequencies from 10 kHz to 1 Hz. Lower frequencies were not included
as their measurement is very time-consuming, and the obtained results
are strongly dependent on ambient temperature [27–30] and relax-
ation [31,32], making their evaluation questionable. Four-point sensing
was enabled by cell holders from Battery Dynamics with spring probe
pins, which allow fast and highly reproducible contacting.
To evaluate structural changes within the cells, μ-Computed To-
mography (CT) scans were performed, with a phoenix nanotom s (GE
Sensing & Inspection Technologies) at 130 kV and 120 μA with 2 ×2
binning, 18.3 μm voxel size, 3 averaged images for each projection
and 1000 projections for every reconstruction. Rather than scanning
the entire cell, each one scan of the positive and the negative pole
was conducted. These settings allowed a short scan time with sufficient
quality. The scans were evaluated with VGSTUDIO (Volume Graphics
GmbH).
2.2. Performed tests
As cells are designed to withstand loads recommended by widely
applied standards, for the first test series it was decided to exceed
loads of automotive standards right away to increase the probability
of provoking damage patterns, which is the purpose of this entire
investigation. Therefore, instead of applying 150 g and 6 ms half-
sine pulses, 300 g and 6 ms half-sine pulses were chosen. Also, the
number of 1000 repetitions was chosen significantly above automotive
Fig. 2. Test setup and pulse form for test series with half sine shocks.
Fig. 3. CT-scans of pristine Samsung INR18650-25RM cell at 0% (left) and 100%
(right) SOC showing a slightly larger gap between jelly roll and case at lower SOC.
After multiple cycles this gap vanishes even at 0% SOC due to jelly roll swelling.
standards, which mostly do not exceed 3 repetitions per orientation.
As cell orientation, axial shocks were chosen as in this direction more
space for relative movement of the jelly roll is available, which was
considered to bear a higher failure potential. For the verification of the
pulse shape, the acceleration was measured by a single axis acceleration
sensor 353B03 from PCB Piezotronics, which has a sensitivity of 10
mV/g, a measurement range of ±500 g and a frequency range from 1 to
7000 Hz. The measurements were logged with a Q.bloxx A101 data log-
ging system from Gantner Instruments. One verification measurement
of a shock pulse is depicted in Fig. 2(b). While the main pulse shows
a peak slightly above the desired 300 g, as well as multiple secondary
pulses, the pulses are well within the acceptable variance recommended
in typical standards [13].
Journal of Energy Storage 43 (2021) 103213
4
M. Spielbauer et al.
Table 3
Cell types for the shock test with 300 g peak acceleration and 6 ms pulse width.
Cell type Number of tested cells Capacity Max. charge current Max. discharge current Max. voltage Min voltage Mandrel
A123 APR18650M1-A 3 1.1 Ah 5 A 30 A 3.6 V 2.0 V Yes
Moli IHR-18650-A 3 2.5 Ah 5 A 20 A 4.2 V 2.0 V Yes
Panasonic UR18650-RX 1 2.05 Ah 1.435 A 10 A 4.2 V 2.75 V No
‘‘Samsung ICR18650-22F’’ 3 2.2 Ah 1 A 1 A 4.2 V 2.75 V No
Samsung INR18650 25RM 3 2.5 Ah 4 A 20 A 4.2 V 2.5 V No
Sony / Murata Konion US18650 VTC4 3 2.1 Ah 4 A 30 A 4.2 V 2.5 V Yes
Fig. 4. Real part of the resistance at 1000 Hz before and after 1000 half sine shocks
with 300 g and 6 ms pulse width.
For this first test, 6 cell types, in total 16 cells, were chosen as
depicted in Table 3. Some of these cells represent state-of-the-art cell
types, others are discontinued models as they had been used in the
literature before [19]. Of these, the Samsung ICR18650-22F is shown
in quotation marks, as it exhibits both different electric characteristics
in the OCV and a varying inner cell structure from other 22F cells. This
gives reason to assume that it is either non-authentic or a much older
version of the cell (production code from 2011). Nevertheless, this cell
type was included in this test, as it was considered a prime example
of weak structural cell design and consequently more susceptible to
mechanical damage. The tests were performed on pristine cells at 0%
SOC, as it was suspected that the low jelly roll swelling and expansion
of the graphite anode [33–35] would result in lower contact forces
between the jelly roll layers and the case and more space for relative
movement, as depicted in Fig. 3, making the cells more susceptible to
shock-induced damages.
2.3. Results
During and after the shock test series, all cells remained electrically
functional and none of the cells showed signs of ISCs or thermal events.
The EIS measurements, conducted before and after the tests at 0% SOC
without intermediate electrical load, exhibit only changes in the real
part of the impedance. For this reason, Fig. 4 only depicts the real part
of the impedance at 1000 Hz and not the entire impedance spectra, as
the latter would offer no additional information. Most cells do not show
any changes in their EIS spectrum. Merely, the discontinued Moli IHR-
18650-A cells, as well as one of the presumably non-authentic Samsung
ICR 18650-22F cells, showed a significant increase of ohmic resistance.
The CT analysis depicted in Fig. 5 revealed no signs of damages
for most cells either. Only the Moli IHR-18650-A cells showed an
imprint of the mandrel in the negative current collector, as had been
observed in literature before [19,26]. The CT scans of the Samsung ICR
18650-22F showed no clear signs of damage but reveal the loose jelly
roll and insufficient fixation in the upper and lower parts of the cell,
Fig. 5. CT-scans of the Moli IHR-18650-A, the Samsung IRC18650-22F and the
Samsung INR18650 25RM after 1000 half-sine shocks with 300 g and 6 ms pulse
width. The CT scans reveal insufficient fixation with spacers, loose anode tabs and loose
packaging of the jelly roll for the Moli IHR-18650-A and the Samsung IRC18650-22F.
The Moli IHR-18650-A shows an imprint of the loose mandrel. The Samsung INR18650
25RM shows no signs of damage due to its tight packaging and the use of spacers at
both ends of the jelly roll.
which allows significant relative movement. State-of-the-art cells, like
the Samsung INR18650 25RM, which is exemplarily depicted, are more
tightly packed and have larger spacers in both the positive and negative
pole directions to prohibit relative movement of the jelly roll and are
therefore more resilient to shock loads.
2.4. Discussion
In this first investigation with loads beyond automotive test stan-
dards for 18650 cells, several damage mechanisms already known from
literature could be observed for the older cell types, such as increases
of ohmic resistance and imprints of loose mandrels, while state-of-the-
art cells showed no indications of damages. Nevertheless, neither for
older nor for state-of-the-art cell types, an ultimate failure occurred
that would allow the assessment of the failure mode. At this point, the
question arises, if loads beyond automotive test standards can occur in
applications (and which acceleration levels they reach) and therefore
if it even makes sense to investigate the failure mode at higher loads,
or if further analysis of the failure mode can be relinquished.
3. Measurements of acceleration loads in an application scenario
To investigate if loads beyond typical standards and regulations
occur and if the failure mode is even practically relevant, it was decided
to measure accelerations in a use case. Therefore, as a high probability
and high load scenario, drops of a battery pack used for power tool
applications, as they might occur e.g. when dropped from scaffolds in
construction sites, were investigated.
3.1. Test setup
For this reason, a drop tower with a pneumatic clutch system for
increased reproducibility was developed, as depicted in Fig. 6(a). The
Journal of Energy Storage 43 (2021) 103213
5
M. Spielbauer et al.
Fig. 6. Test setup for acceleration measurements on cells and modules.
clutch system is attached to a rail, which guides the fall and allows
holding the test object at a defined angle up until releasing it right
before impact by triggering a photoelectric barrier. To cushion the
shocks caused by the impact of the clutch, a hydraulic damper was
attached to the end of the rail. Various impact surfaces can be mounted
to the ground plate with screw joints.
To measure the occurring accelerations, a three axial acceleration
sensor 3313A1 from Dytran, which has a sensitivity of 1 mV/g, a mea-
surement range of ±5000 g, a frequency range from 1.2 to 10000 Hz,
a natural frequency >30 kHz and a weight of 4.1 g, with a high-
frequency data logger from National Instruments (NI 9232 BNC) with
3 channels and 102.4 kHz logging frequency was used with maximum
logging frequency.
3.2. Performed tests
With this test bench, drop tests were performed on a commercial
4s2p battery module for power tools with 18650 cells and a total weight
of 0.8 kg, yet without a power tool attached. To fixate the acceleration
sensor on the module, an adapter was glued onto a cell, onto which
the sensor was mounted by a screw joint, as depicted in Fig. 6(b),
following recommendations in [36]. Due to the tight packaging of the
battery module, a small hole had to be cut out of the module’s case to
create sufficient space for the sensor and the connecting cable. As drop
heights, 0.5 m, 1.0 m, 1.5 m and 2.0 m were chosen, with aluminum
and polyurethane (shore A hardness 90, thickness 5 mm) as impact
surfaces. For the module drop tests, 15 repetitions were performed
for each variation, of which the 10 tests showing least deviation were
evaluated.
3.3. Results
For the quantification of the results of the drop tests, the peak
acceleration values were evaluated. An overview of these is shown in
Table 4.
The results of the drop tests on commercial power tool battery packs
are exemplarily displayed for the shocks with 2.0 m fall height and
impact on aluminum in Fig. 7(a) and for polyurethane in Fig. 7(b).
The figures for all conducted drop tests are displayed in Fig. B.18.
The tests show intricate pulse patterns with good repeatability before
the first peak acceleration value but more deviations and long phases
with high acceleration levels after the first peak. Additionally, multiple
acceleration peaks with lower acceleration levels occur for secondary
impacts, but these are not displayed in these plots as they are much
later than visible on the applied time scale.
For the drop tests on polyurethane, the average measured peak
accelerations ranged from 800 g in the 0.5 m drop up to 1611 g in
the 2.0 m drop, while drops on aluminum ranged from 635 g to 2416
Table 4
Overview of average accelerations measured in drop tests on
single cells and battery modules.
Drop height Impact surface Peak acceleration
0.5 m Polyurethane 800 g
1.0 m Polyurethane 1049 g
1.5 m Polyurethane 1331 g
2.0 m Polyurethane 1611 g
0.5 m Aluminum 635 g
1.0 m Aluminum 1177 g
1.5 m Aluminum 1624 g
2.0 m Aluminum 2416 g
Fig. 7. Acceleration measurements of the impact of a power tool module from a 2.0
m drop. Each 10 repetitions are displayed in the plots.
Fig. 8. Illustration of the test bench for guided drop tests with a 2.0 m long tube with
diameter of 22 mm, 10 cm freefall before impact and exchangeable impact surfaces.
g, with the latter showing significantly larger deviations between the
repetitions. It is noticeable that the resulting acceleration peak values
are larger for the significantly harder aluminum surface for all tests
except the 0.5 m drop test, for which the drops on the aluminum surface
result in smaller peak values.
With this, it must be taken into consideration that not only the peak
value but also the duration and shape of the shock pulse characterize
the impact and that peak acceleration alone is not sufficient to precisely
evaluate the severity of a drop test on its own. While it is difficult to
assess the pulse durations due to their intricate form, at least for the
first peak, durations in the magnitude of 1 ms can be estimated.
3.4. Discussion
The module drop test series has shown that the impact accel-
erations of drops from 2.0 m exceed automotive or transportation
standards by far and with respect to acceleration also exceed the
severe requirements that the NASA/TM-2009-215751 test standard
recommends [12]-despite robust packaging with implemented damping
components and without a power tool attached. Furthermore, even
Journal of Energy Storage 43 (2021) 103213
6
M. Spielbauer et al.
Fig. 9. Changes in the real part of the impedance at 1 kHz for the drop tests with 8 cell types onto aluminum. Cell failure is marked in the plots with a circle. Different 𝑦-axis
scaling is marked red. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
deeper drops from more than 2.0 m are a plausible scenario, which
additionally would result in even higher accelerations and could po-
tentially cause battery failure. In this light, further investigation of the
failure mode appears to be worthwhile and is proceeded in this paper.
Regarding battery abuse testing, it is highly recommended to mea-
sure the actual accelerations in the environment of the specific use case
and to design battery tests based on the results rather than applying
commonly used standards without considering the actual loads. To re-
liably detect such impacts, as it might be desirable for online detection
in power tools, a measurement frequency of at least 5–10 kHz should
be chosen.
4. Analysis of the failure mode
After determining that accelerations in use cases can exceed even se-
vere standards and further testing with loads beyond typical standards
is worthwhile, the following section presents an investigation of the
failure mode by conducting repetitive drop tests in combination with
CT scans and Post Mortem analysis.
4.1. Repeated drop tests on single cells to provoke cell failure
To provoke and analyze the failure mode of 18650 battery cells, a
series of repeated drop tests with variations of the impact surface and
the SOC are presented in the following section.
4.1.1. Test setup
As the achievable accelerations with the pneumatic shock tester
turned out to be too low to provoke cell failure and thereby inves-
tigate the failure mechanism, further tests with this machine were
relinquished. Instead, it was decided to continue the experimental
investigations of the failure with drop tests, which cause significantly
Journal of Energy Storage 43 (2021) 103213
7
M. Spielbauer et al.
Table 5
Cell types for the drop tests from 2.0 m.
Cell type Capacity Max. charge current Max. discharge current Max. voltage Min voltage
A123 APR18650M1-A 1.1 Ah 5 A 30 A 3.6 V 2.0 V
LG ICR18650HE2 2.5 Ah 4 A 20 A 4.2 V 2.5 V
LG INR18650MJ1 3.5 Ah 4 A 10 A 4.2 V 2.5 V
Samsung INR18650-35E 3.4 Ah 8 A 13 A 4.2 V 2.65 V
Samsung INR18650 25RM 2.5 Ah 4 A 20 A 4.2 V 2.5 V
Sony / Murata Konion US18650 VTC4 2.1 Ah 4 A 30 A 4.2 V 2.5 V
Sony / Murata Konion US18650 VTC5 2.6 Ah 4 A 30 A 4.2 V 2.0 V
Sony / Murata Konion US18650 VTC6 3.12 Ah 5 A 15 A 4.25 V 2.0 V
higher accelerations and therefore have a higher chance of provoking
the failure mode. Instead of conducting further tests on modules, a
drop test bench for single cells was constructed, as this made the test
procedure and the evaluation of the cells state during the tests easier
and reduced the complexity of the load condition. Although commonly
applied standards usually recommend free fall testing, a simple test
stand with a 2.0 m tube (inner diameter 22 mm) mounted onto a
wireframe was designed, as depicted in Fig. 8, to guide single cells
during the fall and to enable improved repeatability. To achieve a well
reproducible cell impact angle and undampened cell impact at the same
time, it was decided to position the end of the tube 10 cm above the im-
pact surface. This setup allowed the usage of different impact surfaces,
which were fixated on the underlying surface to reduce bouncing. The
tests were conducted in a room with temperature regulation at ∼22 ◦C.
To avoid any influence on the cell temperature due to their handling,
gloves were used for all drop tests.
4.1.2. Performed tests
For the first set of drop tests, 8 different cell types (Table 5)
were chosen due to their internal cell design (mandrel, tab configura-
tion) [26] and repeatedly dropped on either their positive or negative
pole on aluminum. Each 4 cells for the drop tests and 1 cell for ref-
erence were handled alike. Like the previous tests with the pneumatic
shock tester, these experiments were performed on pristine cells at 0%
SOC. Before the first and after each drop, an EIS measurement was
performed at the same SOC until reaching 50 drops or until no more
valid EIS measurement was possible for the cell.
In the second test series, the Samsung INR18650 25RM was dropped
in negative pole direction onto the previously used, softer polyurethane
surface until cell failure to investigate if the same failure mode occurs
and to analyze if there are more distinct indicators of an imminent cell
failure at lower loads.
The third set of drop tests was conducted on Sony/Murata Konion
US18650 VTC6 cells, for reasons explained later, with the same setup
as the first set of drop tests but a variation of the SOCs with 25%,
50%, 75% and 100% and impact on the positive pole only. As the
batch of cells had a different production code than the cells tested
beforehand, the test with 0% SOC cells was also repeated to analyze
the repeatability of the test series.
4.2. Results
4.2.1. Drop tests with aluminum as impact surface
In the axial drop test series on aluminum with cells with 0% SOC,
none of the cells showed indications of self-heating or any thermal
event. Despite the severity of the shocks, which even lead to case
deformation for the drop tests on the negative pole, all cells withstood
at least 4 drop repetitions than was the case for the older cell type
Sony US18650 VTC4 for the drop on the negative pole. Other cell types
withstood substantially more shocks and some even the maximum of 50
drops.
As the changes of the impedance in this test series were primarily
shifts of the real part of the impedance, the analysis in Fig. 9 depicts
the real part of the impedance 𝑍𝑅𝑒 at 1000 Hz of the cells measured
Fig. 10. Nyquist Plots for the Samsung INR18650 25RM cells dropped onto their
negative pole on polyurethane after every drop test up until cell failure. (For
interpretation of the references to color in this figure legend, the reader is referred
to the web version of this article.)
during the shock tests and displays the numbers of drops until cell
failure (no more EIS measurement possible) or up until 50 drops. A
shorter overview of the results is presented in Table 6. Initial variations
in ohmic resistance can be attributed to cell-to-cell variations [37,38]
and measurement-related differences in the cables connecting the cell
holders to the galvanostats, which influence the measurement despite
four-channel sensing.
Some of the tested cell types, especially in Fig. 9(b),Fig. 9(d),
Fig. 9(o) or Fig. 9(p), exhibit an increasing real part of the impedance
before cell failure. This increase is little for some cells, but significant
for others, like for cell 3 in Fig. 9(b), which shows almost a threefold
increase in the real part of the impedance. Others such as Fig. 9(e) or
Fig. 9(h) show distinct, yet small changes compared to the reference
cells, but without a recognizable trend. Noticeable is that primarily the
cell types that exhibit no cell failure are the ones that also show no
increases in resistance. Nevertheless, these cells likely show an increase
in ohmic resistance and the same failure mode when dropped more
often, as was observed in a pre-test (depicted in Fig. C.19).
4.2.2. Drop tests with polyurethane as impact surface
In the drop tests with the Samsung INR18650 25 RM onto
polyurethane on the negative pole at 0% SOC, the cells showed no
signs of external case deformation and no signs of a thermal event.
Cell failures occurred after 230 drops for the first and after 282 drop
repetitions for the last cell.
The EIS measurements show progressive increases of the real part
of the impedance for all tested cells, with up to four times the initial
value up until cell failure. Cell 1 (Fig. 10(a)) and 2 (Fig. 10(b)) also
Journal of Energy Storage 43 (2021) 103213
8
M. Spielbauer et al.
Table 6
Overview of failed cells for each cell type and changes in 𝑍𝑅𝑒 at 1000 Hz for the failed cells at the last valid measurement
and intact cells after 50 drops.
Cell Type Cell Impact Cells
failed
Average 𝛥𝑍𝑅𝑒
with cell failure
in 𝑚𝛺
Average 𝛥𝑍𝑅𝑒
without cell failure
in 𝑚𝛺
A123 APR18650M1-A Top 2 0.23 0.07
Bottom 4 8.37 –
LG ICR18650HE2 Top 1 1.46 0.26
Bottom 2 14.20 45.77
LG INR18650MJ1 Top 0 – −0.19
Bottom 0 – −0.88
Samsung INR18650-35E Top 0 – 0.02
Bottom 0 – 0.14
Samsung INR18650 25RM Top 0 – 0.29
Bottom 0 – 0.05
Sony / Murata Konion US18650 VTC4 Top 4 0.23 –
Bottom 4 5.79 –
Sony / Murata Konion US18650 VTC5 Top 4 0.23 –
Bottom 1 0.09 −0.16
Sony / Murata Konion US18650 VTC6 Top 4 0.60
Bottom 4 5.63 –
show a shift for high frequencies in the inductive branch right before
the cell failure. While cell 3 in Fig. 10(c) shows progressive increases
in ohmic resistance, cell 4 in Fig. 10(d) shows progressive increases
at first, but after reaching a high damage level (indicated by the
darker red colors), the resistance exhibits volatile fluctuations without
a recognizable pattern up until cell failure.
4.2.3. Drop tests with variation of the SOC
The drop test results of the SOC variation of the Sony/Murata
Konion US18650 VTC6 are depicted in Fig. 11. The repetition of
reference test at 0% SOC in Fig. 11(a) shows similar resilience of the
cells as the tests on the same cell type with a different production code
in Fig. 9(o), which indicates little difference between the cells with
different production codes and good achievable reproducibility with
the drop test setup.
Regarding the impact of the SOC on the resilience of 18650 cells
towards shocks, the tests allow no conclusive assertion, yet indicate
that cells, as assumed, fail slightly faster at lower SOCs. While for this
test series none of the cells exhibited a thermal runaway either, at
higher SOCs some of the cells, e.g. cell 4 for the variation at 100%
SOC in Fig. 11(e), showed self-heating after multiple drops before the
ultimate disruption of the current path. This can also be noticed in the
decrease of the cell impedance, as well as in the Nyquist Plot of the
test in Fig. 12(a), which displays a similar pattern as a cell on which a
temperature variation was conducted in a climate chamber as depicted
in Fig. 12(b). Their comparison indicates a temperature increase of
15–20 ◦C during the drop test.
4.3. CT and post mortem analysis to investigate the failure mode
While the EIS measurements during the drop tests showed increases
in ohmic resistance right before cell failure, this observation does not
allow direct assessment of the failure mode. Therefore, to improve the
understanding of the failure mechanism, CT scans were performed and
Post Mortem analysis was conducted.
4.3.1. CT analysis
For the drop tests on aluminum and polyurethane, CT scans of the
Samsung INR18650 25RM conducted before and after the drop test
series are depicted in Fig. 13. As the cells show similar damage patterns,
the CT scans for the variation at 0% SOC can be found in Fig. D.20. The
CT scans for the cells of SOC variation are not included as they show
no differences in the observable damage patterns.
Fig. 11. Real part of the impedance at 1 kHz for the drop tests with SOC variation of
the Sony/Murata Konion US18650 VTC6 of the positive pole onto aluminum measured
after each drop. Cell failure is marked in the plots with a circle.
For the cells with impact on the positive pole onto the aluminum
surface, severe CID deformation (depicted in Fig. D.20 green) can be
observed, both for cell types with and without mandrel, as the load is
transferred from the cap at the positive pole to the CID components.
Additionally, relative jelly roll movement towards the positive pole
with deformation of the overhanging anodes and separators in the area
Journal of Energy Storage 43 (2021) 103213
9
M. Spielbauer et al.
Fig. 12. Comparison of Nyquist Plots of a Sony US18650 VTC6 cell with self-heating
due to internal short circuit with a temperature variation of the same cell type.
of the casing bead and spacer deformation can be observed (depicted
in Fig. D.20 red). This relative movement causes a similar damage
pattern to the jelly roll as axial compression does [39,40]. For the cells
with negative pole impact, case deformation and displacement of the
jelly roll accompanied by deformation of the overhanging anodes and
separators can be observed. Additionally, the negative tabs and inner
jelly roll layers show significant deformation (Fig. D.20 blue). For the
cell types with mandrels, imprints on the negative current collector tabs
can be reported.
The cells dropped onto polyurethane show no external case defor-
mation and no significant deformation of the CID, but significant jelly
roll relative movement and deformation of the overhanging anodes and
separators.
While the CT scans reveal clear signs of various damage mecha-
nisms, they do not provide evidence of what ultimately makes the cells
fail and interrupts the current path.
4.3.2. Post Mortem analysis
As analysis of the CT scans alone allowed no clear statement re-
garding the failure mechanism, Post Mortem analysis was conducted
on the cells that failed in the drop test series in Section 4.2. Complete
cell dismantling of a cell showed that the current path along the tabs
was still intact for dysfunctional cells but allowed no clear statement
regarding the current path in the positive or negative pole region as its
mechanical integrity was destroyed during the dismantling process.
To further investigate if the current path in the pole regions was
the reason for the cell failures, the caps from the positive pole were
removed, as depicted in Fig. 14, to apply manual pressure with mea-
suring tips to the CID. With this it was possible to measure the original
Fig. 14. Post Mortem analysis by removing the cap of the positive pole and applying
pressure to the CID of a defunct cell.
cell voltages temporarily. After the exertion of more force, which lead
to a slight deformation of the CID, the current path of the cells became
permanently restored. This procedure was conducted for all failed cells,
which revealed that the current path was restorable for all of them,
which is a strong indicator that the primary failure mode for 18650
cells due to high dynamic loads is high ohmic contact loss in the CID
region.
While none of the cells with 0% SOC showed any signs of tem-
perature increase during the tests, this might not have been due to
the absence of an internal short circuit, but due to the low energy
content of the cells and small resulting heat generation in the case of
such an event. For this reason, the cell voltage was measured after
3 months of storage. With this, a voltage drop >0.1 V was chosen
as an indicator of an ISC. However, it has to be mentioned that the
voltage drop could also originate from a defect seal at the positive
pole allowing air humidity to leak into the cell and thereby causing the
voltage drop. According to this criterion, only three of the Sony/Murata
Konion US18650 VTC6 cells that were dropped onto the positive pole
showed signs of an ISC with voltage drops from 2.9 V to 2.70 V, 2.47 V
and 2.07 V, which is also why this cell type was chosen for the test
series with the SOC variation. A potential reason for the occurring short
circuits for this cell type and deformation direction might be the jelly
roll deformation in the area of the bead of cell case at the positive pole,
as depicted in Fig. 15.
4.4. Discussion
The results of the drop tests onto aluminum at 0% SOC paired with
the Post Mortem analysis confirmed that, while there were also other
Fig. 13. CT Scans of the drop test with top and bottom impact of the Samsung INR18650 25RM on aluminum (Al.) and polyurethane (Pol.) at 0% SOC. Deformation of overhanging
anode (red), the CID (green) and current collectors (blue) can be observed. All CT scans from drop tests with 0% SOC can be found in Fig. D.20. (For interpretation of the
references to color in this figure legend, the reader is referred to the web version of this article.)
Journal of Energy Storage 43 (2021) 103213
10
M. Spielbauer et al.
Fig. 15. CT scan of the deformation of the jelly roll for the test of the Sony/Murata
Konion US18650 VTC6 with drop tests in the direction of the positive pole. The
short anode and separator overhang and the consequentially reduced damping effect
facilitates a strong deformation in the area of the cell case bead due to shocks.
damage mechanisms observable, the predominant failure mechanism of
18650 cells due to drop tests is contact loss in the CID region, as had
already been assumed in literature [22,26].
For the conducted drop tests, this failure mode is often precedented
by a progressive increase of ohmic resistance, which is the case for
drops both on aluminum and polyurethane. Thereby, the latter is
showing the progressive increase more clearly as a state of severe
damage is not skipped over as often as in the more damaging drops
onto aluminum. The fluctuations observed for some cells in Fig. 10(d)
might indicate loose contact within the cells.
Despite the differences in the severity of the loads and even the
increasingly deforming nature of the drop tests, the similarity in the
rise of ohmic resistance in the EIS [19,22,24–26] gives reason to
conjecture that the primary failure mechanism is the same for shocks
and vibrations.
The EIS measurements indicate a related damage mechanism, as
Figs. 10(a) and 10(b) show significant changes in the inductive branch
of the Nyquist Plots right before cell failure. A potential explanation
for this behavior could be a change in the current path, as Osswald
et al. [41] observed similar shifts in impedance for measurements on
a 26650 cell when selectively connecting some of the tabs. As not all
cells have multiple anode and cathode tabs [26], this behavior cannot
be expected from every cell type.
A further damage mechanism was observed in the SOC variation
with the Sony/Murata Konion US18650 VTC6, which was deliberately
conducted on the cell type as it also showed signs of ISCs at low SOC.
For this cell, self-heating was observed before the failure of the CID, as
shown in Fig. 12, but without causing a hazardous thermal runaway.
The more distinct deformation of the jelly roll at the bead of the cell,
which is likely the cause of the ISC, is especially severe for this cell type,
as it has a particularly short anode and separator overhangs at the top
(see Table E.8) and the bottom of the jelly roll, which act as dampening
components. This is potentially a result of the reduction of dead mass
and volume to increase the energy content or due to different electrode
balancing. Either way, this cell design adjustment seems to conflict with
the goal of cell safety.
Regarding the influence of the SOC on the resilience against drops,
the test does not allow a definite statement due to the small sample
size, yet indicates that the ultimate failure of the CID might occur
slightly faster at lower SOCs. Such behavior could also be explained
by the previously mentioned expansion of the anode at larger SOCs,
which would result in smaller void spaces in the jelly roll for relative
movement and higher contact forces between the jelly roll and the case.
Therefore, another conflict of goals seems to be between enough space
within the cell to compensate for jelly roll swelling and tight packaging
to avoid relative movement.
In summary, the following macroscopic damage mechanisms from
the drop tests could be observed:
Table 7
Overview of the pre-conditioning of the cells used for the overcharge tests.
Test Cell damage
1 None, Reference Test
2 Drop on negative pole onto aluminum 50 times
3 Drop on positive pole onto aluminum 50 times
4 Drop on positive pole onto polyurethane until 𝛥𝑅𝑖>5𝑚𝛺
5 Drop on negative pole onto polyurethane until 𝛥𝑅𝑖>5𝑚𝛺
•Relative movement of the jelly roll and deformation of the over-
hanging separators and anode
•For some cells: ISC due to deformation of overhanging separators
and anodes
•Deterioration of contacts in the CID and a resulting increase of
ohmic resistance
•For some cells in a small window before cell failure: potential
change in current path due to tab damage
•Ultimate high ohmic failure due to contact loss in the CID area
It has to be kept in mind that while there were no hazardous events
reported in this study, this might be the case when testing a larger
sample size or when testing with even higher loads, which could cause
other failure mechanisms. For this reason, even though the battery cells
showed a high safety level in the tests, occurring loads and appropriate
damping and damage protection should always be taken into account
when designing battery packs. This experiment should also not be
interpreted as a general comparative study of the robustness of the
cells against all types of dynamic loads, as some of the cells that failed
early in this test might cope better with dynamic loads with different
accelerations than others. For a systematic analysis of CID designs, the
reader is referred to Li et al. [42].
The high ohmic cell failure due to contact loss in the CID could
also be regarded as a feature that prohibits the further use of heavily
damaged cells. However, in some applications, damaged cells are dis-
charged by the BMS, which would be prevented by contact loss and
thereby harm the pack safety. Also, manual discharge of damaged cells
to allow safe transport and recycling is hindered.
5. Investigation of the impact of dynamic loads on cell safety
While the previous test series confirmed that the primary failure
mechanism is high ohmic, CT scans of cells in Section 4.3.1 also
revealed that dynamic loads can also result in a significant increase
of ohmic resistance and severe deformation of the CID. This might not
only affect the electrical behavior of the cell, but also the functionality
of the CID and could thereby cause a critical event even days or months
after the initial damaging event. To analyze if damages caused by
dynamic loads can harm the functionality of the CID of 18650 cells
overcharge tests were performed with damaged cells.
5.1. Test setup and performed tests
Therefore new and damaged cells of the type Samsung INR18650
25RM were prepared according to Table 7, with 4 cells overcharged
for each variation. These test settings were chosen to investigate if
either increase in ohmic resistance or visible damages in CT can harm
functionality of the CID. All cells were overcharged with a Gamry
Reference 3000 starting from 0% SOC with 1 C (2.5 A) constant current
(CC) with a test end criterion of 15 V. As a safety measure, the over-
charging was conducted within a self-constructed safety box (255mm
x 235mm x 183mm) with a steel case and an integrated pressure relief
flap for the case of a thermal event but lacking temperature regulation.
Nevertheless, the cell temperature was logged with thermocouples type
K.
Journal of Energy Storage 43 (2021) 103213
11
M. Spielbauer et al.
Fig. 16. Nyquist plots (top) of cells before overcharge tests and voltage over time curves (bottom) during overcharge.
5.2. Results
The Nyquist plots of the reference and the damaged cells are dis-
played in Fig. 16 after pre-conditioning according to Table 7. For the
new cells, deviations in the EIS in the low-frequency range can be
attributed to relaxation effects or temperature variations [27–30]. Some
of the cells dropped on the positive pole onto aluminum (depicted
in Fig. 16(h)), especially cell 2, showed slight increases in ohmic
resistance after the drops. The cells dropped with their negative pole
onto aluminum (depicted in Fig. 16(i)) showed very little change. For
the cells dropped on the positive pole onto polyurethane (depicted
in Fig. 16(j)), three of the cells showed changes in their inductive
behavior, as shown before, and only one changes in the real part of
the impedance. Two cells dropped onto their negative pole (depicted
in Fig. 16(k)) show the same behavior and cell 4 shows an increase
of ohmic resistance far beyond the minimal requirement of resistance
increase but was tested as an example for severe damage anyway.
The conduced overcharge tests showed that the CID consistently
prevents thermal events by interrupting the current path. While none
of the cells exhibited a thermal runaway, for the 2nd cell with the drop
on the positive pole onto the aluminum surface, electrolyte leakage
was observed shortly after the triggering of the CID. While the other
cells did not show any electrolyte leakage in a liquid form, there was
still the smell of electrolyte noticeable in the test chamber, which
indicates gaseous electrolyte leakage. Due to their toxicity and the high
potential of a hazardous event, no CT or Post Mortem analysis could be
performed on these cells.
During the overcharge tests, the voltage curve for all cells shows
typical stages during overcharge above 4.2 V, as described by Wang
et al. and Haung et al. [43,44], from stronger delithiation of the cath-
ode, to a negative potential shift of the anode and related plating, to
accelerated oxidation reaction of the electrolyte and the drop at around
5 V, where a structural change of the cathode active materials and
strong side reactions between the electrolyte and the anode interface
occur. Finally, at around 75 min after the start of the test, the CID
interrupts the current path, causing a voltage spike and the end of the
experiment.
The depiction of the voltage curve of the reference cells in Fig. 16(g)
only shows slight deviations among the overcharge curves. The cells
with positive pole impact on aluminum (depicted in Fig. 16(h)), which
results in intense deformation of the CID, show more distinct devia-
tions, yet the CID interrupted the current path within a window of a
few minutes. Cells dropped on the negative pole (depicted in Fig. 16(i))
show similar behavior but more substantial differences in the voltage
curve at the start of the test. The cells with top impact on polyurethane
with small visible deformation in the CID area but significant resistance
increase show low initial deviations in their voltage curve but more
Fig. 17. Temperature and voltage curves for reference cell 1 (Fig. 16(g)) and cell 4
of the test with impact on polyurethane on the negative pole (Fig. 16(k)).
distinct deviations before the triggering of the CID with several minutes
in between their triggering times, as depicted in Fig. 16(j). Significant
differences can be observed in the test of the cells with bottom impact
onto polyurethane (depicted in Fig. 16(k)), where cell 4, which showed
a significant increase of ohmic resistance, shows a voltage curve around
0.2 mV above all other cells with strong fluctuations. Nevertheless,
even for this cell, the CID interrupted the current path.
As the test chamber offered no temperature control, the temperature
curves recorded show environmental deviations up to 5 ◦C for many
of the tests, there is no unequivocal statement possible if the behavior
occurred due to ambient temperature deviations or mechanical cell
damage. Merely cell 4 for the drop test on polyurethane with negative
pole exhibits a strong correlation in its temperature behavior to the
voltage curve, as depicted in Fig. 17.
5.3. Discussion
Despite the severe mechanical damages, the functionality of the CID
remained intact in all overcharge tests and reliably prevented a thermal
event. With this, it has to be kept in mind, that the chosen voltage test
end criterion could be less critical than a criterion that additionally
allows CV-charging.
Regarding the electrical behavior, the cells with resistance increase
showed deviations from the reference cells in the voltage curves, and,
especially cell 4 of the cells with bottom impact onto polyurethane, also
in the temperature curve due to ohmic losses. This oscillating behavior
in the cell voltage seems to be another sign of loose contact in the CID
region, which can not only be found for mechanically damaged cells
but also right before the interruption of the current path by the CID
as reported in the overheating of 18650 cells by Srinivasan et al. [45].
The observed abnormal thermal and electrical behavior might be an
approach for the detection of damaged cells in the BMS, but considering
Journal of Energy Storage 43 (2021) 103213
12
M. Spielbauer et al.
that the cells with the even stronger deformed CIDs in the drop tests
onto aluminum showed significantly fewer variations to the reference
cells, it is hardly a reliable indicator of mechanical damage caused by
shocks.
6. Conclusion
After the completion of the experimental studies, the results are
concluded in the following section.
The literature analysis showed that 18650 cells subjected to dy-
namic loads may exhibit signs of damage such as an increase in ohmic
resistance or imprints of mandrels in current collector tabs, yet known
publications do not further investigate the ultimate failure mechanism.
Therefore, as a first experiment to potentially provoke mechanical
cell failure, axial shock tests with 1000 repetitions, 300 g acceleration
and 6 ms pulse width were conducted on 6 different cell types. It was
observed, that under these test conditions, older cell types and types
with suboptimal mechanical cell design exhibit damage mechanisms
known from literature, such as an increase of ohmic resistance or
imprints on the current collector tabs due to a loose jelly roll. Still,
no cell failure was reported. On the other hand, state-of-the-art cells
showed no changes in EIS nor in CT scans due to their improved cell
design.
To investigate if higher accelerations occur beyond these standards
and therefore if further testing of the failure mode is practically rel-
evant, acceleration measurements of the impact of commercial power
tool battery packs from different heights and on various impact surfaces
were executed. For this test series, a drop test bench for modules was
used. The results show accelerations above 500 g for a drop test from
0.5 m up to 3000 g for drop tests from 2.0 m, which exceed widely
applied battery test standards by far and even go beyond the test
conditions in high load standards and strengthened the need for further
investigation.
In this further investigation, the failure mode was analyzed using
a guided drop test stand for cells. With this test stand, drop tests with
8 different cell types, different orientations, impact surfaces and SOCs
were conducted. The results showed increases in ohmic resistance and
ultimately high ohmic failure for multiple modern cell types, indicating
that cells that did not fail in the tests likely exhibit the same behavior
after further loading. Also, the variation of the impact surface resulted
in the same failure mechanism. For the cell type used in the SOC
variation, self-heating and voltage drainage due to ISCs were reported,
but no thermal runaway occurred. The tests indicated, that cells with
higher SOCs have a slightly higher resilience against high ohmic failure
but show more intense thermal reactions in the case of an ISC.
CT and Post Mortem analysis revealed that the predominant failure
mechanism under dynamic loads is the loss of contact in the CID
region. For the cell type that exhibited ISCs, their cause was likely
the collision of the jelly roll with the bead of the case. The same
behavior was not reported for the other modern cells as those still
feature somewhat larger anode and separator overhangs, which act
as a dampening component. These were possibly reduced for the cell
type that showed ISCs to increase its energy content or for electrode
balancing. This reveals that for modern battery cells there might be a
rising conflict of objectives in cell design between energy density and
safety.
Finally, potential implications of mechanical damages on the elec-
trical behavior and the functionality of the CID were investigated by
overcharge tests on pre-damaged cells. Despite the severity of the
damages and critical electrical and thermal abnormalities that one cell
showed, the CID remained functional for all tested cells and prevented
thermal runaway.
The conducted tests lead to the conclusion, that while loads in some
applications may significantly exceed widely applied standards, state-
of-the-art 18650 lithium-ion batteries are very resilient against dynamic
loads. Even if a failure occurs, it is likely to be high ohmic and therefore
probably not a direct safety issue, yet may harm safety functions on a
BMS level. However, a more conclusive statement would require larger
sample sizes. In any case, it is recommended that application-specific
loads should always be taken into account when designing battery
packs and defining test conditions for abuse tests.
Regarding the cell design, the following partly conflicting features
for modern cells were derived:
•Tight packaging with dampening components to avoid relative
movement.
•Small separator and anode overhang to increase energy density,
but big enough overhang to dampen impacts.
•Big enough gaps to compensate for jelly roll swelling to enhance
life span, but small gaps between jelly roll and case to prohibit
relative movement between jelly roll and case.
•Use of a mandrel to avoid mechanical jelly roll collapse dur-
ing aging [46] or mandrel-free cell design, as it might damage
components within the cell due to relative movement.
Further research should investigate critical loads for 18650 cells by
conducting experiments with a high acceleration shock tester and cre-
ate a relation between failure probability, shock severity (accelerations
and pulse widths) and repetitions. Investigations could also include
shock tests on aged cells, aging studies of cells after damaging from
dynamic loads, more profound testing regarding the influence of the
SOC and tests with different cell formats (e.g. 21700). While the cells
appear to be very safe, there is little knowledge about potential hazards
on module level (e.g. the resilience of cell connectors).
CRediT authorship contribution statement
Markus Spielbauer: Conceptualization, Methodology, Software,
Validation, Formal analysis, Investigation, Data curation, Writing –
original draft, Writing – review & editing, Visualization. Philipp Berg:
Conceptualization, Methodology, Resources, Writing – review & edit-
ing, Supervision, Funding acquisition. Jonas Soellner: Conceptualiza-
tion, Methodology, Resources, Writing – review & editing. Julia Peters:
Validation, Investigation, Resources, Data curation, Writing – review &
editing. Florian Schaeufl: Software, Validation, Investigation, Writing
– review & editing. Christian Rosenmüller: Validation, Investigation,
Writing – review & editing. Oliver Bohlen: Conceptualization, Method-
ology, Resources, Writing – review & editing, Supervision, Project
administration, Funding acquisition. Andreas Jossen: Conceptualiza-
tion, Methodology, Resources, Writing – review & editing, Supervision,
Project administration, Funding acquisition.
Declaration of competing interest
The authors declare that they have no known competing finan-
cial interests or personal relationships that could have appeared to
influence the work reported in this paper.
Acknowledgments
Funding from the German Federal Ministry for Economic Affairs and
Energy (BMWi) of the project ReVISEDBatt [03ETE004C] and manag-
ing by the Projektträger Jülich (PtJ) is gratefully acknowledged. This
work was financially supported by the Munich University of Applied
Sciences HM and the German Research Foundation (DFG) through the
‘‘Open Access Publishing’’ program. The authors want to express their
gratitude to Robert Stanger, Varnim Goyal and Leonard Janczyk from
Hilti Entwicklungsgesellschaft mbH for fruitful discussions as well as
for supplying the test series with material. The authors also want to
thank Prof. Dr. Gregor Feiertag for supporting this project by granting
access to the CT scanner.
Journal of Energy Storage 43 (2021) 103213
13
M. Spielbauer et al.
Appendix A. Abbreviations
See Table A.1.
Table A.1
Abbreviations.
BMS Battery Management System
CID Current Interruptive Device
EIS Electrochemical Impedance Spectroscopy
ISC Internal Short Circuit
CC Constant Current
CT Computed Tomography
OCV Open circuit voltage
SOC State of Charge
SOH State of Health
Appendix B. Acceleration measurements of module drop tests
See Fig. B.18.
Fig. B.18. Drop tests of a commercial power tool battery module onto different impact surfaces from various heights.
Appendix C. Impedance measurements of a drop test of the Samsung INR18650-35E
See Fig. C.19.
Fig. C.19. Nyquist Plot of a preconducted drop test on a Samsung INR18650-35E onto polyurethane with an EIS measurement every 20 drops.
Journal of Energy Storage 43 (2021) 103213
14
M. Spielbauer et al.
Appendix D. CT scans of drop tests on aluminum
See Fig. D.20.
Fig. D.20. CT scans of drop tests on aluminum. Each time a new cell, a cell with top impact and a cell with bottom impact are displayed. Deformation of overhanging anode
(red), the CID (green) and current collectors (blue) can be observed. (For interpretation of the references to color in this figure legend, the reader is referred to the web version
of this article.)
Appendix E. Anode overhang lengths
See Table E.8.
Table E.8
Lengths of anode overhangs at the positive pole of cells used in the drop test series.
Cell Type A123 APR18650M1-A LG ICR18650HE2 LG INR18650MJ1 Samsung INR18650-35E Samsung INR18650 25RM Sony/Murata Konion US18650 VTC4 Sony/Murata Konion US18650 VTC5 Sony/Murata Konion US18650 VTC6
Min. Overhang Length in mm 0.31 1.06 0.65 0.66 0.51 0.43 0.67 0.38
Max. Overhang Length in mm 1.26 1.26 0.83 0.76 0.65 0.68 0.98 0.54
References
[1] V. Ruiz, A. Pfrang, A. Kriston, N. Omar, P. van den Bossche, L. Boon-Brett, A
review of international abuse testing standards and regulations for lithium ion
batteries in electric and hybrid electric vehicles, Renew. Sustain. Energy Rev. 81
(2018) 1427–1452, http://dx.doi.org/10.1016/j.rser.2017.05.195.
[2] S. Arora, W. Shen, A. Kapoor, Review of mechanical design and strategic
placement technique of a robust battery pack for electric vehicles, Renew.
Sustain. Energy Rev. 60 (2016) 1319–1331, http://dx.doi.org/10.1016/j.rser.
2016.03.013.
[3] UN/ECE-R100.02, Regulation no 100 of the economic commission for europe of
the united nations (UNECE) — Uniform provisions concerning the approval of
vehicles with regard to specific requirements for the electric power train, 2015.
[4] IEC62660-2, Secondary lithium-ion cells for the propulsion of electric road
vehicles - part 2: Reliability and abuse testing, 2018.
[5] ISO16750-3, Road vehicles — Environmental conditions and testing for electrical
and electronic equipment — Part 3: Mechanical loads, 2013.
[6] UL 2580, Batteries for use in electric vehicles, 2013.
[7] UL1642, Standard for safety-lithium batteries, 2012.
[8] SAE J2464, Electric and hybrid electric vehicle rechargeable energy storage
system (RESS) safety and abuse testing, 2010.
[9] SAE J2929, Safety standard for electric and hybrid vehicle propulsion battery
systems utilizing lithium-based rechargeable cells, 2013.
[10] UN38.3, Transport of dangerous goods: Manual of tests and criteria Fifth revised
edition 43–51.
[11] MIL-STD-202G, Department of defense test method standard electronical and
electrical component parts, 2002.
[12] NASA/TM-2009-215751, Guidelines on lithium-ion battery use in space
applications, 2009.
[13] IEC 60068-2-27, Environmental testing - part 2-27: Tests - test ea and guidance:
Shock, 2008.
[14] QC/T 743, Lithium-ion batteries for electric vehicles, 2006.
[15] FreedomCAR, Electrical energy storage system abuse test manual for electric and
hybrid electric vehicle applications, 2006.
[16] MIL-STD-810H, Department of defense test method standard: Environmental
engineering considerations and laboratory tests, 2019.
[17] G. Kjell, J.F. Lang, Comparing different vibration tests proposed for li-ion bat-
teries with vibration measurement in an electric vehicle, in: EVS27 International
Battery, Hybrid and Fuel Cell Electric Vehicle Symposium, 2013.
[18] J.F. Lang, G. Kjell, Comparing vibration measurements in an electric vehicle with
standard vibration requirements for Li-ion batteries using power spectral density
analysis, Int. J. Electr. Hybrid Veh. 7 (3) (2015) 272, http://dx.doi.org/10.1504/
IJEHV.2015.071640.
[19] M.J. Brand, S.F. Schuster, T. Bach, E. Fleder, M. Stelz, S. Gläser, J. Müller, G.
Sextl, A. Jossen, Effects of vibrations and shocks on lithium-ion cells, J. Power
Sources 288 (2015) 62–69, http://dx.doi.org/10.1016/j.jpowsour.2015.04.107.
[20] T. Tsujikawa, K. Yabuta, M. Arakawa, K. Hayashi, Safety of large-capacity
lithium-ion battery and evaluation of battery system for telecommunications,
J. Power Sources 244 (2013) 11–16, http://dx.doi.org/10.1016/j.jpowsour.2013.
01.155.
[21] F. Ebert, G. Sextl, M. Lienkamp, Influence of dynamic mechanical stress on
lithiumion-battery aging, in: 6th Conference on Future Automotive Technology,
2017.
[22] J. Hooper, J. Marco, G. Chouchelamane, C. Lyness, Vibration durability testing of
nickel manganese cobalt oxide (NMC) Lithium-Ion 18,650 battery cells, Energies
9 (1) (2016) 52, http://dx.doi.org/10.3390/en9010052.
[23] T. Bruen, J. Hooper, J. Marco, M. Gama, G. Chouchelamane, Analysis of a battery
management system (BMS) control strategy for vibration aged nickel manganese
cobalt oxide (NMC) lithium-ion 18650 battery cells, Energies 9 (4) (2016) 255,
http://dx.doi.org/10.3390/en9040255.
[24] J.M. Hooper, J. Marco, G.H. Chouchelamane, J.S. Chevalier, D. Williams, Multi-
axis vibration durability testing of lithium ion 18650 NCA cylindrical cells, J.
Energy Storage 15 (2018) 103–123, http://dx.doi.org/10.1016/j.est.2017.11.006.
[25] L. Somerville, J. Hooper, J. Marco, A. McGordon, C. Lyness, M. Walker, P.
Jennings, Impact of vibration on the surface film of lithium-ion cells, Energies
10 (6) (2017) 741, http://dx.doi.org/10.3390/en10060741.
[26] P. Berg, M. Spielbauer, M. Tillinger, M. Merkel, M. Schoenfuss, O. Bohlen,
A. Jossen, Durability of lithium-ion 18650 cells under random vibration load
with respect to the inner cell design, J. Energy Storage 31 (2020) 101499,
http://dx.doi.org/10.1016/j.est.2020.101499.
Journal of Energy Storage 43 (2021) 103213
15
M. Spielbauer et al.
[27] C.T. Love, M.B. Virji, R.E. Rocheleau, K.E. Swider-Lyons, State-of-health monitor-
ing of 18650 4S packs with a single-point impedance diagnostic, J. Power Sources
266 (2014) 512–519, http://dx.doi.org/10.1016/j.jpowsour.2014.05.033.
[28] C. Love, M. Dubarry, T. Reshetenko, A. Devie, N. Spinner, K. Swider-Lyons, R.
Rocheleau, Lithium-ion cell fault detection by single-point impedance diagnostic
and degradation mechanism validation for series-wired batteries cycled at 0 ◦C,
Energies 11 (4) (2018) 834, http://dx.doi.org/10.3390/en11040834.
[29] J.P. Schmidt, S. Arnold, A. Loges, D. Werner, T. Wetzel, E. Ivers-Tiffée,
Measurement of the internal cell temperature via impedance: Evaluation and
application of a new method, J. Power Sources 243 (2013) 110–117, http:
//dx.doi.org/10.1016/j.jpowsour.2013.06.013.
[30] H.P.G.J. Beelen, L.H.J. Raijmakers, M.C.F. Donkers, P.H.L. Notten, H.J. Bergveld,
A comparison and accuracy analysis of impedance-based temperature estimation
methods for li-ion batteries, Appl. Energy 175 (5) (2016) 128–140, http://dx.
doi.org/10.1016/j.apenergy.2016.04.103.
[31] A. Barai, G.H. Chouchelamane, Y. Guo, A. McGordon, P. Jennings, A study on the
impact of lithium-ion cell relaxation on electrochemical impedance spectroscopy,
J. Power Sources 280 (2015) 74–80, http://dx.doi.org/10.1016/j.jpowsour.2015.
01.097.
[32] F.M. Kindermann, A. Noel, S.V. Erhard, A. Jossen, Long-term equalization
effects in li-ion batteries due to local state of charge inhomogeneities and their
impact on impedance measurements, Electrochim. Acta 185 (2015) 107–116,
http://dx.doi.org/10.1016/j.electacta.2015.10.108.
[33] J. Park, S. Kalnaus, S. Han, Y.K. Lee, G.B. Less, N.J. Dudney, C. Daniel, A.M.
Sastry, In situ atomic force microscopy studies on lithium (de)intercalation-
induced morphology changes in LixCoO2 micro-machined thin film electrodes, J.
Power Sources 222 (2013) 417–425, http://dx.doi.org/10.1016/j.jpowsour.2012.
09.017.
[34] B. Rieger, S. Schlueter, S.V. Erhard, J. Schmalz, G. Reinhart, A. Jossen, Multi-
scale investigation of thickness changes in a commercial pouch type lithium-ion
battery, J. Energy Storage 6 (2016) 213–221, http://dx.doi.org/10.1016/j.est.
2016.01.006.
[35] K.-Y. Oh, J.B. Siegel, L. Secondo, S.U. Kim, N.A. Samad, J. Qin, D. Anderson,
K. Garikipati, A. Knobloch, B.I. Epureanu, C.W. Monroe, A. Stefanopoulou,
Rate dependence of swelling in lithium-ion cells, J. Power Sources 267 (2014)
197–202, http://dx.doi.org/10.1016/j.jpowsour.2014.05.039.
[36] T. Kuttner, Praxiswissen Schwingungsmesstechnik, Springer Fachmedien, Wies-
baden, 2015.
[37] S.F. Schuster, M.J. Brand, P. Berg, M. Gleissenberger, A. Jossen, Lithium-ion
cell-to-cell variation during battery electric vehicle operation, J. Power Sources
297 (2015) 242–251, http://dx.doi.org/10.1016/J.JPOWSOUR.2015.08.001.
[38] K. Rumpf, M. Naumann, A. Jossen, Experimental investigation of parametric cell-
to-cell variation and correlation based on 1100 commercial lithium-ion cells, J.
Energy Storage 14 (2017) 224–243, http://dx.doi.org/10.1016/j.est.2017.09.010.
[39] M. Raffler, A. Sevarin, C. Ellersdorfer, S.F. Heindl, C. Breitfuss, W. Sinz, Finite
element model approach of a cylindrical lithium ion battery cell with a focus
on minimization of the computational effort and short circuit prediction, J.
Power Sources 360 (2017) 605–617, http://dx.doi.org/10.1016/j.jpowsour.2017.
06.028.
[40] J. Zhu, X. Zhang, E. Sahraei, T. Wierzbicki, Deformation and failure mechanisms
of 18650 battery cells under axial compression, J. Power Sources 336 (2016)
332–340, http://dx.doi.org/10.1016/j.jpowsour.2016.10.064.
[41] P.J. Osswald, S.V. Erhard, A. Noel, P. Keil, F.M. Kindermann, H. Hoster, A.
Jossen, Current density distribution in cylindrical li-ion cells during impedance
measurements, J. Power Sources 314 (2016) 93–101, http://dx.doi.org/10.1016/
j.jpowsour.2016.02.070.
[42] W. Li, K.R. Crompton, C. Hacker, J.K. Ostanek, Comparison of current interrupt
device and vent design for 18650 format lithium-ion battery caps, J. Energy
Storage 32 (2020) 101890, http://dx.doi.org/10.1016/j.est.2020.101890.
[43] L. Huang, Z. Zhang, Z. Wang, L. Zhang, X. Zhu, D.D. Dorrell, Thermal runaway
behavior during overcharge for large-format lithium-ion batteries with different
packaging patterns, J. Energy Storage 25 (2019) 100811, http://dx.doi.org/10.
1016/j.est.2019.100811.
[44] Q. Wang, P. Ping, X. Zhao, G. Chu, J. Sun, C. Chen, Thermal runaway caused
fire and explosion of lithium ion battery, J. Power Sources 208 (2012) 210–224,
http://dx.doi.org/10.1016/j.jpowsour.2012.02.038.
[45] R. Srinivasan, M.E. Thomas, M.B. Airola, B.G. Carkhuff, L. Frizzell-Makowski,
H. Alkandry, J.G. Reuster, H.N. Oguz, P.W. Green, J. La Favors, L.J. Currano,
P.A. Demirev, Preventing cell-to-cell propagation of thermal runaway in lithium-
ion batteries, J. Electrochem. Soc. 167 (2) (2020) 020559, http://dx.doi.org/10.
1149/1945-7111/ab6ff0.
[46] A. Pfrang, A. Kersys, A. Kriston, D.U. Sauer, C. Rahe, S. Käbitz, E. Figgemeier,
Geometrical inhomogeneities as cause of mechanical failure in commercial 18650
lithium ion cells, J. Electrochem. Soc. 166 (15) (2019) A3745–A3752, http:
//dx.doi.org/10.1149/2.0551914jes.