Journal of Power Sources 535 (2022) 231423

Contents lists available at ScienceDirect

Journal of Power Sources

journal homepage: www.elsevier.com/locate/jpowsour

Avoiding thermal runaway in lithium-ion batteries using ultrasound

detection of early failure mechanisms
Maura C. Appleberry a, Jeffrey A. Kowalski a, *, Steven A. Africk a, Jared Mitchell b,
Thomas C. Ferree a, Vincent Chang b, Vashisth Parekh b, Ziyi Xu b, Ziwen Ye b, Jay F. Whitacre b,
Shawn D. Murphy a
a
Titan Advanced Energy Solutions, Salem, MA, 01970, USA
b
Materials Science and Engineering Department, Carnegie Mellon University, Pittsburgh, PA, 15213, USA

H I G H L I G H T S G R A P H I C A L A B S T R A C T

• Ultrasound can detect lithium-ion bat­

tery failure induced by overcharge.
• Two points of battery failure were
identified including an actionable
warning.
• Using ultrasound, acceleration of failure
conditions may be stopped in real time.

A B S T R A C T

Thermal runaway leading to catastrophic failure has slowed the adoption of lithium-ion batteries, highlighting the need for early warning systems. In this work,
ultrasound is used to detect physical changes in 0.950 Ah Li-ion battery cells by inducing conditions associated with thermal runaway. Ultrasound signal features are
extracted as the batteries are cycled under baseline conditions and subsequently subjected to constant current and voltage overcharges. Two conditions identifying
failure are defined: a warning to detect the start of overcharge and an emergency stop (E-stop), to immediately take the battery out of service. Using changes in
ultrasound signal features, consistent warning and stopping classifications are applied to all failure experiments. On average, a warning notification was issued after
15% of the overcharge time, and an E-stop was triggered after 35% of total overcharge duration before failure. Further testing suggests it is possible to avoid thermal
runaway by reacting to intermediate, warning notifications prior to an immediate E-stop. In other words, not only can detection derived by ultrasound identify
battery failure before a catastrophic event, but can also provide early, actionable warnings so that overcharges can be detected and corrected quickly enough so the
battery does not need to be decommissioned.

1. Introduction
A major challenge of the 21st century is the development of efficient
and sustainable means of energy conversion, distribution, and storage
on a global scale. To enable the widespread integration of renewable
energy sources (e.g., solar photovoltaic, wind, tidal) and in the
improvement of energy efficiency of the electric grid, stationary energy
storage systems (ESSs) are critical [1]. The automotive industry, a
substantial contributor to CO2 emissions, has experienced continued
pressure from governments and consumers to move toward

* Corresponding author.
E-mail address: jeff@titanaes.com (J.A. Kowalski).

https://doi.org/10.1016/j.jpowsour.2022.231423
Received 12 January 2022; Received in revised form 2 April 2022; Accepted 4 April 2022
Available online 23 April 2022
0378-7753/© 2022 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
M.C. Appleberry et al.
electrification, replacing gasoline powered engines with electric ones
[2].
While a range of energy storage options exist to meet the afore­
mentioned services, no single technology (e.g., pumped hydro, electro­
chemical storage) is suitable for all applications [1,3,4]. Electrochemical
energy storage technologies, specifically lithium-ion batteries, have a
number of desirable characteristics including high power/energy den­
sity, high round-trip efficiency, rapid response time, and portability
[5–8]. Recently, these batteries have become the dominant choice due to
their decreasing costs, high cyclability, and technological maturity [3,
6]. However, one major hindrance of widespread lithium-ion battery
adoption is the possibility of a battery/system failure, particularly fires
and explosions, and the dangers and publicity associated with them [2,
9]. Battery safety is imperative for the widespread adoption of electric
vehicles (EVs) to protect consumers. New regulations on EV develop­
ment in the United Nations Global Technical Regulation on the Electric
Vehicle Safety specify that a battery system must provide either a
warning to allow egress or 5 min prior to the presence of a hazardous
condition that could lead to thermal runaway to be compliant [10].
Thermal runaway events, and ensuing fires and explosions, are the
primary cause of catastrophic lithium-ion battery failure [2,11,12].
Elevated ambient temperature (typically above 80 ◦ C) can trigger
additional internal cell heating, leading to a positive feedback cycle,
resulting in an uncontrollable exothermic reaction [2,8,13]. The pri­
mary trigger for this type of event is the generation of heat and/or gas
within the battery cell, commonly through the decomposition of the
electrode materials and/or dissociation of the liquid electrolyte, which is
a flammable carbonate. This can be caused by several factors, including
internal and external short circuiting, electrolyte degradation, increased
cell resistance resulting in cell overheating during higher current use
conditions, and gas generation resulting from either overcharging,
overheating, or over discharging and subsequent electrolyte electrolysis
[11,12,14].
In stationary ESSs, EVs, and consumer electronics devices, internal
and external short circuit hazards tend to be the most prevalent and
cause the majority of thermal runaway events [12,14]. However, the
propensity for internal short circuits to occur cannot be controlled by
electronics or detected in advance with existing technology. Although
internal short circuits will not immediately cause thermal runaway, it is
difficult to detect short circuiting until it is happening, therefore it was
not considered in this work. Gas generation and electrolyte degradation
present impending danger and allow other means of thermal runaway.
During failure, lithium-ion cells move into a hazardous state as they
evolve flammable gases within the battery, which result in internal
pressure and swelling [15]. The main component in a lithium-ion bat­
tery pack responsible for minimizing risks associated with operation is
the Battery Management System (BMS) [16]. A typical BMS has inputs of
battery current measurements, voltage sensors at each cell, and tem­
perature measurements [17]. As exhibited through repeated, historical,
catastrophic events, presently used BMSs are not sufficient for reliably
detecting battery failure [18–22]. In their review of failure mechanisms
of lithium-ion batteries, Wang et al. [22] identify causes of notable
failures since 2010. Other than overheating and short circuiting, com­
mon battery failures include overcharging of the battery due to BMS
failure in applications ranging from electric scooters, EVs, and aircrafts.
In addition, defects leading to thermal runaway and venting were
deduced as the cause of a large ESS fire in Arizona, USA, leading to an
explosion from accumulated gases [18].
Presently, mechanical safeguards such as a current interrupt device
(CID) and safety vents (burst safety disks) to release cell pressure as gas
builds up are installed in most cells to prevent hazards when other safety
methods fail. In addition, considerable work to reduce risk involves
altering anode, cathode, and electrolyte chemistries and improving
technologies that remove flammable electrolyte entirely such as solid-
state batteries [23].
Active areas of research involve means of detecting battery failure
2
Journal of Power Sources 535 (2022) 231423
that fall into four main categories: algorithms based on BMS parameters,
measuring mechanical deformation using creep distance and pressure
sensors, temperature measurements, and gas detection [8,24,25]. Many
algorithm-based detection methods estimate state-of-charge (SoC) and
impedance [26–28], while some employ an equivalent-circuit-model to
detect battery failure by observing contrary measurements of a voltage
response [16,29,30], and some leverage both [31]. However, algo­
rithmic methods that rely on existing voltage, current, and temperature
sensors are cost optimized in practice, but rely on the durability of the
BMS. In pursuit of additional sensors for detection, Koch et al. [32]
reviewed seven sensors including voltage, gas, smoke, creep distance,
temperature, pressure, and force of a cell within a module and deter­
mined that gas, pressure, and force sensors present the most timely
detection warnings, whereas the gas sensor method is able to detect a
thermal runaway event most consistently. However, this work largely
observed that each sensor has challenges and disadvantages [32], and
the optimal solution highly depends on system design. In addition, it is
not practical or possible to add internal thermometers to many batteries
already in use that could be used for energy storage in the near term (i.e.,
second life systems). Moreover, temperature measurements are not
reliable because many temperature sensing detection methods do not
account for increasing impedance [26] and the natural temperature rise
due to aging. Gas sensing is widely applied to various electrochemical
energy storage solutions, however initial gas release at around 120%
SoC is anaerobic with increasing pressure but is difficult to detect [24]. It
is not until a cell vents that this hazardous event may be detected. Such
methods only detect thermal runaway events in progress and therefore
could provide a mere 30 s warning, requiring extremely rapid response
that is often not practical.
Acoustic propagation changes within a battery as its material prop­
erties change due to cycling and aging. Ultrasound has been used by
many disciplines for material characterization, and has recently shown
promise as an additional technique for battery measurements [33–37].
For example, Davies et al. [35] show how changes in SoC may be
detected by differences in ultrasound time of flight. More recently,
Copley et al. [38] created a model to demonstrate how material changes
during cycling affect ultrasound wave reflection and demonstrated how
the measurement of time of flight affects measurements of SoC derived
by ultrasound. Although most work is focused on detecting SoC and
state-of-health (SoH) changes [35,39,40], there is preliminary explora­
tion of failure mechanisms [36,41]. Since faults are known to occur from
dormant defects that worsen through battery use [42], ultrasound is a
reasonable approach to detect the accumulation of these defects sooner
than existing detection methods for more than SoC and SoH changes.
In this work, we show that this ultrasound methodology can detect
potentially dangerous conditions rapidly after they were experimentally
induced, consistent with venting and accidental overcharge in 0.950 Ah
commercial battery cells, demonstrating the feasibility of generating
actionable alarms, well in advance of a catastrophic failure. In slightly
more detail, it was demonstrated that two types of alarms could be
generated, an early indication that a battery has entered an abnormal
state and an emergency-stop to register that a battery has entered a
condition almost certain to lead to a catastrophic failure. Both types of
indications were generated at times earlier than possible with present
technologies.
2. Materials & methods
To collect ultrasound data on cell failure, LG1P-531A, 0.950 Ah
batteries were used for experimentation. Each battery was placed and
centered in an in-house-designed experimental fixture which aligns ul­
trasound transducers (NDT, CHG014) on either side of the battery and
provides sufficient contact pressure between the battery and the ultra­
sound sensor. Approximately ten pounds of force were applied to the
transducers via a compression spring (McMaster-Carr, 9657K338).
Nominally consistent force was ensured by applying consistent
M.C. Appleberry et al. Journal of Power Sources 535 (2022) 231423

Table 1 Table 2
Reference cycling profile. Step 1 shows voltage limits for CC charging at Battery Failure Experiment List. Nine failure experiments consisting of over­
specified current. Step 2 for all experiments is either CC or CV overcharge. charge conditions were explored. Experiment 10 provided reference for com­
Step Description Voltage Current
parison for post-mortem results. The high temperature, CV overcharge
experiment was completed 3 times as it resulted in explosion.
1. Baseline Cycling 2.5 V–4.2 V 0.250 A
2. CC overcharge N/A 0.400 A Test Temperature Overcharge Condition
CV overcharge 5.5 V 5.000 A limit 1 RT, ~23 ◦ C CC overcharge
2 RT, ~23 ◦ C CC overcharge
3 RT, ~23 ◦ C CV overcharge
displacement in each fixture. For each setup, two transducers on oppo­ 4 RT, ~23 ◦ C CV overcharge
site sides of the battery were used such that one transducer transmits an 5 65 ◦ C CV overcharge
ultrasound wave that is received by the opposite transducer, comprising 6 65 ◦ C CV overcharge
7 65 ◦ C CV overcharge
transmission data. To ensure contact between the battery and transducer 8 65 ◦ C CC overcharge
enabling ultrasound transmission, 1/64th inch Viton rubber (McMaster- 9 65 ◦ C CC overcharge
Carr, 1235N11) with 0.0045 inch adhesive (The 3M Company, 9475LE) 10 RT, ~23 ◦ C None, continuous cycling
on one side was used as a couplant between the face of the transducer
and the surface of the battery to facilitate transmission. The adhesive
cycling data and ultrasound data were synchronized such that each ul­
was applied between the Viton rubber and the transducer face. A
trasound capture and temperature had a unique voltage, current,
resistance-based temperature sensor board (Microchip, MCP9808) was
maximum cycle capacity, and SoC. State-of-charge was calculated based
placed on the surface of the battery. Temperature sensors and trans­
on total half cycle.
ducers were connected to an in-house, purpose-designed circuit board
that can generate a reliable ultrasound pulse autonomously and collect Charge/Discharge capacity [Ah]
SoC = × 100 (1)
resultant data rapidly. A diagram of the fixture assembly illustrating the Total half cycle capacity [Ah]
components, as well as images of the testing equipment is included in
To induce battery failure, batteries were moved to a VMP3 Biologic
the SI (Figs. S1–S3). Transducer cables were 1 m in length. When an
ultrasound pulse is initiated, signal gain is applied to maximize signal potentiostat to complete either a CC overcharge at 0.4 A without an
upper voltage limit or a constant voltage (CV) overcharge at 5.5 V until
input on the analog-to-digital converter (ADC) on the board. An ultra­
sound capture is the average response of 16 bipolar square wave pulses the battery vented (Table 2). Each failure experiment was completed at
either room temperature (RT, ~23 ◦ C) or high temperature (~65 ◦ C) in a
(positive first) with a center frequency of 500 kHz and impulse voltage
of 90 V. After a pulse, the response is collected for ~70 μs, and there is a thermal chamber. During each overcharge experiment, ultrasound
captures were collected every 30 s. A full list of experiments is shown in
20 ms rest between each of the 16 pulses. The average of the pulses is
Table 2.
then filtered with a bandpass filter (300 kHz–1.7 MHz) to reduce low
After ultrasound data were collected during the failure, post-mortem
and high frequency noise.
analysis was performed on each of the cells. Scanning electron micro­
The failure experiments consisted of 2 steps: 1) cycling cells under
scopy and electron dispersive x-ray spectroscopy were performed to
normal parameters to establish an ultrasound baseline for at least 30 full
examine the anode and cathode of uncycled and failed batteries to
charge/discharge cycles, and 2) introduction of abnormal conditions via
identify failure mechanisms. Data from Experiment 10 was used to
overcharge until the battery experienced complete failure (e.g., battery
compare failure mechanisms from typical signatures of non-failed
venting, battery explosion). To establish a baseline, batteries were
batteries.
cycled on a Neware BTS4000 series 8-channel 5V6A cycler employing a
In addition to the baseline data collected for each failure (Table 1),
constant current (CC) charge and CC discharge following parameters in
baseline data (step 1) was collected for different temperatures to
Table 1. Baseline cycling was collected at ambient temperature. Since
establish an emperically-derived temperature compensation method for
temperature was measured on the surface of the cells, battery temper­
ultrasound. The physical setup was the same; however, the temperature
ature ranged from approximately 18-28 ◦ C. The average battery tem­
environment varied. The varying temperature test used a CC charge and
perature was 23 ◦ C. Ultrasound captures were collected every 2 min.
a CC discharge at 0.250 A. The temperatures used were 18 ◦ C, 20 ◦ C, and
Koseoglou et al. [43] observed lithium plating at slower rates, some­
22 ◦ C, to establish room temperature behavior, and then 35 ◦ C, 40 ◦ C,
times seen at C/2. To ensure no cell breakdown occurred during baseline
45 ◦ C, 50 ◦ C, and 65 ◦ C, each for three cycles. After 45 ◦ C, 50 ◦ C, and
cycling and to avoid the possibility of lithium plating, the batteries were
65 ◦ C, batteries were cycled at 20 ◦ C.
cycled at ~ C/4. For baseline cycling, 0.250 A was used. Similarly, for
The partial overcharge experiment followed the same cycling profile
CC overcharge, current was below C/2 to ensure progression of cell
as Test 2 from Table 2 (CC overcharge at RT). Baseline data were
overcharge was measured with ultrasound and not fast breakdown due
to extreme heat spikes. collected in the same way. However, the overcharge was stopped before
the battery completely failed as indicated by the ultrasound notification
A capture consisted of the transmission ultrasound signals from a
transducer on one side of the battery and received by another transducer condition determined from the other failure experiments. Failure is
defined as the point at which the battery vents or explodes and is no
on the aligned opposite side of the cell, and so the transducers applied
ultrasound waves perpendicular to battery electrodes. Once all baseline longer capable of holding charge. After the overcharge was stopped, the
battery continued to cycle normally following the baseline procedure (i.
data were collected the cells were moved to an explosion chamber,
where failure was induced. Cycler data included a measure of time e., voltage limits of 2.5 V and 4.2 V and CC charge and discharge at 0.250
A).
(timestamp and experimental time), cell voltage, current, and capacity.
Ultrasound data included time (timestamp), a transmission ultrasound
3. Results & discussion
capture for ~70 μs, and a temperature measurement. Ultrasound data
were captured every 30 s and cycler data were captured within a
All experimental data are preprocessed so that data statistics are
voltage, current, or time tolerance, and paired based on the nearest
timestamp to be within seconds of each measurement. The currents and examined (e.g., change of SoC while cycling) and outlier signals are
removed (e.g., timing is offset or does not match the other ultrasound
data collection frequency used in this study would lead to a maximum
pulses in the same capture). The ultrasound signals are then examined in
error of 0.1% capacity difference between cycler measurement and ul­
the time domain and different transformed domains (i.e., Fourier
trasound signal. Once all data were collected, the timestamps of the

3
M.C. Appleberry et al.
Fig. 1. Representative ultrasound data. Typical ultrasound data through a
battery examined in the time domain (A), the envelope of the time domain
signal determined by a Hilbert transform (B), and the amplitude in the fre­
quency domain determined by a Fourier transform (C). The signal is repre­
sentative of baseline data from Experiment 1.
transform and Hilbert Transform) (Fig. 1). Signals are sectioned into
zones of signal information based on multiples of the time of flight of the
directly transmitted signal. The signal zone consisted of the first 6.5 μs of
the signal and is kept consistent for all experiments. Features calculated
from all domains include relative minima and maxima of the signals,
time/frequency of maxima and minima, inflection points, and zero-
crossings. A set of 43 signal features are extracted from the time-series
signals and transformed signals, and a feature matrix is generated.
Overcharge failure mechanisms are chosen to recreate internal fail­
ure mechanisms consistent with in-field failures [18]. Table 1 describes
the experimentally induced conditions leading to failures. To develop a
detection method, two regions of data are considered: the baseline re­
gion, where a battery operates under baseline cycling conditions; and an
overcharging region, where the battery is overcharged. Time series
signals from Experiment 1 shown in Fig. 2 display the contrast in signal
during baseline (Fig. 2A) and failure (Fig. 2B). Signals change minimally
4
Journal of Power Sources 535 (2022) 231423
Fig. 2. Representative Ultrasound Baseline and Failure. Typical baseline
ultrasound time series captures over a single charge highlighting how the signal
changes at 0% (red), 25% (black), 50% (blue), 75% (purple), and 100% (green)
state of charge (A), and representative time series ultrasound captures over a
battery overcharge (B). The overcharge started at 0 min (red) and continued
with a constant current until the battery failed at 206 min (green) shown as flat
line on x axis. (For interpretation of the references to colour in this figure
legend, the reader is referred to the Web version of this article.)
as a cell cycles compared to the large drop in signal strength as the
battery begins to fail. All ultrasound features are reviewed for the
baseline cycling data to determine how significantly they age with
normal battery cycling/aging. Of these, three features are selected to
build the detectors in this study: one feature from the time series signal,
one from the Hilbert transform, and one from the Fourier transform.
Specifically, the root mean square of the time series signal, the time of
the maximum amplitude from the Hilbert transform (similar to time of
flight), and the amplitude of the maximum frequency from the Fourier
transform are used as the three features. Fig. 3 shows an example of
baseline cycling and the resulting changes in the features. Each of the
features is seen to change with SoC and temperature, however there is
no significant cycle-to-cycle change over the timescale of baseline
cycling in this work.
Although different lithium-ion cell chemistries are studied else­
where, the changes observed with varying SoC in this work are generally
consistent with observations made by Davies et al. [35] in terms of
amplitude and time of flight; as a battery increases in charge, amplitude
first increases and then decreases and signals transmit faster. This work
further displays that the signal feature changes during charge are not
symmetrical to changes on discharge. Instead, features follow the same
general trajectory during charge and discharge, but precise values are
different. This same feature pattern repeats consistently as the batteries
cycle, displayed again for Experiment 1 in Fig. 3.
As shown in Fig. 3, there are changes in the signal as a battery ages.
Specifically, the signal RMS and maximum frequency amplitude in­
crease as a battery ages. Ladpli et al. [39] suggest that this increase in
signal energy is likely due to the viscoelastic properties of battery
M.C. Appleberry et al.
Fig. 3. Baseline Voltage, Temperature, and Ultrasound Data from Exper­
iment 1. Typical baseline voltage (A), temperature (B), and ultrasound signal
features over 362 cycles (83% capacity retention) of constant current charge/
constant current discharge cycling at 0.250 A. The first 5 and the last 5 cycles
are shown to highlight the variation in the ultrasound features over baseline
aging. The 3 ultrasound features examined are the root mean square of the time
series signal (C), the time of the maximum of the Hilbert envelope (D), and the
amplitude of the maximum frequency of the Fourier transform (E).
materials. Due to the ions transferred between the cells that cause
swelling and relaxation, the material changes in damping and attenua­
tion properties [39]. As the cell ages and become more rigid, it possesses
lower attenuation and damping. Contrary to studies examining effect of
time of flight on cell aging [39–41], this study found minimal changes to
signal speed during cycling. Fig. 3 shows that as a battery ages, its mean
relative energy (as represented by signal features such as amplitude and
RMS) and speed of the signal (expressed as time of flight or time of max.)
remain comparably unchanged when compared to exaggerated signal
differences during failure that are shared later in this study.
Signal changes during induced failure are analyzed in the context of
5
Journal of Power Sources 535 (2022) 231423
baseline cycling data. Popp et al. [44] display that temperature differ­
ences in ultrasound create different frequency responses between 0 ◦ C
and 45 ◦ C. They observe that temperature often becomes a dominant
effect over frequency shifts due to SoC changes. Temperature differences
can also alter and even reverse the relationships between SoC, ampli­
tude, and time of flight. Since temperature impacts ultrasound mea­
surements and five failure experiments are performed in a high
temperature chamber, the signal features are corrected such that the
changes in ultrasound features are temperature compensated. To
develop a database to support compensation for high temperature at
some voltages, a cell is cycled for 3 full cycles using baseline cycling
protocol at 18 ◦ C, 20 ◦ C, and 22 ◦ C, and then is moved to an oven to cycle
at five elevated temperatures: 35 ◦ C, 40 ◦ C, 45 ◦ C, 50 ◦ C, and 65 ◦ C. After
45 ◦ C, 50 ◦ C, and 65 ◦ C, the batteries are cycled again at 20 ◦ C to
determine if any degradation occurred during elevated temperature
cycling (Figs. S4 and S5). All batteries reached thermal equilibrium
before cycling. The specific algorithm for temperature correction is
described in the Supplementary Information.
The introduction of abnormal conditions based on the cycler mea­
surements of current and voltage is designated as overcharge beginning
and is the earliest time the signal should deviate from normal as a result
of cell degradation. Therefore, to minimize the number of false positive
reports, only feature changes that occur after the beginning of over­
charge are considered candidates of failure detection signals. To analyze
the changes in the features, two metrics of the baseline data are
compared to the overcharge data:
1. Deviation from the mean by 3.5 standard deviations: The mean and
standard deviation are calculated for all baseline data prior to
overcharge. Magnitude change of a feature by 3.5 standard de­
viations from the mean is outside all reference data and is noted as a
criterion for initial alarm/warning during overcharge (99.98% of
data is within 3.5 standard deviations from the baseline mean).
2. Deviation from the mean by 10 standard deviations: Magnitude
change of a feature by 10 standard deviations from the mean is
outside all baseline data and is noted as a criterion for an emergency
stop during overcharge.
The 3.5 and 10 standard deviations were empirically derived from
the first 10 experiments to avoid false positives and still detect before
critical failure. Once the start of the failure was identified, a detection
method for deviations of the three features is created based on the above
metrics. We use two types of pre-failure notifications:
1) Warning is where ultrasound begins to deviate from baseline cycling
behavior as determined by a change in absolute magnitude below 3.5
standard deviations from the baseline mean, and
2) E-stop (Emergency-stop) is when significant changes in ultrasound
show the battery is on course to catastrophic failure (10 standard
deviations from the baseline mean).
An E-stop notification would prompt a BMS to open the circuit to the
battery and stop current flow. A warning notification may not mean the
battery has reached the point of failure. Once this detection method is
adopted, it is implemented the same way for all experiments.
Detection methods are applied to the overcharge data. For a CC
overcharge, the overcharge begins as the cell voltage increased above
4.2 V, the voltage limit in baseline cycling experiments. For a CV
overcharge, the beginning of the overcharge is marked at the time the
current spiked to its maximum 5 A to bring the battery voltage to 5.5 V.
Fig. 4 shows failure detection of a CC (Experiment 8) and CV
(Experiment 5) overcharge at 65 ◦ C. During the CC overcharge, the
warning criteria is reached 76 min prior to failure coincident with a
sharp increase in the time series feature. E-stop failure criteria is reached
58 min before battery failure. For CC overcharge experiments at high
temperature, batteries expand and vent slightly faster than at RT.
M.C. Appleberry et al.
Table 3
Failure Detection Times. Summary of each experiment and the response time for the warning and E-stop compared to the total time of failure. Warning before failure
time indicates the amount of time between catastrophic failure event and ultrasound warning detected. E-stop before failure time indicates the amount of time prior to
catastrophic failure. The Global Technical Regulation on Electric Vehicle Safety requires a 5 min warning of battery failure [10].
Experiment Induced Failure Warning before failure
1 RT(~23 ◦ C), CC-overcharge 124 min (81%)
2 RT(~23 ◦ C), CC-overcharge 90 min (66%)
3 RT(~23 ◦ C), CV-overcharge 93 min (93%)
4 RT(~23 ◦ C), CV-overcharge 348 min (99%)
5 65 ◦ C, CV-overcharge 21 min (84%)
6 65 ◦ C, CV-overcharge 20 min (65%)
7 65 ◦ C, CV-overcharge 12 min (75%)
8 65 ◦ C, CC-overcharge 76 min (50%)
9 65 ◦ C, CC-overcharge 285 min (99%)
During Experiment 5, the signal changes immediately resulting in a
warning notification 19 min before battery failure. As the battery con­
tinues to degrade, the E-stop threshold is reached 10 min before battery
failure. The interaction between ultrasound energy and gas generation
within a cell may be related to that of gas bubbles, well-studied in the
ultrasound field [45–49]. Bubbles lead to increased time of flight of
ultrasound across the cell (observed in many failure figures,
Figs. S6–S12). Gas-induced battery swelling can also lead to an increase
in time of flight. However, more overwhelmingly, the presence of gas
bubbles causes significant scattering of ultrasound, and the amplitude of
the signals is significantly reduced [45,46]. The scattering and attenu­
ation may increase due to bubble resonance [49] as bubbles increase in
size, which can explain the rapid decline in signal amplitude observed as
batteries approach catastrophic failure.
High temperature, CV overcharge is the only method that leads to an
explosion in this work. Although the detection time for the CV failure is
much closer to the failure event compared to the RT failures, the total
6
Journal of Power Sources 535 (2022) 231423
Fig. 4. High Temperature Battery Failure. A
representative constant current overcharge (Experi­
ment 8) with ultrasound features (left) and constant
voltage overcharge (Experiment 5) with ultrasound
features (right) collected in a controlled temperature
chamber set to 65 ◦ C. Some baseline data and the
points at which overcharge begins (open, orange
circle), the ultrasound algorithm detects the over­
charge (open, orange square), the ultrasound algo­
rithm triggers an emergency stop (filled, red square),
and the battery fails (filled, red circle) are shown. The
voltage profile (A), current profile (F), and tempera­
tures (B, G) are shown along with the RMS of the
ultrasound time series (C, H), the time of the
maximum of the Hilbert envelope of the ultrasound
data (D, I), and the amplitude of the maximum fre­
quency of the Fourier transform of the ultrasound
data (E, J). (For interpretation of the references to
colour in this figure legend, the reader is referred to
the Web version of this article.)
E-Stop before failure Time from overcharge to failure
54 min (35%) 154 min
90 min (66%) 136 min
92 min (92%) 100 min
347 min (99%) 352 min
15 min (60%) 25 min
19 min (61%) 31 min
4 min (25%) 16 min
58 min (38%) 151 min
139 min (48%) 289 min
overcharge time is much shorter. The E-stop issued for the CV failure at
high temperature is 60% of the total overcharge time. Table 3 displays
all experiment warning and E-stop threshold times. The voltage, tem­
perature, and three ultrasound features for all other battery failures are
shown in the supporting information (Figs. S6–S12).
Overcharge duration is proportionally different for distinct induced
failure types. For example, 65 ◦ C, CV-overcharge failure experiments
occur an order of magnitude faster than RT, CV-overcharge experiments.
Still, within similar induced failure types, the time from overcharge to
failure differs between cells. Since cells were cycled under similar con­
ditions, the disparity is likely due to manufacturing differences. Despite
the difference in overall overcharge time, the ultrasound detection
consistently identified failure conditions.
Post-mortem analyses are performed on baseline cycled batteries and
failed batteries to evaluate the difference in physical states caused by
induced failure. The results are compared with those found in ESS fail­
ures, as both are believed to be caused by electrolyte decomposition.
M.C. Appleberry et al. Journal of Power Sources 535 (2022) 231423

Fig. 5. SEM Baseline Comparison to Overcharge. SEM images of the anode (top) and cathode (bottom) of cells before (left) and after (right) a battery failure. The
anode (A) and cathode (C) of a battery only undergo baseline operation, 30 cycles of a constant current charge/discharge at 0.250 A without overcharging
(Experiment 10), compared to the anode after a CV overcharge (B, Experiment 6) and the cathode after a CC overcharge (D, Experiment 8).

Fig. 6. EDX Baseline Comparison to Overcharge. Fluorine (top) and oxygen (bottom) content as determined by EDX of cells before (left) and after (right) a battery
failure. The fluorine (A) and oxygen (C) content of the anode after baseline cycling (Experiment 10) compared to the fluorine (B) and oxygen (D) content of the anode
of a cell after a CV overcharge (Experiment 6).

7
M.C. Appleberry et al.
These comparisons demonstrate strong similarities and suggest that the
experiments are a good model for the in-field failures. Specifically,
scanning electron microscopy (SEM) is used to view surface morphology
of the electrodes and solid electrolyte interphase (SEI) deposition, and
energy-dispersive X-ray spectroscopy (EDX) is used to determine
changes in the elemental components on the anode and cathode due to
electrolyte decomposition. Samples for SEM imaging are taken from the
center of unrolled electrodes for both new and failed batteries. It is
assumed that electrodes are uniform for unused batteries, which is
supported by SEM images across many samples and locations.
In all experiments, batteries swelled and released gas or swelled
before they exploded. Large amounts of swelling are caused by elec­
trolyte degradation leading to gas formation. It is expected that as the
battery fails, the pressure caused by swelling due to outgassing leave
fractures visible on the cathode [11,15,50,51]. This phenomenon may
be observed in the differences between SEM images of the electrodes.
Fig. 5A and C were collected prior to cell failure but after baseline
cycling and Fig. 5B and D were collected after CV and CC overcharge,
respectively. There is clear SEI growth on Fig. 5B when compared to 5A
shown by a coating on the spherical graphite carbon particles [52]. The
SEI, appearing brighter under the microscope since it is non-conductive
8
Journal of Power Sources 535 (2022) 231423
Fig. 7. Cycling and Ultrasound Behavior After
Ultrasound Warning. Overall voltage (A), tempera­
ture (B), RMS of the ultrasound series (C), time of the
maximum Hilbert envelope of the ultrasound data
(D), and the amplitude of the maximum frequency of
the Fourier transform of the ultrasound data (E) for a
battery that was overcharged with constant current
(overcharge starts at the open, orange circle) until the
ultrasound warning algorithm detected overcharge
(open, orange square), and then cycled following the
baseline procedure. The right column highlights the
time surrounding the overcharge specifically for the
voltage (F), temperature (G), RMS of the ultrasound
series (H), time of the maximum Hilbert envelope of
the ultrasound data (I), and the amplitude of the
maximum frequency of the Fourier transform of the
ultrasound data (J). (For interpretation of the refer­
ences to colour in this figure legend, the reader is
referred to the Web version of this article.)
[53], coats the conductive graphite throughout the sample, showing
physical indication of cell failure compared to the baseline cycled
sample. Clear cracking on the cathode between Fig. 5C and D show the
impact of swift swelling as the battery outgassed quickly [51]. The
amount of decomposition products deposited on the surface of the failed
electrodes is significant for the amount of total cycling time, indicating a
rapid failure for both cathode and anode materials.
The formation of the SEI layer is further explored through EDX im­
ages. Fig. 6 shows the elemental mapping of the anodes after baseline
cycling and after battery failure. Fig. 6A and C are the corresponding
elemental compositions for the anode shown in Figs. 5A, 6B and D
correspond to the anode in Fig. 5B. The comparison of the two condi­
tions show evidence of increased deposition of fluorine and oxygen after
the overcharge event. Prior to battery failure fluorine (Fig. 6A) and
oxygen (Fig. 6C) mapping is inhomogeneous, but much less dense than
concentration after CV overcharge (Fig. 6B and D, respectively). Spe­
cifically, the spectrum of the new cell electrolyte consists of 60% carbon,
10% oxygen, and 15% fluorine. After the battery is failed in the high
temperature overcharge experiments, the EDX spectrum indicates 57%
fluorine, 20% oxygen, and 13% carbon on the surface of the anode. After
the battery failure, the relative fluorine abundance has increased by
M.C. Appleberry et al.
Fig. 8. Capacity and Coulombic Efficiency for the Partial Overcharge. The
charging capacity (filled, red circle, left axis), the discharging capacity (open,
red triangle, left axis), and the coulombic efficiency (open, blue square, right
axis) of a battery that was partially overcharged on cycle 22 until the ultra­
sound detection algorithm identified overcharge, thereby providing guidance to
correct the problem and allow the battery to resume normal cycling and
operation. (For interpretation of the references to colour in this figure legend,
the reader is referred to the Web version of this article.)
nearly 4x. The changes in the elemental anode composition are consis­
tent with electrolyte decomposition and reactions induced by high
voltage conditions of failure [11,15,50,54].
Although post-mortem of all experiments shows increases of fluorine
and oxygen in the anode, higher percentages of fluorine are observed in
CV failure experiments, while CC failure anodes had larger oxygen
percentages. The difference is caused by rate changes in the reaction
kinetics induced by high voltage or high current [11,15,50]. As failure
conditions are expected to form gases within the cell, the increase in
fluorine and oxygen after batteries fail suggests that signal differences
observed are a result of expected failure.
The totality of this post-mortem analysis indicates that the failure
mechanism observed in all experiments is electrolyte decomposition
leading to swelling and outgassing. The presence of accumulating gas –
even at small quantities – is the reason ultrasound can readily identify
these conditions. Based on the warning and response triggers, the ul­
trasound detection method can successfully identify and report early
signs of this failure mechanism before potential explosions on the 0.950
Ah cells used in this study.
This work suggests the existence of a “warning” scenario in which a
disruptive overcharge case may be reversed to allow return to normal
battery operation. To test this proposition, one battery is cycled using
baseline cycling parameters (CC at 0.250 A) for 21 cycles to establish
baseline ultrasound features, partially overcharged at CC until it reached
warning criteria, and then returned to original cycling. The partial
overcharge is completed under the same conditions as Experiments 1
and 2. Fig. 7 shows the detection of a the warning after the partial
overcharge began at cycle 22, and the recovery of the magnitude of each
feature after normal cycling continued. After the return to normal
cycling, the battery capacity and coulombic efficiency return to pre-
partial overcharge metrics (Fig. 8).
The results of the partial overcharge experiment suggest the warning
9
Journal of Power Sources 535 (2022) 231423
scenario is a recoverable battery condition and after a return to normal
conditions, its overall capacity retention may not be greatly impacted.
Thus, the distinction between a warning and E-stop may distinguish
between permanent and transient states of concern and provide a
method to avoid failure.
4. Conclusion
It is highly desirable to control ESSs utilizing lithium-ion batteries
with highly effective battery management tools and sensors to avoid
catastrophic events. Improving battery management methods requires
consideration of methodologies not currently in use for measuring the
battery state and promptly detecting conditions leading to battery fail­
ure. This work presents an approach to identifying failure which does
not rely on electrical measurements (i.e., voltage, current), but instead
uses ultrasound to directly measure the battery state by determining
variations in material properties of the cell components. This study has
validated the detection of battery failure using ultrasound by inducing
0.950 Ah batteries to failure while collecting ultrasound data. By
comparing the ultrasound during overcharge (for different types of
overcharging – CC, CV) to baseline ultrasound data collected on this
battery type, two different metrics of failure were determined: a warning
to detect the start of overcharge to identify a dangerous but potentially
reversible change to the battery and an E-stop to identify that the battery
has undergone significant changes implying imminent failure requiring
that it should be immediately taken out of service. For all overcharge
experiments performed, the same warning/stopping criteria provided
similar results, demonstrating the versatility and the adaptability of
ultrasound for failure detection. Both of the metrics were triggered well
before a catastrophic event occurred. Therefore, the implementation of
ultrasound into systems could provide actionable monitoring to drasti­
cally reduce the chance of a serious event.
The results suggest that a warning condition may be implemented by
adjusting BMS limits to regulate current and voltage to avoid acceler­
ating battery degredation and subsequent thermal runaway. An E-stop
notification may be implemented by terminating current to the system
via an inherent contactor or switch, preventing the accumulation of
outgassing leading to failure. Further data collection efforts are required
to calibrate BMS limits in accordance with safe battery control. Since
there are small changes to the ultrasound signal through battery aging,
our approach should be implemented with a moving average of recent
cycling data to establish warning and E-stop thresholds. As a result, the
thresholds are based on changes as a battery’s signal evolves to reduce
false positives.
While this work demonstrates a proof of concept for detecting bat­
tery failure using ultrasound, significant technological advances are
needed to economically implement this technology in the field. Sensors
cannot easily be used to retrofit existing battery packs and modules, and
must be designed into battery cells or groups of cells. The largest barriers
to implementation are sensor thickness and cost, largely being addressed
by the development micro-electromechanical systems (MEMS)
solutions.
Overcharging was primarily explored in this study, future data
collection efforts are required to explore all failure mechanisms and
their corresponding changes in ultrasound signatures. Such experiments
should include thermal cycling, over discharging, and lithium plating.
To prepare a full suite of detection methods for use in a system, similar
data collection methods as in this study must be completed to compare
regular cycling and failure on larger batteries and on modules or packs.
Data availability
Data for this work is unavailable due to confidentiality.
M.C. Appleberry et al.
CRediT authorship contribution statement
Maura C. Appleberry: Conceptualization, Writing – original draft,
Writing – review & editing, Data curation, Formal analysis, Investiga­
tion, Methodology, Visualization. Jeffrey A. Kowalski: Conceptualiza­
tion, Writing – original draft, Writing – review & editing, Data curation,
Project administration, Supervision, Investigation, Methodology, Visu­
alization, Funding acquisition, Formal analysis. Steven A. Africk:
Conceptualization, Funding acquisition, Investigation, Writing – review
& editing. Jared Mitchell: Data curation, Investigation, Methodology,
Writing – review & editing. Thomas C. Ferree: Formal analysis, Writing
– review & editing. Vincent Chang: Data curation. Vashisth Parekh:
Data curation, Writing – review & editing. Ziyi Xu: Data curation.
Ziwen Ye: Data curation. Jay F. Whitacre: Conceptualization, Inves­
tigation, Methodology, Project administration, Supervision, Writing –
review & editing. Shawn D. Murphy: Funding acquisition.
Declaration of competing interest
The authors declare the following financial interests/personal re­
lationships which may be considered as potential competing interests:
Maura C Appleberry reports financial support was provided by US
Department of Energy. Jeffrey A Kowalski reports financial support was
provided by US Department of Energy. Steven A Africk reports financial
support was provided by USDepartment of Energy. Thomas C Ferree
reports financial support was provided by US Department of Energy.
Shawn D Murphy reports financial support was provided by US
Department of Energy. Jeffrey A Kowalski has patent
#US20210249702A1 issued to Titan Advanced Energy Solutions Inc.
Shawn D Murphy has patent #US20210249702A1 issued to Titan
Advanced Energy Solutions Inc. Steven A Africk has patent
#US20200106137A1 issued to Titan Advanced Energy Solutions Inc.
Shawn D Murphy has patent #US20200106137A1 issued to Titan
Advanced Energy Solutions Inc.
Acknowledgements
This material is based upon work supported by the U.S. Department
of Energy, Office of Science, under Award Number(s) DE SC0020735.
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.
org/10.1016/j.jpowsour.2022.231423.
References
[1] S. Chu, A. Majumdar, Opportunities and challenges for a sustainable energy future,
Nature 488 (2012) 294–303, https://doi.org/10.1038/nature11475.
[2] 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,
https://doi.org/10.1016/j.jpowsour.2012.02.038.
[3] I. Gyuk, M. Johnson, J. Vetrano, K. Lynn, W. Parks, R. Handa, L. Kannberg,
S. Hearne, K. Waldrip, R. Braccio, Grid Energy Storage, U.S. Department of Energy,
2013. https://www.energy.gov/sites/prod/files/2014/09/f18/Grid%20Energy%
20Storage%20December%202013.pdf.
[4] L. Kishter, J. Izevbigie, A. Best, P. Siebenlist, J. Homer, J. Taft, M. Weimar, P. De
Martini, Smart Grid System Report, U.S. Department of Energy, 2018. https://
www.energy.gov/sites/prod/files/2019/02/f59/Smart%20Grid%20System%20
Report%20November%202018_1.pdf.
[5] T.B. Reddy, in: Linden’s Handbook of Batteries, fourth ed., McGraw-Hill Education,
2011. https://www.accessengineeringlibrary.com/content/book/9780071624213.
(Accessed 30 November 2021).
[6] Z. Zhang, S.S. Zhang, Rechargeable Batteries: Materials, Technologies and New
Trends, Springer International Publishing, 2015.
[7] R.M. Darling, K.G. Gallagher, J.A. Kowalski, S. Ha, F.R. Brushett, Pathways to low-
cost electrochemical energy storage: a comparison of aqueous and nonaqueous
flow batteries, Energy Environ. Sci. 7 (2014) 3459–3477, https://doi.org/10.1039/
C4EE02158D.
[8] Z. Liao, S. Zhang, K. Li, G. Zhang, T.G. Habetler, A survey of methods for
monitoring and detecting thermal runaway of lithium-ion batteries, J. Power
Sources 436 (2019) 226879, https://doi.org/10.1016/j.jpowsour.2019.226879.
10
Journal of Power Sources 535 (2022) 231423
[9] P. Sun, R. Bisschop, H. Niu, X. Huang, A review of battery fires in electric vehicles,
Fire Technol. 56 (2020) 1361–1410, https://doi.org/10.1007/s10694-019-00944-
3.
[10] Global technical regulation on the electric vehicle safety (EVS) (n.d.), https://u
nece.org/fileadmin/DAM/trans/main/wp29/wp29wgs/wp29gen/wp29registr
y/ECE-TRANS-180a20e.pdf. (Accessed 30 November 2021).
[11] D. Belov, M.-H. Yang, Failure mechanism of Li-ion battery at overcharge
conditions, J. Solid State Electrochem. 12 (2008) 885–894, https://doi.org/
10.1007/s10008-007-0449-3.
[12] C. Hendricks, N. Williard, S. Mathew, M. Pecht, A failure modes, mechanisms, and
effects analysis (FMMEA) of lithium-ion batteries, J. Power Sources 297 (2015)
113–120, https://doi.org/10.1016/j.jpowsour.2015.07.100.
[13] H. Yang, H. Bang, K. Amine, J. Prakash, Investigations of the exothermic reactions
of natural graphite anode for Li-ion batteries during thermal runaway,
J. Electrochem. Soc. 152 (2004) A73, https://doi.org/10.1149/1.1836126.
[14] Q. Yuan, F. Zhao, W. Wang, Y. Zhao, Z. Liang, D. Yan, Overcharge failure
investigation of lithium-ion batteries, Electrochim. Acta 178 (2015) 682–688,
https://doi.org/10.1016/j.electacta.2015.07.147.
[15] B. Rowden, N. Garcia-Araez, A review of gas evolution in lithium ion batteries,
Energy Rep. 6 (2020) 10–18, https://doi.org/10.1016/j.egyr.2020.02.022.
[16] M. Ouyang, M. Zhang, X. Feng, L. Lu, J. Li, X. He, Y. Zheng, Internal short circuit
detection for battery pack using equivalent parameter and consistency method,
J. Power Sources 294 (2015) 272–283, https://doi.org/10.1016/j.
jpowsour.2015.06.087.
[17] L. Lu, X. Han, J. Li, J. Hua, M. Ouyang, A review on the key issues for lithium-ion
battery management in electric vehicles, J. Power Sources 226 (2013) 272–288,
https://doi.org/10.1016/j.jpowsour.2012.10.060.
[18] McMicken battery energy storage system event technical analysis and
recommendations, Arizona public service. https://coaching.typepad.com/files/m
cmicken.pdf, 2020.
[19] S. Graham, Another Devastating Chevy Bolt Fire Just Days after Fix Announced,
Electrek, 2021. https://electrek.co/2021/05/03/another-devastating-chevy-b
olt-fire-just-days-after-fix-announced/. (Accessed 30 November 2021).
[20] M. Wayland, GM to Spend $1 Billion to Expand Chevy Bolt EV Recall Due to Fires,
CNBC, 2021. https://www.cnbc.com/2021/08/20/gm-to-spend-1-billion
-to-expand-chevy-bolt-ev-recall-due-to-fires.html. (Accessed 30 November 2021).
[21] C. Isidore, P. Valdes-Dapena, Hyundai’s recall of 82,000 electric cars is one of the
most expensive in history - CNN. https://www.cnn.com/2021/02/25/tech/hyun
dai-ev-recall/index.html, 2021. (Accessed 30 November 2021).
[22] Q. Wang, B. Mao, S.I. Stoliarov, J. Sun, A review of lithium ion battery failure
mechanisms and fire prevention strategies, Prog. Energy Combust. Sci. 73 (2019)
95–131, https://doi.org/10.1016/j.pecs.2019.03.002.
[23] A.A. Tidblad, K. Edström, G. Hernández, I. de Meatza, I. Landa-Medrano, J. Jacas
Biendicho, L. Trilla, M. Buysse, M. Ierides, B.P. Horno, Y. Kotak, H.-G. Schweiger,
D. Koch, B.S. Kotak, Future material developments for electric vehicle battery cells
answering growing demands from an end-user perspective, Energies 14 (2021)
4223, https://doi.org/10.3390/en14144223.
[24] Y. Fernandes, A. Bry, S. de Persis, Identification and quantification of gases emitted
during abuse tests by overcharge of a commercial Li-ion battery, J. Power Sources
389 (2018) 106–119, https://doi.org/10.1016/j.jpowsour.2018.03.034.
[25] W. Zhirong, Y. Yun, T. Xuan, C. Changan, L. Kai, J. Mingju, Li-ion batteries piles
thermal runaway autoalarm and its monitoring method based on gas-monitoring,
CN108008083A, https://patents.google.com/patent/CN108008083A/en, 2018.
(Accessed 30 November 2021).
[26] M.A. Roscher, R.M. Kuhn, H. Döring, Error detection for PHEV, BEV and stationary
battery systems, Control Eng. Pract. 21 (2013) 1481–1487, https://doi.org/
10.1016/j.conengprac.2013.07.003.
[27] M.H. Parekh, B. Li, M. Palanisamy, T.E. Adams, V. Tomar, V.G. Pol, In situ thermal
runaway detection in lithium-ion batteries with an integrated internal sensor, ACS
Appl. Energy Mater. (2020), https://doi.org/10.1021/acsaem.0c01392.
[28] X. Lai, W. Yi, X. Kong, X. Han, L. Zhou, T. Sun, Y. Zheng, Online detection of early
stage internal short circuits in series-connected lithium-ion battery packs based on
state-of-charge correlation, J. Energy Storage 30 (2020) 101514, https://doi.org/
10.1016/j.est.2020.101514.
[29] X. Feng, C. Weng, M. Ouyang, J. Sun, Online internal short circuit detection for a
large format lithium ion battery, Appl. Energy 161 (2016) 168–180, https://doi.
org/10.1016/j.apenergy.2015.10.019.
[30] J. Klink, J. Grabow, N. Orazov, R. Benger, A. Börger, A. Ahlberg Tidblad, H. Wenzl,
H.-P. Beck, Thermal fault detection by changes in electrical behaviour in lithium-
ion cells, J. Power Sources 490 (2021) 229572, https://doi.org/10.1016/j.
jpowsour.2021.229572.
[31] M. Seo, M. Park, Y. Song, S.W. Kim, Online detection of soft internal short circuit in
lithium-ion batteries at various standard charging ranges, IEEE Access 8 (2020)
70947–70959, https://doi.org/10.1109/ACCESS.2020.2987363.
[32] S. Koch, K.P. Birke, R. Kuhn, Fast thermal runaway detection for lithium-ion cells
in large scale traction batteries, Batteries 4 (2018) 16, https://doi.org/10.3390/
batteries4020016.
[33] K. Rhodes, N. Dudney, E. Lara-Curzio, C. Daniel, Understanding the degradation of
silicon electrodes for lithium-ion batteries using acoustic emission, J. Electrochem.
Soc. 157 (2010) A1354, https://doi.org/10.1149/1.3489374.
[34] J. Bernard, Method and system for diagnosis of the internal state of a battery
through acoustic emission, US8984944B2, https://patents.google.com/patent/US8
984944/en, 2015. (Accessed 30 November 2021).
[35] G. Davies, K.W. Knehr, B.V. Tassell, T. Hodson, S. Biswas, A.G. Hsieh, D.
A. Steingart, State of charge and state of health estimation using electrochemical
M.C. Appleberry et al.
acoustic time of flight analysis, J. Electrochem. Soc. 164 (2017) A2746, https://
doi.org/10.1149/2.1411712jes.
[36] C. Bommier, W. Chang, J. Li, S. Biswas, G. Davies, J. Nanda, D. Steingart, Operando
acoustic monitoring of SEI formation and long-term cycling in NMC/SiGr
composite pouch cells, J. Electrochem. Soc. 167 (2020), 020517, https://doi.org/
10.1149/1945-7111/ab68d6.
[37] T. Hodson, S. Patil, D.A. Steingart, A initial exploration of coupled transient
mechanical and electrochemical behaviors in lithium ion batteries, J. Electrochem.
Soc. 168 (2021), 070515, https://doi.org/10.1149/1945-7111/ac0f86.
[38] R.J. Copley, D. Cumming, Y. Wu, R.S. Dwyer-Joyce, Measurements and modelling
of the response of an ultrasonic pulse to a lithium-ion battery as a precursor for
state of charge estimation, J. Energy Storage 36 (2021) 102406, https://doi.org/
10.1016/j.est.2021.102406.
[39] P. Ladpli, F. Kopsaftopoulos, F.-K. Chang, Estimating state of charge and health of
lithium-ion batteries with guided waves using built-in piezoelectric sensors/
actuators, J. Power Sources 384 (2018) 342–354, https://doi.org/10.1016/j.
jpowsour.2018.02.056.
[40] A.G. Hsieh, S. Bhadra, B.J. Hertzberg, P.J. Gjeltema, A. Goy, J.W. Fleischer, D.
A. Steingart, Electrochemical-acoustic time of flight: in operando correlation of
physical dynamics with battery charge and health, Energy Environ. Sci. 8 (2015)
1569–1577, https://doi.org/10.1039/C5EE00111K.
[41] Y. Wu, Y. Wang, W.K.C. Yung, M. Pecht, Ultrasonic health monitoring of lithium-
ion batteries, Electronics 8 (2019) 751, https://doi.org/10.3390/
electronics8070751.
[42] G.-H. Kim, K. Smith, J. Ireland, A. Pesaran, Fail-safe design for large capacity
lithium-ion battery systems, J. Power Sources 210 (2012) 243–253, https://doi.
org/10.1016/j.jpowsour.2012.03.015.
[43] M. Koseoglou, E. Tsioumas, D. Ferentinou, N. Jabbour, D. Papagiannis,
C. Mademlis, Lithium plating detection using dynamic electrochemical impedance
spectroscopy in lithium-ion batteries, J. Power Sources 512 (2021) 230508,
https://doi.org/10.1016/j.jpowsour.2021.230508.
[44] H. Popp, G. Glanz, K. Alten, I. Gocheva, W. Berghold, A. Bergmann, Mechanical
frequency response analysis of lithium-ion batteries to disclose operational
parameters, Energies 11 (2018) 541, https://doi.org/10.3390/en11030541.
11
Journal of Power Sources 535 (2022) 231423
[45] C. Devin, Survey of thermal, radiation, and viscous damping of pulsating air
bubbles in water, J. Acoust. Soc. Am. 31 (1959) 1654–1667, https://doi.org/
10.1121/1.1907675.
[46] V.C. Anderson, Sound scattering from a fluid sphere, J. Acoust. Soc. Am. 22 (1950)
426–431, https://doi.org/10.1121/1.1906621.
[47] M.A. Ainslie, T.G. Leighton, Review of scattering and extinction cross-sections,
damping factors, and resonance frequencies of a spherical gas bubble, J. Acoust.
Soc. Am. 130 (2011) 3184–3208, https://doi.org/10.1121/1.3628321.
[48] G. Kapodistrias, P.H. Dahl, On scattering from a bubble located near a flat
air–water interface: laboratory measurements and modeling, J. Acoust. Soc. Am.
110 (2001) 1271–1281, https://doi.org/10.1121/1.1390335.
[49] H. Medwin, C.S. Clay, Chapter 8 - bubbles, in: H. Medwin, C.S. Clay (Eds.),
Fundam. Acoust. Oceanogr., Academic Press, San Diego, 1998, pp. 287–347,
https://doi.org/10.1016/B978-012487570-8/50010-6.
[50] Q. Yuan, F. Zhao, W. Wang, Y. Zhao, Z. Liang, D. Yan, Overcharge failure
investigation of lithium-ion batteries, Electrochim. Acta 178 (2015) 682–688,
https://doi.org/10.1016/j.electacta.2015.07.147.
[51] A. Mukhopadhyay, A. Tokranov, X. Xiao, B.W. Sheldon, Stress development due to
surface processes in graphite electrodes for Li-ion batteries: a first report,
Electrochim. Acta 66 (2012) 28–37, https://doi.org/10.1016/j.
electacta.2012.01.058.
[52] T. Waldmann, M. Wilka, M. Kasper, M. Fleischhammer, M. Wohlfahrt-Mehrens,
Temperature dependent ageing mechanisms in Lithium-ion batteries – a Post-
Mortem study, J. Power Sources 262 (2014) 129–135, https://doi.org/10.1016/j.
jpowsour.2014.03.112.
[53] P. Verma, P. Maire, P. Novák, A review of the features and analyses of the solid
electrolyte interphase in Li-ion batteries, Electrochim. Acta 55 (2010) 6332–6341,
https://doi.org/10.1016/j.electacta.2010.05.072.
[54] K. Leung, S.B. Rempe, M.E. Foster, Y. Ma, J.M.M. del la Hoz, N. Sai, P.B. Balbuena,
Modeling electrochemical decomposition of fluoroethylene carbonate on silicon
anode surfaces in lithium ion batteries, J. Electrochem. Soc. 161 (2013) A213,
https://doi.org/10.1149/2.092401jes.