Professional Documents
Culture Documents
CrossMark
View Export
Online Citation
AFFILIATIONS
Oak Ridge National Laboratory, 1 Bethel Valley Road, Oak Ridge, Tennessee 37830, USA
a)
Author to whom correspondence should be addressed: sunh1@ornl.gov
ABSTRACT
This study applied the nonlinear ultrasonic method, second harmonic generation, to precisely estimate the state of charge (SoC) in lithium-
ion batteries. The second harmonic of the longitudinal wave is generated on a pouch cell battery at 5 MHz with a through-transmission setup.
The relative nonlinear parameter b0 is determined by analyzing the amplitudes at the fundamental and second harmonic frequencies. To
enhance the nonlinear parameter’s measurement accuracy, multiple excitation amplitudes are employed. Two separate charge/discharge tests
(four-cycle and eight-cycle) are conducted on the battery at a rate of C/10. The nonlinear parameter is measured periodically during the
The lithium-ion battery (LIB) stands as a widely embraced energy and is insensitive to the initial SoC value. However, the precision
storage solution, making its presence felt across diverse industries and and complexity of the battery model are very critical to the estima-
in our daily routines. For the optimal performance and lifetime of tion accuracy. The neural network model method, a data-driven
LIBs, a battery management system (BMS) is required to regulate their method, exhibits a versatile capability for real-time SoC estimation
operation. The state of charge (SoC) is one of the critical parameters across diverse battery types while demanding extensive training
for battery management. Diverse methods have been implemented to using a substantial dataset. Recently, the acoustic/ultrasonic nonde-
estimate SoC, including the discharge test method, the ampere-hour structive evaluation (NDE) of LIBs has gathered increasing attention
integral (coulomb counting) method,1 the open-circuit voltage (OCV) on the estimation of SoC and state of health (SoH) as well as moni-
method,2 the battery model-based approach,3 and the neural network toring structural changes within LIBs.7–9 The ultrasonic NDE tech-
model method.4 However, each method presents unique advantages nique exhibits several advantages for SoC estimation in LIBs. It is a
and challenges. The discharge test, while simple and reliable, proves nondestructive method that can provide rapid results through
impractical for real-time SoC estimation due to its time-intensive open-circuit measurements, a feature especially advantageous for
nature.5 The ampere-hour integral method’s precision is notably influ- inspecting second-life batteries. Its uncomplicated setup and high
enced by the initial SoC and the battery operating states (SoC, temper- repeatability further facilitate real-time monitoring of battery states.
ature, current, etc.). The OCV method, while simple, confronts Furthermore, the LIB operates as a composite material system com-
challenges with hysteresis during charge/discharge processes, particu- prising multiple layers of anodes and cathodes. The underlying prin-
larly affecting specific battery types (e.g., C/LiFePO4);6 additionally, it ciple of ultrasonic NDE for LIBs lies in its ability to detect the
mandates the battery to be in an equilibrium state for accurate real- physical changes in the material properties and internal structure of
time SoC estimation. The battery model-based method (equivalent the battery. Therefore, this approach can offer a deeper and more
circuit model and electrochemical model) does not require rest time fundamental understanding of the battery’s condition than
Appl. Phys. Lett. 124, 073902 (2024); doi: 10.1063/5.0187829 124, 073902-1
Published under an exclusive license by AIP Publishing
Applied Physics Letters ARTICLE pubs.aip.org/aip/apl
traditional electrochemistry methods for an enhanced determina- Considering 1D wave propagation in an isotropic solid,
tion of SoC, SoH, and battery failure.
During a LIB’s charge and discharge cycles, the mechanical prop- @ 2 u1 2 @ u1
2
@u1
¼ c 1 þ b ; (1)
erties, such as modulus, density, and Poisson’s ratio, undergo changes @t 2 @X12 X1
as part of the delithiation and lithiation processes. In ultrasonic testing, where u1 is the displacement, t is the time, and c is the longitudinal
the stress wave generated from the ultrasonic transmitter propagates wave velocity. b is the nonlinearity parameter and X1 is the coordinate.
through the thickness of the battery and carries information about the The solution to this equation can be given by considering a time-
material properties and even the changes in the internal structure. harmonic plane wave A1 cosðkX1 xtÞ,
Correlations could be established between the battery’s SoC or SoH and
the various wave features. Among these, time of flight (TOF) is a prom- 1
u1 ¼ bk2 A21 X1 þ A1 cosðkX1 xtÞ
inent parameter that signifies the time it takes for a wave to travel 8
between the transmitter and receiver, typically across the battery’s 1
þ bk2 A21 X1 cos½2ðkX1 xtÞ
thickness.7,10 The signal amplitude,9,11 wave attenuation,12,13 and wave 8
velocity13,14 are also utilized as crucial indicators. In practice, these fea- ¼ A0 þ A1 cosðkX1 xtÞ þ A2 cos½2ðkX1 xtÞ; (2)
tures are often combined to enhance the accuracy of the SoC estimate.
However, to effectively estimate SoC, it is desirable to have an ultrasonic where A1 and A2 represent the absolute amplitudes of the funda-
wave feature that exhibits a linear relationship with the SoC. mental harmonic and second harmonic, respectively, in the fre-
Unfortunately, many studies have shown that TOF and signal ampli- quency spectral. x is the wave propagation distance, and k is the
tude do not exhibit a perfect linear relationship with battery SoC, par- wave number. The nonlinear parameter b can be calculated using
ticularly in the SoC range below 20%. Moreover, these studies have these two amplitudes:
found hysteresis behavior during the charge and discharge pro- 8A2
cesses.7,13,15 This behavior implies that the same SoC may yield signifi- b¼ : (3)
A21 xk
cantly different wave feature values during battery charging and
discharging. These nonlinear behaviors contribute to a substantial mar- However, the solution in Eq. (3) only applies to the longitudinal wave.
gin of error in the battery’s SoC estimation. For the Raleigh wave, a different solution of b is used.16 Furthermore,
These methods, employing various ultrasonic wave features, fall the amplitudes of the fundamental and second harmonics are absolute
Appl. Phys. Lett. 124, 073902 (2024); doi: 10.1063/5.0187829 124, 073902-2
Published under an exclusive license by AIP Publishing
Applied Physics Letters ARTICLE pubs.aip.org/aip/apl
Appl. Phys. Lett. 124, 073902 (2024); doi: 10.1063/5.0187829 124, 073902-3
Published under an exclusive license by AIP Publishing
Applied Physics Letters ARTICLE pubs.aip.org/aip/apl
FIG. 4. Correlation between the SoC and nonlinear parameter b0 for the four-cycle
and eight-cycle charge/discharge tests.
closely aligns with the correlation curve from the four-cycle test, dis-
playing a similar approximately linear relationship. This robust repeat-
FIG. 3. (a) SoC history for a four-cycle charge/discharge and (b) corresponding ability indicates that the nonlinear parameter b0 maintains a consistent
nonlinear parameter b0 measured. linear relationship with the battery SoC across multiple test cycles.
The linear relationship was fitted using all the data points gath-
increased [blue curve in Fig. 3(b)], escalating as the SoC advanced ered from both the four-cycle and eight-cycle tests. The fitted relation-
from 0% to 100%. Conversely, in the discharge phase, b0 decreased ship (magenta curve) can be expressed as b0 ¼ ð0:0208 SoC
with the SoC’s return to 0%. During charge, lithium ions are driven þ 4:562Þ 106 with a high goodness of fit R2 ¼ 0:984 and small
from the cathode to the anode. This migration causes structural root mean square error (RMSE). However, the data points falling below
Appl. Phys. Lett. 124, 073902 (2024); doi: 10.1063/5.0187829 124, 073902-4
Published under an exclusive license by AIP Publishing
Applied Physics Letters ARTICLE pubs.aip.org/aip/apl
measured by the potentiostat. The predicted SoC curves closely match Energy (DOE). The U.S. government retains and the publisher, by
the actual SoC values, demonstrating the viability of employing the accepting the article for publication, acknowledges that the U.S.
nonlinear parameter b0 to predict SoC. When using the linear fit, the government retains a nonexclusive, paid-up, irrevocable, worldwide
mean absolute error (MAE) of the predicted SoC was 3.1% compared license to publish or reproduce the published form of this
with the actual SoC measured by the potentiostat, and the predicted R2 manuscript or allow others to do so, for U.S. government purposes.
value was 0.981, indicating a high level of prediction accuracy. DOE will provide public access to these results of federally
However, a relatively larger prediction error, approximately 10%, sponsored research in accordance with the DOE Public Access Plan
appeared as the battery neared 100%. Using the second-order polyno- (http://energy.gov/downloads/doe-public-access-plan).
mial model, the MAE improved to 2.8%, and the R2 increased slightly
to 0.985. Most notably, the maximum absolute error in prediction near AUTHOR DECLARATIONS
100% SoC was reduced to 6%. Thus, the second-order polynomial Conflict of Interest
model appears to offer greater accuracy for SoC prediction than the
linear model, particularly in SoC regions near 0% and 100%. In con- The authors have no conflicts to disclose.
clusion, the proposed method employing the nonlinear parameter b0 ,
measured via SHG, effectively predicts LIB SoC with an accuracy of Author Contributions
less than 3%. Further improvements may be achieved by using nonlin- Hongbin Sun: Conceptualization (equal); Data curation (equal);
ear models, particularly in SoC regions close to 0% and 100%. Formal analysis (equal); Funding acquisition (equal); Investigation
In this work, a nonlinear ultrasonic method, SHG, was used to (equal); Methodology (equal); Validation (equal); Writing – original
estimate the SoC of LIBs. The nonlinear parameter b0 was measured draft (equal); Writing – review & editing (equal). Pradeep Ramuhalli:
with the wave amplitudes at 2.5 and 5 MHz using a through- Funding acquisition (equal); Methodology (equal); Project administra-
transmission test setup. This nonlinear parameter was continuously tion (equal); Supervision (equal); Writing – review & editing (equal).
monitored throughout multiple charge and discharge cycles. A robust Ruhul Amin: Conceptualization (equal); Investigation (equal);
linear relationship was observed between the nonlinear parameter b0 Resources (equal); Software (equal); Writing – original draft (equal);
and actual SoC. To model this relationship, both a linear and a Writing – review & editing (equal). Ilias Belharouak: Funding acquisi-
second-order polynomial model were employed, and the two fitted tion (equal); Project administration (equal); Supervision (equal);
models were applied to predict SoC in a second battery based on the Writing – review & editing (equal).
Appl. Phys. Lett. 124, 073902 (2024); doi: 10.1063/5.0187829 124, 073902-5
Published under an exclusive license by AIP Publishing
Applied Physics Letters ARTICLE pubs.aip.org/aip/apl
16 21
K. H. Matlack, J.-Y. Kim, L. J. Jacobs, and J. Qu, J. Nondestruct. Eval. 34, 273 (2015). V. E. Nazarov and A. M. Sutin, J. Acoust. Soc. Am. 102, 3349 (1997).
17 22
S. Gebrekidan, T. Kang, H.-J. Kim, and S.-J. Song, Ultrasonics 85, 23 (2018). J.-Y. Kim, L. J. Jacobs, J. Qu, and J. W. Littles, J. Acoust. Soc. Am. 120, 1266
18
B. Fuchs, J. Qu, J.-Y. Kim, K. A. Unocic, Q. Guo, P. Ramuhalli, and L. J. Jacobs, (2006).
23
J. Appl. Phys. 130, 165102 (2021). G. Dace, R. B. Thompson, L. J. Brasche, D. K. Rehbein, and O. Buck, in Review
19
A. Bellotti, J.-Y. Kim, J. E. Bishop, B. H. Jared, K. Johnson, D. Susan, P. J. Noell, of Progress in Quantitative Nondestructive Evaluation (Springer, 1991), pp.
and L. J. Jacobs, J. Acoust. Soc. Am. 149, 158 (2021). 1685–1692.
20 24
A. G. Hsieh, B. J. Van Tassell, R. C. Mohr, A. Wilkinson, J. Ajo-Franklin, and G. Dace, R. Thompson, and O. Buck, in Review of Progress in Quantitative
S. Biswas, “Nonlinear acoustic resonance spectroscopy (NARS) for determining Nondestructive Evaluation (Plenum Press, 1992), Vol. 11, pp. 2069–2076.
25
physical conditions of batteries,” U.S. patent 10,502,793 (2019). P. Hess, A. M. Lomonosov, and A. P. Mayer, Ultrasonics 54, 39 (2014).
Appl. Phys. Lett. 124, 073902 (2024); doi: 10.1063/5.0187829 124, 073902-6
Published under an exclusive license by AIP Publishing