Literature review: Data Analytics in

battery cell production for automotive

applications

Fran Šaban

Zagreb, 2024.
Fran Šaban

1. Introduction

The demand for electric vehicles (EVs) as a sustainable and efficient mode of transportation has been
steadily increasing the past few years. In 2023 there have been 3.5 million more electric car sales than in
2022, marking a 35% year-on-year increase (IEA, 2024). High demand for EVs as well as other electrical
appliances such as portable electronics reflects increasing demand for battery cells that power them.
While the production quality of lithium-ion battery (LIB) cells has improved considerably the past few
years, the production process of LIB battery cells still remains highly complex with a large number of
consecutive process steps. Material properties, electrode compositions and the design of the battery cell
significantly impacts the production process and the quality of the produced battery cell. The complexity
of LIB production also sparked interest in the development of new types of battery cells, such as solid
state batteries which could offer much higher energy density than the typical LIBs. Alongside the constant
improvement of current battery technologies in terms of the materials used and the manufacturing
process, the advent of data analytics provides a convenient way of analyzing data provided by the
production process, giving insights into what areas of the process can be optimized.

This review will provide an overview of current battery cell technologies, their production processes as
well as emerging battery cell technologies. It will also focus on the role of data analytics and its usage in
the optimization of the production process.

1
Fran Šaban

2. Classification of current battery cell technologies

To be able to discern the advantages and disadvantages of current battery technologies, key parameters
must be identified which indicate battery performance. The most important parameters to judge battery
performance (Budde-Meiwes et al., 2013) are given below.

● Power density is one of the most important parameters and is defined as the ratio of available
power in a battery cell to its weight or volume.
● Alongside power, energy density is also important as it indicates how much energy is contained in
the battery per unit volume or unit mass.
● Energy efficiency is the ratio of the discharged energy to charged energy, and as such it gives a
measure of energy losses during the charge-discharge cycle.
● Calendar lifetime is the battery lifetime until failure while it is not in use. This describes the
influence of environmental factors such as temperature on battery health during storage.
● As opposed to calendar lifetime, cycle lifetime describes how many cycles the battery can perform
until it fails. It depends on the average state of charge of the battery, depth of discharge, current
rate and other factors.
● The final parameter is the costs associated with a given battery type. While this is not a direct
measure of battery performance, it provides useful data on the profitability of a given battery
technology.

Batteries can also be classified into two groups. Primary (single-use) batteries are used once and
discarded since the discharge process permanently alters the electrode materials. That makes them
unsuitable for automotive applications due to their unsustainability and high costs. Secondary
(rechargeable) batteries can be charged and discharged multiple times using an applied electric current.
Rechargeability of the batteries makes them applicable in automotive applications. Only secondary
battery types will be further considered.

2.1. Lithium-ion batteries

Lithium-ion battery cells have become widely used as a means of portable energy storage in portable
electronics, EVs and grid storage because of their high energy and power density and long cycle life.

2
Fran Šaban

Lithium-ion battery cells are complex electrochemical systems consisting of an anode, cathode,
electrolyte, separator and housing which includes the battery terminals. First commercial lithium-ion
battery was patented by Akira Yoshino in 1987 using carbon as an anode and lithium cobalt oxide
(LiCoO2), developed by John B. Goodenough. Since the 1990s, much progress has been made in terms of
increasing power and energy density, cycle life, and improving safety while decreasing costs (Liu et al.,
2021). LIBs can be manufactured to optimize energy or power density by varying the electrode materials
and design. They offer the highest available power and energy densities out of all other battery types, as is
shown on Figure 1. The term lithium-ion battery encompasses a wide range of batteries that use lithium
salts, with the most prominent being lithium nickel manganese cobalt oxide (Lithium-NMC), lithium iron
phosphate (LFP) and lithium-titanium-oxide (LTO) batteries.

Figure 1. Power and energy density of various battery types

Cathode materials that are used in automotive LIB cells are electrochemically active materials such as
layered oxides (LiNiCoMnO2, LiNiCoAlO2), olivines (LiFePO4) and spinels (LiMn2O4). The anode is
made out of either synthetic or natural graphite. Small amounts of silicon (~5%) blended with graphite
can be used to increase the energy density. The addition of conductive additives (conductive graphite, CG;
carbon fibers; carbon nanotubes CNT) and polymer binders enhance electrical contact and provide low
impedance while enhancing mechanical integrity in terms of adhesion, stability and elasticity (Kwade et
al., 2018).

3
Fran Šaban

Figure 2. Components of LIB cell

The current collectors for the anode are made out of copper, while those for the cathode are made out of
aluminum. They should exhibit high conductivity and tensile strength to facilitate manufacturing. Cell
separators are made out of single- or multi-layer polymers, and can be coated with a thin layer of ceramic
materials which lower the risk of penetration by particles or swarfs and their heat-during cycling,
reducing the chance of a thermal runaway reaction resulting in the flaming of the cell.

Liquid electrolytes typically contain two or more carbonate solvents, one of which is ethylene carbonate
(EC) mixed with vinylene carbonate (VC). Their addition into the liquid electrolyte enables the formation
of a protective layer between the liquid electrolyte and the graphite particles of the anode, due to their
film-forming properties. The formed layer is called the solid electrolyte interphase (SEI). Lithium salts
are also incorporated into the liquid electrolyte mixture to enhance ion conductivity.

Today, LIBs are offered in two major production lines: high-energy and high-power batteries. The relative
volumes of metal foils and the separator (or electrolyte) are higher compared with the volume of active
mass in high-power batteries, so to reach the same energy capacity as high-energy batteries, high-power
batteries need more material which increases production costs. In addition to being able to specialize the
cells into high-power or high-energy depending on the use case, LIB cells also have high-capacity
utilization at high current rates. This is why they are especially suitable for use in high current
applications such as EVs. Currently, there are three different cell designs being produced and developed:
cylindrical, prismatic and pouch cells. Cylindrical cell design provides advantages among other cell
designs because of its long experience in production. Moreover, the cell container is dense and can
withstand a certain amount of inner pressure arising from heat generation during the charge cycle. On the
other hand, its disadvantages are lower packaging density and lower heat transportation from the cell to
the outside.

4
Fran Šaban

Figure 3. Cell designs for lithium-ion batteries: (a) cylindrical, (b) prismatic, (c) pouch (Magne et al., 2015.)

Prismatic cells are constructionally similar to cylindrical cells. The strength of the cell container makes
them the preferred choice in automotive applications. Pouch cells suffer the issue of welding quality even
after years of development, but it is theorized that they might be able to achieve lower production costs
and a slightly higher energy density as the case foils are lighter than the cylindrical case containers.

Mass production of LIBs for automotive use is still feasible because of the high costs associated. Cells
that are currently in mass production status for portable consumer electronic applications (mobile phones,
laptops, digital cameras, etc.) do not fulfill the demands for automotive applications, particularly with
respect to cell safety and lifetime. One more problem that faces mass production of LIBs is limited
availability of lithium, with the largest lithium deposits being limited to a few countries with the core area
being in South America.

2.2. Sodium-ion batteries

Sodium-ion batteries (SIB) are rechargeable batteries that use sodium ions (Na+) as their charge carriers.
The working principle of SIBs is similar to LIBs, but uses sodium as the intercalating ion instead of
lithium. Big advantage of using sodium instead of lithium is its natural abundance, especially in saltwater.
Cobalt, copper and nickel are not required for production of sodium based batteries, with iron-based
materials working well as a replacement (Zhao et al., 2017). Development of SIBs started in the 1970s

5
Fran Šaban

but was quickly overshadowed by a rising commercial interest in LIB technology. However, research
once again accelerated in recent years due to potentially lower production costs associated with SIB
technology and the possibility of mass production due to the abundance of sodium in the environment.

Design of SIBs is similar to LIBs, consisting of two electrode materials (anode and cathode), a porous
composite polymer membrane acting as a separator, and an electrolyte which contains sodium salt
dissolved in an organic liquid solvent. While SIBs are easier to manufacture than LIBs, the differences
between sodium and lithium in terms of their electrochemical potential, ionic radius, atomic weight and
their coordination modes hinder the current development and use of SIBs for use in high energy density
and efficiency applications such as the automotive industry. Today, SIBs are mainly used for fixed
electrical grid storage.

2.3.Solid-state batteries

Solid-state batteries (SSBs) use solid electrolytes instead of liquid ones. Using solid electrolytes offers a
few key advantages over its liquid counterparts. Firstly, battery crosstalk which occurs when soluble
components diffuse across the battery onto the other electrode is minimized, improving battery life and
safety (Janek et al., 2016.). Additionally, bulk polarization of both lithium ions and anions cause severe
concentration gradients during current flow limiting cell current. In SSBs, only lithium ions are mobile
which prevents bulk polarization and allows for higher current densities and quicker charging times.
Secondly, some inorganic solid electrolytes are stable at elevated temperatures with their conductivity
increasing with temperature, so their conductivity may reach that of liquid electrolytes (Janek and Zeier,
2016).

Electrolytes in SSBs can be divided into two groups: organic solid polymers and inorganic solids
(crystalline, glass or glass-ceramic). Polymer based electrolytes can compensate for volume changes of
electrodes by their ability to undergo plastic and elastic deformation. However, the conductivity of lithium
ions in polymer electrolytes is currently too low for battery operation at room temperature. On the other
hand, inorganic solid electrolytes (SEs) such as quaternary sulfides and thiophosphates (Li7P3S11 and
Li10GeP2S12) have been reported to have higher room temperature conductivities as opposed to liquid
electrolytes. Most solid electrolytes are easily reduced at low potentials and oxidized at intermediate
potentials. It is crucial to include protective interphases to stabilize electrolyte-electrode contact.

6
Fran Šaban

Inorganic SEs offer the highest lithium ion conductivity in the solid state, with phosphates and oxides also
showing great potential.

To be commercially viable, SSBs need to provide a significant performance jump, especially in terms of
energy density which is considered the top priority. The volumetric energy density of liquid electrolytes in
LIBs and SE are equal. If the liquid electrolyte in LIB is replaced by a denser SE, the gravimetric energy
density would be lower. To recover the lost energy density, the cell voltage would have to be increased
from 4.2V to 5V, raising the energy density by 20%. The use of SE is then only viable if it allows the use
of either high-voltage cathodes or high-capacity materials that cannot be in conjunction with liquid
electrolytes. In addition to energy density, long-term stability, long cycle and calendar life are also key
requirements that may be lowered by the mechanical strain caused by volume changes of the electrodes
during cycling. Moreover, high power and energy density requirements to outmatch LIBs may cause
problems during development of SSBs. All of these challenges need to be overcome to see the widespread
use of SSBs in EVs.

7
Fran Šaban

3. Production process of lithium-ion batteries

The production of lithium-ion batteries is a complicated process that requires multidisciplinary knowledge
of electrochemistry, mechanical engineering, data science and many others. Since the battery cells
undergo multiple production steps before being usable, tight process control and monitoring must be used
to ensure required energy and power density, stability, calendar lifetime, cycle lifetime and safety of the
produced battery cells.

Figure 4. Production process of LIBs

As is shown on Figure 4. the LIB production process can be divided into 5 key areas:

● material production and conditioning,

● electrode production,
● cell production,
● cell conditioning,
● assembly and lifecycle.

8
Fran Šaban

Battery materials are generally produced separately from the rest of the battery components and are as
such out of scope of this review.

3.1. Electrode production

Modern electrode and separator production processes are based mostly on liquid suspension and web
coating. A number of suspension strategies are formulated for the dry and wet mixing processes to blend
and structure the active material with the conducting agents and, sometimes, the binder. Two common
types of such processes are low-intensity dry homogenization and intensive dry mixing. The polymer
binder may alternatively be dissolved in solvent before conductive additives are dispersed in the binder
matrix solution structure. Using a planetary, or other type of high-intensity mixer, the mixtures are usually
fed to a batch wet dispersion process. That process ensures that the conductive additives are efficiently
structured and homogeneously dispersed. To finalize the process, the suspension is degassed and filtered
to achieve a stable, homogenous suspension. The suspension is then applied as a coating to the electrode
either continuously or in patterns. The application process of the coating depends on the electrode, current
collector, targeted cell geometry, throughput, calendaring technology, and minimization of production
rejects.

To enhance energy density and other physical properties, electrodes are compressed in a two-roller
calender. Thickness tolerances are restricted to ±2 µm which requires precise and sophisticated machine
design capable of high roll accuracy and process control. After calendaring (or sometimes before), the
electrodes are slit to width and length and dried before transfer into the dry room. The drying process
minimizes the water content inside the electrode improving cell lifetime. During every step of the process,
effective quality control of electrode qualities is required in conjunction with laboratory testing. Electrode
qualities that are monitored and tested during the production procedure are: thickness consistency, exact
edge geometry, low porosity, no defects, low tortuosity, high adhesion to the substrat, sufficient
mechanical coating stability/flexibility of the particulate-binder composite, low electrical resistance and
impedance.

9
Fran Šaban

3.2. Cell production

To begin the cell production process, the dried electrodes and separator are typically transferred into a dry
room. At present, there are three prevailing processes for production of electrode-separator assembly
(ESA). The first is winding, commonly used for small circular cell formats and has been transferred to
prismatic cells for EVs.

The limiting factors of electrode format and energy density are bending stress and crack formation, which
increase with electrode thickness. The second process of cell production is stacking which is superior over
winding in terms of applying uniform mechanical load to electrode sheets so that binding with the
electrode is less crucial. This allows for production of much thicker electrodes (thickness > 100 µm)
without the fear of coating failure, which improves the ratio of active material to cell volume and in turn
increases the energy density of the cell. While stacking offers production of electrodes with higher
thickness, it lacks the precision and productivity that winding provides.

Figure 5. Cell production processes

10
Fran Šaban

The third process is Z-folding which continually feeds the separator and layers it into a continuous zigzag
shape. The discrete electrode sheets are stacked in between the separator. This means less item handling,
faster processing and more precise alignment of ESA.

The ESA is contacted internally, usually by ultrasonic or laser welding and then inserted into a housing.
The housing is then filled with electrolyte under weak vacuum in extremely dry conditions and tightly
controlled temperatures. After filling, the cells are closed and the formation process is carried out,
followed by and aging procedure which may take two to three weeks. A light charging rate is applied
throughout the process to facilitate the creation of SEI layer. The most crucial quality parameters of cell
production are low remaining moisture within electrode and separator material, high accuracy of surface
alignment between the anode and cathode, low tab resistance after welding, and consistent wetting of the
ESA.

3.3. Module and pack assembly

Battery cells in automotive battery packs are arranged in battery modules. Each module is a separate
sub-system with internal electrical circuits. The bus bars that connect the cells to achieve pack voltage are
screwed or spot-welded to adjacent cell tabs. Since mechanical constraints are introduced in the system
during the process of connecting battery cells, the process must be carried out to ensure that no local
mechanical stress is introduced in the assembly. This also applies to the process of stacking cells and
modules using roll-and-eject and pick-and-place handling systems. The battery modules, battery
management system, wiring and, optionally, the cooling system, are inserted into a pack housing made of
aluminum or composite material to protect against temperature and mechanical shocks.

11
Fran Šaban

4. Data Analytics in the battery production process

In recent years, the advancements in data analytics and machine learning provided a novel way of
optimizing the production lines of various advanced technological components. Production process of
LIBs is, as was previously shown, a highly complex process containing manifold process-product
interdependencies and high sensitivity to ambient conditions. This is particularly evident in the high scrap
rates of up to 10% in series production and even higher values in the start-up phases of battery production
(Kampker et al., 2024). Using a data driven approach to collect data during the production and testing
phases, machine learning algorithms can identify patterns and correlations in the data, providing valuable
insights. Combining automated and manual data acquisition can format the data for easier manual or
automated analysis, management and visualization.

Batteries can make 30% of the cost of an EV, 30% of its weight and 50% of the CO2-emissions during the
manufacturing process (Turettsky et al., 2019). Being able to adjust the production process to meet the
desired economic or environmental goals is of great importance to manufacturers. That is possible only
through comprehension of the whole production system and its interdependencies.

Data analytics used to collect all the available data comprises of four levels of analytics, which are

● Descriptive Analytics analyzes historical data to identify trends, patterns and relationships,
● Predictive Analytics makes predictions about the future events or trends based on historical and
real-time data
● Diagnostic Analytics focuses on understanding the underlying factors that contribute to a
particular outcome or event,
● Preventive Analytics identifies potential problems before they occur.

4.1.Prediction and optimization in battery production

Predictive approaches offer the possibility of making a comprehensive concept that can enhance
production quality, based on the distinction that can be made along the production process between
process and machine parameters. Those parameters, such as web speed or dryer temperature, can be
directly influenced in the production process and resulting intermediate products (wet film thickness,
coating viscosity) can be evaluated. To enable data-driven solutions, the correlation of process and

12
Fran Šaban

intermediate product information is done by traceability systems as well as data aggregation in digital
twins. The digital twin serves as a digital model for the abstract description of a class of physical units
and its information as well as the simulation for the prediction and optimization of performance
characteristics behavior (Ludwigs et al., 2023) .The characteristics of the intermediate products can then
be predicted based on the process parameters. After predicting the intermediate characteristics, quality
characteristics (i.e. final cell properties) can also be predicted and integrated in feedback mechanisms
allowing for real-time optimization.The overarching concept for predictive quality is shown on Figure 6.

Figure 6. Feedback-loop diagram for optimization of battery production process (Kampker et al., 2024)

4.2.Data Acquisition

To be able to discern what data is relevant in the production process, the relation between intermediate
products and final product needs to be understood. Preliminary analysis of the data basis needs to be
conducted to identify different data sources and data types. The potential data for acquisition can be
gathered and evaluated manually, through expert interviews with scientific staff and analysts. As an
example, the result of expert interviews and analytical methods done by Turetskyy et al. (2019) during the
lithium-ion battery cell production in Battery LabFactory Braunschweig identified 16 process steps, 33

13
Fran Šaban

production machines and 43 different analytical devices and methods in order to analyze the intermediate
and final product properties. The final data basis contains the following data:

● Process/machine data describes all relevant adjustable process parameters and state variables,
without taking ambient conditions and technical building services into account. It also contains
information on how the production machines are operating.
● Energy demands data gives information about the energy demand of the equipment and machines
used in the production line. Energy demand measuring devices are typically installed separately
since most machines do not have that kind of measuring capability.
● Technical building service data includes the state of ambient conditions (temperature, humidity,
etc.).
● Intermediate product analytics addresses intermediate products using in-line or off-line analytical
methods to determine their structure or properties.
● Final product analysis provides diagnostic and analytic information of the battery cell on the
electrochemical scale (c-rates, capacity, etc.) helping to assess the quality of the final product.
● Operational data describes data relevant to planning and executing production campaigns such as
production times and the amount of material needed. This data is of vital importance since it
describes how to allocate all the other data in the production campaign.

Data generated during the production process can be collected with the use of sensors, PLCs, production
machines, technical building services and environmental conditions. Much of the data can be collected
automatically, but some part of the data, such as simulation results or final product analysis in a
laboratory, must be collected manually.

4.3.Data Mining

Data mining is the process of applying statistical methods or machine learning models to large data sets,
which are too large to be processed manually, to uncover trends and correlations. One of the most
common process definitions used in the industry today is the Cross Industry Standard Process for Data
Mining (CRISP DM). This is a recursive process, containing feedback loops and iterative steps to ensure
all knowledge and understanding derived from the process can be applied to increase power of the
algorithm. The six steps of the process visible on Figure 7 are:

14
Fran Šaban

● Business understanding is the first step in which the understanding of the business (system) is
gained and it underpins requirements and restrictions needed for further steps like data
preparation and modeling.

Figure X, CRISP DM cycle (Schnell et al., 2019)

● The next step is data understanding, where the initial data is gathered and explored using visual
and explorative analytic methods (scatter plot, heat maps, correlation analysis, etc.). This step
offers information needed to adjust the data mining goals set in the business understanding step.
● Data preparation step prepares the data for modeling. It is one of the most time intensive steps
since the quality of the input data is greatly reflected in the outcome of the modeling.
● The modeling step consists of selecting the appropriate modeling technique in terms of the data
mining goals. It also includes the building of the initial model, typically using machine learning,
and its assessment to adjust the quality or the form of the input data to improve the model further.
● In the evaluation step, the results of the model are evaluated, the data mining procedure is
reviewed and further steps are defined.
● The final step is the deployment where the obtained and evaluated results are visualized and
presented. This step may generate new knowledge about the investigated business/system, and
strategies for further improvement, adjustment or investigation of the business/system can be
derived.

15
Fran Šaban

Data mining methods can be divided into two categories: predictive and descriptive data mining.
Descriptive data mining is an unsupervised process that derives new nontrivial information from a given
data set, while predictive data mining is its supervised counterpart that uses models built based on training
data. The trained models can control machine parameters, provide continuous condition monitoring, or be
used for quality management (Thiede et al., 2019).

4.4. Machine learning in battery production process

Machine learning (ML) is the branch of artificial intelligence that deals with the development of
algorithms and models that can learn patterns and correlations in the given data alongside being capable
of unmonitored task execution. ML is especially useful in exploratory problems with exponentially
complex solutions and multiple input data streams. This makes ML suitable for use in much of the battery
production process, from the creation of novel electrolyte materials to the optimization of the production
process (Chen et al., 2020). Bhowmik et al. (2019) used deep learning neural networks (subset of ML) to
form an inverse-design process for understanding and controlling the formation process of SEIs,
combining datasets generated by multi-scale computer simulations, high-throughput synthesis and
laboratory testing.

A large body of literature exists on using ML models for adjusting manufacturing conditions to enhance
LIB electrode production. In a mapping study Haghi et al. (2023) showed that the electrode coating
process is the most investigated step in the process chain, while the drying process is in the minority.
Duquesnoy et al., (2021) analyzed key parameters from the early stage of the manufacturing process such
as the amount of active material in the slurry, liquid-to-solid ratio and the coating gap which provided
them with the ability to automate the characterization of coated NMC811 electrodes as heterogeneous or
homogeneous. Good results from the model allowed the researchers to graphically disclose the
relationship between manufacturing conditions and heterogeneity probabilities.

Liu et al. 2022 employed ML models to predict battery capacity and evaluated them on a dataset collected
from the coating stage of a real battery manufacturing chain. A regression ML model collected three key
coating parameters (mass, thickness and porosity) and predicted battery cell capacity, as well as
gravimetric and volumetric cell capacity if cell weight and dimension measurements were given. Their
underlying parameter dynamical dependencies and interactions were also analyzed. Battery deterioration

16
Fran Šaban

can also be modeled as is shown by Sharma and Bora (2023), who modeled battery deterioration using the
estimates of the remaining usable life of the battery. Samanta, Chowdhuri and Williamson (2021)

17
Fran Šaban

Literature

[1] IEA (2024) Trends in electric cars, Available at:

https://www.iea.org/reports/global-ev-outlook-2024/trends-in-electric-cars

[2] Budde-Meiwes, H., Drillkens, J., Lunz, B., Muennix, J., Rothgang, S., Kowal, J., & Sauer, D. U.
(2013). A review of current automotive battery technology and future prospects. Proceedings of the
Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering, 227(5), 761–776.
doi:10.1177/0954407013485567

[3] Liu, Y., Zhang, R., Wang, J., & Wang, Y. (2021). Current and future lithium-ion battery
manufacturing. iScience, 24(4), 102332. doi:10.1016/j.isci.2021.102332

[4] Kwade, A., Haselrieder, W., Leithoff, R., Modlinger, A., Dietrich, F., & Droeder, K. (2018). Current
status and challenges for automotive battery production technologies. Nature Energy, 3(4), 290–300.
doi:10.1038/s41560-018-0130-3

[5] Zhao, C., Lu, Y., Li, Y., Jiang, L., Rong, X., Hu, Y.-S., … Chen, L. (2017). Novel Methods for
Sodium-Ion Battery Materials. Small Methods, 1(5), 1600063. doi:10.1002/smtd.201600063

[6] Janek, J., & Zeier, W. G. (2016). A solid future for battery development. Nature Energy, 1(9), 16141.
doi:10.1038/nenergy.2016.141

[7] Bilgin, Berker & Magne, Pierre & Malysz, Pawel & Yang, Yinye & Pantelic, Vera & Preindl, Matthias
& Korobkine, Alexandre & Jiang, Weisheng & Lawford, Mark & Emadi, Ali. (2015). Making the Case
for Electrified Transportation. IEEE Transactions on Transportation Electrification. 1. 1-1.
10.1109/TTE.2015.2437338.

[8] Turetskyy, A., Thiede, S., Thomitzek, M., von Drachenfels, N., Pape, T., & Herrmann, C. (2019).
Towards data‐driven Applications in Lithium‐Ion Battery Cell Manufacturing. Energy Technology.
doi:10.1002/ente.201900136

[9] Thiede, S., Turetskyy, A., Kwade, A., Kara, S., & Herrmann, C. (2019). Data mining in battery
production chains towards multi-criterial quality prediction. CIRP Annals.
doi:10.1016/j.cirp.2019.04.066

18
Fran Šaban

[10] Bhowmik, A., Castelli, I. E., Garcia-Lastra, J. M., Jørgensen, P. B., Winther, O., & Vegge, T. (2019).
A perspective on inverse design of battery interphases using multi-scale modelling, experiments and
generative deep learning. Energy Storage Materials. doi:10.1016/j.ensm.2019.06.011

[11] Duquesnoy, M., Boyano, I., Ganborena, L., Cereijo, P., Ayerbe, E., & Franco, A. A. (2021). Machine
learning-based assessment of the impact of the manufacturing process on battery electrode heterogeneity.
Energy and AI, 5, 100090. doi:10.1016/j.egyai.2021.100090

[12] A. Kampker, H. Heimes, P. Lingohr, J. Schmied, H. Clever, B. Dorn (2024) Developing a concept for
the implementation of predictive quality in battery production. Procedia CIRP 126, 823–828.

[13] K. Liu, M. Faraji-Niri, G. Apachitei, M. Lain, D. Greenwood, J. Marco. (2022). Interpretable

machine learning for battery capacities prediction and coating parameters analysis. Control Engineering
Practice. 124. 105202. 10.1016/j.conengprac.2022.105202.

[14] S. Haghi, M. F. V. Hidalgo, M. F. Niri, R. Daub, J. Marco (2023) Machine Learning in Lithium-Ion
Battery Cell Production: A Comprehensive Mapping Study. Batteries & Supercaps, 6, e202300046.

[15] P. Sharma, B. J. Bora. (2023). A Review of Modern Machine Learning Techniques in the Prediction
of Remaining Useful Life of Lithium-Ion Batteries. Batteries 9, no. 1: 13.
https://doi.org/10.3390/batteries9010013

[16] C. Chen, Y. Zuo, W. Ye, X. Li, Z. Deng, S. P. Ong (2020). A Critical Review of Machine Learning of
Energy Materials. Advanced Energy Materials, 10(8), 1903242. doi:10.1002/aenm.201903242

[17] R. Ludwigs, J. Schmied, H. Clever, H. Heimes, A. Kampker. (2023). Digital Twin in the Battery
Production Context for the Realization of Industry 4.0 Applications. 10.15488/13433.

[18] A. Samanta, S. Chowdhuri, SS. Williamson. (2021) Machine Learning-Based Data-Driven Fault
Detection/Diagnosis of Lithium-Ion Battery: A Critical Review. Electronics. 10(11):1309.
https://doi.org/10.3390/electronics10111309

[19] E. Rohkohl, M. Schönemann, Y. Bodrov, C. Herrmann. (2022) A data mining approach for
continuous battery cell manufacturing processes from development towards production, Advances in
Industrial and Manufacturing Engineering, Volume 4, 100078, ISSN 2666-9129,
https://doi.org/10.1016/j.aime.2022.100078.

19
Fran Šaban

[20] J. Meiners, A. Fröhlich, K. Dröder. (2022) Potential of a machine learning based cross-process
control in lithium-ion battery production, Procedia CIRP, Volume 112, p. 525-530, ISSN 2212-8271,
https://doi.org/10.1016/j.procir.2022.09.093.

[21] Schnell, J., Nentwich, C., Endres, F., Kollenda, A., Distel, F., Knoche, T., & Reinhart, G. (2019).
Data mining in lithium-ion battery cell production. Journal of Power Sources, 413, 360–366.
doi:10.1016/j.jpowsour.2018.12.062

20