Keywords

1 Introduction

1.1 Traceability in Lithium-ion Battery Production

Traceability not only plays an important role in production but also over the entire life cycle of a battery cell. Thus, the EU Commission anchors transparency along the entire supply and value chain in its proposal for a regulation concerning batteries and waste batteries. In addition to the CO2 footprint during production, information about the origin and composition of the materials, repair or reuse, and end-of-life options such as recovery and recycling processes should also be provided [1].

In lithium-ion battery production, using a traceability system is considered a promising approach to reduce scrap rates, lower costs, and enable more efficient production [2]. It allows identifying possible problems or defects at an early stage of production. A traceability system collects information from trace objects, e.g. a single part or a segment of a continuous product during different phases of the product life cycle or at different production steps. It enables the assignment of process information based on tracing data for these objects. Generally, such systems consist of core elements, such as identification, data acquisition, data linking, and communication, as shown in Fig. 1 [3].

Fig. 1.
figure 1

Traceability system overview, adopted from [4].

Full size image

In industrial battery production, an approach for cell-specific identification of electrode segments has not yet been established. Serialization becomes almost impossible when splitting the coil into subcoils or cutting out additional segments. Initial cell-specific serialization usually takes place after electrode production during cell assembly. Identifying cell-specific electrode segments within the electrode production process is currently one of the biggest challenges in cell production of lithium-ion batteries. The reasons can be summarized as follows:

  • Change of batch structure (Raw materials → electrode paste → electrode coating → electrode sheets → cell assembly) [5].

  • Manual removal of electrode segments due to product defects or to perform quality control.

  • Prevention of material modifications or entry of any contamination due to the application of markings (e.g. ink).

1.2 Identification Techniques

Various techniques were investigated to select suitable identification solutions for lithium-ion battery production. Basically, trace objects are distinguished by their indirect and direct object features.

In the case of identification through indirect object features, the identification data is placed either on the workpiece itself or on workpiece carriers. The applied identifier is detected either contactless or tactile [6]. Optical markings belong to the group of contactless identification techniques. For this purpose, plain text, symbols, or an n-dimensional code such as a 1D bar code, a 2D Data Matrix code (DMC) or a color code is applied to the trace object and detected optically via digital image processing. Further options include electromagnetic or radio-based approaches. Trace objects could be equipped with active or passive radio modules (RFID, Bluetooth, WLAN modules), which transmit their identity signals. Tactile techniques require direct contact between identifier and sensor. For example, this includes devices that read the integrated circuits of smart cards.

Identification by direct object features eliminates the need for new identifiers to be applied to the trace object. Thus, a trace object is identified by its inherent characteristics, or the identity is calculated using logical models. For example, with some object materials it is possible to identify a trace object by its surface characteristics [7, 8] or to calculate e.g., the localization of a part or object segment based on process parameters via the use of physical-logical relations of the production sequence [9].

2 Cell-Specific Identification of Electrode Segments

Based on the described identification techniques and the requirements for electrode production, all labeling options have restrictions or limitations. Printed identifiers such as 1D barcodes or 2D Data Matrix code transmit printing ink [10], which may influence the battery’s performance. Laser-marked identifiers may cause material lesions due to the laser process, and mathematical models have gaps once a part of the electrode is removed. Therefore, a marker-free technique is currently being researched to enable cell-specific traceability.

2.1 Marker-Free Identification Using Surface Microstructures

In the joint project DigiBattPro 4.0, scientists from Fraunhofer IPA and Fraunhofer IPM are working on marker-free identification via existing surface microstructures or direct object features in the area of individual electrode segments.

A camera-based system was designed to uniquely identify individual components without labeling, using only the individual microstructure of the surface. Laboratory tests showed that aluminum and copper electrodes are suitable for marker-free identification, as shown in Fig. 2.

Fig. 2.
figure 2

Laboratory setup for the evaluation of electrode surface microstructures

Full size image

The marker-free identification system is based on the Track & Trace Fingerprint Technology developed by Fraunhofer IPM. It is a tracking system for mass-produced parts, enabling individual serialization and authentication. The technology is driven by the fact that many semi-finished goods or components have an individually distinct microscopic surface or interwoven color structure. An industrial camera takes high-resolution images of selected areas on the component’s surface. The specific structural pattern captured by the image and how they are positioned relative to each other is used to generate a numerical identification code, the so-called fingerprint. It is stored in a database and linked with other process and measurement data via a unique part ID. This process can be repeated to identify the component at a later stage in production.

Figure 3 shows an exemplary histogram on the fingerprint recognition of searched components or parts by comparing all fingerprints in a database. The x-axis shows the normalized Hamming distance, a measure of the relative similarity of two compared fingerprints. The smaller the value determined, the more similar the fingerprints are. The y-axis shows the number of fingerprints with the same degree of similarity. A typical successful comparison is shown in Fig. 3, where the sought-after part is sufficiently well separated from all other parts to make a statistically safe and unique identification. The distance between the center of the distribution of all other parts and the searched part in units of the standard deviation of the spread σ is given as a measure of the recognition reliability. When a comparison falls below a certain threshold of these measures, it is regarded as statistically unsafe and labeled as such. Thus, misidentifications are avoided. The value of these thresholds depends on the amount of fingerprints to be compared. A threshold value of 7 σ is sufficient to identify a single fingerprint out of a seven-digit number of fingerprints with statistical certainty.

Fig. 3.
figure 3

Exemplary histogram with statistical reliability of fingerprint signature recognition. The colors on the x-axis indicate if the sought-after part is identified with statistical certainty (green), identified but with insufficient confidence (yellow) or not identified at all (red).

Full size image

Up to now, the technology has been implemented in industrial production lines to track single objects [11, 12]. However, the existing identification solution for discrete production objects or pieces does not meet the requirements for continuous electrode production. Therefore, further development of the fingerprint system is necessary to close the existing information gaps between the individual processes of electrode production and increase the granularity of serialization (from coil to cell-specific tracing of electrode segments), as well as to prevent material entry into battery cells by marker-free tracing. In continuous material production, a challenge for marker-free identification is the enormous amount of image data that must be dealt with when registering an object for the first time. Fingerprints have to be recorded and stored so that identification is possible, even when taking the image anywhere on the electrode. The developed technology can generate fingerprints within a few hundreds of milliseconds.

To identify cell-specific electrode segments, two cases have to be distinguished. In the first case, identification takes place prior to cutting the electrode. Here, identification can be enhanced by consecutive sequencing of individual cell-specific electrode segments. Furthermore, the system needs to identify if individual segments have been cut out, e.g., due to quality aspects. In the second case, when cutting is already done, it may occur that the order of the single electrodes is lost. As a result, the number of possible candidates is much larger than in the first case. The proposed solution is to compare and process the captured fingerprints on high-performance graphics processing units (GPUs), as this process can be largely parallelized. With recent improvements in GPU performance, processing such data volumes becomes feasible. Thus, the required detection time for identifying fingerprints depends on the hardware system and software configuration. The detection time will be validated and optimized within the planned operation of the fingerprint demonstrator system, enabling the traceability of materials used in continuous processes such as coils or foils.

2.2 Approach for Proving Concept Implementation

The following chapter describes current developments and the implementation of a traceability system for marker-free identification in continuous electrode production. To develop and evaluate the fingerprint demonstrator system, it is integrated into a roll-to-roll coating system operating at conveyor speeds between 0.5 and 20 m/min (see Fig. 4). Substrate foils and electrodes up to 200 mm in width can be handled and treated with infrared radiators and inline calender. The equipment enables the development and testing according to current specifications on lithium-ion production processes defined in advance within the project.

Fig. 4.
figure 4

Roll-to-roll coating system with inline calender as basis for the fingerprint demonstrator.

Full size image

The implementation of the Track & Trace Fingerprint technology consists of at least two reader systems: One reader system scans and serializes the electrode and adds fingerprints to the database (to make them known to the system). The second reader system captures the fingerprints (tracing them in the database) and identifies the electrode segments. Consequently, a fingerprint reader system was developed (see Fig. 5) and two identical systems were set up from industrial computer vision components (see one of them in Fig. 6).

A reader system includes four high-resolution cameras with corresponding lenses, LED bars as illumination devices, and an IT interface to the central fingerprint management software. Stiff mechanical mounts were designed for capturing images without vibrations. Four cameras are needed to track four parallel tracks of electrodes. The sampling of the optical system is limited by the pixel size and amounts to 26,2 µm per pixel. The measuring field of a single image is 70 mm x 21 mm, while operating distance is about 230 mm. The actual area used for fingerprint generation is even smaller at 16 mm x 1 mm, resulting in a memory footprint of less than 1 KB per fingerprint. In the case of adding and tracing discrete parts, a control system such as a programmable logic controller (PLC) sends add or trace commands to the fingerprint system.

Fig. 5.
figure 5

Design of the two fingerprint reading systems in CAD, where System A has two possible mounting positions (Pos. 1 and Pos. 2).

Full size image

When applying in combination with endless material, the reader system itself has to keep track of the coil position and add fingerprints at a fixed distance from each other automatically, independent of the electrode segment length. This is done by image processing of the detected surface microstructure and necessary to ensure the generation of at least one fingerprint in the captured tracing area. The distance between the individual fingerprints varies depending on the camera setup and field of view, fingerprint dimensions, as well as processing times.

A coil with 800 m length requires around 30 million fingerprint comparisons for a single request. As the comparisons can be computed in a highly parallel manner, a GPU acceleration is promising to speed up the process to the required identification rate. Early tests on a low-end mobile GPU already show a doubling in performance compared to the previous system. The objective is to identify cell-specific electrode segments in production cycle time.

A first test setup of the developed system including the lighting system was validated and will be finally implemented on a coating machine. In a first step, the uncoated area of the electrode segment was successfully serialized, identified, and recognized by the developed Track & Trace Fingerprint-based system. In general, all technical surfaces that exhibit an individual microstructure are suitable for the technology. Based on this, further tests will be carried out on different electrode materials together with different coatings (e.g., Cu and Al substrates and graphite coating). In the current state, the system can support up to a 10 m/min feed rate but is projected to reach 20 m/min or more with further optimization. For further use of the fingerprint serialization within subsequent production processes, it is necessary to transfer the fingerprint information from continuous electrode production to discrete cell manufacturing and housing. From the point of cell assembly, this is currently realized by conventional tracking codes such as DMC.

Fig. 6.
figure 6

Prototype setup of the demonstrator on a coating machine.

Full size image

3 Conclusion and Outlook

Traceability of individual lithium-ion battery cells within electrode production is challenging, since there is currently no reliable way of implementing continuous traceability at single-cell level, thus linking electrode production and process data on a cell-specific basis. To meet this challenge and eliminate the existing identification and information gaps between process clusters, this paper established and presented a Track & Trace Fingerprint-based approach for continuous single-cell identification. Based on the newly designed Track & Trace Fingerprint system, a prototype setup was integrated and successfully implemented on a coating machine at the Center for Digitalized Battery Cell Manufacturing (ZDB) of Fraunhofer IPA. The hardware and software system will be further adapted and optimized in the next development steps to optimize the marker-free identification of continuous material. After testing precoated substrates or finished electrode rolls on the current coating and calendaring machine, a further expansion of the Track & Trace Fingerprint system is planned. The integration into additional coating lines allows evaluation of further aspects, such as modified material systems, residual moisture in the coating, and other influences accompanying the wet coating process.

The new marker-free identification approach enables the tracing of electrode segments without material changes and contamination due to the application of printed or lasered identifiers. Even if electrode segments are cut out, it will be possible to allocate electrode process parameters and data obtained during the coating process to the individual battery at cell level. This will provide a more in-depth understanding of the battery production process, a rapid localization of defects in the production chain, and consequently will improve product quality. Quality improvements and determining influential process factors will become possible by a more granular analysis. Besides, new high-quality data sets allow for data-driven production analysis and optimization. With the developed Track & Trace Fingerprint system, cell-specific traceability of lithium-ion battery components and process steps to the finished product becomes possible. In the future, the technology can be used for other sectors and continuous manufacturing processes such as the production of continuous materials.