Method for identifying the position of a corner region of an electrode composite stack

Abstract

To provide a method and a program for determining a position a corner of a polygonal electrode sheet in at least one corner region of an electrode composite stack (ESV).SOLUTION: A method includes a step of: 3D image generating and shooting a corner region of an ESV in an imaging region by using an image generation process. As a result, a data set is created that includes 3D position information on an electrode sheet in the corner region of the ESV relative to a carrier 5 or relative to a marker arranged in the imaging region. The method also includes a step of determining, from the data set, a first edge extension and a second edge extension of the edges framing the corner region of each electrode sheet. A position of the corner of each electrode sheet is determined on the basis of the edge extensions. The electrode sheets and the edges of the electrode sheets can be determined via a neural network system in this case.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a method for identifying the position of a corner region of an electrode composite stack according to claim 1. The present invention further relates to a computer program for implementing the method.

[0002] When manufacturing an electrode composite stack, also referred to as an ESV (electrode separator composite), a plurality of electrode sheets, i.e., anodes and cathodes, are stacked one above the other in an alternating order. Between these electrode sheets, one separator layer each for electrically insulating the electrodes from each other is arranged.

[0003] The deposition accuracy of the electrode sheets is both a quality criterion for the process capability of the lamination process and a product characteristic related to the safety and function of the ESV. The aim is that all electrode sheets are deposited within a defined interval and a defined tolerance range. The higher the electrode coverage in the ESV, the higher the electrochemical performance. In many ESVs, the anode electrode sheet is larger by 1 millimeter to several millimeters in the circumferential direction in order to be able to completely cover the cathode electrode sheet. The smaller the protruding peripheral edge of the anode sheet in the circumferential direction, the more precisely the stacking of the stack has to be carried out. However, at the same time, an improvement in the manufacturing speed is also desired, which gets in the way of accurate deposition. For this reason, it is necessary to reliably, quickly, and accurately identify the deposition accuracy of the electrode sheets in the ESV.

[0004] In the prior art, various different methods for identifying deposition accuracy are known. Typically, X-ray radiation is used for image generation photography. However, the 3D image data obtained from such photography often has a poor signal-to-noise ratio (depending on the photography speed), and therefore the image data is often evaluated by an expert in image generation software in order to identify corresponding parameters such as the position of the corners of the electrode sheets.

[0005] Furthermore, in the prior art, it is possible to identify the rotation of the electrode sheet, and thereby determine the deposition accuracy, but it is not always possible to derive means for correcting the improved deposition.

[0006] Accordingly, an object of the present invention is to provide a method for automatically, robustly, and accurately determining the position of a corner region of an electrode sheet in an ESV.

[0007] The object of the present invention is solved by the method according to claim 1. Advantageous embodiments of the present invention are described in the dependent claims and will be explained below.

[0008] According to the present invention, there is provided a method, especially a computer-implemented method, for identifying the position of at least one corner region of an individual polygonal, especially rectangular, electrode sheet in an electrode composite stack, the method comprising: - generating a 3D image of a corner region of an electrode sheet of an electrode composite stack within a photographing region using an image generation method, especially a computed tomography-based image generation method, thereby generating a data set containing 3D position information of the electrode sheet within the corner region of the electrode composite stack relative to a support and / or relative to markers arranged within the photographing region of the image generation method; - identifying a first edge extension and a second edge extension of an edge bounding the corner region among the respective electrode sheets of the electrode composite stack from the data set, and identifying the position of the corner of each electrode sheet based on the corrected edge extensions, for example by extrapolation of the corrected edge extensions; including - First, the first edge extension and the second edge extension are identified by the following steps, namely: - generating a plurality of xz cross-sectional images and yz cross-sectional images; - identifying the electrode sheet in each of the xz cross-sectional images and each of the yz cross-sectional images; ○In each xz cross-sectional image and each yz cross-sectional image, and for each identified electrode sheet in the xz cross-sectional image and the yz cross-sectional image, ·Using a first neural network system, identifying the electrode sheet extension from a reference position region in the electrode composite stack to the electrode edge position of the electrode sheet; ·Adapting the electrode sheet extension such that the path length of the electrode sheet extension is maintained invariant and the electrode sheet extension extends from the reference region to the corrected electrode edge position at the height of the reference position; Steps to be performed; ○For each electrode sheet, a step of respectively identifying a straight line extending along the corrected electrode edge position along the first edge or the second edge, the straight line corresponding to the first edge extension and the second edge extension; A method is provided which is identified by.

[0009] The first convolutional neural network system is a trained convolutional neural network system configured to recognize, among other things, the electrode sheet extension.

[0010] Neural network is called neuronales Netz in the German-speaking region. In that case, convolutional neural network would roughly correspond to konvolutionales neuronales Netz, but this is not common in the German-speaking region, and for this reason, the term convolutional neural network, which is used worldwide, is used in this specification.

[0011] A coordinate system, in particular a Cartesian coordinate system, can be associated with a dataset, especially the photographed area. In this case, preferably, the z-axis is perpendicular to the electrode sheet, and the x-axis and y-axis are each oriented such that they are perpendicular to the z-axis. When the electrode sheet is a rectangular electrode sheet, the x-axis and y-axis are preferably oriented along the desired edge direction of the electrode sheet.

[0012] The coordinate system is preferably associated with markers and / or supports. The y-axis extends, for example, along the first extension direction of the support, and the x-axis extends along the second extension direction of the support.

[0013] In this context, it is noted that the terms "xz cross-sectional image" and "yz cross-sectional image" do not necessarily have to be understood in relation to a Cartesian coordinate system. First and foremost, they are used to conceptually distinguish multiple cross-sectional images. However, the xz cross-sectional image and the yz cross-sectional image do not extend parallel to each other. Preferably, the cross-sectional plane, i.e., the cross-sectional plane along which the cross-sectional image is generated, is oriented such that it extends substantially or exactly perpendicular to the first edge extension or the second edge extension of the ideally oriented electrode sheet. Especially when the support reflects the geometry of the electrode sheet along the first edge and the second edge, the cross-sectional image can also extend perpendicular to the first edge and the second edge of the support.

[0014] Here, for better understanding, it is pointed out that the term "edge extension" represents the entire edge of the electrode sheet, while the term "electrode edge position" represents, in particular, the position of the electrode edge within the cross-sectional image. From the multiple electrode edge positions within the cross-sectional image, the edge extension of the electrode sheet can be specified in 3D for each respective electrode sheet.

[0015] The position of the electrode edge corresponds, in this case, in particular to the position of the edge section located outside the stack / the periphery of the electrode sheet among the electrode sheets.

[0016] According to one embodiment of the present invention, the corrected electrode edge position is determined by specifying the electrode sheet position at a plurality of sample points. The first sample point corresponds to the electrode edge position. The reference position of the electrode sheet in the electrode composite stack is detected by at least one further sample point. The sample points are specified by a first convolutional neural network system. The corrected electrode edge position is achieved by displacing at least the first sample point to the height of the reference position, and the path length between the first sample point and at least one further sample point is maintained constant so that the path length of the electrode sheet extension remains unchanged.

[0017] From the sample points within each cross-sectional image and within the electrode sheet, a path having a corresponding path length can be specified. The path length can be specified based on the distance between the sample points. Alternatively, it is also possible to specify the path length of the path from the path (for example, the spline segment of the sample points) specified by the sample points.

[0018] One important aspect in specifying the corrected electrode edge position is that the electrode sheet extension or the path length of the path remains constant. This achieves that the corrected electrode edge position does not suggest distortion or elongation of the electrode sheet.

[0019] In particular, within each cross-sectional image and for each electrode sheet, all the specimen points of the electrode sheet are displaced so as to form a single linear path within the cross-sectional image, in which case these specimen points are displaced to the height of the reference position. Thereby, one edge extension is identified that would correspond to an electrode sheet that extends completely in one plane. In many cases, some of the plurality of electrode sheets in the stack are bent or folded at the edges, which makes it difficult to identify the position of the corner region. Displacing the specimen points to the height of the reference position solves this problem.

[0020] The identification of the specimen points is used, in particular, to check the electrode sheet extension within the cross-sectional image, and the corrected edge extension is achieved by performing a "linear bending" of the electrode sheet extension within the cross-sectional image, if this is required.

[0021] From the corrected electrode edge positions, for each electrode sheet, the corrected edge extension can be identified. If the electrode sheets are arranged completely flat within the stack, from this corrected electrode edge extension, the position of the corner region or the position of the corners of the electrode sheets can be identified by extrapolation or other appropriate measures. From this information alone, the above-mentioned quality criteria and / or performance criteria can be reliably derived.

[0022] The term "neural network system" includes, in particular, preferably a convolutional neural network system, and for this reason is also abbreviated as a CNN system.

[0023] A neural network system is, in the context of this specification, always a trained system that has been trained to identify or classify corresponding features based on a correspondingly labeled dataset.

[0024] The corresponding training methods are known to those skilled in the art.

[0025] The CNN system includes, among other things, a plurality of neural networks, among other things, a plurality of convolutional neural networks, each of which is trained and designed for a specific task.

[0026] According to one embodiment of the present invention, each xz cross-sectional image extends along, is shifted parallel to, and is generated along a plane including a first cutting direction (x) that extends perpendicular to the stacking direction (z) of the electrode composite stack and a first edge extension direction transmitted by the support and / or marker, and / or each yz cross-sectional image extends along, is shifted parallel to, and is generated along a plane including a second cutting direction (y) that extends perpendicular to the stacking direction (z) of the electrode composite stack and a second edge extension direction transmitted by the support or marker.

[0027] According to a further embodiment of the present invention, the following steps, namely - For each electrode sheet, identifying a first straight line corresponding to a corrected first edge extension passing through a corrected electrode edge position identified from the xz cross-sectional image, - For each electrode sheet, identifying a second straight line corresponding to a corrected second edge extension passing through a corrected electrode edge position identified from the yz cross-sectional image are provided.

[0028] According to a further embodiment of the present invention, for each electrode sheet, the intersection of the first edge extension and the second edge extension is identified, and in particular, the identification of the intersection includes the intersection with respect to the projection of the edge extension along the stacking direction, i.e., onto one plane, and the position of the corner of each electrode sheet is associated with the intersection.

[0029] In the electrode sheet, the corner region may be rounded or damaged, so this embodiment can be used to very accurately identify the position of the corner of the electrode sheet.

[0030] According to a further embodiment of the present invention, the method is carried out on two or more corner regions of the electrode sheet of the electrode composite stack, and the positions of two or more corners of the electrode sheet are determined.

[0031] The plurality of corner regions may be carried out in parallel during a single photographing of the electrode composite stack, or may be carried out sequentially by different photographings of each corner region.

[0032] By photographing and evaluating a plurality of corner regions, and thus a plurality of corner positions, it becomes possible to more accurately determine the posture of the electrode sheet in the electrode composite stack.

[0033] In particular, it is envisaged to determine the positions of the corners of two corner regions facing each other diagonally in the electrode sheet. This is particularly advantageous in the case of a rectangular electrode sheet geometry.

[0034] According to a further embodiment of the present invention, the electrode sheet is identified in the xz cross-sectional image and in the yz cross-sectional image by a further neural network system for identifying the electrode sheet, for example by a second trained, in particular convolutional neural network system, and in particular the electrode sheet is identified in the region of the electrode composite stack that includes the reference position.

[0035] The reference position is located, in particular, in the region of the stack where the stack is framed and pressed by the support so that the spacing and orientation of the plurality of electrode sheets are substantially the same. In particular, in this case, the orientation along the cross-sectional image is carried out parallel to the support, whereby the plurality of electrode sheets will, in particular, extend horizontally parallel to each other and along the x-direction or the y-direction.

[0036] With this second CNN system, it becomes possible to identify individual electrode sheets in the cross-sectional image, and for example, it also becomes possible to determine a sequence, such as in the form of numbering the electrode sheets. Thereby, the relative positions of the electrode sheets in the stack can be specified. The plurality of electrode sheets may each be different with respect to the gray values in the cross-sectional image. In particular, the gray value of the electrode sheet within the region of the electrode sheet varies according to the position within the electrode sheet.

[0037] Regardless of this, the CNN system according to this embodiment enables robust and accurate identification of the electrode sheets.

[0038] The second CNN system can include, for example, three convolutional neural networks (CNNs). The first CNN (CNN3) of the second CNN system is configured to identify the region in the cross-sectional image where the electrode sheet extends between the upper part and the lower part of the support. In this case, the electrode sheet is pressed between the upper part and the lower part of the support.

[0039] This region particularly includes a reference position and particularly refers to a region where there is no intermediate space between the electrode sheets. This region is also referred to as the reference position region in the context of this specification.

[0040] The second CNN (CNN6) of the second CNN system is configured to identify, for example, the first type of electrode sheet, such as the anode, within the reference position region. The third CNN (CNN7) of the second CNN system is configured to identify, for example, the second type of electrode sheet, such as the cathode, within the reference position region.

[0041] In particular, for the CNN, the number of the first type and the second type of electrode sheets is set as a boundary condition (or is already implicitly created in the training data set), which brings about an improvement in robustness when identifying the electrode sheets.

[0042] In this way, based on the second CNN system, the electrode sheet can be identified, especially according to the type and with respect to the stacking order within the reference position area.

[0043] According to a further embodiment of the present invention, based on the corrected first edge extension and the corrected second edge extension, and the position of at least one corner of each electrode sheet, the posture of each electrode sheet with respect to the support and / or the marker is determined.

[0044] In particular, when the geometry of the electrode sheet is known, those skilled in the art can determine the position and orientation of the electrode sheet, which is also known as the posture. Thereby, for example, the positions of all the corners of the electrode sheet can be specified, in which case only one corner region is photographed and evaluated.

[0045] According to a further embodiment of the present invention, based on the specified posture, which is also referred to as the actual posture in the context of this specification, for each electrode sheet, the deviation from a predetermined posture, for example from a target posture, with respect to the support and / or the marker is specified. In particular, the deviation of the posture is implemented by data regarding the instantaneous center for each electrode sheet.

[0046] In particular, when describing the deviation of the posture of the electrode sheet, not only linear translational motion but also rotational motion can be described by the instantaneous center through the rotation center and the rotation angle of the electrode sheet. The deviation of the posture can be completely described using only two values (rotation center and rotation angle) by the instantaneous center.

[0047] When the deviation is determined, for example, when a systematic deviation of the posture is specified, it becomes possible to optimize the manufacturing process.

[0048] According to a further embodiment of the present invention, during the manufacturing method for a further electrode composite stack, the deposition posture of the electrode sheets in the electrode composite stack is adapted based on a specified deviation so that the deviation in the further electrode composite stack becomes smaller.

[0049] Thereby, during the manufacturing process, optimization of the deposition of the electrode sheets can already be carried out.

[0050] According to a further embodiment of the present invention, for each electrode sheet, from a specified position of the corner, the positions of the remaining corners of each electrode sheet that have not yet been specified are determined, and the determination of the remaining corners is carried out from the dimensions of the electrode sheets stored in the database.

[0051] According to this embodiment, overall information regarding the posture, geometry, and position of all corners of the electrode sheet can be specified, which makes it possible to improve the degree of robustness and accuracy of the method.

[0052] According to a further embodiment of the present invention, the plurality of electrode sheets includes a plurality of electrode sheets of a first type and a plurality of electrode sheets of a second type.

[0053] The electrode sheets of the first type may be, for example, anodes as already described in the preceding paragraph, and the electrode sheets of the second type may be, for example, cathodes.

[0054] According to a further embodiment of the present invention, the area of the electrode sheets of the first type is larger than that of the electrode sheets of the second type, and the electrode sheets of the first type and the electrode sheets of the second type are alternately laminated in the electrode composite stack such that the electrode sheets of the first type protrude beyond the electrode sheets of the second type at least on one side, and particularly on all sides, within the electrode composite stack. In particular, these electrode sheets are laminated concentrically.

[0055] In this context, the term "concentric" means that the centers of gravity of the respective electrode sheets are ideally located on the same axis parallel to the stacking direction, and that the plurality of electrode sheets have the same orientation.

[0056] Between these electrode sheets, separator layers are arranged as already described in the previous paragraph.

[0057] In particular, the first type of electrode sheet is several micrometers or millimeters larger in the circumferential direction in order to be able to completely cover the second type of electrode sheet. In particular, the first type of electrode sheet may deviate from the reference position within the circumferentially protruding region, for example, it may be folded or bent. When the electrode sheet protrudes in this way, the reference position region can be defined as the region where the electrode sheet is deposited on top of the second type of electrode sheet without protruding. Within this region, typically the degree of parallelism between the electrode sheets is high, so this region is provided as a reference position, in particular as a reference height, and along this reference position, in particular along this reference height, the corrected electrode sheet extension extends.

[0058] According to a further embodiment of the invention, the first type of electrode sheet has a different gray value range in the cross-sectional image than the second type of electrode sheet, so that the plurality of electrode sheets can be distinguished based on different gray value ranges in the data set.

[0059] The gray value in the cross-sectional image is determined, for example, from the absorption coefficient of the electrode sheet, in particular from the absorption coefficient in the X-ray region of a computed tomography-based image generation method.

[0060] In this case, the term "gray value" also includes the color in the cross-sectional image, in particular the pseudo-color. Thus, the electrode sheets of the first and second electrode types can also differ from each other with respect to their respective colors in the cross-sectional image.

[0061] According to a further embodiment of the present invention, in the first image processing step, a third trained neural network system for recognizing the provisional electrode edge position of the electrode sheet tentatively identifies the electrode edge position of the electrode sheet in each xz cross-sectional image and in each yz cross-sectional image and for each electrode sheet, and further, in each xz cross-sectional image and in each yz cross-sectional image and for each electrode sheet, a reference position is identified, and the reference position is identified from a region of the electrode composite stack in which the electrode sheets are stacked at regular intervals, especially without gaps along the stacking direction.

[0062] If the electrode composite stack has first and second type electrode sheets and the first type electrode sheet protrudes beyond the second type electrode sheet in the circumferential direction, the following CNN system architecture may be advantageous.

[0063] In this case, the third CNN system includes, for example, a first CNN (CNN2), and the first CNN (CNN2) is trained to identify a region of the electrode composite stack that includes the electrode edge position of the second type electrode sheet, i.e., for example, the (smaller) cathode, in each cross-sectional image.

[0064] Furthermore, the third CNN system includes a second CNN (CNN5), and the second CNN (CNN5) is trained to detect the electrode edge position of the second type electrode sheet. In this case, this electrode edge position is treated only as a provisional result, which is expressed by the term "tentatively identified". To enable reliable provisional identification, the number of electrode edge positions to be identified can be set (e.g., implicitly by the training set or by additional boundary conditions). By limiting the region for recognizing the electrode edge position, an improvement in robustness and reliability is achieved.

[0065] The position of the electrode edge of the first type is also tentatively identified in the same way. That is, the third CNN system further includes a third CNN (CNN1), and the third CNN (CNN1) identifies a region that includes the position of the electrode edge of the first type in each cross-sectional image. Then, within this region, a fourth CNN (CNN4) tentatively identifies the position of the electrode edge of the first type of electrode sheet.

[0066] To enable reliable identification, the number of electrode edge positions to be identified can be set (implicitly, for example, by a training set or additional boundary conditions). By limiting the region for recognizing the electrode edge position, an improvement in robustness and reliability is achieved.

[0067] In the features (type of electrode sheet, position of the electrode sheet in the electrode composite stack, reference position region, tentative electrode edge position) recognized by this third CNN system and possibly the second CNN system, the first CNN system can identify, for example, a first sample point and possibly additional sample points, thereby identifying a path that contains a plurality of sample points, connects the plurality of sample points, and reproduces the extension of the electrode sheet in the cross-sectional image. In particular, in this case, the first sample point corresponds to the (final) electrode edge position, that is, it no longer corresponds to the tentatively identified edge position.

[0068] The creation of a path passing through the sample points can also be implemented using a CNN.

[0069] According to a further embodiment of the present invention, in the second image processing step, the first trained CNN system identifies a plurality of sample points within each xz cross-sectional image and within each yz cross-sectional image, and for each electrode sheet. The identification for both the first type of electrode sheet and the second type of electrode sheet is respectively within the first type or the second type of electrode sheet, the innermost in the direction of the center of the electrode composite stack, i.e., farthest from the temporarily identified electrode edge position located at the electrode edge, and the outermost of the first type or the second type of electrode sheet, i.e., farthest from the center of the electrode composite stack. It is limited to the region set by the temporarily identified electrode edge position.

[0070] This limitation enables reliable identification of sample points by the first CNN system.

[0071] According to a further embodiment of the present invention, for each xz cross-sectional image and each yz cross-sectional image, and for each electrode sheet, a path and a path length associated with the path corresponding to the path lengths of the plurality of sample points are identified. For each xz cross-sectional image and each yz cross-sectional image, and for each electrode sheet, the identified sample points are displaced by conversion to a straight-line path perpendicular to the height of the electrode composite stack. Specifically, they are displaced so that the path length is maintained constant. As a result, the height of the first sample point within each cross-sectional image, especially the height of all sample points of each electrode sheet, is adjusted to a reference position.

[0072] According to a further embodiment of the present invention, the support has a corner region at its own corners, and at least the corner region of the support has a lower absorption coefficient compared to the central region of the support. Thus, the X-ray radiation of the image generation method for photographing the electrode composite stack is less absorbed by the corner region of the support than by the central region of the support.

[0073] The corner region of the support body corresponds, in particular, to the corner region of the electrode composite stack.

[0074] According to a further aspect of the invention, there is provided a computer program comprising computer-readable computer code which, when executed on a computer, implements the method according to one of the above-described embodiments.

[0075] The computer program can be stored as a computer program product on a computer-readable non-transitory storage medium.

[0076] According to a further aspect of the invention, there is provided a support body for an electrode composite stack, the electrode composite stack having a central region framed by the edges of the support body, the support body further having a plurality of corner regions positioned at the ends of the edges of the support body, the support body having a lower absorption rate in the corner regions compared to the central region, and thus the corner regions being transparent to X-ray radiation, i.e., absorbing less than 20% of the radiation, and the central region not being transparent to this radiation, in particular the central region absorbing more than 80% of the radiation.

[0077] In particular, the central region of the support body has special steel.

[0078] The support body can be used as a support body in the method according to the invention.

[0079] In particular, the corner regions are made of a material having a low absorption coefficient. Alternatively, the thickness of the corner regions can be made thinner than that of the central region so that the absorption rate is low.

[0080] According to a further embodiment of the support body, the corner regions are made of aluminum, carbon, or plastic, in particular the plastic containing ABS, POM.

[0081] In particular, the mass attenuation coefficients in the corner regions are 2.62 and 0.011 m for photon energies of 10 keV and 300 keV. 2 / kg range.

[0082] According to a further embodiment of the present invention, the mass attenuation coefficients in the corner regions are 23.7 and 0.011 m for photon energies of 10 keV and 300 keV. 2 / kg range.

[0083] In particular, the imaging area includes a portion of the support, which can therefore be used as a reference system for determining the position or orientation of the corners of the electroded sheet.

[0084] According to a further embodiment of the present invention, marker spheres are arranged in a pre-specified three-dimensional arrangement within the imaging area, thereby allowing a uniquely defined coordinate system to be identified for each imaging session, thereby allowing the position or orientation of the corners of the electrode sheet relative to this coordinate system to be identified and suggested.

[0085] Further features and advantages of the invention will be explained below on the basis of the description of the drawings of exemplary embodiments. [Brief description of the drawings]

[0086]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

Figure 17

[0087] Figure 1 shows a schematic diagram of a computed tomography-based image generation system for capturing a dataset of 3D position information of electrode sheets in an electrode composite stack (ESV) relative to a support. The image generation system includes an X-ray light source 6 and an X-ray imaging device 7 (detector) configured to create a computed tomography photograph 8 of at least one corner region of the ESV. The ESV is held for this purpose on a support 5, by which it is pressed along the stacking direction of the ESV (here, without loss of generality, along the z-axis of the Cartesian coordinate system KS). The support 5 has receiving means that guarantee a reproducible accommodation of the support 5 via a centering bush. The support 5 is clamped in a holding device 3 so as to be rotatable about a first axis of rotation 500 such that the electrode sheets in the ESV can be completely captured in three dimensions, at least from the corner regions of the ESV. The holding device 3 itself can be fixed on a turntable 1 and can be supported so as to be rotatable relative to the turntable 1 about a further axis of rotation 501.

[0088] Based on the reproducibility of the position of the coordinate system, the position of the ESV, the position of the X-ray light source (or the X-ray radiation direction), and the position of the detector, the inspection space, i.e., the volume from which the photograph is captured, will always be in the same position and have the same orientation. In this way, it becomes possible, for example, to uniquely describe later the orientation of the electrode sheets relative to the support or relative to markers present in the inspection space.

[0089] FIG. 2 (similarly for FIG. 5) shows one embodiment of the support 5. The support 5 includes two support plates SD and SB, and an ESV is sandwiched between these two support plates along the stacking direction (z-axis). The support plates have a central region that covers the ESV in the circumferential direction, and at the edges of this central region, spaces and openings for connection elements (e.g., screws or bolts) for the two support plates are arranged. The support 5 further has a region 5-2, and this region 5-2 is made of a material that is weakly absorbent within the region of CT radiation or is made thinner than the central region 5-1 of the support 5. The weakly absorbent region 5-2 is located at the corners of the electrode sheet, thus ensuring an improvement in the imaging quality of the electrode sheet in the regions of the corners E1, E2, E3, E4.

[0090] The electrode sheet of the ESV in this embodiment is rectangular, and the edges of the long sides of the electrode sheet extend parallel to the edges of the support, and thus, the edge direction of the electrode sheet is transmitted through the support. The ESV has four corner regions E1, E2, E3, E4, and in these four corner regions E1, E2, E3, E4, the corner regions and corners of the electrode sheet are arranged. Conductor lugs 9 for the cathode and anode are provided on the end faces of the ESV, respectively.

[0091] In this embodiment, the supports 5, 5-2 overlap the electrode sheet of the ESV only in the central region 5-1 in the lateral direction, and in this case, the corner regions of the electrode sheet are exposed. Alternatively, the support 5 in this region 5-2 may be made of a material that is (X-ray radiation) transmissive, which improves the stability and robustness of the composite against impacts.

[0092] Figures 3 and 4 show cross-sectional views of an idealized ESV. The ESV consists of a plurality of electrode sheets of a first type and a second type, which are alternately spaced apart and insulated from each other along the stacking direction (z-axis) of the ESV by a separator layer S and are bundled into the ESV, and is pressed by a support 5 including support portions SD and SB, at least in the central region of the support 5. The area to be imaged is preferably covered only by the area 5-2 of the support, and thus, radiographic imaging can be efficiently performed. In both these areas 5-2 and the area 5-1, the electrode sheets extend parallel to the support plate of the support portion, and this direction / plane corresponds to, for example, the x-y direction / plane transmitted by the support 5.

[0093] The electrode sheets of the first type correspond to the anode A, and the electrode sheets of the second type correspond to the cathode K. As can be particularly seen from FIG. 4, the anode is larger than the cathode in the circumferential direction, and thus protrudes beyond the cathode in the direction of the x-axis along the illustrated cross-section (spacing d AK (x)). That is, the anode also protrudes beyond the cathode along the y-axis (not shown). The separator layer itself protrudes beyond the anode in the circumferential direction (spacing d SA (x)). The latter is useful for the process reliability of the ESV, while on the other hand, the anode protrudes beyond the cathode to ensure that the cathode is completely covered, thereby ensuring the performance of the ESV.

[0094] These electrodes and separator layers can be identified as functions of their respective positions (A(z), K(z), S(z)) within the ESV.

[0095] In most cases, the separator layer cannot be discerned in a computed tomography image.

[0096] Therefore, the rectangular electrode sheet in this example has four corner regions E1 to E4, and at these corner regions E1 to E4, the edges of the electrode sheet intersect each other (FIG. 5).

[0097] In the manufacturing process of the ESV, a plurality of electrode sheets are sequentially stacked vertically. In this case, the electrode sheet or the corners of the electrode sheet must be positioned within a predetermined tolerance range along the x-direction and the y-direction. This is schematically shown in FIG. 6. Therefore, when the positions of a plurality of corners are determined, or when the edges forming the corners and their extending directions are determined, it becomes possible to estimate the posture of the electrode sheet based on the known geometric shape of the electrode sheet.

[0098] The geometric shape of each electrode sheet can be determined by an optical method, for example, before laminating it onto the ESV. As a result, the geometric shape, particularly the dimensions, are known for each electrode sheet. For example, optical detection can be performed by a camera system connected to an evaluation unit, and thereby the geometric shape can be specified and associated with each electrode sheet.

[0099] The corners of the electrode sheet are defined particularly by the extension lines of the edges forming the corners. This is because the electrode sheet may be partially rounded or chipped at the corners.

[0100] Figure 6 shows, for one corner region, within what tolerance ranges the corners of all the electrode sheets in the stack should be located. In this case, the anode, the cathode, and (also the separator layer) are distinguished. In this case, the position of the anode corner in the x-y plane is indicated, for example, in the form of A(x) (x coordinate) and A(y) (y coordinate), while on the other hand, the position of the cathode corner in the x-y plane is indicated in the form of K(x) (x coordinate) and K(y) (y coordinate). The tolerance ranges are shown in Figure 6 as corridors along the respective axes. For the electrode edges to be located inside these corridors, the electrode sheets should perform only very slight rotational and translational movements relative to the ideal centroid position of the electrode sheets.

[0101] By specifying the posture of the corner and the posture of the edge of that corner, it is possible to confirm whether each electrode sheet is located within a predetermined tolerance corridor. According to the present invention, this should be specified for each electrode sheet within the ESV.

[0102] The tolerance ranges are typically within the submillimeter range. On the other hand, the interval d AK (x) in the x direction and the interval d AK (y) in the y direction are also within the submillimeter range. That is, the anode protrudes only slightly, which requires high precision when manufacturing the stack.

[0103] Generally speaking, in the case of an ESV stack containing N (for example, 20) cathode sheets, it can be said that a plurality of check features can be specified. In this case, the check features include, for example, the following check criteria: ·All four corners of the 2N + 1 separator layers must be located within the deposition plane defined for the separator in the x direction and the y direction. ·All four corners of the N + 1 anodes must be located within the deposition plane defined for the anode in the x direction and the y direction. ·All four corners of the N cathodes must be located within the deposition plane defined for the cathodes in the x- and y-directions. ·Each different electrode sheet and separator layer must be deposited (alternately) in the correct order, i.e., S1 - A1 - S2 - K1 - S3 - A2 - S4 - K2 - S5 ······ A N+1 - S 2N+1 in the order of. ·The distance between all separators and all anodes in the x- and y-directions must be greater than the minimum distance d SA (x) or d SA (y). ·The distance between all anodes and all cathodes in the x- and y-directions must be greater than the minimum distance d AK (x) or d AK (y). ·For the separators, anodes, and cathodes, the relative position of the ESV with respect to the support can be centered so that the optimal straight lines in the x- and y-directions can be defined at all corners.

[0104] Figure 7 shows a flowchart showing the general flow of the method according to the present invention. By data acquisition, a dataset of two corner regions facing each other on the diagonal of the ESV, including 3D position information of the rectangular electrode sheet, is generated in this embodiment. After this data acquisition, for each of these corner regions and for each electrode sheet, data processing and data analysis are performed so that the corner positions are determined by being decomposed according to the type and position in the stack of the electrode sheet.

[0105] From these positions, it is possible to determine with what accuracy / tolerance the electrodes in the ESV are arranged. For this purpose, the orientation within the stack can be determined for each individual electrode sheet. Within the framework of result integration, for example, the relative rotational and translational movements with respect to the target orientation are determined via the known geometry (rectangle) of the electrode sheet and the identified corner positions facing each other on the diagonal of the electrode sheet, whereby the orientation of each electrode sheet, i.e., the actual orientation (direction and position), is determined. This can advantageously be carried out by determining the instant center that indicates and quantifies the deviation between the target orientation and the actual orientation using the rotation angle and the center of rotation. For example, within the framework of result utilization, based on this deviation, further deposition of the electrode sheets can be corrected so that this deviation is minimized. This deviation may also be used for the quantification of performance or manufacturing tolerance.

[0106] In this context, it is noted that the position and orientation are determined, as described above, based on a coordinate system transmitted, for example, by a support or by markers arranged within the space to be examined. Thereby, the actual orientation of the electrode sheet can be accurately and repeatedly determined from the data.

[0107] In FIG. 8, the evaluation principle of result extraction for the corner region E1 of the ESV is schematically shown. In this case, the 3D positions of the corners of the electrode sheet are determined by being decomposed according to the type from the captured data for each individual electrode sheet. Regarding the position along the z-axis, this position along the z-axis can also be indicated by an index representing the position within the stack. The x- and y-coordinates of the corner, i.e., the corner position, are determined very accurately by the method described below. The corner positions thus determined can be shown using a graph in the coordinate system (right panel of FIG. 8). In this embodiment, it can be seen that the corners of the electrode sheet are located within a specific region of the tolerance range of the deposition position according to their respective positions within the stack.

[0108] Therefore, the corner positions of the first eight anode sheets and cathode sheets are concentrated within the range represented by "×", and the corner positions of the subsequent six anode sheets and cathode sheets are concentrated within the range marked with a square. The remaining anode and cathode sheets are concentrated within the range marked with a circle. All corners are located within a predetermined tolerance range. This analysis can be performed on at least one additional corner region, thereby identifying the actual orientation of the electrode sheet. This is also shown in FIG. 9. Here, the clusters are distributed somewhat differently (see subscripts K1-K8 or A1-A 10 etc.). Nevertheless, all corners are located within the tolerance range.

[0109] Once the positions of all corners are determined, the geometry of the electrode sheet can be calculated backwards.

[0110] FIG. 10 shows how the results are integrated for each electrode sheet to identify the actual orientation, based on the corner positions of the electrode sheets identified in FIGS. 8 and 9 and the known rectangular geometry.

[0111] FIG. 11 shows an xz cross-sectional view of a computer tomography image of one corner region of the ESV. In this cross-sectional view, it can be seen that the electrode sheet is distinguishable into an anode (protruding in dark gray) and a cathode (shorter in light gray). The electrode sheets are stacked alternately up and down along the stack height in the z direction. The separator layers respectively arranged between the electrode sheets cannot be seen.

[0112] In the reference position region between line L1 and line L2, the electrode sheets are arranged substantially parallel and without gaps within the ESV. This is ensured by a support (not visible in the photograph) that presses the electrode sheets against each other along the z direction.

[0113] Therefore, the reference position region is suitable for identifying the ideal "horizontal direction", that is, the electrode sheet extension along the x-y plane.

[0114] That is, as can also be seen from FIG. 11, the electrode sheet of the anode A is bent in some regions at its own electrode edge, which is also referred to as the electrode edge position in the context of this specification. Therefore, it protrudes along the z direction from the x-y plane defined by the reference position of each electrode sheet. This is also shown in FIG. 12 for one section.

[0115] In order to identify the persuasive corner position of the anode, correction must be made for such deviations.

[0116] In FIG. 11, a region between line L2 and line L3 is further defined, and all electrode edge positions (that is, electrode sheet edges) corresponding to the cathode in the cross-sectional image are located within this region. In FIG. 11, a region between line L3 and line L4 is further defined, and all electrode edge positions of the anode are located within this region.

[0117] The discrimination between the reference position region 100 and the region of the electrode edge position can be implemented using a convolutional neural network specially trained for this purpose. For this reason, a second CNN system can be provided. Subsequently, within this region of the electrode edge position, the electrode edge position of the electrode sheet, that is, the position of the electrode sheet edge in the cross-sectional image, can be identified. This is implemented by the first CNN system according to the present invention.

[0118] The identification of the electrode sheet is also carried out using a specially trained convolutional network. In particular, the electrode sheet within the reference position area 100 and the provisional electrode edge position are also identified (the latter using, for example, a third CNN system), and subsequently, robust detection is carried out by pattern recognition for each segment in regions L2 and L3 or in regions L4 and L5 in the direction of the electrode edge position. In this case, since the gray values within the electrode sheet vary, it is particularly advantageous to use a CNN in that it would be disadvantageous, for example, for histogram-based identification. The task of identifying the electrode edge is carried out by a first CNN system according to the present invention. Pattern recognition can also be carried out by a CNN, and it is also possible to use different CNNs for each of the two different types of electrode sheets. In that case, these CNNs are part of the first CNN system.

[0119] To ensure reliable identification, the number of stacked electrode sheets within the ESV may be set as a boundary condition.

[0120] The aim is to identify the electrode edge position of each electrode sheet in each cross-sectional image.

[0121] Pattern recognition is carried out, for example, in a plurality of segments of the same size, and these segments are arranged one after the other in the x or y direction (in the case of an xz cross-sectional image, i.e., when these segments are arranged in the direction of the x-axis), which is shown, inter alia, in FIG. 14.

[0122] Here, in each segment where the electrode sheet is recognized by pattern recognition, it is possible to identify a sample point, that is, the centroid position, of a predetermined pattern (similar to one section of the electrode sheet). As a result, in each cross-sectional view and for each electrode sheet, a plurality of sample points are identified, and the electrode sheet extends along these sample points. With the sample points of the electrode sheet, it is possible to draw one path connecting these sample points within each cross-sectional image. This is shown in FIG. 12. In this case, the sample point corresponding to the electrode edge position is located on the outermost side of the ESV and is also referred to as the first sample point in the context of this specification.

[0123] In this case, the identification of the sample points is performed by the first CNN system.

[0124] For each electrode within the ESV and in all xz cross-sectional images and yz cross-sectional images, the first sample point, that is, the electrode edge position, is identified respectively.

[0125] Here, in order to identify the corner positions for each electrode sheet, the first specimen point and, optionally, further specimen points are displaced to the height of the reference position of each electrode sheet. Specifically, they are displaced such that the paths passing through these specimen points extend along one straight line along the x-direction. In this case, as the height of the reference position, specimen points located within the reference position region 100 are used. From the positions of these specimen points, for example, the average height can be calculated. In this case, this height corresponds to the height of the reference position 101 of the electrode sheet. This process is shown in FIG. 12. The displaced first specimen point 301 and the resulting corrected electrode edge position 302 are plotted with respect to the electrode sheet. The corrected electrode edge position 302 corresponds to the electrode edge position in the case of an electrode sheet that would be perfectly oriented on the extension line of the reference position 101 without bending or folding. In particular, this corrected electrode edge position 302 is located further outside than the uncorrected first specimen point. In this transformation, it is important that the path length of the path passing through the specimen point 300 remains constant even if it is transformed (i.e., even if the specimen point is displaced).

[0126] These specimen points 300 can be connected, for example, using a cubic spline to form one path 303.

[0127] FIG. 13 shows the positions and arrangements of the xz cross-sectional image and the yz cross-sectional image in the cross-section along the x-y plane passing through the ESV. These xz cross-sectional image and yz cross-sectional image are specified along the dashed line within the corner region of the ESV. Further, the corrected first specimen points for the anode and the cathode are plotted.

[0128] Here, in order to identify the corner positions of the two electrode sheets, one straight line is transmitted for each electrode sheet with respect to a first edge extension (along the y-axis) passing through the corrected first specimen point and a second edge extension (along the x-axis). In this case, only the x-y position of the first specimen point is substantially relevant. Then, according to the present invention, the corner of each electrode sheet is located at the intersection of the straight line passing through the first edge extension in the x-y plane and the straight line passing through the second edge extension.

[0129] FIG. 14 shows means for identifying the provisional electrode edge positions of the electrode sheets for each electrode sheet included in the xz cross-sectional image and based on the xz cross-sectional image.

[0130] In this embodiment where the electrode composite stack has the first type and the second type of electrode sheets, the following CNN system architecture may be advantageous.

[0131] For this purpose, the third CNN system introduced in the preceding paragraph is used. In this case, the third CNN system includes, for example, a first CNN (CNN2). The first CNN (CNN2) is trained to identify, in each cross-sectional image, the region 420 that includes the electrode edge positions of the second type of electrode sheets in the electrode composite stack, that is, for example, the smaller cathode sheets. In this case, this identification is used only for finding the region where the electrode edges of all the second type of electrode sheets are present, rather than targeting individual electrode sheets.

[0132] Furthermore, the third CNN system includes a second CNN (CNN5), and the second CNN (CNN5) is trained to detect the electrode edge positions of the second type of electrode sheet within the region identified by the first CNN (CNN2). In this case, these electrode edge positions are treated only as provisional results, which is expressed by the term "provisionally identified". To enable reliable provisional identification, the number of electrode edge positions to be identified can be set (implicitly, for example, by a training set or by additional boundary conditions). By limiting the region for recognizing the electrode edge positions, an improvement in robustness and reliability is achieved.

[0133] The electrode edge positions of the first type are also provisionally identified in the same way. That is, the third CNN system further includes a third CNN (CNN1), and the third CNN (CNN1), similar to CNN2, identifies in each cross-sectional image the region 430 (see FIG. 15) that includes the electrode edge positions of the first type, i.e., for example, the anode A. Then, within this region, a fourth CNN (CNN4) provisionally identifies the provisional electrode edge positions 433 of the first type of electrode sheet.

[0134] To enable reliable identification, the number of electrode edge positions to be identified can be set (implicitly, for example, by a training set or by additional boundary conditions). By limiting the regions 420, 430 for recognizing the electrode edge positions, an improvement in robustness and reliability is achieved.

[0135] In this case, the CNN5 for identifying the provisional electrode edge position of the cathode K is trained to recognize, among other things, a feature 421 that not only includes the region of the cathode K but also includes the anode sheet A at the periphery (the feature 421 is surrounded by a rectangular frame in FIG. 14). That is, the feature 421 is higher than the expected cathode sheet thickness 422 in the image, which ensures that a part of the adjacent anode sheet A can further contribute for stabilization. Furthermore, it is assumed that the features 421 do not overlap or cross each other. The feature 421 is substantially rectangular and has at its center an electrode edge position 423 (shown as a circle) corresponding to the provisional electrode edge position 423 associated with this feature.

[0136] The provisional electrode edge position 423 of the cathode can be further refined, for example, by a further CNN as described below.

[0137] Similarly, in the pre-identified region 430 of the anode edge, a provisional electrode edge position can be identified using the CNN4 for the anode A (see FIG. 15). The feature 431 for which the CNN4 is trained is different from the feature 421 of the CNN5. This feature 431 is also rectangular but includes a different pattern. A position 433 corresponding to the provisional edge position of the anode A is also associated with this feature 431. In this case, the size (height, width) of the feature 431 plays an important role, and this size is preferably selected such that the feature is twice as narrow in the x or y direction as in the z direction and higher than the anode sheet thickness 432. In this case, the features may overlap each other. The circle at the center of the rectangular feature indicates the identified provisional electrode edge position 433 of the anode A.

[0138] To identify the electrode sheet within the reference area 100, other features and other CNNs (CNN6 or CNN7) can also be used here. Advantageously, the features 621, 631 each include adjacent electrode sheets A, K (see FIG. 16).

[0139] In this recognized feature (type of electrode sheet, position of the electrode sheet within the electrode composite stack, reference position area, provisional electrode edge position), the first CNN system can identify, for example, the first sample point 101 and, optionally, additional sample points 101, thereby identifying a path that includes a plurality of sample points, connects the plurality of sample points, and reproduces the extension of the electrode sheet within the cross-sectional image. In particular, in this case, the first sample point corresponds to the (final) electrode edge position, i.e., no longer corresponds to the provisionally identified edge positions 423, 433.

[0140] The creation of a path passing through the sample points can also be implemented using a CNN.

[0141] FIG. 17 shows how the first CNN system identifies the electrode edge position for the electrode sheet of the anode A.

[0142] For this purpose, for example, the region 430 between L4 and L5 where the provisional electrode edge 433 of the anode A is determined to be present is divided into a plurality of segments 400. Subsequently, using a CNN trained for the feature recognition of the feature 410 shown in FIG. 17, this feature is searched for in each segment 400. It should be noted that the non-horizontal extension of the anode sheet A is also reliably recognized. This is because the segments 400 are selected to be small so that the appearance shape of the electrode sheet hardly changes within the segments 400. In the first hatched segment 400, the CNN recognizes the corresponding feature 410 from all anode sheets A and determines its position. The location corresponding to the feature can be used as the sample point 300 for each electrode sheet A.

[0143] Only three further electrode sheets of the anode A extend completely through the second hatched segment, while one electrode sheet only extends into the segment 400 with a part 412 thereof. Therefore, the CNN recognizes only the electrode sheets having the complete feature 410.

[0144] For the electrode sheet that only extends into the segment with a part 412, the position of the electrode edge (i.e., the electrode edge position), i.e., the first specimen point 301, can be associated with the preceding segment. The error of this association regarding the actual electrode edge is negligible and is within the positioning tolerance range due to the size of the segment 400 and is very accurate.

[0145] Similarly, the edge position of the cathode K is specified using other CNNs of the first CNN system (not shown).

[0146] Here, a plurality of specimen points 300 obtained by pattern recognition in the segment 400 are connected to each other using, for example, a spline path. The length of the path is calculated, and then the path is transformed so as to extend along a straight line and have the same length as described above. Thereby, in some cases, the position (height, and X or Y direction) of the first specimen point 301 is changed. This new position corresponds to the corrected electrode edge position, and then a plurality of straight lines are transmitted by this corrected electrode edge position, and these straight lines enable very accurate positioning of the corners of the electrode sheet at the intersections of these straight lines (see FIG. 13).

[0147] By the method according to the present invention, the deposition accuracy of the electrode sheet in the ESV can be efficiently and reliably specified, and in some cases, the correction or improvement of the deposition of the electrode sheet can also be specified from the specified posture of the electrode sheet.

Explanation of Signs

[0148] 1 turntable 3 Holding device 5 Support 6 X-ray light source 7 Detector 8 CT image 9 Conductor lug of electrode sheet 10x xz cross-sectional image / cross-sectional plane 10y y-z cross-sectional image / cross-sectional plane 20x Edge extension x along the x direction 20y Edge extension along the y direction 21x Electrode sheet edge along the x direction 21y Electrode sheet edge along the y direction 100 Reference position area 101 Reference position 300 Specimen point 301 First specimen 302 Corrected electrode edge position 303 Path 304 Transformed straight path 400 Segment 410 Feature 412 Incomplete feature 420 Region of cathode edge 421 Provisional cathode edge feature 422 Cathode sheet thickness 423 Provisional electrode edge position of cathode 430 Region of anode edge 431 Provisional anode edge feature 432 Anode sheet thickness 433 Provisional electrode edge position of anode 500 Axis of rotation 501 Axis of rotation 621 Cathode feature in the reference region 631 Anode feature in the reference region A Electrode sheet of anode K Electrode sheet of cathode E1, E2, E3, E4 Corners of the electrode sheet SD, SB Two opposing plates of the support 5 Direction of the x,y coordinate system Direction of the z stacking direction, z-axis of the coordinate system L1, L2, L3, L4, L5 region boundary lines KS coordinate system 5-1 Central region 5-2 Region with weak absorbency

Claims

1. A method for identifying the positions of the corners of a polygonal electrode sheet within at least one corner region of an electrode composite stack (ESV), the method comprising: - generating a 3D image of the corner region of the electrode composite stack (ESV) within the imaging region using an image generation method, thereby generating a data set containing 3D position information of the electrode sheet (A, K) within the corner region of the electrode composite stack (ESV) relative to a support (5, SD, SB) or relative to markers arranged within the imaging region; - identifying, from the data set, a first edge extension and a second edge extension (20x, 20y) of the edges (21x, 21y) bounding the corner region for each electrode sheet (A, K), wherein the positions of the corners (E1, E2, E3, E4) of each electrode sheet (A, K) are identified based on the edge extensions (20x, 20y); comprising: - the first edge extension and the second edge extension (20y, 20x) are identified by the following steps, namely: ○ generating a plurality of xz cross-sectional images (10x) and yz cross-sectional images (10y); ○ identifying the electrode sheet (A, K) within each of the xz cross-sectional images (10x) and within each of the yz cross-sectional images (10y); ○ within each of the xz cross-sectional images (10x) and within each of the yz cross-sectional images (10y), and for each identified electrode sheet (A, K) within the xz cross-sectional images (10x) and within the yz cross-sectional images (10y): ・ identifying, by a first neural network system, the electrode sheet extension from a reference position region within the electrode composite stack to the electrode edge position of the electrode sheet; ・ adapting the electrode sheet extension such that the path length of the electrode sheet extension remains unchanged and the electrode sheet extension extends at the height of a reference position (101) from the reference position region to the corrected electrode edge position; performing; A step of respectively specifying a straight line (20x, 20y) extending along the corrected electrode edge position (302) along the first edge or the second edge (21x, 21y) for each of the electrode sheets (A, K), wherein the straight lines (20y, 20x) correspond to the first edge extension (20y) and the second edge extension (20x), and the step Specified by Method **Claim 2** The corrected electrode edge position is determined by specifying the electrode sheet positions at a plurality of sample points (301, 300), The first sample point (301) corresponds to the electrode edge position, The reference position (101) of the electrode sheet (A, K) in the electrode composite stack (ESV) is detected by at least one additional sample point (300), The sample points (300, 301) are specified by the first neural network system, The corrected electrode edge position (302) is achieved by displacing at least the first sample point (301) to the height of the reference position (101), The path length between the first sample point (301) and the at least one additional sample point (300) is maintained constant so that the path length of the electrode sheet extension remains unchanged, The method according to claim 1 **Claim 3** For each of the electrode sheets (A, K), an intersection point (E1) between the first edge extension (20y) and the second edge extension (20x) is specified, The position of the corner of each of the electrode sheets (A, K) is associated with the intersection point (E1), The method according to claim 1 **Claim 4** The method is implemented for two or more corner regions of the electrode sheet (A, K) of the electrode composite stack (ESV), The positions of two or more corners (E1, E2, E3, E4) of the electrode sheet (A, K) are determined, The method according to claim 1 **Claim 5** The electrode sheet (A, K) is identified by a further neural network system for identifying the electrode sheet in the xz cross-sectional image (10x) and in the yz cross-sectional image (10y), the method according to claim 1 **Claim 6** Based on the corrected first edge extension and the corrected second edge extension (20y, 20x) and the position of the at least one corner (E1) of each of the electrode sheets (A, K), the posture of each of the electrode sheets (A, K) with respect to the support (5) and / or the marker is determined, the method according to claim 1.

7. Based on the posture, for each of the electrode sheets (A, K), a deviation from a predetermined posture with respect to the support (5) and / or the marker is identified, the method according to claim 6.

8. During the manufacturing method for a further electrode composite stack, based on the identified deviation, the deposition posture of the electrode sheets in the electrode composite stack is adapted so that the deviation in the further electrode composite stack is smaller, the method according to claim 7.

9. For each of the electrode sheets (A, K), from the identified positions of the corners (E1, E3), the positions of the remaining corners (E2, E4) of each of the electrode sheets (A, K) that have not yet been identified are determined, The determination of the remaining corners is carried out from the dimensions of the electrode sheets stored in the database, the method according to claim 1.

10. The plurality of the electrode sheets includes a plurality of first-type electrode sheets (A) and a plurality of second-type electrode sheets (K), the method according to claim 1.

11. The first-type electrode sheets (A) have a different gray value range in the cross-sectional images (10x, 10y) than the second-type electrode sheets (K) so that the plurality of the electrode sheets (A, K) can be distinguished based on different gray value ranges in the dataset. the method according to claim 10.

12. In a first image processing step, a neural network system for recognizing the electrode edge positions of the electrode sheets identifies the electrode edge positions of the electrode sheets in each xz cross-sectional image (10x) and in each yz cross-sectional image (10y) and for each of the electrode sheets, furthermore, the reference positions (101) are identified in each xz cross-sectional image (10x) and in each yz cross-sectional image (10y) and for each of the electrode sheets, The reference position (101) is specified from a region (100) of the electrode composite stack (ESV) in which the electrode sheets (A, K) are stacked at regular intervals. The method according to claim 9.

13. In the second image processing step, the first neural network system identifies a plurality of sample points (300, 301) in each xz cross-sectional image (10x) and in each yz cross-sectional image (10y), and for each of the electrode sheets (A, K). The plurality of electrode sheets include a plurality of first-type electrode sheets (A) and a plurality of second-type electrode sheets (K). The identification of the plurality of sample points (300, 301) for both the first-type electrode sheet and the second-type electrode sheet is respectively limited to a region defined by the innermost electrode edge position among the first-type or second-type electrode sheets and the outermost electrode edge position among the first-type or second-type electrode sheets. The method according to claim 12.

14. For each xz cross-sectional image (10x) and each yz cross-sectional image (10y), and for each electrode sheet (A, K), a path (303) and a path length associated with the path, corresponding to the path length of the plurality of sample points (300, 301), are identified. For each xz cross-sectional image (10x) and each yz cross-sectional image (10y), and for each electrode sheet (A, K), the identified sample points (300, 301) are displaced by transformation to a straight-line path (304) perpendicular to the height (z) of the electrode composite stack (ESV), specifically, displaced so that the path length is maintained constant. Thereby, the height of the sample point (301) is adjusted to the reference position (101). The method according to claim 13.

15. A computer program including computer program code for implementing the method according to any one of claims 1 to 14 when executed on a computer.

Images