Battery cell inspection system and method

Abstract

Provided are a battery cell inspection system and method capable of tracking an end point of an electrode in a battery cell image using an artificial intelligence learning model and automatically determining whether the electrode is defective based on the tracked end point.SOLUTION: The battery cell inspection system includes a photographing unit 200 configured to acquire images of electrodes of a C1 and a C2 of a battery cell, a memory 310 configured to store a learning model learned to track end points of electrodes in an image of electrodes of an arbitrary battery cell, and a controller 320 configured to track the end points of the electrodes in the image of the electrodes of the battery cell using the learning model and determine whether the electrodes of the battery cell are defective based on coordinate information of the end points of the electrodes.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention claims the benefit of priority based on Korean Patent Application No. 10-2021-0185151, filed December 22, 2021, and all contents disclosed in the documents of this Korean patent application are incorporated herein by reference. SUMMARY OF THE INVENTION The embodiments disclosed herein relate to battery cell inspection systems and methods. [Background technology]

[0002] In recent years, with the rapid development of the electrical, electronic, communication, and computer industries, the demand for high-performance, high-safety batteries has been gradually increasing. Lithium polymer batteries, a type of secondary battery, are batteries that can be repeatedly charged and discharged by reversible interconversion between chemical energy and electrical energy, and have the advantages of long life and large capacity, and are widely used in portable electronic devices, electric vehicles, and the like.

[0003] Typically, the electrode assembly of a lithium polymer battery is constructed with multiple positive and negative electrodes alternately stacked with a separator between them. To prevent lithium precipitation, the negative electrode is made large enough to cover the positive electrode, but the protruding portion of the negative electrode is called an overhang. If the overhang is bent, protrudes excessively, or is missing due to manufacturing process issues, it can lead to battery defects. Therefore, an internal CT scan is required to inspect the electrodes. However, existing battery cell inspection systems require inspectors to visually examine the image of the test piece to determine defects, which is time-consuming and makes full inspection difficult. Sampling inspections have limitations in reducing the defect rate. Summary of the Invention [Problem to be solved by the invention]

[0004] An objective of the embodiments disclosed herein is to provide a battery cell inspection system and method that can use an artificial intelligence learning model to track electrode end points in a battery cell image and automatically determine whether the electrodes are defective based on the tracked end points.

[0005] The technical problems of the embodiments disclosed in this document are not limited to the technical problems mentioned above, and other technical problems not mentioned will be clearly understood by those skilled in the art from the following description. [Means for solving the problem]

[0006] A battery cell inspection system according to one embodiment includes an imaging unit that acquires electrode images of a battery cell, a memory that stores a learning model that has been trained to track electrode endpoints in the electrode image of an arbitrary battery cell, and a controller that uses the learning model to track the electrode endpoints in the electrode image of the battery cell and determines whether or not the electrodes of the battery cell are defective based on coordinate information of the electrode endpoints.

[0007] In one embodiment of the battery cell inspection system, the learning model may be generated by receiving input learning data including electrode images of any battery cell, labeling the learning data, extracting features based on the labeled learning data, and structuring a database in which the features are recorded.

[0008] In one embodiment of the battery cell inspection system, the imaging unit can include a radiation source that outputs radiation toward a region of interest of the battery cell, and a radiation detector that detects radiation that has passed through the region of interest of the battery cell.

[0009] In one embodiment of a battery cell inspection system, the imaging unit acquires a three-dimensional image of a region of interest of the battery cell, the controller generates cross-sectional images in two or more directions based on the three-dimensional image, and tracks the end points of the electrodes of the battery cell for each of the cross-sectional images, and the learning model can be trained using the cross-sectional images in two or more directions generated based on any three-dimensional image as learning data.

[0010] In one embodiment of a battery cell inspection system, the electrodes of the battery cell have a structure in which multiple positive electrodes and negative electrodes are alternately stacked with separators between them, and the controller can determine that the electrodes of the battery cell are defective in at least one of the following cases: when a portion of the positive electrode or negative electrode is bent; when a portion of the positive electrode or negative electrode is missing; when the gap between the end points of adjacent positive electrodes and negative electrodes is equal to or greater than a reference value; when the insertion order of the positive electrodes or negative electrodes is incorrect; when the gap between the end points of adjacent positive electrodes or the gap between the end points of adjacent negative electrodes is equal to or greater than a reference value; when a parameter indicating the alignment state of the positive electrodes and negative electrodes over a certain section is equal to or less than a reference value; and when a portion of the positive electrode or negative electrode is broken.

[0011] A battery cell inspection system according to one embodiment may further include a tray for loading two or more battery cells, and a transport unit that transports the tray with the battery cells loaded to an imaging point of the imaging unit and, once imaging is complete, transports the tray out of the imaging point.

[0012] In one embodiment of the battery cell inspection system, the transport unit simultaneously transports two or more trays loaded with the battery cells to the imaging position, and the two or more trays may be positioned such that the radiation output from the radiation source simultaneously passes through the regions of interest of the battery cells loaded on all of the trays.

[0013] In one embodiment of the battery cell inspection system, the two or more trays may be arranged such that an edge of the battery cells loaded on each tray faces an edge of the battery cells loaded on another tray at the imaging point.

[0014] In one embodiment of the battery cell inspection system, the transfer unit rotates the tray on which the battery cells are loaded around an axis at the imaging point, and the imaging unit can acquire a three-dimensional image of the battery cells based on two-dimensional radiation images acquired while the tray is rotating.

[0015] A battery cell inspection management method according to one embodiment includes the steps of training a learning model to track electrode endpoints in an electrode image of an arbitrary battery cell, acquiring an electrode image of the battery cell, using the learning model to track the electrode endpoints in the electrode image of the battery cell, and determining whether or not there are defects in the electrodes of the battery cell based on coordinate information of the electrode endpoints.

[0016] In one embodiment of the battery cell inspection method, the learning model may be generated by receiving input learning data including electrode images of any battery cell, labeling the learning data, extracting features based on the labeled learning data, and structuring a database in which the features are recorded.

[0017] In one embodiment of a battery cell inspection method, the step of acquiring an electrode image of the battery cell may include the steps of emitting radiation toward a region of interest of the battery cell and detecting the radiation that has passed through the region of interest of the battery cell.

[0018] In one embodiment of a battery cell inspection method, the step of acquiring electrode images of the battery cell includes acquiring a three-dimensional image of a region of interest of the battery cell, and the step of tracking the electrode endpoints of the battery cell includes generating cross-sectional images in two or more directions based on the three-dimensional image and tracking the electrode endpoints of the battery cell for each of the cross-sectional images, and the learning model can be trained using the cross-sectional images in two or more directions generated based on any three-dimensional image as training data.

[0019] In one embodiment of a battery cell inspection method, the electrodes of the battery cell have a structure in which multiple positive electrodes and negative electrodes are alternately stacked with separators interposed therebetween, and the step of determining whether the electrodes of the battery cell are defective can include determining that the electrodes of the battery cell are defective in at least one of the following cases: when a portion of the positive electrode or negative electrode is bent; when a portion of the positive electrode or negative electrode is missing; when the gap between end points of adjacent positive electrodes and negative electrodes is equal to or greater than a reference value; when the insertion order of the positive electrodes or negative electrodes is incorrect; when the gap between end points of adjacent positive electrodes or between end points of adjacent negative electrodes is equal to or greater than a reference value; when a parameter indicating the alignment state of the positive electrodes and negative electrodes over a certain section is equal to or less than a reference value; and when a portion of the positive electrode or negative electrode is disconnected.

[0020] A battery cell inspection method according to an embodiment may further include the steps of loading two or more battery cells onto a tray, transporting the tray with the loaded battery cells to an imaging location, and, once imaging is complete, transporting the tray out of the imaging location.

[0021] In one embodiment of a battery cell inspection method, in the step of transporting the tray to the imaging position, two or more trays loaded with the battery cells may be simultaneously transported to the imaging position, and the two or more trays may be positioned such that the radiation simultaneously passes through the regions of interest of the battery cells loaded on all of the trays.

[0022] In one embodiment of the battery cell inspection method, the two or more trays may be arranged such that an edge of the battery cells loaded on each tray faces an edge of the battery cells loaded on another tray at the imaging point.

[0023] According to an embodiment, the battery cell inspection method may further include rotating the tray on which the battery cells are loaded around an axis at the imaging point, and acquiring an electrode image of the battery cell may include acquiring a three-dimensional image of the battery cell based on the two-dimensional radiation image acquired while the tray is rotating. [Effects of the Invention]

[0024] According to one embodiment, a battery cell inspection system and method are provided that can use an artificial intelligence learning model to track electrode endpoints in a battery cell image and automatically determine whether the electrodes are defective based on the tracking.

[0025] The proposed system and method can simultaneously acquire electrode images of multiple loaded battery cells and automatically inspect the electrodes of each battery cell using a learning model, significantly reducing the time required for inspection. Therefore, unlike existing sampling inspections, 100% inspection is possible, significantly reducing the risk of product defects. In addition, this document may provide a variety of other benefits that may be perceived directly or indirectly. [Brief explanation of the drawings]

[0026] In order to more clearly describe the embodiments disclosed in this document or the technical solutions of the prior art, drawings necessary for describing the embodiments are briefly introduced below. It should be understood that the following drawings are only for describing the embodiments of the present specification and are not intended to limit the same. In addition, for the sake of clarity, the representation of some components in the drawings may be exaggerated or omitted.

[0027] [Figure 1] 1 illustrates the configuration and operation process of a battery cell inspection system according to one embodiment. [Figure 2] 1 illustrates an imaging region of interest of a battery cell and inspection directions along the x-axis / y-axis according to one embodiment. [Figure 3a] This shows a two-dimensional image of the electrodes inside a battery cell obtained by X-ray photography. [Figure 3b] This shows a cross-sectional image generated based on a three-dimensional image of the electrodes inside a battery cell obtained by X-ray photography. [Figure 4a] This shows a case where a battery cell electrode is determined to be defective. [Figure 4b] This shows a case where a battery cell electrode is determined to be defective. [Figure 4c] This shows a case where a battery cell electrode is determined to be defective. [Figure 4d] This shows a case where a battery cell electrode is determined to be defective. [Figure 4e] This shows a case where a battery cell electrode is determined to be defective. [Figure 4f] This shows a case where a battery cell electrode is determined to be defective. [Figure 4g] This shows a case where a battery cell electrode is determined to be defective. [Figure 5] 1 shows a tray and battery cell arrangement structure for simultaneously testing multiple battery cells. [Figure 6] 1 is a flowchart illustrating a battery cell inspection method according to an embodiment. [Figure 7] 10 is a flowchart illustrating a battery cell inspection method according to another embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0028] Hereinafter, the embodiments disclosed herein will be described in detail with reference to the accompanying drawings. When assigning reference numerals to components in each drawing, it should be noted that the same reference numerals are assigned to the same components in other drawings whenever possible. Furthermore, when describing the embodiments disclosed herein, if a detailed description of related known structures or functions is deemed to hinder understanding of the embodiments disclosed herein, such detailed description will be omitted.

[0029] The terms used in this document have been selected to the extent possible based on current widespread and general terms, taking functionality into consideration. However, these may vary depending on the intentions or practices of engineers in the relevant field or the emergence of new technologies. In addition, in certain cases, the applicant may arbitrarily select terms, and in such cases, their meanings will be described in the description section of the specification. Therefore, it is clear that the terms used in this document should be interpreted based on the substantive meaning of the terms and the overall content of this document, rather than simply the name of the terms. Furthermore, the terms used in this document are used merely to describe specific embodiments and are not intended to limit the scope of other embodiments. A singular term may include a plural term unless the context clearly dictates otherwise.

[0030] 1 shows the configuration and operation process of a battery cell inspection system according to one embodiment. Referring to FIG. 1, the battery cell inspection system according to one embodiment includes trays 101 and 102 for loading two or more battery cells C1 and C2, a transport unit 110 for transporting the trays 101 and 102, an imaging unit 200 for capturing electrode images of the battery cells C1 and C2, and an inspection unit 300 for tracking the electrode end points of the battery cells C1 and C2 using a learning model and determining whether or not the electrodes are defective based on the tracking.

[0031] A battery cell is the smallest unit constituting a rechargeable secondary battery, and can be manufactured through a process of packaging an electrode assembly, an activation process of charging and discharging the packaged battery to activate it, a degassing process of discharging gas from the battery cell, and a folding process of cutting a portion of the pouch and then sealing it.

[0032] An electrode assembly can be manufactured through a notching process to create positive and negative electrode tabs, a stacking process to stack battery materials, and a tab welding process to connect the positive and negative electrode tabs. A typical electrode structure involves alternating stacking of one or more positive electrode sheets, each with a positive electrode active material formed on a positive current collector, and one or more negative electrode sheets, each with a negative electrode active material formed on a negative current collector, with the positive and negative electrode sheets facing each other with a separator sheet in between. To prevent lithium deposition, the negative electrode is typically large enough to cover the positive electrode. However, if the protruding portion of the negative electrode is bent, if the gap between the negative electrode and the positive electrode is too large, or if part of the electrode is missing, this can lead to battery cell failure. Therefore, a separate inspection process is required to determine whether the electrode is defective.

[0033] Two or more battery cells C1, C2 can be loaded on each of the trays 101, 102. The battery cells are loaded in a state where they are precisely stacked on top of each other so that the edge portions of the multiple battery cells C1, C2 can be inspected simultaneously, and the trays 101, 102 have loading spaces that match the sizes of the battery cells C1, C2, respectively.

[0034] The transfer unit 110 transfers the trays 101 and 102 loaded with the battery cells C1 and C2 to a photographing location, and once photographing is complete, transfers the trays 101 and 102 out of the photographing location. According to one embodiment, the transfer unit 110 may connect the trays 101 and 102 to fastening portions (not shown) of the trays 101 and 102 by vacuum suction (or using a gripper), and then transfer the trays 101 and 102 to a designated location. According to another embodiment, the transfer unit 110 may be a conveyor belt for transporting the trays 101 and 102. In this case, the trays 101 and 102 are not separately connected, but are placed on a moving stage of the conveyor belt and can be transported at a predetermined speed and direction.

[0035] The imaging unit 200 acquires an electrode image by imaging the ROI of the battery cells C1 and C2 transferred by the transfer unit 110. Here, the ROI may be an edge portion where the electrodes of each battery cell are located. According to an embodiment, the imaging unit 200 may include a radiation source 210 that outputs radiation toward the ROI of the battery cell, and a radiation detector 220 that detects radiation that has passed through the ROI.

[0036] The radiation source 210 may be, for example, an X-ray tube having a capacity of 160 to 240 kV. The X-ray tube may be, for example, a hot cathode X-ray tube that uses a heated tungsten filament to generate electron beams that move at high speed and then collide with a material to emit X-rays.

[0037] The radiation detector 220 detects X-rays and can reconstruct an image using the X-ray information. As shown in Fig. 1, the radiation detector 220 is located on the opposite side of the radiation source 210 with respect to the battery cell to be inspected, detects X-rays that have passed through the region of interest (ROI) of the battery cell, and reconstructs the image.

[0038] The battery cells C1 and C2 are transported to an imaging position located between the radiation source 210 and the radiation detector 220 (i.e., a position where radiation emitted from the radiation source 210 passes through the edge electrode portions of the battery cells). According to one embodiment, the position of the transport unit 110 coupled to the trays 101 and 102 may be finely adjusted manually or automatically so that the battery cells C1 and C2 are positioned at the correct imaging position. According to another embodiment, the positions of the trays 101 and 102 and the transport unit 110 may be fixed, and the position of the radiation source 210 may be finely adjusted manually or automatically.

[0039] 2 shows the imaging regions of interest for a pouch-type battery cell according to one embodiment, and the inspection directions along the x-axis and y-axis. In the case of a pouch-type battery cell, an overhang, where the negative electrode protrudes more than the positive electrode, can be observed at the edge, so the edge portions can be set as regions of interest (ROI 1, ROI 2). Although not shown, in the case of a prismatic or cylindrical battery cell, other portions where the overhang can be observed can be set as regions of interest to acquire electrode images.

[0040] The electrode images of the battery cells C1 and C2 acquired by the imaging unit 200 are transmitted to the inspection unit 300. Referring to Fig. 1, the inspection unit 300 may include a memory 310 that stores a learning model trained to track electrode endpoints in the electrode images of the battery cells, and a controller 320 that determines whether or not there is a defect in the electrodes using the learning model.

[0041] The memory 310 may be various storage media, such as a semiconductor memory such as a RAM, a ROM, or a flash memory, a magnetic disk, an optical disk, etc. The learning model stored in the memory 310 is an artificial intelligence machine learning model trained to track the electrode endpoints in an electrode image of an arbitrary battery cell, and may be, for example, a deep learning model having a multi-layer artificial neural network structure.

[0042] The process of generating and training a learning model according to one embodiment is as follows: Learning data including an electrode image of a battery cell is received as input, and the learning data is labeled as a basic learning task for tracking the electrode endpoints. Labeling refers to the process of inputting various information appropriate for a purpose into each piece of data, such as an image, video, or text, so that an AI can learn from the data. Each piece of labeled data is stored as a point, forming the learning data. Features are extracted based on the labeled learning data, and a database containing the features is structured so that it can be matched with the features of the input data, thereby generating a learning model. The learning model receives an electrode image of a battery cell as input data, extracts and structures the features of each piece of input data, and matches them with the features of a pre-generated database to optimize objects and shapes with a high matching rate.

[0043] The controller 320 tracks the electrode endpoints in the electrode image of the battery cell using the learning model stored in the memory 310, and determines whether or not the electrode of the battery cell is defective based on the coordinate information of the electrode endpoints. The controller 320 may include a processor, such as a CPU, an MPU, or an MCU, that processes data and performs calculations.

[0044] An exemplary electrode endpoint tracking process using the learning model is as follows: First, input image data is preprocessed. For example, the input image can be converted into a high-resolution image using a deep learning-based high-resolution algorithm. Next, in the shape detection step, the high-resolution image is segmented to detect the negative and positive electrode endpoints. Next, a skeleton model is extracted from the endpoint shapes detected through segmentation, and disconnected and connected portions are post-processed to extract contour information. Through this series of processes, the controller 320 can track the coordinates of the electrode endpoints in the input image and determine whether the electrodes are defective based on the coordinate information.

[0045] According to one embodiment, the battery cell inspection system can perform inspections based on 3D images of battery cell electrodes. According to the embodiment, the transfer unit 110 rotates a tray loaded with battery cells to be inspected 360 degrees around an axis at an imaging point, and the imaging unit 200 generates a 3D image based on tomographic images captured while the tray and battery cells are rotating. The 3D image capturing method can utilize computed tomography (CT) technology, in which X-rays are transmitted through the rotating battery cells to capture tomographic images, and the 3D image can be obtained by reconstructing the multiple tomographic images.

[0046] According to an embodiment, the controller 320 can generate cross-sectional images in two or more directions (e.g., a cross-sectional image in the x-axis direction and a cross-sectional image in the y-axis direction) based on the three-dimensional image, and track the electrode end points for each of the cross-sectional images using a learning model. In this case, the learning model is trained using the cross-sectional images in two or more directions (e.g., the x-axis direction and the y-axis direction) generated based on any three-dimensional image as training data.

[0047] FIG. 3a shows a two-dimensional image of an electrode inside a battery cell acquired by radiography, and FIG. 3b shows a cross-sectional image generated based on a three-dimensional image of the electrode inside the battery cell acquired by radiography. In the two-dimensional image shown in FIG. 3a, the electrodes are not clearly distinguished due to low resolution, making it difficult to track the electrode endpoints. In contrast, when a three-dimensional image is acquired, a high-resolution cross-sectional image can be obtained, as shown in FIG. 3b, and the accuracy of tracking the endpoints can be improved by examining the cross-section in two or more directions (e.g., the x-axis and y-axis directions).

[0048] The controller 320 determines whether an electrode is defective using the electrode tracking coordinates. According to an embodiment, the controller 320 may determine that an electrode of the battery cell is defective in at least one of the following cases: when a portion of a positive electrode or a negative electrode is bent; when a portion of a positive electrode or a negative electrode is missing; when a gap between end points of adjacent positive electrodes and negative electrodes is equal to or greater than a reference value; when the insertion order of the positive electrodes or negative electrodes is incorrect; when a gap between end points of adjacent positive electrodes or between end points of adjacent negative electrodes is equal to or greater than a reference value; when a parameter indicating the alignment state of the positive electrodes and negative electrodes over a certain section is equal to or less than a reference value; and when a portion of the positive electrode or negative electrode is disconnected.

[0049] 4a to 4g are diagrams showing cross-sectional images of the electrodes of a battery cell, each of which shows a case where an electrode can be determined to be defective. FIG. 4a shows a case where a portion of the negative electrode A is bent in an electrode structure in which negative electrodes A and positive electrodes B are alternately stacked. The controller 320 uses a learning model to track the endpoint coordinates (x1, y1) and (x2, y2) of the negative electrode, and can calculate how much the negative electrode A is bent based on the coordinate information. For example, if the endpoint coordinates of the negative electrode satisfy the following formula, it can be determined that excessive deformation has occurred in the negative electrode. Here, the reference value X for determining defects is ref , Y ref can be set arbitrarily.

[0050] [Formula 1]

number

[0051] Alternatively, when the points traced from the bottom of the negative electrode are aligned sequentially by position, it can be assumed that the negative electrode is bent if the y-value of the intermediate coordinate is the largest. If the negative electrode is bent in this way, it may come into contact with other electrodes, causing a short circuit, and the battery cell is therefore determined to be defective.

[0052] 4b shows a case where a portion of a positive electrode B' is missing from an electrode structure in which negative electrodes A and positive electrodes B are alternately stacked. The controller 320 can track the electrode end points to check the number of positive and negative electrode stacks, and if the checked number of stacks does not match the specified number, it determines that a portion of the electrode is missing. In this case, the negative electrodes or positive electrodes may come into contact with each other, causing a short circuit, and the battery cell is determined to be defective.

[0053] 4c shows an example where the gap between the end points of adjacent negative electrodes A and positive electrodes B in an electrode structure in which negative electrodes A and positive electrodes B are alternately stacked is equal to or greater than a reference value. The controller 320 uses a learning model to track the end point coordinates (x1, y1), (x3, y3) of the negative electrodes and the end point coordinates (x2, y2) of the positive electrodes, and can measure the size of the gap between the electrodes based on the coordinate information. If the gap size is equal to or greater than the reference value, there is a high possibility of a short circuit due to electrode warping or contact between the electrodes, and the battery cell is therefore determined to be defective.

[0054] FIG. 4d illustrates a case where the insertion order of positive or negative electrodes is incorrect in a structure in which negative and positive electrodes are alternately stacked. Normally, positive electrodes should be inserted at positions A and A', and negative electrodes should be inserted at position B. However, if a negative electrode is inserted instead of a positive electrode or a positive electrode is inserted instead of a negative electrode, as shown in FIG. 4d, a fault is determined. For example, the controller 320 tracks the endpoint coordinates (x1, y1), (x2, y2), and (x3, y3) of three consecutive electrodes in the bi-cell order, compares the heights (y1, y2, y3) of each endpoint, and determines that a fault is due to incorrect bi-cell insertion if y1 and y3 are higher than y2. Alternatively, a fault can be determined if, after tracking the heights of each electrode, there are consecutive high electrodes or consecutive low electrodes without alternating high and low.

[0055] FIG. 4e shows a case where a foreign object (e.g., tape, scrap, etc.) has entered between the positive and negative electrodes in a structure where negative and positive electrodes are alternately stacked, or the separator has been folded, causing the gap between the end points of adjacent electrodes to be equal to or greater than a reference value. For example, the controller 320 may calculate the end point coordinates (x n , y n ) satisfies the following formula, it can be determined that a defect has occurred. Here, the reference value X ref can be set arbitrarily.

[0056] [Formula 2]

number

[0057] FIG. 4f shows a case where a parameter indicating the alignment state of the positive and negative electrodes is below a reference value in a certain section. This is a case where AC overhang does not occur, but the alignment state of the positive and negative electrodes is not good and excessive dispersion is observed. In this case, the controller 320 can set a reference line (temporary line) indicating the alignment standard of the positive and negative electrodes, and set the difference from the reference line for each positive and negative electrode as a parameter. For example, if the reference line for the positive or negative electrode is set to AVG, the endpoint coordinate (x n , y n ) satisfies the following formula, it can be determined that a defect has occurred. ref can be set for each of the positive and negative electrodes.

[0058] [Formula 3]

number

[0059] Figure 4g shows a case where a portion of the positive or negative electrode is broken. Referring to Figure 4g, A shows a case where the electrode is completely broken, and B shows a case where the electrode is partially broken. The controller 320 can determine a complete break when the length of the electrode traced from the bottom end is shorter than that of the adjacent electrode, and a partial break when the trace is interrupted at the tab portion. When the tab is broken in this way, safety is reduced, and the battery cell can be determined to be defective.

[0060] Although the cases in which an electrode is determined to be defective have been described above with reference to FIGS. 4a to 4g, these are merely examples, and various cases in which an electrode can be determined to be defective based on the coordinates of its endpoints can be added to the embodiments.

[0061] According to one embodiment, the controller 320 may determine whether an electrode is defective based on a rule-based algorithm, either in addition to or in addition to the artificial intelligence learning model. The rule-based algorithm determines whether an electrode is defective based on rules pre-specified by the user for an input electrode image. This method has the advantage of automatically determining whether a defect is present without a separate machine learning process, but accuracy may be reduced due to the inconsistent shape of the electrodes in a battery cell. Therefore, it is preferable to simultaneously apply a deep learning-based learning model and a rule-based algorithm to complement each other.

[0062] 5 shows a tray and battery cell arrangement structure for simultaneously inspecting multiple battery cells. As shown in FIG. 1, the transport unit 110 can simultaneously transport two or more trays 101 and 102 to the imaging position. Each tray 101 and 102 is loaded with two or more battery cells C1 and C2.

[0063] According to one embodiment, the trays 101 and 102 transported to the imaging location may be positioned so that radiation output from the radiation source 210 simultaneously passes through the regions of interest R1 and R2 of the battery cells C1 and C2 loaded on all of the trays 101 and 102. For example, as shown in FIG. 5 , the trays may be positioned so that the edge of the battery cell C1 loaded on the tray 101 faces the edge of the battery cell C2 loaded on the other tray 102. The radiation that passes through the regions of interest R1 and R2 of each battery cell is detected by the radiation detector 220, and the imaging unit 200 can acquire electrode images for the multiple battery cells C1 and C2 using this information. By simultaneously inspecting multiple battery cells in this manner, the time required for inspection can be significantly reduced.

[0064] 6 is a flowchart illustrating a battery cell inspection method according to one embodiment. Referring to FIG. 6, first, a step (S610) of training a learning model to track electrode endpoints in an electrode image of a battery cell is performed. According to one embodiment, the learning model may be generated and trained using a segment labeling method. The generation and training process of the learning model has been described above, so a detailed description thereof will be omitted. The learning model may be stored in memory 310 of the inspection unit 300.

[0065] Next, step S620 of acquiring an electrode image of the battery cell is performed. This step S620 may be performed by the imaging unit 200 including a radiation source 210 that outputs radiation (e.g., X-rays) toward a region of interest (ROI) of the battery cell, and a radiation detector 220 that detects the radiation that has passed through the region of interest, as shown in FIG. 1 . According to one embodiment, step S620 may include the steps of outputting radiation toward the region of interest of the battery cell using the radiation source 210 and detecting the radiation that has passed through the region of interest of the battery cell using the radiation detector 220.

[0066] Next, a step (S630) of tracking electrode endpoints in the electrode image of the battery cell using the learning model is performed. This step (S630) can be performed by the controller 320 included in the inspection unit 300. According to one embodiment, the controller 320 can generate cross-sectional images in two or more directions (e.g., a cross-sectional image in the x-axis direction and a cross-sectional image in the y-axis direction) based on the 3D image, and track the electrode endpoints for each cross-sectional image using the learning model trained based on the 3D image data.

[0067] Next, a step (S640) is performed to determine whether or not the electrode of the battery cell is defective based on the coordinate information of the electrode end points. This step (S640) can be performed by controller 320. The presence or absence of a defect in the electrode of the battery cell can be determined in at least one of the following cases: when a portion of the positive or negative electrode is bent; when a portion of the positive or negative electrode is missing; when the gap between the end points of adjacent positive and negative electrodes is equal to or greater than a reference value; when the insertion order of the positive or negative electrodes is incorrect; when the gap between the end points of adjacent positive electrodes or the end points of adjacent negative electrodes is equal to or greater than a reference value; when a parameter indicating the alignment state of the positive and negative electrodes over a certain section is equal to or less than a reference value; or when a portion of the positive or negative electrode is disconnected (see FIGS. 4a to 4g).

[0068] 7 is a flowchart illustrating a battery cell inspection method according to another embodiment. Referring to FIG. 7, first, a step (S710) is performed in which a learning model is trained to track electrode endpoints in an electrode image of a battery cell. This step (S710) is substantially the same as step (S610) described with reference to FIG. 6, and therefore a detailed description thereof will be omitted.

[0069] Next, a step of loading two or more battery cells onto a tray (S720) and a step of transporting the tray with the loaded battery cells to a photographing position (S730) are performed. As described with reference to Fig. 1, two or more battery cells C1, C2 can be loaded onto each of the trays 101, 102. The battery cells are loaded in a state where they are precisely overlapped with each other so that the edge portions of the plurality of battery cells C1, C2 can be inspected simultaneously.

[0070] Next, the following steps are performed in sequence: acquiring an electrode image of the battery cell (S740), tracking the electrode endpoints in the electrode image of the battery cell using the learning model (S750), and determining whether or not the battery cell has an electrode defect based on the coordinate information of the electrode endpoints (S760). Steps (S740 to S760) are substantially the same as steps (S620 to S640) described with reference to Figure 6, so detailed description will be omitted.

[0071] After the electrodes have been photographed, the tray is transported out of the photographing location (S770). This step (S770) may be performed at any time after the electrode image acquisition step (S740) is completed. The battery cell and tray transport steps (S730, S770) may be performed by the transport unit 110 (see FIG. 1).

[0072] According to one embodiment, when a three-dimensional image of a battery cell is to be acquired, the transport unit 110 may further perform a step of rotating the tray on which the battery cells are loaded around an axis at the imaging point, and in this case, the imaging unit 200 may further perform a step of acquiring a three-dimensional image of the battery cell based on the two-dimensional radiation image acquired while the tray is rotating.

[0073] At least some steps of the battery cell testing method according to the above-described embodiments may be implemented as an application or as program instructions executable by various computer components and recorded on a computer-readable recording medium, which may include program instructions, data files, data structures, and the like, singly or in combination.

[0074] The battery cell inspection system and battery cell inspection method described above can use an artificial intelligence learning model to track electrode endpoints in battery cell images and automatically determine whether or not the electrodes are defective based on the tracking. Furthermore, by simultaneously acquiring electrode images of multiple loaded battery cells and automatically inspecting the electrodes of each battery cell using the learning model, the time required for inspection can be significantly reduced. Therefore, unlike existing sampling inspections, 100% inspection is possible, significantly reducing the risk of product defects.

[0075] Although all components constituting the embodiments have been described above as being combined or operating in combination, this does not necessarily mean that the embodiments are limited to such embodiments, and all components may be selectively combined and operate in one or more combinations within the intended scope. Furthermore, unless otherwise specified, the terms "include," "comprise," "have," and the like used above mean that the component in question can be contained within them, and therefore should be interpreted as not excluding other components but as including other components.

[0076] The above description is merely an illustrative example of the technical ideas disclosed in this document, and various modifications and variations may be made by a person having ordinary skill in the art to which the embodiments disclosed in this document pertain without departing from the essential characteristics of the embodiments disclosed in this document.

[0077] Therefore, the embodiments disclosed in this document are intended to illustrate, not limit, the technical ideas disclosed in this document, and such embodiments do not limit the scope of the technical ideas disclosed in this document. The scope of protection of the technical ideas disclosed in this document should be interpreted according to the claims set forth below, and all technical ideas within the scope equivalent thereto should be interpreted as being included in the scope of rights of this document. [Explanation of symbols]

[0078] C1, C2: Battery cells 101, 102: Tray 110:Transfer section 200: Photography Department 210:Radiation source 220: Radiation detector 300: Inspection Department 310: Memory 320: Controller

Claims

1. an imaging unit for capturing an electrode image of a battery cell; a memory for storing a learning model trained to track electrode endpoints in the electrode image of any battery cell; a controller that tracks the electrode end points in the electrode image of the battery cell using the learning model and determines whether or not the electrodes of the battery cell are defective based on coordinate information of the electrode end points; A battery cell inspection system comprising:

2. 2. The battery cell inspection system of claim 1, wherein the learning model is generated by receiving input learning data including electrode images of any battery cell, labeling the learning data, extracting features based on the labeled learning data, and structuring a database in which the features are recorded.

3. The imaging unit is a radiation source that outputs radiation toward a region of interest of the battery cell; a radiation detector that detects the radiation that has passed through a region of interest of the battery cell.

4. the imaging unit acquires a three-dimensional image of a region of interest of the battery cell; The controller generates cross-sectional images in two or more directions based on the three-dimensional image, and tracks end points of the electrodes of the battery cell for each of the cross-sectional images; The battery cell inspection system according to claim 3 , wherein the learning model is trained using, as training data, cross-sectional images in two or more directions generated based on any one of the three-dimensional images.

5. The electrodes of the battery cell have a structure in which a plurality of positive electrodes and negative electrodes are alternately stacked with separators interposed therebetween, 2. The battery cell inspection system of claim 1, wherein the controller determines that the electrode of the battery cell is defective in at least one of the following cases: when a portion of a positive electrode or a negative electrode is bent; when a portion of the positive electrode or a negative electrode is missing; when a gap between end points of adjacent positive electrodes and negative electrodes is equal to or greater than a reference value; when the insertion sequence of the positive electrodes or negative electrodes is incorrect; when a gap between end points of adjacent positive electrodes or between end points of adjacent negative electrodes is equal to or greater than a reference value; when a parameter indicating an alignment state of the positive electrodes and negative electrodes over a certain section is equal to or less than a reference value; and when a portion of the positive electrode or negative electrode is disconnected.

6. a tray for loading two or more battery cells; 4. The battery cell inspection system according to claim 3, further comprising: a transport unit that transports the tray on which the battery cells are loaded to an imaging point of the imaging unit, and transports the tray out of the imaging point when imaging is completed.

7. the transport unit simultaneously transports two or more trays on which the battery cells are loaded to the photographing location; The battery cell inspection system according to claim 6 , wherein the two or more trays are arranged such that the radiation output from the radiation source simultaneously transmits the regions of interest of the battery cells loaded on all of the trays.

8. 8. The battery cell inspection system according to claim 7, wherein the two or more trays are arranged such that an edge of the battery cells loaded on each tray at the photographing location faces an edge of the battery cells loaded on another tray.

9. the transfer unit rotates the tray on which the battery cells are loaded around an axis at the photographing point; The battery cell inspection system according to claim 6 , wherein the imaging unit acquires a three-dimensional image of the battery cell based on a two-dimensional radiation image acquired while the tray is rotating.

10. training a learning model to track electrode endpoints in an electrode image of any battery cell; acquiring an electrode image of a battery cell; tracking electrode endpoints in an electrode image of the battery cell using the learning model; determining whether or not the electrodes of the battery cell are defective based on the coordinate information of the end points of the electrodes; A battery cell inspection method comprising:

11. 11. The battery cell inspection method of claim 10, wherein the learning model is generated by receiving input learning data including electrode images of any battery cell, labeling the learning data, extracting features based on the labeled learning data, and structuring a database in which the features are recorded.

12. The step of acquiring an electrode image of the battery cell includes: outputting radiation toward a region of interest of the battery cell; Detecting radiation transmitted through a region of interest of the battery cell.

13. acquiring an electrode image of the battery cell includes acquiring a three-dimensional image of a region of interest of the battery cell; the step of tracking the end points of the electrodes of the battery cell includes the steps of generating cross-sectional images in two or more directions based on the three-dimensional image, and tracking the end points of the electrodes of the battery cell for each of the cross-sectional images; The battery cell inspection method according to claim 12 , wherein the learning model is trained using cross-sectional images in two or more directions generated based on an arbitrary three-dimensional image as learning data.

14. The electrodes of the battery cell have a structure in which a plurality of positive electrodes and negative electrodes are alternately stacked with separators interposed therebetween, 11. The battery cell inspection method according to claim 10, further comprising: determining that an electrode of the battery cell is defective in at least one of the following cases: when a portion of a positive electrode or a negative electrode is bent; when a portion of a positive electrode or a negative electrode is missing; when a gap between end points of adjacent positive electrodes and negative electrodes is equal to or greater than a reference value; when the insertion sequence of the positive electrodes or negative electrodes is incorrect; when a gap between end points of adjacent positive electrodes or between end points of adjacent negative electrodes is equal to or greater than a reference value; when a parameter indicating an alignment state of the positive electrodes and negative electrodes over a certain section is equal to or less than a reference value; and when a portion of the positive electrode or negative electrode is disconnected.

15. loading two or more battery cells onto a tray; transporting the tray on which the battery cells are loaded to a photographing location; The battery cell inspection method according to claim 12 , further comprising the step of transporting the tray out of the photographing location after the photographing is completed.

16. In the step of transporting the tray to the photographing location, two or more trays loaded with the battery cells are simultaneously transported to the photographing location; The battery cell inspection method according to claim 15 , wherein the two or more trays are arranged so that the radiation passes through regions of interest of the battery cells loaded on all of the trays simultaneously.

17. The battery cell inspection method according to claim 16 , wherein the two or more trays are arranged such that an edge of the battery cells loaded on each tray faces an edge of the battery cells loaded on another tray at the imaging point.

18. The method further includes rotating the tray on which the battery cells are loaded around an axis at the photographing point; 16. The battery cell inspection method of claim 15, wherein the step of acquiring an electrode image of the battery cell includes the step of acquiring a three-dimensional image of the battery cell based on a two-dimensional radiation image acquired while the tray is rotating.

Images