Method for manufacturing a battery element

Through the combination of eddy current sensors and artificial intelligence, the problem of difficulty in online detection of welding connection quality of battery pack components is solved, and the optimization and quality control of the welding process are achieved, reducing defective rate and production costs are reduced.

CN114571167BActive Publication Date: 2024-11-26POWERCO SE
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
CN202111440492.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Priority Date
2020-11-30
Filing Date
2021-11-30
Publication Date
2024-11-26
Estimated Expiration
2041-11-30

AI Technical Summary

Technical Problem

The prior art is difficult to effectively check the quality of the welding connection of battery pack components online, resulting in high defect rate and increased production costs during the welding process.

Method used

The eddy current sensor is used in combination with artificial intelligence to detect the eddy current data of the weld during the welding process, conduct online quality inspection and process adjustment to ensure the optimization and quality control of welding connections.

Benefits of technology

It realizes lossless, fast and automated inspection of welding connections, reduces defective rates and production costs, and improves welding quality and production efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114571167B_ABST
    Figure CN114571167B_ABST
Patent Text Reader

Abstract

A method for producing a battery element (2), wherein at least two battery components (4, 6) are joined in a materially bonded manner by means of at least one welded connection (18) during a welding process, wherein at least one first eddy current sensor (20) is provided for generating and detecting eddy currents in the welded connection (18), wherein the first eddy current sensor (20) is guided through the at least one welded connection (18) at least once, wherein the first eddy current sensor (20) generates eddy current data (26) when guided through the welded connection (18), wherein a first output variable (28) characterizing the welded connection (18) is generated based on the eddy current data (26) and a trained artificial intelligence, wherein a quality check is performed based on the first output variable (28), and wherein the welding process is controlled and / or regulated based on the result of the quality check.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to a method for producing a battery element, in which at least two battery components stacked one on top of the other are joined in a materially bonded manner by means of at least one welded connection during a welding process. The invention also relates to a device and software for carrying out the method. Background Art

[0002] Electrically driven or electrically drivable motor vehicles, such as electric vehicles or hybrid vehicles, usually have an electric machine as a drive motor, which is coupled to an electric (high-voltage) energy storage device inside the vehicle for supplying electrical energy. Such an energy storage device is implemented, for example, in the form of a (vehicle) battery pack.

[0003] In this case, an electrochemical battery is to be understood in particular as a so-called secondary battery of a motor vehicle. The chemical energy consumed can be restored by means of a charging process. Such batteries are in particular designed as electrochemical accumulators, for example as lithium-ion accumulators. In order to generate or provide a sufficiently high operating voltage, such batteries usually have a plurality of individual battery cells which are connected in a modular manner.

[0004] Batteries or battery cells are made, for example, from foil-shaped battery electrodes, which usually have active material coatings applied to foil materials, in particular copper or aluminum foils.

[0005] During the manufacturing process, it is necessary to electrically connect the battery electrodes or coated foil materials of the battery cells or batteries to the current collectors or conductive connections (busbars, conductive sheets). In this case, the connection is usually established by means of a welding process, for example by means of a laser or ultrasonic welding process, via one or more weld seams.

[0006] The quality and quality of welded connections or weld seams on battery components are important for the quality of the manufactured batteries or battery cells. Therefore, in order to reduce rejects and to improve process understanding, it is desirable to investigate and / or test welded connections or weld seams produced on battery components as non-destructively as possible and preferably in-line, i.e. during the manufacturing process in a manufacturing or production line.

[0007] For example, the characterization, i.e. inspection or investigation and / or testing of the welded connection can be achieved by X-ray analysis, in particular by means of computed tomography (CT) or ultrasonic measurement. Disadvantageously, X-ray measurements are relatively time-consuming and costly, so that these X-ray measurements cannot be performed online. In the case of ultrasonic measurement, in particular, the need for a coupling medium is problematic, making it essentially impossible to use online. In addition, for the investigation and / or testing of welded connections, it is conceivable, for example, to measure the resistance via the weld. However, the resistance measurement is too insensitive for investigating welds in battery pack assemblies. In addition, such a resistance measurement has only a small information value regarding the mechanical load capacity of the weld.

[0008] In order to evaluate the mechanical stability of the welded connection, a pull-out test can be carried out, for example. In the case of such a pull-out test, for example, the (tensile) strength and the deformation properties of the weld until breakage are determined. However, the welded connection and therefore the battery cell or battery are destroyed in this way, so that the battery cell or battery cannot be used as a defective product.

[0009] For the reasons mentioned above, welds are usually not characterized online during the manufacture of battery components. Therefore, the safety factor is usually designed to be large enough when designing welds so that the quality of the welds is ensured even with only sporadic inspections or characterizations. As a result, economic disadvantages and an increase in the number of defective products after the production or manufacture of battery components are usually accepted. Summary of the invention

[0010] The invention is based on the object of specifying a method which is particularly suitable for producing battery elements. In particular, the method should be suitable for optimizing the battery cell production process. Preferably, it should be possible to perform a simple and online investigation and / or testing of welded connections between components, preferably between battery components. The invention is also based on the object of specifying a particularly suitable device and a particularly suitable software.

[0011] According to the invention, this object is achieved with the features according to the invention with respect to the method, with the features according to the invention with respect to the device, and with the features according to the invention with respect to the software. Advantageous embodiments and extensions are the subject of the following description.

[0012] By contrast, the advantages and configurations mentioned with respect to the method can also be transferred to the device and / or the software, and vice versa. The conjunction "and / or" is to be understood here and hereinafter to mean that the features connected by means of this conjunction can be constructed not only together but also as alternatives to each other.

[0013] The method according to the invention is provided for manufacturing battery elements, in particular battery cells, and is suitable and designed for this. According to the method, at least two battery assemblies are joined in a materially compatible manner by means of at least one welded connection during a welding process. For example, the battery assemblies are stacked in this case in an overlapping manner, which means that at least two battery assemblies are arranged partially overlapping, wherein a welded connection is formed in particular in the overlapping area. Alternatively, it is also conceivable, for example, that the battery assemblies are arranged edge-to-edge or side-by-side (butt-jointed). The battery assemblies that are joined or welded are in particular foil elements of battery elements, such as cathode current collectors made of aluminum or anode current collectors made of copper. During the welding process, for example, one or more battery electrodes are connected and connected to the current collectors or conductive wiring parts or busbar contacts.

[0014] During the welding process, at least one weld seam is produced as a welded connection between the battery components. Here and below, a “weld seam” is to be understood in particular as the spatial extent and / or spatial course of the welded connection or the material connection caused by the welding process.

[0015] Here and hereinafter, "material fit" or "material fit connection" between at least two parts connected to each other is understood in particular to mean that the parts connected to each other are bonded at their contact surfaces by material combination or crosslinking (for example due to atomic or molecular bonding forces), if necessary with the action of additives.

[0016] According to the method, a first eddy current sensor is guided through at least one welded connection at least once, wherein a data set of the detected measurement signals is generated by the first eddy current sensor as eddy current data. Then, a first output quantity characterizing the welded connection is generated based on the eddy current data using a trained artificial intelligence. A quality check is performed based on the first output quantity, wherein the welding process is controlled and / or regulated based on the result of the quality check. In other words, the welded connection is checked, wherein the welding process is controlled and / or regulated for subsequent battery elements based on the results of the check. As a result, a method that is particularly suitable for manufacturing battery elements is achieved. Therefore, in particular, a welding process that is as optimized and matched as possible to the battery assembly is always ensured.

[0017] In a preferred embodiment, the characterization of the at least one weld seam, i.e. the eddy current inspection, is performed online. This means that the eddy current measurement with regard to the production of the battery element is carried out in a process-integrated manner. That is, the eddy current measurement is performed during the production or production line, for example "on-the-fly", thereby enabling a particularly uniform production and manufacturing of the battery element.

[0018] By means of online eddy current measurement for characterizing or checking weld seams, rejects are reduced, thereby advantageously subsequently reducing the production and manufacturing costs of the battery element. In particular, complete automation of investigations and / or tests, preferably carried out during the production cycle, is thus possible.

[0019] In addition, online eddy current measurement enables automatic feedback into process parameters and improved process control by weld characterization. This makes it possible, for example, to implement a target-actual comparison for checking the dimensional accuracy or shape changes of the weld produced, for example by comparison with a reference data set. This also improves the process understanding in producing the welded connection, which helps to improve the quality of the weld and thus the quality of the battery element.

[0020] According to the method, the welded connection or weld seam or each welded connection or weld seam is thus characterized non-destructively by means of eddy current measurement. In the case of such eddy current measurement, use is made of the effect that dirt and damage, i.e. defects or faults, in the electrically conductive weld seam influence its electrical conductivity or magnetic permeability. Eddy currents will be generated in the weld seam due to the induced voltage by the alternating electromagnetic field of the eddy current sensor which is guided through the weld seam as described above. In this case, the sensor acts, for example, as a transmitter and as a receiver at the same time, so that the influence of the above-mentioned defects in the weld seam on the induced eddy currents can be detected, thereby enabling a simple, cost-effective and non-destructive investigation and / or inspection of the or each weld seam.

[0021] According to the invention, for eddy current measurement, at least one eddy current sensor is provided, which can be moved along at least one weld seam and during this movement generates a data set of detected measurement signals. For example, the eddy current sensor generates an alternating electromagnetic field, wherein a voltage and subsequently eddy currents are induced in the material of the weld seam by the movement of the eddy current sensor along the weld seam. These eddy currents are detected as measurement signals or black signal levels by means of the eddy current sensor, and a data set is generated as eddy current data from this. To detect these measurement signals, the eddy current sensor has, for example, a detection coil, which detects the influence of the structure of the weld seam on the induced current.

[0022] In this case, the data set includes, for example: real data, which are in phase with the AC excitation; and imaginary data, which are detected with a 90° phase shift or perpendicular to the AC excitation. In this case, the depth or Z information of the welded connection, i.e. the depth resolution, is contained in the ratio of the real and imaginary data. Due to the real and imaginary data, the eddy current data have a relatively low interpretability compared to other reference methods, in particular imaging methods or topographic methods. Therefore, for safe and reliable quality inspection, the eddy current data are evaluated with the help of trained artificial intelligence (KI) and a (first) output is generated. This output ensures the simplest and most cost-effective (automatic or manual / by eye) interpretability of the quality of the welded connection during the quality inspection (quality analysis). As a result, reliable weld inspection in battery cell production is ensured.

[0023] Artificial intelligence or machine learning is based on computational models and algorithms for classification, bundling, regression and dimensionality reduction. Machine learning methods or machine learning are to be understood in particular as the generation of knowledge from experience, in which an artificial system learns from examples and can generalize these examples after the end of this learning phase. This means that the examples are not learned from memory, but the system determines patterns and regularities in the learning or training data. In this case, the machine learning method used to train the artificial intelligence is implemented, for example, as a learning or prediction algorithm, which is implemented by means of a controller.

[0024] In particular, a neural network is used as artificial intelligence, in which the evaluation or conversion of the eddy current data is integrated into the output variable. The evaluation or conversion is then not stored in a memory, but integrated in the topology of the neural network. Preferably, a convolutional neural network (CNN or CovNet) is used in this case.

[0025] In a preferred embodiment of the method, the KI generates a prediction data set of a reference method and / or a quality parameter of the welded connection as a first output variable. In other words, the measurement results of a measuring method (eddy current measurement) are brought close to the measurement data of another measuring method (reference method) by means of the trained KI. That is, the corresponding or corresponding image data of the imaging reference method are estimated based on the eddy current data. Additionally or alternatively, it is also possible that the KI is trained in such a way that it directly predicts or predicts the quality parameters required for the quality inspection based on the eddy current data. This makes it possible to achieve particularly reliable quality inspections and thus process control / regulation.

[0026] Here and hereinafter, a quality parameter is to be understood in particular as a parameter or quantity which characterizes the quality or properties of a welded connection. In this case, KI can, for example, estimate from eddy current data: mechanical quality parameters, such as the breaking force; and / or electrical quality parameters, such as the contact resistance between the welded battery components; and / or topological quality parameters, such as the surface properties or surface contour of the welded connection.

[0027] In an advantageous extension, a tomography method, in particular a computed tomography method (CT), is used as a trained reference method. The measurement results of eddy current measurements are very convincing, but the interpretability is low. Artificial intelligence is preferably used to approximate the measurements of the reference method based on eddy current signals. Computed tomography provides detailed, easy-to-interpret images, but is not attractive for use in production due to the long measurement time and its procurement and maintenance costs. By using a correspondingly trained artificial intelligence, predictions for the CT data are created based on the eddy current data. This makes it possible to simulate easy-to-interpret measurement methods and use them in production. However, in this case, the method according to the invention is not limited to the eddy current-CT pairing, but can also be used for other measurement methods, such as X-rays or ultrasound.

[0028] In order to train the KI, in particular the CNN, for example, a plurality of samples are characterized and analyzed. Battery components welded to one another are used as samples. For example, 100 welded (cathode / anode) current collectors are measured not only by means of eddy current measurements but also by means of a reference method to be reproduced or trained, for example by means of CT. These measurement data are used as training data, wherein training is performed according to back propagation, wherein the weights in the CNN are corrected in the gradient descent method until a prediction or reproduction of the CT data that is as accurate as possible is achieved based on the eddy current data.

[0029] Correspondingly, the KI can also be trained to determine the quality parameters directly from the eddy current data. If there is a sufficiently large database containing eddy current measurements and the corresponding quality parameters (e.g. electrical resistance and tensile strength), a convolutional neural network can be trained which automatically extracts important features from the eddy current signal. This means, for example, that the eddy current data are classified with respect to different quality parameters by means of a classifier. In the case of this classification, the quality parameters are determined from the eddy current data by investigating whether the eddy current data have previously known features that belong to previously known quality parameters.

[0030] In one possible embodiment of the method, eddy current data are measured at high eddy current power, i.e. high excitation amplitude, of the first eddy current sensor and at low eddy current power, i.e. low excitation amplitude, of the first eddy current sensor. By means of the eddy current data at high excitation and low excitation, even smaller defects or influences can be distinguished in the measurement data with the aid of the trained KI and can be taken into account for determining the output variable.

[0031] An additional aspect or another aspect of the present invention provides that a second eddy current sensor is provided, which is arranged upstream of the first eddy current sensor. In other words, the welded connection is first measured using the second eddy current sensor and then using the first eddy current sensor. In this case, the second eddy current sensor is configured as an inductive proximity sensor with a reflection geometry. In this case, the second eddy current sensor is provided and established for characterizing the topology of the welded connection. That is, using the second eddy current sensor, the surface roughness or surface properties of the welded connection are first checked. If, for example, sharp protruding parts of the welded connection are detected in this case, which may cause damage to the first eddy current sensor, the joined battery pack assembly can be rejected as defective before measurement with the first eddy current sensor. Preferably, KI is not used to evaluate or process the eddy current data of the second eddy current sensor. Thereby, the online use of the method is further improved.

[0032] In a conceivable development of the method, in addition to the trained KI, eddy current data of the first eddy current sensor are also analytically evaluated and a second output variable characterizing the welded connection is generated, wherein the analytically determined (second) output variable is compared with the (first) output variable of the KI. In particular, in a suitable embodiment, a threshold range is determined as a function of the second output variable, wherein a quality check is only performed if the first output variable is within the threshold range.

[0033] This takes into account the following situation: it is difficult to rely solely on the use of KI in the context of large-scale production, because the rationality of the results of KI cannot be checked in advance for every arbitrary scenario that may occur. Therefore, a two-stage evaluation of the eddy current data is usually set up. First, the results are evaluated using classical analysis methods. Usually, this analysis method does not provide results that are as accurate as a trained KI, but the results, i.e. the second output quantity, are easier to understand. In a second step, the first output quantity is created based on the accurate KI prediction. The results of the two evaluations are compared with each other and it is checked whether the result of KI is within the tolerance range of the less accurate result of the analysis method. Therefore, unpredictable outbreaks or deviations of the first output quantity can be reliably identified. The use of KI in production is thus guaranteed by the analysis component.

[0034] For example, in battery cell production, the method predicts the number of defects in welds based on eddy current data and KI. First, the number of circular structures in the eddy current data is roughly determined analytically in the background. Although the exact number of defects cannot be determined, the results of KI can be cross-checked in this regard and serious misbehavior of KI can therefore be identified early.

[0035] For analytical evaluation, for example, an amplitude analysis of the eddy current data, that is, an analysis or determination of the minimum and maximum values, can be envisaged. Alternatively, the analytical evaluation described in the German application with application number 102019219743.4 filed by the applicant on December 16, 2019, which was not yet published at the time of application, can be used. The disclosure of the German application, in particular the claims of the German application (and the related explanations), are hereby explicitly included in the present application. In particular, reference is made to Figure 5 and the related explanations, in particular on pages 12 to 14. In this case, the curve is fitted to the eddy current data, wherein the position and size of defects in the welded connection can be determined analytically, for example, based on the obtained fitting parameters.

[0036] The device according to the invention is suitable for and is configured for manufacturing battery elements, in particular batteries or battery cells. In this case, the device has a welding device for a material-matched welding connection of at least two battery components. The welding device is suitable for and is configured for ultrasonic welding or laser beam welding, for example. The battery components to be joined or welded are, for example, battery electrodes and current collectors or conductive sheets or bus bars. The device also has a (first) eddy current sensor for generating and detecting eddy currents in the region of at least one weld produced by the welding device. The welding device and the eddy current sensor are coupled to a controller, that is to say to a control unit. Thus, a device that is particularly suitable for manufacturing battery elements is realized.

[0037] In this case, the controller is usually set up - in a program and / or circuit technology manner - to carry out the method or use according to the invention described above. Therefore, the controller is specifically set up to: after the welding process of the welding device, by means of an eddy current sensor, induce eddy currents in the weld seam or in the area of ​​the weld seam, and detect and evaluate these eddy currents or the distribution and intensity of these eddy currents in the weld seam or in the area. In particular, the trained KI is stored or integrated in the controller. In addition, the controller is particularly suitable and set up to: control and / or regulate the manufacturing or production process of the battery pack element, in particular the welding process, based on the characterized weld seam.

[0038] In a preferred embodiment, the controller is formed at least in its core by a microcontroller having a processor and a data memory, wherein the functionality for executing the method according to the invention is implemented in a program-technical manner in the form of operating software (firmware), so that when the operating software is executed in the microcontroller, the method is automatically executed, optionally interacting with a user of the device. However, within the scope of the invention, the controller can alternatively also be formed by a non-programmable electronic component, such as an application-specific integrated circuit (ASIC), in which the functionality for executing the method according to the invention is implemented by means of circuit technology.

[0039] The first eddy current sensor can be implemented, for example, in reflection geometry. However, in a preferred embodiment, the first eddy current sensor is implemented as a fork sensor in transmission geometry, that is, as a transmission fork sensor. This achieves a particularly suitable eddy current sensor.

[0040] In this case, the fork-shaped sensor has, for example, an approximately horseshoe-shaped or U-shaped (sensor) housing, wherein a transmitting coil and an oppositely arranged receiving coil are arranged in the region of the free ends of the vertical U-shaped legs. In this case, the coil or coil pair is appropriately connected to the controller using the transformer principle. In this case, the excitation frequency of the coils, i.e. the frequency of the magnetic field generated when these coils are energized, is selected in such a way that sufficient penetration of the conductive material of the weld connection is ensured.

[0041] In order to investigate the welded connection, the battery element is introduced between the legs in the region of the fork (opening). In this case, due to the fork shape, the eddy current sensor can be easily moved and / or pivoted along the welded connection.

[0042] Advantageously, the eddy current sensor is designed in this case in a manner that meets the inspection requirements, i.e., it is matched to the battery element with regard to its size and design, so that a reliable and simple characterization of the joint connection can be achieved online. In other words, the sensor geometry is designed with regard to the product design, i.e., according to the type of joint (edge ​​joint, butt joint), the component geometry, the electrical characteristic parameters and the shape of the joint connection.

[0043] During the method, a first eddy current sensor is guided linearly across the weld seam. This traversal generates, for example, a data set with image data in the form of (two-dimensional) sectional images or sectional photographs of the corresponding cross section (XZ). In this case, the abscissa axis (X axis, X direction) is oriented, for example, along the measuring direction (travel / movement direction) of the first eddy current sensor or along the longitudinal direction of the weld seam, and the ordinate axis (Y axis, Y direction) is oriented along a transverse direction oriented transversely thereto, and the application axis (Z axis, Z direction) is oriented along the height of the weld seam.

[0044] This means that for three-dimensional resolution or characterization of the welded joint, a (first) eddy current sensor equipped with a single coil pair is moved across the weld seam multiple times, wherein for the detection of eddy current data in the Y direction, for example, multiple consecutive measurements arranged offset in the Y direction are performed along the X direction. In other words, multiple traversals of the weld seam are performed using the first eddy current sensor.

[0045] In order to optimize the measuring time, in an advantageous embodiment it is provided that an array of a plurality of coil pairs is created, which are arranged adjacent to one another in the Y direction on the eddy current sensor, for example, so that the eddy current sensor detects the required eddy current data in a single pass through the welded connection. This significantly reduces the required measuring time, thereby improving the suitability of the system or method in production. This makes it possible, in particular, to achieve faster investigations and / or tests, thereby advantageously reducing the cycle time during characterization and during quality checks.

[0046] An additional aspect or another aspect of the present invention provides for a software on a medium or a data carrier for executing or implementing the method described above. This means that the software is stored on a data carrier and is configured to implement the method described above when the software is run on a computer, and is suitable for and designed for this. In this case, the software in particular has a KI trained according to the method, such as a CNN. As a result, a software that is particularly suitable for manufacturing battery components is realized, with which the functionality for executing the method according to the present invention is implemented in a program-technical manner. Therefore, the software is in particular operating software (firmware), wherein the data carrier is, for example, a data memory of a controller. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Subsequently, embodiments of the present invention will be described in more detail with reference to the accompanying drawings. In which a simplified and schematic diagram is shown:

[0048] Figure 1 An apparatus for manufacturing battery elements is shown in top view;

[0049] Figure 2 An eddy current sensor for investigating a weld seam, implemented as a fork-shaped sensor, is shown in a perspective view; and

[0050] Figure 3 Eddy current data of an eddy current sensor and CT data generated with the aid of a trained KI as well as measured CT data are shown. DETAILED DESCRIPTION

[0051] exist Figure 12 shows a schematic and simplified battery element 2 of a vehicle battery (not shown in greater detail). The battery element 2 has two battery assemblies 4 , 6 as components.

[0052] The battery element 2 is, for example, a battery cell, wherein the battery component 4 in this case is embodied in particular as a battery electrode with a coating 8 and with a current collector or collector foil 10, wherein the battery component 6 is embodied in particular as a conductive sheet. For an improved connection in the region of the battery component or conductive sheet 6 and in a housing part of the battery element 2, which is not shown in greater detail, a so-called pre-sealing strip 12 is arranged in the region of the battery component 6. With the aid of the pre-sealing strip 12, a minimum distance for electrical insulation between the live conductive sheet 6 and the housing part is achieved.

[0053] exist Figure 1 In the drawing, a device 14 for producing a battery element 2 is also shown in a schematic and greatly simplified illustration.

[0054] In this case, the device 14 has a welding device 16 for welding the battery assemblies 4 and 6 stacked at least partially overlapping in a material-fitting manner. The welding device 16 is suitable and configured, for example, for ultrasonic welding or laser beam welding. In this case, the welding device 16 is moved, for example, in a joining direction F and joins the battery assemblies 4, 6 in a material-fitting manner by means of at least one welded connection or weld seam 18.

[0055] The device 14 also has an eddy current sensor 20 for characterizing or checking the at least one weld seam 18 . In this case, the eddy current sensor 20 is suitable and designed to generate and detect eddy currents in the region of the at least one weld seam 18 .

[0056] The welding device 16 and the eddy current sensor 20 are coupled to a control device 22 , ie to a control unit.

[0057] exist Figure 1 In the embodiment of the battery element 2, the battery element 2 has a single, elongated weld seam 18. In other words, the battery components 4 and 6, in particular the conductive sheet 6 and the collector foil 10, are welded to each other by means of the weld seam 18. However, for example, elongated weld seams arranged distributed along the joining direction F or a plurality of approximately point-shaped weld seams (weld spots) which are arranged distributed perpendicular to the joining direction F in two rows are also conceivable.

[0058] exist Figure 2 The eddy current sensor 20 shown in greater detail in FIG. 2 is embodied, for example, in the form of a transmissive fork sensor.

[0059] In this case, the fork-shaped sensor 20 has, for example, an approximately horseshoe-shaped or U-shaped (sensor) housing 24, wherein a transmitting coil and an oppositely arranged receiving coil are arranged in the region of the vertical U-shaped legs. In this case, the coils, which are not shown in greater detail, are connected to the controller using the transformer principle. In this case, the excitation frequency of these coils is selected so that sufficient penetration of the conductive material of the welding connection 18 is ensured. Preferably, a coil array is provided, i.e. a plurality of adjacently arranged coil pairs, which can respectively generate and receive an excitation field.

[0060] The controller 22 controls and / or regulates the production process of the battery element 2, in particular the welding process, i.e. the welding device 16, based on the quality inspection or characterization of the produced weld seam 18. In this case, it is possible to characterize the battery element 2 as defective if, for example, a serious defect in the weld seam 18 is detected based on the quality inspection.

[0061] A method for producing a battery element 2 is explained in more detail below, which method is carried out, for example, by means of software of a control unit 22 .

[0062] According to the method, the battery components 4, 6 are welded in a material-bonded manner in a welding process or welding procedure by means of a welding device 16. An eddy current sensor 20 is then passed through the welded connection or weld seam 18, wherein a data set of the detected measurement signals is generated by the eddy current sensor 20 as eddy current data 26. The eddy current data 26 are sent to a controller 22.

[0063] The trained artificial intelligence is integrated in the control unit 22, in particular as a convolutional neural network. In this case, the artificial intelligence or CNN is trained in such a way that it determines output variables 28 that characterize the welded connection 18 based on the eddy current data 26. Based on the output variables 28, a quality check of the welded connection 18 is automatically performed, for example, by the control unit 22, wherein the welding process, i.e. the welding device 16, is controlled and / or regulated based on the result of the quality check.

[0064] In this case, the KI can be trained so that a quality parameter such as the breaking force or the contact resistance is directly output as an output quantity 28. Thus, the output quantity 28 itself can be used to directly implement a quality check. Additionally or alternatively, the KI can be trained so that a predicted or estimated data set is generated as an output quantity 28, which corresponds to the measurement of the reference method. In other words, the measurement results of the measurement method (eddy current measurement) are brought close to the measurement data of another measurement method (reference method) by the trained KI. This improves the interpretability, in particular when the quality check is performed manually or by eye by the user of the device 14.

[0065] In one embodiment, which is not shown in greater detail, a second eddy current sensor is provided, for example, which is arranged upstream of the eddy current sensor 20. In other words, the welded connection 18 is first measured with the second eddy current sensor and then with the eddy current sensor 20. In this case, the second eddy current sensor is designed as an inductive proximity sensor in reflection geometry and is provided and designed to characterize the topology of the welded connection 18. In this case, the data of the second eddy current sensor are appropriately also transmitted to the controller 22 and taken into account, for example, in a quality check.

[0066] Subsequently, based on Figure 3 Let’s elaborate on the predictions of the reference method in more detail. Figure 3 The diagram of comprises eight horizontal, superimposed sections 30 , 32 , 34 , 36 , 38 , 40 , 42 , 44 .

[0067] In this case, sections 30, 32, 34 and 36 show eddy current data 26 of the measured welded connection 18. Sections 38 and 42 show measurement data of a reference method, in particular a computed tomography measurement, wherein sections 40 and 44 show corresponding CT data, which are predicted by the trained KI as output variables 28 from the eddy current data 26. In this case, sections 30 to 44 each show a two-dimensional cross-sectional image through the welded connection 18.

[0068] In sections 30 to 44, the measured distance Mx along the joining direction F is plotted horizontally, i.e. along the abscissa axis (X axis), for example in millimeters (mm). In this case, in sections 30 to 40, the measured distance My along the transverse direction is shown, for example in millimeters (mm), along the vertical ordinate axis (Y axis). Along the vertical ordinate axis (Y axis), sections 42 and 44 show the measured distance Mz along the height direction of the welded connection 18, for example in millimeters (mm). This means that sections 30 to 40 each show a cross-section in a transverse XY plane of the welded connection 18, wherein sections 42 and 44 show a cross-section in an XZ plane oriented perpendicularly thereto.

[0069] The eddy current data 26 shown in parts 30 to 36 are preferably recorded using an eddy current sensor 20 having a coil array. In particular, each measurement line is recorded using a coil pair along the measuring distance Mx. For example, the eddy current sensor 20 has twelve adjacent coil pairs in this case.

[0070] Part 30 shows real data, which are detected when the excitation power of the eddy current sensor 20 is low, wherein part 32 shows the corresponding imaginary data. Part 34 shows real data, which are detected when the excitation power of the eddy current sensor 20 is high, wherein part 36 shows the corresponding imaginary data.

[0071] The trained KI determines the corresponding CT data (portions 40, 44) from the recorded eddy current data 26 of portions 30 to 36. Figure 3 As can be seen in FIG. 1 , the CT data generated from the eddy current data 26 by means of the trained KI has a high degree of consistency with the actually measured CT data (parts 38 , 42 ).

[0072] In one conceivable development of the method, in addition to the trained KI, the eddy current data 26 of the eddy current sensor 20 are also evaluated analytically and a second output variable 46 characterizing the welded joint is generated. A threshold range is determined as a function of the output variable 46, wherein a quality check is only performed if the output variable 28 is within the threshold range. In other words, the trained KI and the result of the analytical evaluation are compared with each other and a check is performed to see whether the result of the KI is within the tolerance range of the result of the analytical method.

[0073] The claimed invention is not limited to the embodiments described above. Rather, further variants of the invention can also be derived therefrom by a person skilled in the art within the framework of the disclosed claims without departing from the subject matter of the claimed invention. Furthermore, all individual features described in conjunction with the various embodiments can also be combined in other ways within the framework of the disclosed claims without departing from the subject matter of the claimed invention.

[0074] List of Reference Numerals

[0075] 2 Battery Pack Components

[0076] 4.6 Battery Pack Components

[0077] 8 Coating

[0078] 10 Collector foil

[0079] 12 Pre-sealing tape

[0080] 14 Equipment

[0081] 16 Welding equipment

[0082] 18 Welding connection

[0083] 20 Eddy current sensor / fork sensor

[0084] 22 Controller

[0085] 24 Housing

[0086] 26 Eddy Current Data

[0087] 28 Output

[0088] 30...44 parts

[0089] 46 Output

[0090] F Joint direction

[0091] Mx, My, Mz measurement distance

Claims

1. A method for producing a battery element (2), the method comprising a trained artificial intelligence in the form of a neural network, the neural network being trained in such a way that the neural network estimates corresponding image data of a tomography method from eddy current data as output quantity, in, For a number of battery pack components welded to each other, eddy current data and tomography image data are used as training data. - at least two battery components (4, 6) are joined in a materially bonded manner by means of at least one welded connection (18) during a welding process, - at least one first eddy current sensor (20) is provided for generating and detecting eddy currents in the welded connection (18), wherein the first eddy current sensor is guided past the at least one welded connection (18) at least once, - wherein the first eddy current sensor (20) generates eddy current data (26) when guided through the weld connection (18), - wherein a first output (28) characterizing the welded connection (18) is generated based on the eddy current data (26) and the trained artificial intelligence, - wherein a quality check is performed on said first output quantity (28), and - wherein the welding process is controlled and / or regulated as a function of the result of the quality check.

2. The method according to claim 1, It is characterized in that In addition to the image data of the tomographic method, a quality parameter of the welded connection is generated as a first output variable (28).

3. The method according to claim 1 or 2, It is characterized in that The eddy current data (26) are measured in the case of a high eddy current power of the first eddy current sensor (20) and in the case of a low eddy current power of the first eddy current sensor.

4. The method according to claim 1 or 2, It is characterized in that - providing a second eddy current sensor as an inductive proximity sensor, said second eddy current sensor being arranged upstream of said first eddy current sensor (20), and - characterizing the topology of the welded connection (18) using the second eddy current sensor.

5. The method according to claim 1 or 2, It is characterized in that - the eddy current data (26) are evaluated analytically and a second output quantity (46) is generated which characterizes the welded connection (18), and - comparing the first and second output quantities (28, 46) with one another.

6. The method according to claim 5, It is characterized in that - determining a threshold range based on said second output quantity (46), and - If the first output quantity (28) is within the threshold range, performing the quality check.

7. A device (14) for producing a battery element (2), the device comprising: - a welding device (16) for a material-fitting welding connection of at least two battery components (4, 6); and a first eddy current sensor (20) for generating and detecting eddy currents in the region of at least one welded connection (18) produced by the welding device, in, The first eddy current sensor (20) is movable along the welded connection (18) and generates a data set as eddy current data (26) based on the detected measurement signals during such movement; and - a controller (22) coupled to the welding device (16) and the first eddy current sensor (20), for executing the method according to any one of claims 1 to 6, - wherein a trained artificial intelligence in the form of a neural network is integrated into the controller (22), - wherein the neural network is trained in such a way that it estimates corresponding image data of a tomographic method from the eddy current data as an output variable, wherein eddy current data and image data of a tomographic method are used as training data for a plurality of battery components welded to one another, - wherein the controller (22) generates an output (28) based on the eddy current data (26) by means of the neural network and performs a quality check, and - wherein the controller (22) controls and / or regulates the welding device (16) as a function of the result of the quality check.

8. The device according to claim 7, It is characterized in that The first eddy current sensor (20) is embodied as a fork sensor in transmission geometry.

9. The device according to claim 7 or 8, It is characterized in that The first eddy current sensor (20) has a coil array having a plurality of coils, by means of which eddy currents can be generated and detected respectively.

10. A computer program product on a data carrier for carrying out the method according to any one of claims 1 to 6 when the computer program product is run on a computer.

Citation Information

Patent Citations

  • Method for characterizing a weld

    DE102019219743A1

  • Eddy current sensor type automatic weld tracking control method

    CN101837504A

  • Method for eddy current sensing welding seam automatic following control of front dual probes

    CN104070267A

  • Edge micro-plasma arc welding forming control method based on identification of dynamical system

    CN105478976A

  • Exciting coil weld joint path deviation-rectifying device and method adopting double Hall sensors

    CN108247251A