Lithium battery pole piece detection system and method

Abstract

The invention discloses a lithium battery pole piece detection system and method, the system comprises an image acquisition device, an industrial personal computer and a marking machine, the image acquisition device is used for acquiring positive and negative pole piece images and uploading the images to the industrial personal computer, the industrial personal computer analyzes and processes the uploaded images through a built-in detection algorithm, and the marking machine is used for marking the positive and negative pole piece images. And the processing result is fed back to the marking machine, and the marking machine carries out marking processing on the position area, where the flaws are located, of the pole piece. According to the scheme, the structure is simple, operation is convenient and fast, and the detection method based on the improved YOLOv7-tiny is provided. And classifying the flaws through labeling. Through early-stage verification, the deep learning defect detection algorithm of the roller press image has relatively good performance; compared with a traditional method, the method has higher feature learning ability. The deep learning model can automatically learn and understand complex features, so that the defect that the contrast ratio is not strong can be detected more effectively.

Description

Technical Field

[0001] The present invention relates to the field of lithium ion battery detection, and in particular to a lithium ion battery pole piece detection system and method. Background Art

[0002] As an important part of the lithium battery production process, unqualified pole pieces will lead to safety hazards after the lithium battery leaves the factory. Lithium battery pole piece defect detection has become an important part of lithium battery production. Traditional pole piece defect detection is carried out through machine learning. Machine learning requires the definition of the pixels, aspect ratio, object outline, qualified parameter standards, etc. of the image, and it is impossible to perform multiple pole piece defect detection. The traditional defect detection algorithm currently used is not effective and has a high proportion of overkill. Key defects that affect quality need to be strictly detected. . Summary of the invention

[0003] In order to solve the existing problems, the present invention provides a lithium-ion battery electrode detection system and method, the specific scheme is as follows

[0004] A lithium battery pole piece detection system includes an image acquisition device, an industrial computer and a marking machine. The image acquisition device is used to acquire images of positive and negative pole pieces and upload them to the industrial computer. The industrial computer analyzes and processes the uploaded images through a built-in detection algorithm and feeds back the processing results to the marking machine. The marking machine marks the position area on the pole piece where the defect is located.

[0005] Preferably, the system further comprises a linear array light source, and the linear array light source is used to illuminate the material by using the light source to make the image of the inspected material clear.

[0006] Preferably, the image acquisition device includes CCD cameras for front and back surface defect detection arranged at the upper and lower ends of the pole piece, and CCD cameras for width detection arranged at the left and right sides of the pole piece.

[0007] Preferably, the front and back surface defect detection CCD camera adopts a 16K camera, whose line scanning frequency is 85000mm / 60s / 700mm*16384=33.2KHz, and the defect detection accuracy is 700 / 16384=0.0427mm / pixel. The maximum limit line frequency of the 16K camera is 80KHz, which can meet the on-site pole defect detection requirements.

[0008] Preferably, the width detection CCD camera performs width detection on the slit pole piece, the maximum detection width of the pole piece is 650mm, the effective size of the roller is 700mm, and the maximum detection speed is 85m / min; the width measurement CCD camera is provided with 2 sets of 8K cameras on the left and right sides of the pole piece, and the line scanning frequency is 85000mm / 60s / 350mm*8192=33.2KHz, the width detection accuracy is 350 / 8192=0.0427mm / pixel, and the maximum limit line frequency of the 8K camera is 110KHZ, which can meet the width measurement requirements.

[0009] Preferably, the marking machine adopts a high-speed labeling machine with a labeling speed of 15 labels per second and a labeling accuracy of ±4mm, which can meet the marking requirements of defective electrodes on a high-speed production line.

[0010] Preferably, the industrial computer includes detection software and a detection interface; the detection software adopts the YOLOv7-tiny target detection algorithm, and the detection head part of the YOLOv7-tiny model adopts the ConvMixer architecture to enhance the small target detection performance; the detection interface sets three levels of management authority - level 1 - super administrator, level 2 - workshop administrator / technician, level 3 - operator, and controls the functions that can be used in the system according to the job responsibilities of employees.

[0011] Preferably, the process steps of the YOLOv7-tiny target detection algorithm include:

[0012] S1', build the training data set of YOLOv7 network;

[0013] S2', improved YOLOv7-tiny network;

[0014] S3', train the YOLOv7-tiny network to obtain the target detection model;

[0015] S4', using the target detection model to detect the target to be detected;

[0016] S5', obtaining target space information and classification information.

[0017] Preferably, a lithium battery pole piece detection method based on any of the above systems comprises the following steps:

[0018] S1, after logging into the software interface, the operator needs to click on the loop run, and then click on the run interface to complete the software startup and the system initialization;

[0019] S2, fix the coated electrode to the unwinding mechanism, pass the electrode through the gap between the two rollers correctly, connect the winding system and set the unwinding speed, and perform edge trimming and heating after unwinding;

[0020] S3, the roller press rolls the incoming lithium battery pole pieces. The compaction density of the active material in the pole piece directly affects the energy density and power density of the battery;

[0021] S4, the front and back defect detection CCD cameras located on the upper and lower sides of the material area select and merge the cross-type defects;

[0022] S5, the cutter cuts the lithium battery electrode according to the set number of cutting strips;

[0023] S6, the width detection CCD cameras located on the left and right sides of the material area detect the width of the electrode after cutting;

[0024] S7, the marking machine is linked with the defect detection CCD camera, and various types of defects are calibrated and labeled after algorithm processing;

[0025] S8, the winding mechanism is connected to the unwinding mechanism to wind the material, and the process ends after the reel is completed.

[0026] The beneficial effects of the present invention are:

[0027] The scheme has a simple structure and convenient operation. It proposes a detection method based on improved YOLOv7-tiny. Defects are classified by annotation. Through preliminary verification, the deep learning defect detection algorithm of roller press images has relatively good performance; compared with traditional methods, it has more powerful feature learning ability. The deep learning model can automatically learn and understand complex features, so as to more effectively detect defects with weak contrast. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0029] Figure 1 It is a schematic diagram of the YOLOv7-tiny network model structure of the present invention;

[0030] Figure 2 It is the YOLOv7-tiny algorithm flow chart of the present invention;

[0031] Figure 3 It is a schematic diagram of the system structure of the present invention;

[0032] Figure 4 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0033] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0034] A lithium battery pole piece detection system includes an image acquisition device, an industrial computer and a marking machine. The image acquisition device is used to acquire images of positive and negative pole pieces and upload them to the industrial computer. The industrial computer analyzes and processes the uploaded images through a built-in detection algorithm and feeds back the processing results to the marking machine. The marking machine marks the position area on the pole piece where the defect is located.

[0035] The system also includes a linear array light source, which is used to utilize the light source to illuminate so that the image of the inspected material is clear.

[0036] The image acquisition device includes front and back defect detection CCD cameras arranged at the upper and lower ends of the pole piece, and width detection CCD cameras arranged at the left and right sides of the pole piece. The width detection CCD cameras include a left sub-roll width measurement CCD camera and a right sub-roll width measurement CCD camera.

[0037] The CCD camera for front and back defect detection uses a 16K camera with a line scan frequency of 85000mm / 60s / 700mm*16384=33.2KHz. The defect detection accuracy is 700 / 16384=0.0427mm / pixel. The maximum limit line frequency of the 16K camera is 80KHz, which can meet the on-site electrode defect detection requirements.

[0038] The width detection CCD camera detects the width of the slit pole piece. The maximum detection width of the pole piece is 650mm, the effective size of the roller is 700mm, and the maximum detection speed is 85m / min. The width measurement CCD camera is equipped with 2 sets of 8K cameras on the left and right sides of the pole piece, and its line scanning frequency is 85000mm / 60s / 350mm*8192=33.2KHz. The width detection accuracy is 350 / 8192=0.0427mm / pixel. The maximum limit line frequency of the 8K camera is 110KHZ, which can meet the width measurement requirements.

[0039] The marking machine adopts a high-speed labeling machine with a labeling speed of 15 per second and a labeling accuracy of ±4mm, which can meet the marking requirements of defective electrodes on high-speed production lines.

[0040] The industrial computer includes detection software and a detection interface.

[0041] The detection software includes an operator set and detection processes corresponding to each image acquisition device. The operator set includes a logical control sequence, and the detection process corresponding to the image acquisition device refers to converting the grayscale value of the electrode to be tested to obtain the bright text value under multiple spectra, and using a multi-spectral algorithm to obtain the true temperature and spectral emissivity of the target.

[0042] The detection software adopts the YOLOv7-tiny target detection algorithm, and the detection head part of the YOLOv7-tiny model adopts the ConvMixer architecture to enhance the small target detection performance. The ConvMixer in the prediction head helps to capture the spatial and channel relationships in the features passed to the prediction head. The deep learning model can automatically learn and understand complex features, so as to more effectively detect defects with weak contrast. And the deep learning model can be continuously iterated through large-scale data set training, and has rich learning experience of defect samples and scene changes. Such training makes the system more robust and can adapt to the detection needs under various actual working conditions. Complex pattern recognition and defect detection problems are often involved in industrial production. By learning complex patterns in large-scale data sets, the deep learning model can more accurately perform defect detection, product quality control and anomaly detection.

[0043] Specifically, Figure 1 , the YOLOv7-tiny network model structure is divided into four parts: Input, Backbone, Neck and Head. The input part uses mosaic data enhancement and adaptive anchor box calculation to preprocess the input image; the Backbone part is composed of several CBL modules, ELAN layers and MP layers. The CBL module is composed of Conv layer, BN layer and Silu function, and the ELAN layer is composed of multiple CBL modules. The MP layer is composed of CBL modules and Maxpool respectively; the Neck part uses SPPCSPC and PAN structures to fuse the features of each layer to detect targets of different scales. The structural connection is composed of CBL module, MP layer, SPPCSPC module and ELAN+. The SPPCSPC module connects Backbone and Neck, which is composed of multiple CBL modules and three Maxpools. The difference between ELAN+ and ELAN is only the number of outputs selected during the connection of CBS modules. The Head part uses the Rep layer and CBM module. The grid structure of the Rep module is different during training and inference, and the CBM module is composed of the Conv layer, BN layer and Sigmoid function. In the Head layer stage, the output image is passed through three Rep layers and CBM layers to output three prediction results of different scales.

[0044] like Figure 2 , the process steps of the YOLOv7-tiny target detection algorithm include:

[0045] S1', build the training data set of YOLOv7 network;

[0046] S2', improved YOLOv7-tiny network;

[0047] S3', train the YOLOv7-tiny network to obtain the target detection model;

[0048] S4', using the target detection model to detect the target to be detected;

[0049] S5', obtaining target space information and classification information.

[0050] The detection interface can display image parameters and non-image parameters as well as account management authority items in real time. The image parameters and non-image parameters refer to the collected values ​​that meet the on-site process requirements. Specifically, it includes production data, virtual images of material rolls, defect images, defect information, as well as material re-inspection and setting defect standard functions. The image is analyzed and calculated to obtain the surface defects and size information of the pole piece. The system can provide output pole piece slitting width information, calculate the pole piece slitting width value, and then use the CCD unit to feedback to the industrial computer. The detection interface sets three levels of management authority: Level 1-Super Administrator, Level 2-Workshop Administrator / Technician, Level 3-Operator, and controls the functions that can be used in the system according to the job responsibilities of employees.

[0051] The inspection software reads the relevant product information of the client MES system and organizes and feeds back the inspection results in the form of customer needs. The defective pictures are saved by category number, film roll number, and work number according to customer needs, and the defective pictures and information of each roll in actual production can be queried; the inspection data, picture information, etc. are displayed in real time on the display screen, and the corresponding data and pictures are saved in real time on the industrial computer according to needs, and the information retention period is ≥90 days.

[0052] The detection software uses Labelimg annotation tool to annotate the defect images collected on site, and the annotation format is XML format. After defect data screening, according to the defect data distribution, fifteen types of defects are selected, including bubbles, gaps, powder loss, deep scratches, shallow scratches, black spots, film area leakage, convex hulls, convex spots, white spots, dark spots, tab damage, tab wrinkles, edge foil leakage, and pole piece damage. The original image is cropped and images (training set, verification set, test set) are selected for training and verification. The results are uploaded to the marking machine for standard processing.

[0053] like Figure 3The front defect detection CCD camera and the back defect detection CCD camera are located at the upper and lower ends of the pole piece respectively. The industrial computer connects the front defect detection CCD camera and the back defect detection CCD camera via USB. The camera uploads the collected image information to the industrial computer, and the industrial computer analyzes and processes the uploaded image information through the built-in algorithm. The cutter located at the upper end of the pole piece performs one-out-six processing on the incoming pole piece according to the process requirements. The width detection CCD cameras located on the left and right sides of the pole piece detect the width of the pole piece after slitting (the maximum detection width of the pole piece is 650mm, the effective size of the roller is about 700mm, and the maximum detection speed is 85m / min). The marking machine is linked with the defect detection CCD camera. The industrial computer feeds back the defect information collected by the defect detection CCD camera to the marking machine, and the marking machine marks the area where the defect is located.

[0054] Specifically, Figure 4 A lithium battery pole piece detection method based on any of the above systems comprises the following steps:

[0055] S1, after logging into the software interface, the operator needs to click on the loop run, and then click on the run interface to complete the software startup and the system initialization;

[0056] S2, fix the coated electrode to the unwinding mechanism, pass the electrode through the gap between the two rollers correctly, connect the winding system and set the unwinding speed, and perform edge trimming and heating after unwinding;

[0057] S3, the roller press rolls the incoming lithium battery pole pieces. The compaction density of the active material in the pole piece directly affects the energy density and power density of the battery;

[0058] S4, front and back defect detection CCD cameras located on the upper and lower sides of the material area select and merge cross-type defects; a total of 6 categories: ① metal leakage; ② streaks; ③ black spots / black spots; ④ white spots / white spots; ⑤ uncoated / tape; ⑥ unrolled

[0059] S5, the cutter cuts the lithium battery electrode according to the set number of cutting strips;

[0060] S6, the width detection CCD cameras located on the left and right sides of the material area detect the width of the electrode after cutting;

[0061] S7, the marking machine is linked with the defect detection CCD camera, and various types of defects are calibrated and labeled after algorithm processing;

[0062] S8, the winding mechanism is connected to the unwinding mechanism to wind the material, and the process ends after the reel is completed.

[0063] The present invention has a simple structure and convenient operation, and can classify defects by marking. Through preliminary verification, the deep learning defect detection algorithm for roller press images has relatively good performance; compared with traditional methods, it has more powerful feature learning capabilities. The deep learning model can automatically learn and understand complex features, so as to more effectively detect defects with weak contrast. Collect production line defect samples, train the deep learning model, and then perform preliminary processing on the production line data. After adopting deep learning technology + traditional defect algorithm, the accuracy of defect detection is greatly improved, which can reduce the risk of defective outflow and avoid excessive overkilling and false detection.

[0064] Those skilled in the art will further appreciate that the various illustrative logic blocks, modules, circuits, and algorithm steps described in conjunction with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or a combination of the two. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps are generally described above in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the specific application and the design constraints imposed on the overall system. The technician may implement the described functionality in different ways for each specific application, but such implementation decisions should not be interpreted as resulting in a departure from the scope of the present invention.

[0065] The previous description of the disclosure is provided to enable any person skilled in the art to make or use the disclosure. Various modifications to the disclosure will be apparent to those skilled in the art, and the general principles defined herein may be applied to other variations without departing from the spirit or scope of the disclosure. Thus, the disclosure is not intended to be limited to the examples and designs described herein, but should be granted the widest scope consistent with the principles and novel features disclosed herein.

[0066] Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent substitutions for some of the technical features therein; and these modifications or substitutions do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A lithium battery pole piece detection system, characterized in that: It includes an image acquisition device, an industrial computer and a marking machine. The image acquisition device is used to acquire images of positive and negative electrodes and upload them to the industrial computer. The industrial computer analyzes and processes the uploaded images through a built-in detection algorithm and feeds the processing results back to the marking machine. The marking machine marks the location area on the electrode where the defect is located.

2. The system according to claim 1, characterized in that: It also includes a linear array light source, which is used to utilize the light source to illuminate so that the image of the detected material is clear.

3. The system according to claim 1, characterized in that: The image acquisition device includes CCD cameras for detecting front and back defects arranged at the upper and lower ends of the pole piece, and CCD cameras for detecting width arranged at the left and right sides of the pole piece.

4. The system according to claim 3, characterized in that: The front and back defect detection CCD camera adopts a 16K camera, whose line scanning frequency is 85000mm / 60s / 700mm*16384=33.2KHz, and the defect detection accuracy is 700 / 16384=0.0427mm / pixel. The maximum limit line frequency of the 16K camera is 80KHz, which can meet the on-site electrode defect detection requirements.

5. The system according to claim 3, characterized in that: The width detection CCD camera detects the width of the slit electrode piece, the maximum detection width of the electrode piece is 650mm, the effective size of the roller is 700mm, and the maximum detection speed is 85m / min; the width measurement CCD camera is equipped with 2 sets of 8K cameras on the left and right sides of the electrode piece, and its line scanning frequency is 85000mm / 60s / 350mm*8192=33.2KHz, the width detection accuracy is 350 / 8192=0.0427mm / pixel, and the maximum limit line frequency of the 8K camera is 110KHZ, which can meet the width measurement requirements.

6. The system according to claim 1, characterized in that: The marking machine adopts a high-speed labeling machine with a labeling speed of 15 per second and a labeling accuracy of ±4mm, which can meet the marking requirements of defective electrodes on high-speed production lines.

7. The system according to claim 1, characterized in that: The industrial computer includes detection software and a detection interface; the detection software adopts the YOLOv7-tiny target detection algorithm, and the detection head part of the YOLOv7-tiny model adopts the ConvMixer architecture to enhance the small target detection performance; the detection interface sets three levels of management authority - level 1 - super administrator, level 2 - workshop administrator / technician, level 3 - operator, and controls the functions that can be used in the system according to the job responsibilities of employees.

8. The system according to claim 1, characterized in that: The process steps of the YOLOv7-tiny target detection algorithm include: S1', build the training data set of YOLOv7 network; S2', improved YOLOv7-tiny network; S3', train the YOLOv7-tiny network to obtain the target detection model; S4', using the target detection model to detect the target to be detected; S5', obtaining target space information and classification information.

9. A lithium battery pole piece detection method based on the system according to any one of claims 1 to 8, characterized in that: The following steps are involved: S1, after logging into the software interface, the operator needs to click on the loop run, and then click on the run interface to complete the software startup and the system initialization; S2, fix the coated electrode to the unwinding mechanism, pass the electrode through the gap between the two rollers correctly, connect the winding system and set the unwinding speed, and perform edge trimming and heating after unwinding; S3, the roller press rolls the incoming lithium battery pole pieces. The compaction density of the active material in the pole piece directly affects the energy density and power density of the battery; S4, the front and back defect detection CCD cameras located on the upper and lower sides of the material area select and merge the cross-type defects; S5, the cutter cuts the lithium battery electrode according to the set number of cutting strips; S6, the width detection CCD cameras located on the left and right sides of the material area detect the width of the electrode after cutting; S7, the marking machine is linked with the defect detection CCD camera, and various types of defects are calibrated and labeled after algorithm processing; S8, the winding mechanism is connected to the unwinding mechanism to wind the material, and the process ends after the reel is completed.

Images