Evaluation method of pole piece appearance online vision measurement system
By collecting unqualified sample images in the online visual detection system for pre-processing and feature analysis, combined with large sample evaluation of the production process, the repetitive testing problems of the online detection system are solved, the misjudgment rate and missed detection rate are controlled, and the quality of the electrode plate and the performance of the lithium battery are improved.
Patent Information
- Application Number
- CN202411782873.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-06
- Publication Date
- 2025-08-15
AI Technical Summary
The existing online visual inspection system for the appearance of lithium battery pole plates is difficult to perform effectiveness analysis under high-speed production conditions, and it is impossible to achieve repeatability testing, resulting in the inability to truly reflect the detection capabilities, affecting the quality of the pole plates and the performance of lithium battery.
By preparing detection programs on visual inspection equipment, unqualified sample images are collected for pre-processing, defects are analyzed using digital signals and characteristic parameters, abnormal areas are automatically taken to take photos, and the system effectiveness is evaluated through misjudgment rates and missed detection rates, and large-sample analysis is performed in combination with the production process.
The ability to accurately evaluate the detection equipment under actual production speed and vibration conditions is achieved, the error judgment rate and leakage detection rate are reduced, the electrode sheet quality is ensured, and the safety and reliability of lithium battery production are improved.
Smart Images

Figure CN120490107A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of automated visual inspection of lithium batteries, and in particular to an evaluation method for an online visual measurement system for pole piece appearance. Background Art
[0002] Modern factories are becoming increasingly automated, requiring more and more reliable equipment. Analyzing measurement system capabilities is essential for mass production. Automated lithium battery production runs at high speeds, typically exceeding 100 parts per million (PPM). This type of camera can achieve 100% inspection at high speeds, making it particularly important for companies to analyze the effectiveness of such visual inspection systems. Established analytical methods exist for appearance-based counting inspection systems. For unmanned inspection systems, a common approach involves selecting 50 products (both qualified and rejected) covering the entire process range and performing repeated testing nine times to analyze the effectiveness of the measurement system. This type of equipment is typically housed in a separate space, making it easier to achieve repeatable product measurements. However, in-line inspection equipment is embedded within production equipment, connecting upstream and downstream processing steps. The state and shape of the product in the downstream process often change compared to the previous process. Furthermore, due to the compact design and short distances between upstream and downstream processes, the final product often arrives at the visual inspection system after the previous one has already been processed. The processed product cannot be restored to its original state for repeatability testing, making repeatability analysis difficult with conventional measurement systems. Furthermore, the electrode product is very thin, only at the micron level, and the coating is prone to shedding. Repeated testing requires repeated rewinding from one roll to another. If shedding occurs during the rewinding process, the electrode's appearance will also change, interfering with the measurement system's visual inspection. The results do not reflect the true detection capabilities. Furthermore, normal production equipment does not have rewinding capabilities, requiring additional costly modifications to implement them. For these reasons, current online visual inspection systems for this type of test are generally difficult to perform conventional measurement system analysis.
[0003] The coating quality of lithium battery electrodes can affect their performance. For example, incomplete electrode coating or damage to the coated area during electrode movement during production can lead to low capacitance or lithium deposition in the battery, shortening its lifespan. Repeated charging of lithium batteries can also lead to capacity drops or fires. Electrode appearance defects in lithium batteries typically include foil leaks, splices, and cracks. Visual inspection equipment installed on slitting machines can identify and mark these electrodes with foil leaks, splices, or cracks. This allows these defective electrodes to be removed from individual rolls during the winding process, preventing them from being released and potentially causing safety issues. Therefore, evaluating the effectiveness of the measurement system is crucial. However, due to the high production speed and continuous operation of slitting machines, each electrode is very long. By the time the last electrode reaches the visual inspection equipment for inspection, the previous electrode has already reached the next station for slitting. This altered shape after slitting prevents repeatable testing using conventional methods. Summary of the Invention
[0004] In response to the above problems, the present invention provides an evaluation method for an online visual measurement system for pole piece appearance, which determines the correct detection rate of the pole piece appearance visual measurement system to improve the system and enhance product quality.
[0005] To achieve the above object, the technical solution adopted by the present invention is: The evaluation method of the electrode appearance online visual measurement system of the present invention comprises the following steps: Compile a test program on the visual inspection equipment, collect images of various unqualified samples as reference images, pre-process the reference images into black and white binarization, obtain the morphological information of the photographed target, and convert the morphological information into a digital signal; The image processing system of the visual inspection equipment calculates these digital signals to extract the digital features of the target and extracts parameters from the captured digital feature areas. Each defect type corresponds to a set of digital feature parameters. When the inspection result can match a certain feature parameter, it can be determined what type of defect the appearance belongs to.
[0006] Start the equipment and run the finished electrode to the inspection area of the visual inspection equipment, and then automatically take pictures of the front and back. When the grayscale value in a certain area of the finished electrode differs from the grayscale value in the normal area by more than the set value, the area is marked as an abnormal area. The system automatically detects and calculates the grayscale, brightness, area, width, length, roundness and other parameter information of the marked area. When all the above information meets a certain preset defect type, the area is marked and fed back to the marking machine for marking. Otherwise, it is judged as a normal electrode. The number of normal electrodes is counted as x, and the number of electrodes marked with defects is y. Dissect the defective electrode, visually inspect the surface appearance of the peeled electrode, and determine whether there are real defects. If there are defects, it means that the actual situation is consistent with the judgment of the detection system and there is no misjudgment. If there are no defects on the electrode, it means that the appearance detection system has misjudged it. The number of misjudged electrodes is counted and recorded as ; Check all the images of normal electrodes to determine whether there are any defects that are judged as qualified by the system. If there are no defects, it means that the system has not missed any inspections. If there are defects, it means that the visual system has missed any inspections. The number of electrodes that have been missed is counted and recorded as y1; Calculation of false positive rate, false positive rate = , represents the ratio of the number of misjudgments to the number of qualified opportunities, missed detection rate = , represents the ratio of the number of missed inspections to the number of unqualified opportunities, Result judgment: If the misjudgment rate is ≤5%, the missed detection rate is ≤2%, the measurement system is qualified, otherwise it is unqualified.
[0007] According to the evaluation method of the electrode appearance visual measurement system, the characteristic content also includes parameters such as area, brightness, shape and grayscale.
[0008] According to the evaluation method of the electrode appearance visual measurement system, step 3) also includes the following steps: after the rolled large electrode roll is loaded onto the equipment loading position, the equipment is started and the equipment is automatically operated. When the electrode runs to the detection area of the visual inspection equipment, the front and back sides are automatically photographed, and each electrode corresponds to a photo. According to the pre-entered slitting width and number of slitting strips, the visual analysis software divides the large electrode into n strip areas along the electrode running direction. When the grayscale value in a strip area differs from the grayscale value of the normal area by more than a set value, the area is marked as an abnormal area. The system automatically detects the grayscale, brightness, area, width, length, roundness and other parameter information of the marked area. When all of the above information meets a certain preset defect type, the mark point turns red and feedback defect NG information is fed back to the marking machine of the corresponding number of strips, and the marking machine marks it. When the grayscale, brightness, area, width, length, roundness and other parameter information detected in the marked area does not match any defect type, the area is still determined to be a normal electrode and no information is fed back to the marking machine.
[0009] According to the evaluation method of the electrode appearance visual measurement system, in step 3), the normal electrodes and defective electrodes are counted. After the electrode passes the appearance inspection, it continues to move forward to the cutter, which cuts the large piece into n strips longitudinally. The n cut strips correspond one to one with the n strips divided by the inspection system. After cutting, all small strips pass through the marking machine. The marking machine that receives the NG information automatically affixes a bad label on the small strip. The marking machine that does not receive the NG information does not mark. The small strips after cutting are each rolled onto the small roll at the corresponding position. This process is continued until all the large rolls are cut. A total of n small rolls are cut. The visual inspection system classifies and saves the OK and NG images of all the electrodes in the roll. The number of OK and NG is recorded as x and y respectively. The number of x+y is the total number of pieces.
[0010] According to the evaluation method of the electrode appearance visual measurement system, the number of misjudged electrodes in step 3) is counted, and one roll is selected from the n rolled small rolls to track to the next winding process. The winding machine will identify the bad labels during the winding process. The electrodes with bad labels are directly rolled up separately without being rewound to avoid waste. After all the winding is completed, all the single roll cores with bad labels are collected and dissected. The surface appearance of the peeled electrode is visually inspected to determine whether there are real defects. If there are, it means that the actual situation is consistent with the judgment of the detection system and there is no misjudgment. If there are no defects in the electrode, it means that the appearance detection system has misjudged it. If there are defects but they are inconsistent with the NG type determined by the system, it is also classified as a misjudgment. The number of misjudged electrodes is counted and recorded as x1. If there is no misjudgment, x1=0.
[0011] According to the evaluation method of the electrode appearance visual measurement system, the number of electrodes missed in step 3) is counted. Since normal electrodes must be scrapped after dissection, and it is not practical to disassemble and inspect all normal cores, a visual inspection of the OK images stored in the inspection system is used instead. A total of x electrodes are inspected to determine whether there are defects that are judged as qualified by the system. If there are no defects, it means that the system has not missed any defects. If there are defects, it means that the visual system has missed any defects. The number of missed electrodes is counted and recorded as y1. If there are no missed electrodes, y1 = 0.
[0012] The advantages of the present invention are that: instead of directly performing repeatability testing on the measurement system, the method relies on the production process and replaces the analysis of multiple repeated results of small samples with a single result of a large sample to calculate the false positive rate and missed detection rate of the measurement equipment, and uses the false positive rate and missed detection rate to measure the effectiveness of the measurement system. This method does not require additional equipment investment and can truly reflect the measurement capability of the detection equipment under actual production speed and vibration conditions.
[0013] This method can be used directly in production, eliminating the need for additional cost to implement evaluation. By utilizing information from the production process for analysis, the sample size is sufficiently large, offering advantages over smaller samples. The data obtained is based on dynamic conditions, making it more realistic and reflecting the true performance of the measurement system.
[0014] The results of this invention provide a basis for actual specification setting. The false positive rate and missed detection rate often affect each other. Reducing one may lead to an increase in the other. Therefore, it is necessary to balance the relationship between the two according to the specific application scenario and requirements. BRIEF DESCRIPTION OF THE DRAWINGS
[0015] Figure 1 Analysis table of the CCD measurement system for the appearance of the positive electrode piece in Example 1; Figure 2 CCD measurement system analysis table of negative electrode appearance. DETAILED DESCRIPTION
[0016] The specific content of the present invention will be further described below: Example 1: Analysis and evaluation process of the CCD measurement system of the positive electrode slitting machine.
[0017] The following steps are included: A testing program is developed on the testing equipment. Images of various unqualified samples are collected in advance as reference images. These reference images are preprocessed to obtain morphological information of the captured target. Based on pixel distribution, brightness, color, and other information, these images are converted into digital signals. The image processing system performs various operations on these signals to extract the target's features. The captured feature areas are analyzed for parameters such as area, brightness, shape, and grayscale. Each defect type corresponds to a set of characteristic parameters, and the defect type can be determined from the results of these operations.
[0018] After loading the rolled large roll of electrode to the loading position of the equipment, enter the product batch information and start the equipment. The equipment will run automatically. When the electrode runs to the detection area of the visual inspection equipment, the front and back sides will be automatically photographed. Since the area of each electrode is large, the inspection system will take photos multiple times and then automatically stitch the multiple photos into one large piece. According to the pre-entered slitting width and number of slitting strips, the visual analysis software will divide a large piece of electrode into 11 strip areas along the running direction of the electrode. When the grayscale value in a certain strip area differs from the grayscale value of the normal area by more than the set value, the area is marked as an abnormal area. The system automatically detects the grayscale, brightness, area, width, length, roundness and other parameter information of the marked area. When all of the above information meets a preset defect type, the mark point turns red and feedback defect NG information is fed back to the PLC. The PLC then feeds back the NG information to the marking machine of the corresponding number of strips. When the grayscale, brightness, area, width, length, roundness and other parameter information detected in the marked area do not match any defect type, the area is still judged as a normal electrode and no information is fed back to the marking machine.
[0019] After visual inspection, the electrode continues forward to the cutter, which cuts the large sheet longitudinally into 11 strips. The 11 strips actually cut correspond one-to-one with the 11 strips detected by the inspection system. After slitting, all the strips pass through the marking machine. The marking machine that receives an "OK" signal automatically affixes a "bad" label to the strip. Marking machines that do not receive an "OK" signal do not mark the strip. The cut strips are each rolled onto the corresponding small roll. This process continues until all the large rolls are cut, resulting in a total of 11 small rolls. The visual inspection system automatically saves the "OK" and "NG" images of all the large strips on the roll. There are 1171 "OK" images, represented by "x", and 94 "NG" images, represented by "y".
[0020] One of the 11 rolled reels is selected and tracked to the next winding process. During the winding process, the winding machine identifies defective labels and winds the electrodes with them directly into individual coils. After winding is complete, the cores of the single reels with defective labels are collected and dissected. The peeled electrode surface is visually inspected to determine if any defects are present. After all reels are disassembled, the number of electrodes without defects is counted as 1, denoted by x1.
[0021] Copy the OK pictures of the large film saved in the third step, and visually check them all. The number of actual defects that are judged as qualified by the system is 1, which is recorded as y1. The data statistics obtained in the above steps are as follows Figure 1 As shown in Table 1.
[0022] Calculation: False positive rate = =0.1%, which represents the ratio of the number of misjudgments to the number of qualified opportunities.
[0023] Missed detection rate = =1.1%, which represents the ratio of the number of missed inspections to the number of unqualified opportunities.
[0024] The statistical process is shown in Figure 1 Table 2.
[0025] Result determination: The equipment standard requirements are that the false positive rate is ≤5%, the missed detection rate is ≤2%, and the measurement system is qualified.
[0026] Example 2: Analysis and evaluation of the CCD measurement system of the negative electrode slitting machine, A testing program is developed on the testing equipment. Images of various unqualified samples are collected in advance as reference images. These reference images are preprocessed to obtain morphological information of the captured target. Based on pixel distribution, brightness, color, and other information, these images are converted into digital signals. The image processing system performs various operations on these signals to extract the target's features. The captured feature areas are analyzed for parameters such as area, brightness, shape, and grayscale. Each defect type corresponds to a set of characteristic parameters, and the defect type can be determined from the results of these operations.
[0027] After loading the rolled large roll of electrode to the loading position of the equipment, enter the product batch information and start the equipment. The equipment will run automatically. When the electrode runs to the detection area of the visual inspection equipment, the front and back sides will be automatically photographed. Since the area of each electrode is large, the inspection system will take photos multiple times and then automatically stitch the multiple photos into one large piece. According to the pre-entered slitting width and number of slitting strips, the visual analysis software will divide a large piece of electrode into 11 strip areas along the running direction of the electrode. When the grayscale value in a certain strip area differs from the grayscale value of the normal area by more than the set value, the area is marked as an abnormal area. The system automatically detects the grayscale, brightness, area, width, length, roundness and other parameter information of the marked area. When all of the above information meets a preset defect type, the mark point turns red and feedback defect NG information is fed back to the PLC. The PLC then feeds back the NG information to the marking machine of the corresponding number of strips. When the grayscale, brightness, area, width, length, roundness and other parameter information detected in the marked area do not match any defect type, the area is still judged as a normal electrode and no information is fed back to the marking machine.
[0028] After visual inspection, the electrode continues forward to the cutter, which cuts the large sheet longitudinally into 11 strips. The 11 strips actually cut correspond one-to-one with the 11 strips detected by the inspection system. After slitting, all strips pass through the marking machine. Marking machines that receive a "good" signal automatically apply a "bad" label to the strips. Marking machines that do not receive a "bad" signal do not mark the strips. The strips are then rolled onto the corresponding small rolls. This process continues until all the large rolls have been cut, resulting in a total of 11 small rolls. The visual inspection system automatically saves images of all the large strips on the roll, both OK and NG. There are 1428 "OK" images, represented by x, and 38 "bad" images, represented by y.
[0029] One of the 11 rolled reels is selected and tracked to the next winding process. During the winding process, the winding machine identifies defective labels and winds the electrodes with them directly into individual coils. After winding is complete, the cores of the single reels with defective labels are collected and dissected. The peeled electrode surfaces are visually inspected to determine if any defects are present. After disassembly, the number of defect-free electrodes is actually two, represented by x1.
[0030] Copy the OK pictures of the large piece saved in the third step, and visually check them all. The number of pictures that actually have defects but are judged as qualified by the system is 1, which is recorded as y1.
[0031] Calculation: False positive rate = =0.1%, which represents the ratio of the number of misjudgments to the number of qualified opportunities.
[0032] Missed detection rate = =2.7%, which represents the ratio of missed inspections to the number of unqualified opportunities. See the calculation statistics for details. Figure 2 Table 2.
[0033] Result determination: The equipment standard requirements are that the false positive rate is ≤5%, the missed detection rate is ≤2%, and the measurement system is qualified.
[0034] Through measurement system analysis, measurement equipment errors can be quantitatively described, guiding adjustments to equipment parameters based on the actual product conditions. Severe defects require stricter controls, with the missed detection rate adjusted to 0%. For minor defects, testing parameters should be adjusted to achieve a balance between false positives and missed detections, taking into account both cost and product performance. Actual implementation requirements for various types of testing equipment vary, and actual needs will ultimately guide the approach.
[0035] When the measurement system analysis results in a failure, a detailed analysis is conducted based on the false positive rate and missed detection rate. These rates are highly dependent on the product being inspected. If a batch of products contains a wide variety of cosmetic defects and the boundary between defective and substandard products is blurred, different people visually inspecting the same sample will produce different evaluations. In this case, the visual inspection equipment's parameter settings must be carefully considered. To ensure that only good products are considered good, stricter settings are required. While this eliminates missed detections, the false positive rate will increase. Similarly, in less demanding applications, the inspection equipment software settings may be relaxed, resulting in a low false positive rate. However, some defective products may be present among the good products, leading to missed detections. Therefore, achieving both a 0% false positive rate and a missed detection rate is difficult. Both rates range from 0 to 1, with lower values indicating stronger detection capabilities. The false positive rate and missed detection rate often interact; reducing one may increase the other. Therefore, a balance between these two rates is crucial based on the specific application scenario and requirements.
Claims
1. A method for evaluating an online visual measurement system for electrode appearance, characterized in that: The following steps are included: 1) Prepare a detection program on the visual inspection equipment, collect images of various unqualified samples as reference images, pre-process the reference images into black and white binarization, obtain the morphological information of the photographed target, and convert the morphological information into a digital signal; 2) The image processing system of the visual inspection equipment calculates these digital signals to extract the digital features of the target and extracts parameters from the captured digital feature areas. Each defect type corresponds to a set of digital feature parameters. When the inspection result can match a certain feature parameter, it can determine the type of defect in the appearance. 3) Start the equipment and run the finished electrode to the inspection area of the visual inspection equipment, and then automatically take pictures of the front and back sides. When the grayscale value in a certain area of the finished electrode differs from the grayscale value in the normal area by more than the set value, the area is marked as an abnormal area. The system automatically detects and calculates the grayscale, brightness, area, width, length, roundness and other parameter information of the marked area. When all the above information meets a certain preset defect type, the area is marked and fed back to the marking machine for marking. Otherwise, it is judged as a normal electrode. The number of normal electrodes is counted as x, and the number of electrodes marked with defects is y; 4) Dissect the defective electrode and visually inspect the surface appearance of the peeled electrode to determine whether there are real defects. If there are defects, it means that the actual situation is consistent with the judgment of the detection system and there is no misjudgment. If there are no defects on the electrode, it means that the appearance detection system has misjudged it. The number of misjudged electrodes is counted and recorded as x1; 5) Check all the images of normal electrodes to determine whether there are any defects that are judged as qualified by the system. If there are no defects, it means that the system has not missed any inspections. If there are defects, it means that the visual system has missed any inspections. The number of electrodes that have been missed is counted and recorded as y1; 6) Calculation of misjudgment rate, Represents the ratio of the number of misjudgments to the number of qualified opportunities, Represents the ratio of the number of missed inspections to the number of unqualified opportunities, Result judgment: If the misjudgment rate is ≤5%, the missed detection rate is ≤2%, the measurement system is qualified, otherwise it is unqualified.
2. The evaluation method of the electrode appearance visual measurement system according to claim 1 is characterized in that the characteristic content Includes parameters area, brightness, shape and grayscale.
3. The evaluation method of the electrode appearance visual measurement system according to claim 1, characterized in that: Step 3) also includes the following steps: after the rolled large roll electrode is loaded onto the equipment loading position, the equipment is started and the equipment runs automatically. When the electrode runs to the detection area of the visual inspection equipment, the front and back sides are automatically photographed. Each electrode corresponds to a photo. According to the pre-input slitting width and the number of slitting strips, the visual analysis software divides a large electrode into n strip areas along the running direction of the electrode. When the grayscale value in a certain strip area differs from the grayscale value in the normal area by more than the set value, the area is marked as an abnormal area. The system automatically detects the grayscale, brightness, area, width, length, roundness and other parameter information of the marked area. When all of the above information meets a certain preset defect type, the marking point turns red and the defect NG information is fed back to the corresponding number of marking machines, and the marking machine marks it. When the grayscale, brightness, area, width, length, roundness and other parameter information detected in the marked area do not match any defect type, the area is still determined to be a normal electrode, and no information is fed back to the marking machine.
4. The evaluation method of the electrode appearance visual measurement system according to claim 3, characterized in that: In step 3), the normal and defective electrodes are counted. After the electrode passes the appearance inspection, it continues to move forward to the cutter. The cutter cuts the large piece into n strips longitudinally. The n strips cut correspond one to one with the n strips divided by the inspection system. After cutting, all small strips pass through the marking machine. The marking machine that receives the NG information automatically affixes a bad label on the small strip. The marking machine that does not receive the NG information does not mark. The small strips after cutting are each rolled onto the small roll at the corresponding position. This process continues until all the large rolls are cut. A total of n small rolls are cut. The visual inspection system classifies and saves the OK and NG images of all the electrodes in the roll. The number of OK and NG is recorded as x and y respectively, and the number of x+y is the total number of pieces.
5. The evaluation method of the electrode appearance visual measurement system according to claim 3, characterized in that: In step 3), the number of misjudged electrodes is counted, and one roll is selected from the n small rolls to be tracked to the next process for winding. The winding machine will identify bad labels during the winding process, and the electrodes with bad labels are directly rolled up separately without being rewound to avoid waste. After all the winding is completed, all the single-roll cores with bad labels are collected and dissected. The surface appearance of the electrode after peeling is visually inspected to determine whether there are real defects. If there are, it means that the actual situation is consistent with the judgment of the detection system and there is no misjudgment. If there is no defect in the electrode, it means that the appearance detection system has misjudged it. If there is a defect but it is inconsistent with the NG type determined by the system, it is also classified as a misjudgment. The number of misjudged pole pieces is counted and recorded as x1. When there is no misjudgment, x1=0.
6. The evaluation method of the electrode appearance visual measurement system according to claim 3, characterized in that: In step 3), the number of electrodes that were missed is counted. Since normal electrodes can only be scrapped after dissection, it is impossible to disassemble all normal cores for inspection. Therefore, the OK image saved in the visual inspection detection system is used instead. A total of x electrodes are checked to determine whether there are defects that are judged as qualified by the system. If there are no defects, it means that the system has no missed inspections. If there are defects, it means that the visual system has missed inspections. The number of missed electrodes is counted and recorded as y1. If there is no missed inspection, y1=0.
Citation Information
Patent Citations
On-line lithium battery pole piece surface defect detection sorting system and method
CN103785627A
Polyimide film defect point detection system and method
CN107300559A
Pole piece die cutting method and system
CN114473249A
Fabric defect intelligent identification method fusing semantic segmentation network and fuzzy decision
CN117474844A