Method for on-line detection of pole piece strip burrs

Abstract

The application discloses a kind of accurate burr online detection methods, comprising the following steps: step one, using burr online detection device to the cutting edge of pole piece material belt is photographed, obtains the end surface image data and side image data of cutting edge;Step two, using FPGA acquisition card to identify end surface image data and side image data, by calculating the vertical distance of exceeding burr vertex to coating edge or foil edge, then screen different kinds of burr defects, and form burr image;By whether detecting foil starting coordinate, end coordinate, and whether the continuous column value of fracture defect exceeds the fracture length threshold value, judge whether metal foil exists fracture defect;Step three, using traditional algorithm and AI model detection to burr image again detection, finally determine the kind of burr.The application can improve burr type detection precision.

Description

Technical Field

[0001] This invention relates to the field of burr detection in electrode strips, and more particularly to an online method for detecting burrs in electrode strips. Background Technology

[0002] The main product structure for detecting electrode burrs currently on the market involves a fixed camera, high-brightness light source, and lens. This structure utilizes the automatic zoom principle of a liquid lens to detect burrs on the electrode end face and longitudinal side surfaces. The disadvantages of this structure include:

[0003] a. Within the working distance of the liquid lens depth of field, and close to the limit working distance, the electrode edge tabs are prone to shake, causing the electrode position to shift, resulting in blurred imaging;

[0004] b. High-brightness light source products have higher costs;

[0005] c. Offline electrode burr detection requires the overall structure to travel the entire length of the electrode burr product. High-brightness light source products have thicker wiring harnesses connecting to the light source controller and larger structural movement volumes, making them unsuitable for detecting products with long burrs.

[0006] To address this, online burr detection vision mechanisms for lithium battery electrode sheets have been publicly disclosed. For example, the invention patent published on December 30, 2022, with publication number CN115541500A, discloses a camera with a telecentric lens that dynamically zooms and takes pictures of a running material strip. The camera is driven by a servo motor to move back and forth in the direction of approaching or moving away from the material strip, enabling real-time zooming and taking pictures of the material strip while it is running at high speed. By judging the distance of the offset of the reference edge of the material strip in real time, the servo motor is then controlled to drive the camera and the telecentric lens to move a corresponding distance, ensuring that the image of the material strip captured is clear.

[0007] However, when detecting burrs, it is also necessary to determine the type of burr and ultimately, based on the requirements, whether it is a type of burr that needs to be removed. Currently, burr type detection only compares the length of the detected burr with a reasonable threshold to determine its category, but this judgment is not accurate and can lead to false positives. Summary of the Invention

[0008] To overcome the above-mentioned shortcomings, the purpose of this invention is to provide an online burr detection method for electrode strips that improves the accuracy of burr type detection.

[0009] To achieve the above objectives, the technical solution adopted by the present invention is: a precise online burr detection method, characterized by comprising the following steps:

[0010] Step 1: Use an online burr detection device to take end face and side view photos of the slit edge of the electrode strip to obtain end face image data and side view image data of the slit edge.

[0011] Step 2: Use the FPGA acquisition card to identify end face image data and side image data. Calculate the vertical distance from the burr vertex to the coating edge or foil edge to filter different types of burr defects and form burr images. Determine whether the metal foil has fracture defects by detecting the foil's start and end coordinates and whether the continuous column values ​​of the fracture defects exceed the fracture length threshold.

[0012] Step 3: Use traditional algorithms and AI models to detect burrs in the image again, and finally determine the type of burrs.

[0013] Preferably, the burrs have type A burrs, type B burrs, and type C burrs;

[0014] The Class A burrs are defined as follows: the burrs extend beyond the coating, and the vertical distance from the burr apex to the coating edge is greater than the Class A burr setting value (e.g., 7 μm).

[0015] The Class B burrs are defined as follows: the burrs do not extend beyond the coating, and the vertical distance from the apex of the burr to the edge of the foil reaches the Class B burr setting value (e.g., 20-35um).

[0016] The C-type burr is defined as follows: the burr does not extend beyond the coating, and the vertical distance from the burr vertex to the base weight edge reaches the C-type burr setting value (e.g., 10-20 μm), where the C-type setting value is less than the B-type setting value.

[0017] Preferably, the algorithm logic for the FPGA acquisition card to identify Class A glitch is as follows:

[0018] Collect the starting coordinates of the foil, the starting coordinates of the coating, the number of consecutive columns of Class A burr defects, and the set value of Class A burrs;

[0019] Compare the starting coordinates of the foil and the starting coordinates of the coating, and compare the number of consecutive columns of Class A burr defects with the set value of Class A burrs;

[0020] If the starting coordinate of the foil is less than the starting coordinate of the coating, and the number of consecutive columns of Class A burr defects is greater than or equal to the Class A burr setting value, then it is determined to be a Class A burr.

[0021] Preferably, the algorithm logic for identifying Class B glitch by the FPGA acquisition card adopts the neighborhood analysis method, the method of which is as follows:

[0022] Collect the starting coordinates, starting average coordinates, ending coordinates, and ending average coordinates of the foil (or substrate).

[0023] And compare the magnitude between |Foil Start Coordinates - Foil Start Average Coordinates| and the Class B Burr Setting Value, or compare the magnitude between |Foil End Coordinates - Foil End Average Coordinates| and the Class B Burr Setting Value;

[0024] If |starting coordinate of foil - average starting coordinate of foil| ≥ the set value of type B burr, or |ending coordinate of foil - average ending coordinate of foil| ≥ the set value of type B burr, then it is a type B burr.

[0025] Preferably, the algorithm logic for the FPGA acquisition card to identify Class C burrs is as follows: if |starting coordinate of foil - average starting coordinate of foil| < Class B burr setting value, or |ending coordinate of foil - average ending coordinate of foil| < Class B burr setting value, then it is a Class C burr.

[0026] Preferably, the logic of the traditional algorithm is as follows:

[0027] First, for the burr images filtered out by the FPGA acquisition card, the upper and lower boundaries of the coating are found by two straight-line measurement edge finding methods.

[0028] Several upper and lower boundary points are obtained by measuring the upper and lower edges of the coating. The upper and lower boundary points are then fitted with straight lines to output two straight lines.

[0029] The upper and lower boundary points are traversed separately, and the perpendicular distance from the boundary point to the line is calculated. If the distance exceeds the threshold, it is classified as a Class A burr.

[0030] Cut out the image between the two straight lines, then perform two more straight line measurements to measure the upper and lower boundaries of the foil, thereby calculating whether there are Class B burrs.

[0031] Preferably, the logic for AI model detection includes:

[0032] The model training steps include collecting end face image data and side face image data, labeling the data, configuring training parameters for model training, evaluating the model accuracy, and if the evaluation meets the standard, exporting the model for deployment and inference. If the evaluation does not meet the standard, analyzing the evaluation results and re-collecting end face image data and side face image data until the evaluation meets the standard.

[0033] The deployment thrust step involves importing the model and parameter configuration into the AI ​​detection model. The AI ​​detection model reads the burr image to be detected. The burr image is generated by the online burr detection device. The GPU performs computation to extract the positioning box, detection score, and detection type. Based on the size of the positioning box, it determines whether the burr is of type A, B, or C. Based on the detection type, it determines whether the image is taken from the end face or the side.

[0034] Preferably, the process further includes step four: using a laser marking machine to mark the tabs corresponding to different types of burrs and metal foil fracture defects. A tab positioning fiber optic sensor is installed at the front end of the tab strip in the running direction before the burr online detection device, and an encoder and laser marking machine are installed at the rear end of the tab strip in the running direction before the burr online detection device. The PLC collects the encoder signal, triggers a high frame rate area array camera to collect images at a fixed distance, and takes approximately 7mm of images at a time. There is overlap between the images taken at different times. The correction sensor corrects the deviation accordingly. The fiber optic sensor detects the tabs to distinguish the start and end of the tabs. The results of multiple images are combined using EA (Each Acquisition) (EA refers to the process of integrating, analyzing, and verifying the multiple independent image results obtained from multiple images). The vision software integrates the detection results and outputs the marking signal with a delay distance.

[0035] Preferably, the online burr detection device includes:

[0036] The base is fixedly installed on one side of the slit edge of the electrode strip;

[0037] The end face inspection camera assembly is driven by a servo motor to move back and forth in the direction of approaching or moving away from the cutting edge of the electrode strip. The servo motor is fixedly mounted on the base. The end face inspection camera assembly is used to take pictures of the side end face of the running electrode strip.

[0038] A correction sensor assembly is fixedly installed on the upper surface of the base. The correction sensor assembly is located on one side of the end face detection camera assembly. The electrode strip passes through the correction sensor assembly. When the correction sensor assembly detects the electrode strip offset, it drives the servo motor through the controller to adjust the distance between the end face detection camera assembly and the side end face of the electrode strip to perform offset compensation.

[0039] A side-view camera assembly is fixedly mounted on the base. The side-view camera assembly is used to take pictures of the lower end face of one side of the running electrode strip.

[0040] Two smoothing components are provided, each of which is mounted on a base along the width of the electrode strip, the electrode strip passing over the two smoothing components.

[0041] Preferably, each of the smoothing components further includes:

[0042] A smoothing base is fixedly installed on one side of a base by bolts, and an adjustment groove is provided on the outer side of the smoothing base;

[0043] A smoothing height adjustment plate is set perpendicular to the electrode strip. The bottom of the smoothing height adjustment plate is provided with a longitudinal height adjustment bolt. The bolt head of the height adjustment bolt is limited in the adjustment groove. The smoothing height adjustment plate is provided with at least two first smoothing longitudinal adjustment waist holes. The smoothing base is provided with at least two first smoothing limiting bolts that pass through the first smoothing longitudinal adjustment waist holes on one side.

[0044] An angle adjustment block is horizontally fixedly installed on the upper outer side of the smoothing height adjustment plate. The angle adjustment block is threaded with two longitudinally distributed angle limiting bolts and fixed by nuts. The smoothing height adjustment plate above the angle adjustment block is provided with multiple smoothing limiting bolt holes.

[0045] The mounting bracket is provided with a number of longitudinal slots for front and rear angle adjustment that are greater than the number of smoothing limit bolt holes. The smoothing limit bolt holes and some of the longitudinal slots for front and rear angle adjustment are locked together by a second smoothing limit bolt. The top of each angle limit bolt abuts against the bottom of the mounting bracket.

[0046] A smoothing roller, which has a tubular structure, has its two ends rotatably mounted on a guide shaft via bearings, and the two ends of the guide shaft are fixedly mounted on a mounting bracket.

[0047] Preferably, the end face detection camera assembly, the correction sensor assembly, and the side detection camera assembly are mounted on the front and rear adjustment base plate. The base is fixedly mounted with a front and rear adjustment block. The front and rear adjustment block is provided with a horizontally distributed front and rear adjustment bolt. The base plate is fixedly mounted with a front and rear connecting block. The front and rear adjustment bolts are threaded through the front and rear connecting block and locked by nuts.

[0048] The front and rear adjustment base plate is equipped with a plurality of front and rear limit bolt holes. The base has a plurality of front and rear adjustment waist holes corresponding to the front and rear limit bolt holes. The front and rear limit bolt holes and the front and rear adjustment waist holes are fixed together by front and rear limit bolts.

[0049] Preferably, the correction sensor assembly includes a housing, a correction sensor, a sensor bracket, and a sensor adjustment bracket. The sensor bracket is fixed to the upper surface of the front and rear adjustment base plate by bolts. The upper surface of the front and rear adjustment base plate is also provided with a guide groove. The bottom of the sensor bracket has a sensor support plate with a width matching the width of the guide groove and a length shorter than the length of the guide groove. The sensor support plate is provided with a sensor front and rear adjustment hole. A sensor front and rear limiting bolt for locking the sensor support plate in the guide groove is installed in the sensor front and rear adjustment hole. The sensor adjustment bracket is provided with a sensor adjustment bolt running in a front and rear direction. The end of the sensor adjustment bolt is threaded to the side wall of the sensor bracket. The correction sensor is fixedly installed on the top of the sensor bracket, and the housing is installed outside the correction sensor.

[0050] Preferably, the end-face detection camera assembly includes:

[0051] End face base, which is fixed on the front and rear adjustment base plate;

[0052] An electric slide table, the fixed part of which is mounted on the end face base, and the sliding part of which is driven by a servo motor to move back and forth on the end face base;

[0053] An end face support plate is fixedly installed on the sliding part of the electric slide table;

[0054] The end face camera, end face lens, and end face light source are mounted on the end face support plate. The bottom of the end face light source is locked and fixed to the end face support plate through a front-to-back oriented nipple and a matching limiting bolt. The end face camera is a high frame rate area array camera.

[0055] Preferably, the side detection camera assembly includes:

[0056] A support base plate is fixed to the front and rear adjustment base plate by means of left and right oriented base plate strip-shaped waist holes and matching limit bolts. The front and rear adjustment base plate is provided with left and right oriented side adjustment bolts, and the side adjustment bolts are threaded through the support base plate.

[0057] The side camera is fixed to the rear of the support base plate by means of a camera strip-shaped waist hole running front and back and matching limit bolts.

[0058] Side-mounted camera, which is mounted on a supporting base plate;

[0059] A prism, which is mounted on a support base plate at the lens end of a side lens;

[0060] A side light source is mounted on a supporting base plate and is horizontally positioned above a prism.

[0061] The beneficial effects of this invention are:

[0062] Firstly, images are acquired using an online burr detection device. Burrs are first screened out by an FPGA, and the burr images in the acquisition card are read. The burr images are then re-detected using traditional algorithms and AI models, ultimately accurately detecting the burrs.

[0063] Secondly, FPGAs can achieve processing speeds of over 500 FPS, completing image processing within the image acquisition interval, making them more adaptable to high-speed production environments than pure AI or traditional algorithms. FPGAs can quickly filter out burr images, replacing the pre-processing work of PCs, which only perform high-precision measurements and burr type detection. Using FPGAs combined with traditional / AI methods can ensure both high-speed production and detection accuracy.

[0064] Third, the position of the electrode strip is sensed by a visual correction sensor, which feeds back the offset value of the electrode strip and guides the sliding part of the electric slide to move by the servo motor. Then, the distance between the end face camera and the electrode strip (this distance is perpendicular to the running direction of the electrode strip) is adjusted to achieve focusing. The follow-up correction method avoids exceeding the depth of field of the lens when passing the electrode tab, which would cause the imaging effect to be blurry and the correction response to be slow, requiring multiple electrode tabs to pass before correction and focusing.

[0065] Fourth, in addition to detecting burrs, this application can also detect fracture defects in electrode strips;

[0066] Fifth, the smoothing component, the correction sensor component, the end face detection camera component, and the side detection camera component are adjustable in multiple directions, which can realize the burr detection of different electrode strips. Attached Figure Description

[0067] Figure 1 This is a flowchart of steps one through three of this embodiment;

[0068] Figure 2 This embodiment defines burrs of types A, B, and C.

[0069] Figure 3 This is the algorithm logic for determining foil breakage in this embodiment;

[0070] Figure 4 This is the algorithm logic for determining type A burrs in this embodiment;

[0071] Figure 5 This is the algorithm logic for determining type B burrs in this embodiment;

[0072] Figure 6This is an interface diagram of the repeated measurement of the upper cutting plane in this embodiment; the plane is the cutting action area parallel to the upper and lower surfaces of the electrode;

[0073] Figure 7 This is an interface diagram of the upper slit end face for repeated measurement in this embodiment; the end face is the slit side area perpendicular to the upper and lower surfaces of the electrode.

[0074] Figure 8 This is an interface diagram of the cutting plane under repeated measurements in this embodiment;

[0075] Figure 9 This is an interface diagram of the slitting end face under repeated measurements in this embodiment;

[0076] Figure 10 This is a flowchart of steps one through four of this embodiment;

[0077] Figure 11 This is a flowchart of the model training process in this embodiment;

[0078] Figure 12 Flowchart for deploying the AI ​​model inference in this embodiment;

[0079] Figure 13 This is a perspective view of the online burr detection device in this embodiment;

[0080] Figure 14 This is a perspective view of the smoothing component in this embodiment;

[0081] Figure 15 This is an exploded view of the smoothing component in this embodiment;

[0082] Figure 16 This is an exploded view of the front and rear adjustment base plates and the base in this embodiment;

[0083] Figure 17 This is a perspective view of the correction sensor assembly in this embodiment;

[0084] Figure 18 This is an exploded view of the correction sensor assembly in this embodiment;

[0085] Figure 19 This is a perspective view of the end-face detection camera assembly in this embodiment;

[0086] Figure 20 This is a perspective view of the side-detection camera assembly in this embodiment. Detailed Implementation

[0087] The preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby providing a clearer and more explicit definition of the scope of protection of the present invention.

[0088] See Figures 1 to 20 As shown in the figure, this embodiment discloses a precise online burr detection method, including the following steps:

[0089] Step 1: Use an online burr detection device to take end face and side face photos of the slit edge of the electrode strip (the slit edge is the side without the electrode tabs) to obtain end face image data and side face image data of the slit edge.

[0090] Step 2: Use the FPGA acquisition card to identify end face image data and side image data. Calculate the vertical distance from the burr vertex to the coating edge or foil edge to filter different types of burr defects and form burr images. Determine if the metal foil has a fracture defect by detecting the foil's start and end coordinates and whether the continuous column values ​​of the fracture defect exceed the fracture length threshold. If the foil's start and end coordinates are not detected and the continuous column values ​​of the fracture defect exceed the fracture length threshold, the metal foil is considered to have a fracture defect.

[0091] Step 3: Use traditional algorithms and AI models to detect burrs in the image again, and finally determine the type of burrs.

[0092] Burrs are categorized into Class A burrs, Class B burrs, and Class C burrs.

[0093] Type A burrs are defined as burrs extending beyond the coating area, with the vertical distance from the burr vertex to the coating edge exceeding the Type A burr setting value (e.g., 7μm). The coating refers to a continuous area near the coating area within the uncoated metal foil substrate region of the electrode strip. The size and position of this continuous area perfectly match the coverage of the separator during battery assembly, used to isolate the metal edges of the positive and negative electrodes. The coating is the pixel coordinate of the inner edge of the separator closest to the center of the electrode on the electrode image. This set coordinate value is pre-calibrated in the FPGA system using a defect-free electrode + separator bonding sample.

[0094] Type B burrs are defined as burrs that do not extend beyond the coating and whose vertical distance from the burr tip to the foil edge reaches the set value for Type B burrs (e.g., 20-35µm).

[0095] Type C burrs are defined as burrs that do not extend beyond the coating layer, and the vertical distance from the burr vertex to the base edge reaches the Type C burr setting value (e.g., 10-20µm). The Type C setting value is less than the Type B setting value.

[0096] The algorithm logic for the FPGA acquisition card to identify Class A glitch is as follows:

[0097] The data collection process includes collecting the starting coordinates of the foil, the starting coordinates of the coating, the number of consecutive columns of Class A burr defects, and the set value for Class A burrs. The number of consecutive columns refers to the number of times the same defect is detected in multiple consecutive longitudinal scan columns. Because anomalies in the coordinates of a single column may be image noise or tiny debris on the electrode surface, only when this feature appears in multiple consecutive columns does it indicate that the protrusion is continuous and has a certain length, rather than random interference. In addition, the coating is only applied to the middle area of ​​the foil, leaving a blank area on both sides of the foil without coating. This blank area is used for subsequent electrode tab fabrication.

[0098] Compare the starting coordinates of the foil and the starting coordinates of the coating, and compare the number of consecutive columns of Class A burr defects with the set value of Class A burrs;

[0099] If the starting coordinate of the foil is less than the starting coordinate of the coating, and the number of consecutive columns of Class A burr defects is greater than or equal to the Class A burr setting value, then it is determined to be a Class A burr.

[0100] The algorithm logic for identifying Class B glitch in the FPGA acquisition card adopts the neighborhood analysis method, which is as follows:

[0101] Collect the starting coordinates of the foil, the starting average coordinates of the foil (the starting average can be understood as the upper edge of the foil), the ending coordinates of the foil, and the ending average coordinates of the foil (the ending average can be understood as the lower edge of the foil).

[0102] And compare the magnitude between |Foil Start Coordinates - Foil Start Average Coordinates| and the Class B Burr Setting Value, or compare the magnitude between |Foil End Coordinates - Foil End Average Coordinates| and the Class B Burr Setting Value;

[0103] If |starting coordinate of foil - average starting coordinate of foil| ≥ the set value of type B burr, or |ending coordinate of foil - average ending coordinate of foil| ≥ the set value of type B burr, then it is a type B burr.

[0104] The algorithm logic for the FPGA acquisition card to identify Class C burrs is as follows: if |starting coordinates of foil - average starting coordinates of foil| < Class B burr setting value, or |ending coordinates of foil - average ending coordinates of foil| < Class B burr setting value, then it is a Class C burr.

[0105] The logic of the traditional algorithm in this embodiment is as follows:

[0106] First, for the burr images filtered by the FPGA acquisition card, the upper and lower boundaries of the coating are found by two straight-line measurements. The measurement principle is as follows: for the upper straight line, the change point from dark to light in the image is found by using a small measurement box perpendicular to the straight line with the maximum number of image columns. In order to avoid noise interference, the image inside the caliper needs to be smoothed and filtered. For the lower straight line, the change point from dark to light in the image is found from the opposite direction of the image.

[0107] Several upper and lower boundary points are obtained by measuring the upper and lower edges of the coating. The upper and lower boundary points are then fitted with straight lines to output two straight lines.

[0108] The upper and lower boundary points are traversed separately (e.g., if the measurement box corresponding to the maximum number of image columns is 1000 columns, there are 1000 measurement boxes and 1000 boundary points). The vertical distance from the boundary point to the line is calculated. If it exceeds the threshold, it is a type A spur.

[0109] Cut out the image between the two straight lines, then perform two more straight line measurements to determine the upper and lower boundaries of the foil, thereby calculating the type of burr.

[0110] The algorithm flow for burr detection using AI model detection is as follows: 1. Locate the burr using AI detection algorithm; 2. Locate the burr using machine vision image algorithm simultaneously; 3. Verify the location using AI and traditional machine vision algorithms, and measure the size of the burr after determining its position.

[0111] The specific logic of AI model detection includes:

[0112] The model training steps include collecting end face image data and side face image data, labeling the data, configuring training parameters for model training, evaluating the model accuracy, and if the evaluation meets the standard, exporting the model for deployment and inference. If the evaluation does not meet the standard, analyzing the evaluation results and re-collecting end face image data and side face image data until the evaluation meets the standard.

[0113] The deployment of thrust involves importing the model and parameter configuration into the AI ​​detection model. The AI ​​detection model reads the burr image to be detected. The burr image is generated by the online burr detection device. The GPU performs computation to extract the positioning box, detection score, and detection type. Based on the size of the positioning box, it determines whether the burr is of type A, B, or C. Based on the detection type, it determines whether the image was taken from the end face or the side.

[0114] The process also includes step four, which involves using a laser marking machine to mark the tabs corresponding to different types of burrs and metal foil fracture defects. A tab positioning fiber optic sensor is installed at the front end of the tab strip in the running direction before the online burr detection device, and an encoder and laser marking machine are installed at the rear end of the tab strip in the running direction before the online burr detection device. The PLC collects the encoder signal, triggers a high frame rate area array camera to collect images at a fixed distance, and takes approximately 7mm of images at a time. There is overlap between the images taken at different times. The correction sensor corrects the deviation accordingly. The fiber optic sensor detects the tabs to distinguish the start and end of the tabs. The results of multiple images are combined using the EA (Eye Analysis) method. The multiple independent image results obtained from multiple shots are integrated, analyzed, and verified. The vision software integrates the detection results and outputs the marking signal with a delay distance.

[0115] The online burr detection device includes:

[0116] The base 100 is fixedly installed on one side of the slitting edge of the electrode strip;

[0117] The end face inspection camera assembly 200 is driven by a servo motor 210 to move back and forth in the direction of approaching or moving away from the cutting edge of the electrode strip. The servo motor 210 is fixedly mounted on the base 100. The end face inspection camera assembly 200 is used to take pictures of the side end face of the running electrode strip.

[0118] The correction sensor assembly 300 is fixedly installed on the upper end face of the base 100. The correction sensor assembly 300 is located on one side of the end face detection camera assembly 200. The electrode strip passes through the correction sensor assembly 300. When the correction sensor assembly 300 detects the deviation of the electrode strip, it drives the servo motor 210 through the controller to adjust the distance between the end face detection camera assembly 200 and the side end face of the electrode strip to perform deviation compensation.

[0119] The side detection camera assembly 400 is fixedly mounted on the base 100. The side detection camera assembly 400 is used to take pictures of the lower end face of one side of the running electrode strip.

[0120] Two smoothing components 500 are mounted on the base 100 along the width direction of the electrode strip, and the electrode strip passes over the two smoothing components 500.

[0121] Each smoothing component 500 also includes:

[0122] The smoothing base 510 is fixedly installed on one side of the base 100 by bolts, and the smoothing base 510 has an adjustment groove 511 on its outer side;

[0123] A smoothing height adjustment plate 520 is set perpendicular to the electrode strip. A longitudinal height adjustment bolt 512 is provided at the bottom of the smoothing height adjustment plate 520. The bolt head of the height adjustment bolt 512 is limited in the adjustment groove 511. At least two first smoothing longitudinal adjustment waist holes 521 are provided on the smoothing height adjustment plate 520. At least two first smoothing limiting bolts are provided on one side of the smoothing base 510, which pass through the first smoothing longitudinal adjustment waist holes 521 respectively. When it is necessary to adjust the height of the smoothing height adjustment plate 520, the first smoothing limiting bolts are loosened, the height adjustment bolts 512 are adjusted, and the first smoothing longitudinal adjustment waist holes 521 of the smoothing height adjustment plate 520 are guided by the first smoothing limiting bolts for lifting and lowering adjustment until the appropriate position is reached. Then, the first smoothing limiting bolts are tightened.

[0124] Angle adjustment block 530 is horizontally fixed on the outer side of the upper end of the smoothing height adjustment plate 520. Two longitudinally distributed angle limiting bolts 531 are threaded on the angle adjustment block 530 and fixed by nuts. Multiple smoothing limiting bolt holes 522 are provided on the smoothing height adjustment plate 520 above the angle adjustment block 530.

[0125] Mounting frame 540 has a number of longitudinal slots 541 for adjusting the front and rear angles, which are greater than the number of smoothing limit bolt holes 522. The smoothing limit bolt holes 522 and some of the longitudinal slots 541 for adjusting the front and rear angles are locked together by a second smoothing limit bolt. The top of each angle limit bolt 531 abuts against the bottom of the mounting frame 540. When the angle of the mounting frame 540 needs to be adjusted, the second smoothing limit bolt is loosened, and the height of the two angle limit bolts 531 is adjusted to achieve the angle adjustment of the mounting frame 540. After the angle reaches the appropriate position, the second smoothing limit bolt is tightened. When the left and right position of the mounting frame 540 needs to be adjusted, it is locked together with some of the longitudinal slots 541 for adjusting the front and rear angles. In this embodiment, the front and rear directions are perpendicular to the direction of the conveyor belt.

[0126] The smoothing roller 550 has a tubular structure. Both ends of the smoothing roller 550 are rotatably mounted on the guide shaft 552 via bearings 551. Both ends of the guide shaft 552 are fixedly mounted on the mounting bracket 540.

[0127] The smoothing roller 550 uses a carbon fiber material roller and an internal bearing structure to rotate, which can reduce the damage to the electrode strip surface during bonding and tensioning. Two sets of carbon fiber roller assemblies are used to bond with the moving electrode strip surface to ensure the tension of the electrode strip in the detection area, and to ensure that the detection of end face and side electrode burrs is not affected by the tightness of the strip.

[0128] The end face inspection camera assembly 200, the correction sensor assembly 300, and the side inspection camera assembly 400 are mounted on the front and rear adjustment base plate 600. The base 100 has a front and rear adjustment block 110 fixedly mounted on its bottom. The front and rear adjustment block 110 is provided with horizontally distributed front and rear adjustment bolts 120. The front and rear adjustment base plate 600 has a front and rear connecting block 610 fixedly mounted on its bottom. The front and rear adjustment bolts 120 are threaded through the front and rear connecting block 610 and are locked in place by nuts. By adjusting the front and rear position of the front and rear adjustment base plate 600, the distance between the end face inspection camera assembly 200, the correction sensor assembly 300, the side inspection camera assembly 400 and the electrode strip can be adjusted.

[0129] The front-to-back adjustment base plate 600 is equipped with multiple front-to-back limiting bolt holes 620. The base 100 has multiple front-to-back adjustment waist holes 130 on both sides corresponding to the front-to-back limiting bolt holes 620. The front-to-back limiting bolt holes 620 and the front-to-back adjustment waist holes 130 are fixed together by front-to-back limiting bolts. By loosening these limiting bolts, the front-to-back limiting bolt holes 620 are positioned corresponding to the front-to-back adjustment waist holes 130, thereby adjusting the front-to-back position.

[0130] The correction sensor assembly 300 includes a housing 310, a correction sensor 320, a sensor bracket 330, and a sensor adjustment bracket 340. The sensor bracket 330 is fixed to the upper surface of the front and rear adjustment base plate 600 by bolts. The upper surface of the front and rear adjustment base plate 600 is also provided with a guide groove 630. The bottom of the sensor bracket 330 has a sensor support plate 331 with a width matching the width of the guide groove 630 and a length shorter than the length of the guide groove 630. The sensor support plate 331 is provided with a sensor front and rear adjustment hole 332. A sensor front and rear limiting bolt for locking the sensor support plate 331 in the guide groove 630 is installed in the sensor front and rear adjustment hole 332. The sensor adjustment bracket 340 is provided with a sensor adjustment bolt 341 with a front and rear direction. The end of the sensor adjustment bolt 341 is threaded to the side wall of the sensor bracket 330. The correction sensor 320 is fixedly installed on the top of the sensor bracket 330, and the housing 310 is installed on the outside of the correction sensor 320. When it is necessary to adjust the front and rear position of the sensor support plate 331, loosen the front and rear limit bolts of the sensor in the front and rear adjustment hole 332, but do not disengage it from the guide groove 630, and rotate the sensor adjustment bolt 341 to adjust the front and rear distance of the sensor bracket 330 in the guide groove 630.

[0131] The end-face inspection camera assembly 200 includes:

[0132] The end face base 220 is fixed on the front and rear adjustment base plate 600;

[0133] The electric slide table 230 has a fixed part 231 mounted on the end face base 220, and a sliding part 232 of the electric slide table 230 is driven by the servo motor 210 to move back and forth on the end face base 220.

[0134] The end support plate 240 is fixedly installed on the sliding part 232 of the electric slide table 230;

[0135] The end face camera 250, end face lens 260, and end face light source 270 are mounted on the end face support plate 240. The bottom of the end face light source 270 is locked and fixed to the end face support plate 240 through the front and rear facing end face strip waist hole 271 and the matching limit bolt. In this embodiment, the front and rear position of the end face light source 270 on the end face support plate 240 is adjustable. The end face camera 250 is a high frame rate area array camera.

[0136] The side detection camera assembly 400 includes:

[0137] The support base plate 410 is fixed to the front and rear adjustment base plate 600 by means of the bottom plate strip waist hole 411 in the left and right direction (the running direction of the electrode material strip) and the matching limit bolt. The front and rear adjustment base plate 600 is provided with side adjustment bolts 640 in the left and right direction, and the side adjustment bolts 640 are threaded through the support base plate 410.

[0138] The side camera 420 is fixed to the rear of the support base plate 410 by a camera strip-shaped waist hole 421 running in the front and back direction and a matching limit bolt.

[0139] Side lens 430 and side camera 420 are mounted on support base plate 410;

[0140] A prism is mounted on a support base plate 410 at the lens end of the side lens 430.

[0141] The side light source 440 is mounted on the support base plate 410 and is horizontally positioned above the prism.

[0142] The above embodiments are only for illustrating the technical concept and features of the present invention. Their purpose is to enable those skilled in the art to understand the content of the present invention and implement it. They should not be used to limit the scope of protection of the present invention. All equivalent changes or modifications made in accordance with the spirit and essence of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A precise online burr detection method, characterized in that: Includes the following steps: Step 1: Use an online burr detection device to take end face and side view photos of the slit edge of the electrode strip to obtain end face image data and side view image data of the slit edge. Step 2: Use the FPGA acquisition card to identify end face image data and side image data. Calculate the vertical distance from the burr vertex to the coating edge or foil edge to filter different types of burr defects and form burr images. Determine whether the metal foil has fracture defects by detecting the foil's start and end coordinates and whether the continuous column values ​​of the fracture defects exceed the fracture length threshold. Step 3: Use traditional algorithms and AI models to detect burrs in the image again, and finally determine the type of burrs.

2. The precise online burr detection method according to claim 1, characterized in that: The burrs are classified into Class A burrs, Class B burrs, and Class C burrs; The Class A burrs are defined as: burrs extending beyond the coating, with the vertical distance from the burr vertex to the coating edge greater than the Class A burr setting value; The B-type burr is defined as follows: the burr does not extend beyond the coating, and the vertical distance from the apex of the burr to the edge of the foil reaches the set value for the B-type burr. The C-type burr is defined as follows: the burr does not extend beyond the coating, and the vertical distance from the burr vertex to the base weight edge reaches the C-type burr setting value, wherein the C-type setting value is less than the B-type setting value.

3. The precise online burr detection method according to claim 2, characterized in that: The algorithm logic for the FPGA acquisition card to identify Class A glitch is as follows: Collect the starting coordinates of the foil, the starting coordinates of the coating, the number of consecutive columns of Class A burr defects, and the set value of Class A burrs; Compare the starting coordinates of the foil and the starting coordinates of the coating, and compare the number of consecutive columns of Class A burr defects with the set value of Class A burrs; If the starting coordinate of the foil is less than the starting coordinate of the coating, and the number of consecutive columns of Class A burr defects is greater than or equal to the Class A burr setting value, then it is determined to be a Class A burr.

4. The precise online burr detection method according to claim 2, characterized in that: The algorithm logic for identifying Class B glitch in the FPGA acquisition card adopts the neighborhood analysis method, which is as follows: Collect the starting coordinates, starting average coordinates, ending coordinates, and ending average coordinates of the foil. And compare the magnitude between |Foil Start Coordinates - Foil Start Average Coordinates| and the Class B Burr Setting Value, or compare the magnitude between |Foil End Coordinates - Foil End Average Coordinates| and the Class B Burr Setting Value; If |starting coordinate of foil - average starting coordinate of foil| ≥ the set value of type B burr, or |ending coordinate of foil - average ending coordinate of foil| ≥ the set value of type B burr, then it is a type B burr.

5. The precise online burr detection method according to claim 4, characterized in that: The algorithm logic for the FPGA acquisition card to identify Class C burrs is as follows: if |starting coordinate of foil - average starting coordinate of foil| < Class B burr setting value, or |ending coordinate of foil - average ending coordinate of foil| < Class B burr setting value, then it is a Class C burr.

6. The precise online burr detection method according to claim 1, characterized in that: The logic of the traditional algorithm is as follows: First, for the burr images filtered out by the FPGA acquisition card, the upper and lower boundaries of the coating are found by two straight-line measurement edge finding methods. Several upper and lower boundary points are obtained by measuring the upper and lower edges of the coating. The upper and lower boundary points are then fitted with straight lines to output two straight lines. The upper and lower boundary points are traversed separately, and the perpendicular distance from the boundary point to the line is calculated. If the distance exceeds the threshold, it is classified as a Class A burr. Cut out the image between the two straight lines, then perform two more straight line measurements to determine the upper and lower boundaries of the foil, thereby calculating the type of burr.

7. The precise online burr detection method according to claim 1, characterized in that: The logic for AI model detection includes: The model training steps include collecting end face image data and side face image data, labeling the data, configuring training parameters for model training, evaluating the model accuracy, and if the evaluation meets the standard, exporting the model for deployment and inference. If the evaluation does not meet the standard, analyzing the evaluation results and re-collecting end face image data and side face image data until the evaluation meets the standard. The deployment of thrust involves importing the model and parameter configuration into the AI ​​detection model. The AI ​​detection model reads the burr image to be detected. The burr image is generated by the online burr detection device. The GPU performs computation to extract the positioning box, detection score, and detection type. Based on the size of the positioning box, it determines whether the burr is of type A, B, or C. Based on the detection type, it determines whether the image was taken from the end face or the side.

8. The precise online burr detection method according to claim 1, characterized in that: The process also includes step four, which involves using a laser marking machine to mark the tabs corresponding to different types of burrs and metal foil fracture defects. A tab positioning fiber optic sensor is installed at the front end of the tab strip in the running direction before the online burr detection device, and an encoder and laser marking machine are installed at the rear end of the tab strip in the running direction before the online burr detection device. The PLC collects the encoder signal, triggers a high frame rate area array camera to collect images at a fixed distance, and takes approximately 7mm of images at a time. There is overlap between the images taken at different times. The correction sensor corrects the deviation accordingly. The fiber optic sensor detects the tabs to distinguish the start and end of the tabs. The results of multiple images are combined using the EA (Expert Analyzer) method. The vision software integrates the detection results and outputs the marking signal with a delay distance.

9. The precise online burr detection method according to claim 1, characterized in that: The online burr detection device includes: The base (100) is fixedly installed on one side of the slit edge of the electrode strip; The end face inspection camera assembly (200) is driven by a servo motor (210) to move back and forth in the direction of approaching or moving away from the cutting edge of the electrode strip. The servo motor (210) is fixedly mounted on the base (100). The end face inspection camera assembly (200) is used to take pictures of the side end face of the running electrode strip. A correction sensor assembly (300) is fixedly installed on the upper end face of the base (100). The correction sensor assembly (300) is located on one side of the end face detection camera assembly (200). The electrode strip passes through the correction sensor assembly (300). When the correction sensor assembly (300) detects the electrode strip offset, it drives the servo motor (210) through the controller to adjust the distance between the end face detection camera assembly (200) and the side end face of the electrode strip to perform offset compensation. A side detection camera assembly (400) is fixedly mounted on the base (100) and is used to take pictures of the lower end face of one side of the electrode strip in operation. Two smoothing components (500) are mounted on a base (100) along the width direction of the electrode strip, the electrode strip passing over the two smoothing components (500).

10. The precise online burr detection method according to claim 9, characterized in that: Each of the smoothing components (500) also includes: A smoothing base (510) is fixedly installed on one side of the base (100) by bolts, and an adjustment groove (511) is provided on the outer side of the smoothing base (510). A smoothing height adjustment plate (520) is set perpendicular to the electrode strip. A longitudinal height adjustment bolt (512) is provided at the bottom of the smoothing height adjustment plate (520). The bolt head of the height adjustment bolt (512) is limited in the adjustment groove (511). At least two first smoothing longitudinal adjustment waist holes (521) are provided on the smoothing height adjustment plate (520). At least two first smoothing limiting bolts are provided on one side of the smoothing base (510) respectively passing through the first smoothing longitudinal adjustment waist holes (521). An angle adjustment block (530) is horizontally fixedly installed on the upper outer side of the smoothing height adjustment plate (520). The angle adjustment block (530) is threaded with two longitudinally distributed angle limiting bolts (531) and fixed by nuts. The smoothing height adjustment plate (520) above the angle adjustment block (530) is provided with multiple smoothing limiting bolt holes (522). Mounting bracket (540) is provided with a number of longitudinal waist holes (541) for front and rear angle adjustment that are greater than the number of smoothing limit bolt holes (522). The smoothing limit bolt holes (522) and a portion of the longitudinal waist holes (541) for front and rear angle adjustment are fixed by a second smoothing limit bolt. The top of each angle limit bolt (531) abuts against the bottom of the mounting bracket (540). The smoothing roller (550) has a tubular structure. Both ends of the smoothing roller (550) are rotatably mounted on the guide shaft (552) via bearings (551). Both ends of the guide shaft (552) are fixedly mounted on the mounting bracket (540).

11. The precise online burr detection method according to claim 9, characterized in that: The end face detection camera assembly (200), the correction sensor assembly (300), and the side detection camera assembly (400) are mounted on the front and rear adjustment base plate (600). The base (100) is fixedly mounted with a front and rear adjustment block (110). The front and rear adjustment block (110) is provided with a horizontally distributed front and rear adjustment bolt (120). The front and rear adjustment base plate (600) is fixedly mounted with a front and rear connecting block (610). The front and rear adjustment bolt (120) is threaded through the front and rear connecting block (610) and fixed by a nut. The front and rear adjustment base plate (600) is equipped with a plurality of front and rear limit bolt holes (620). The base (100) has a plurality of front and rear adjustment waist holes (130) on both sides corresponding to the front and rear limit bolt holes (620). The front and rear limit bolt holes (620) and the front and rear adjustment waist holes (130) are fixed together by front and rear limit bolts.

12. The precise online burr detection method according to claim 11, characterized in that: The correction sensor assembly (300) includes a housing (310), a correction sensor (320), a sensor bracket (330), and a sensor adjustment bracket (340). The sensor bracket (330) is fixed to the upper surface of the front and rear adjustment base plate (600) by bolts. The upper surface of the front and rear adjustment base plate (600) is also provided with a guide groove (630). The bottom of the sensor bracket (330) has a sensor support plate (331) with a width matching the width of the guide groove (630) and a length shorter than the length of the guide groove (630). The sensor support plate (331) is provided with... The sensor has a front and rear adjustment hole (332), and a sensor front and rear limiting bolt is installed in the sensor front and rear adjustment hole (332) to lock the sensor support plate (331) in the guide groove (630). The sensor adjustment bracket (340) is provided with a sensor adjustment bolt (341) with a front and rear direction. The end of the sensor adjustment bolt (341) is threaded to the side wall of the sensor bracket (330). The correction sensor (320) is fixedly installed on the top of the sensor bracket (330). The housing (310) is installed on the outside of the correction sensor (320).

13. The precise online burr detection method according to claim 11, characterized in that: The end-face detection camera assembly (200) includes: An end face base (220) is fixed on the front and rear adjustment base plate (600); An electric slide (230) has a fixed part (231) mounted on an end face base (220), and a sliding part (232) of the electric slide (230) is driven by a servo motor (210) to move back and forth on the end face base (220). An end face support plate (240) is fixedly installed on the sliding part (232) of the electric slide table (230); The end face camera (250), end face lens (260), and end face light source (270) are installed on the end face support plate (240). The bottom of the end face light source (270) is locked and fixed to the end face support plate (240) through the front and rear oriented end face nipple (271) and the matching limit bolt. The end face camera (250) is a high frame rate area array camera.

14. The precise online burr detection method according to claim 11, characterized in that: The side detection camera assembly (400) includes: A support base plate (410) is fixed to the front and rear adjustment base plate (600) by means of a left-right oriented base plate strip waist hole (411) and a matching limiting bolt. The front and rear adjustment base plate (600) is provided with a left-right oriented side adjustment bolt (640), and the side adjustment bolt (640) is threaded through the support base plate (410). The side camera (420) is fixed to the tail of the support base plate (410) by a camera strip waist hole (421) running in the front and back direction and a matching limiting bolt. Side lens (430), the side camera (420) is mounted on the support base plate (410); A prism, which is mounted on a support base plate (410) at the lens end of a side lens (430); A side light source (440) is mounted on a support base plate (410) and is horizontally positioned above a prism.

Images