A lithium battery assembly and detection system for power tools

Abstract

The present application relates to lithium battery assembly and detection technical field, specifically to a kind of lithium battery assembly and detection system for electric tool, including connecting piece feeding module, for automatically supplying connecting piece;Connecting piece positioning placement module is used to place connecting piece in the positive and negative electrode lead-out end surface of battery group;Connecting piece welding module is used to weld connecting piece with battery electrode, and the welding parameter of the device in the connecting piece welding module is adjustable;Welding quality on-line detection module includes image acquisition unit and image processing and analysis unit;The image processing and analysis unit is configured as: before welding, according to the two-dimensional image and three-dimensional point cloud data, determine the actual position offset of connecting piece relative to battery pole, and guide the connecting piece welding module to carry out adaptive positioning compensation.

Description

Technical Field

[0001] This invention relates to the field of lithium battery assembly and testing technology, specifically to a lithium battery assembly and testing system for power tools. Background Technology

[0002] In existing manufacturing processes, lithium-ion batteries for power tools are typically assembled from multiple individual cells welded together using series and parallel connecting tabs. These connecting tabs (usually nickel or copper tabs) must be precisely placed on the positive and negative terminals of the cell assembly and form a reliable electrical connection with the cell terminals using methods such as laser welding. The welding quality directly affects the battery pack's conductivity, mechanical strength, and cycle life, making it a critical step in the battery assembly process.

[0003] In existing lithium battery assembly systems, the feeding, positioning, and welding of connector pieces have been automated to a certain extent. A typical assembly process is as follows: the connector feeding module delivers the connector piece to a predetermined position, the positioning module picks up the connector piece and places it on the surface of the cell terminal, and then the laser welding device completes the welding according to a preset trajectory. However, the existing system still has the following main shortcomings.

[0004] Existing systems often rely on mechanical positioning fixtures to ensure the relative position of the connecting piece and the pole, which makes it difficult to compensate for the dimensional deviation of the connecting piece, the machining tolerance of the pole, and random offsets during placement. There is often a deviation between the welding focus and the ideal position, which can easily lead to welding misalignment or incomplete welding.

[0005] Therefore, this invention proposes a lithium battery assembly and testing system for power tools. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a lithium battery assembly and testing system for power tools, including a connecting piece feeding module, a connecting piece positioning and placement module, a connecting piece welding module, an online welding quality detection module, and a rework robot.

[0007] Connecting piece feeding module, used to automatically supply connecting pieces;

[0008] The connector positioning and placement module is used to place the connector on the positive and negative lead-out end faces of the battery cell assembly;

[0009] A connecting piece welding module is used to weld connecting pieces to the cell electrodes, and the welding parameters of the device in the connecting piece welding module are adjustable;

[0010] The online welding quality inspection module includes an image acquisition unit and an image processing and analysis unit;

[0011] The image acquisition unit is equipped with a polarizer, a narrowband filter and a structured light projector to acquire clear two-dimensional images and three-dimensional point cloud data of the pad area before and after welding.

[0012] The image processing and analysis unit is configured as follows:

[0013] Before welding, the actual position offset of the connecting piece relative to the cell electrode is determined based on the two-dimensional image and three-dimensional point cloud data, and the connecting piece welding module is guided to perform adaptive positioning compensation.

[0014] The three-dimensional point cloud data is also used to determine whether there is an oxide layer or foreign matter on the surface of the battery cell electrode before welding.

[0015] After welding, the weld morphology is identified based on the three-dimensional point cloud data to determine whether there are defects such as incomplete welding, missing welding, porosity, or welding deviation. When it is determined that there are repairable defects, the spatial coordinates of the defective weld point are calculated to guide the rework robot to perform fixed-point repair welding. After the repair welding is completed, the image acquisition unit is controlled to acquire three-dimensional point cloud data again and perform re-inspection. The rework robot is used to perform fixed-point repair welding on weld points with repairable defects.

[0016] Preferably, the image acquisition unit acquires clear two-dimensional images and three-dimensional point cloud data of the pad area before and after welding, specifically as follows:

[0017] S201. Control the structured light projector to project structured light with a preset coding pattern onto the pad area. The preset coding pattern is a stripe pattern or a pseudo-random speckle pattern.

[0018] S202. After the structured light is projected onto the surface of the connecting piece and the battery cell electrode, it deforms due to the change in surface height. The camera in the image acquisition unit is controlled to synchronously acquire the structured light image that has been deformed after being modulated by the surface of the pad area.

[0019] S203. The image processing and analysis unit compares the deformed structured light image with the pre-calibrated reference plane image to obtain the pixel displacement of each pixel and converts the pixel displacement into a physical offset using the pixel size. Then, using the triangulation principle, the three-dimensional coordinates of each point in the pad area are calculated based on the physical offset to generate the three-dimensional point cloud data.

[0020] Preferably, the image acquisition unit further includes a polarizer and a narrowband filter;

[0021] The polarizer is mounted on the front end of the camera lens, and its polarization direction is configured to be orthogonal to the main polarization direction of the light reflected by the mirror surface of the battery cell electrode or connecting piece, so as to suppress the brightness saturation of the highly reflective area.

[0022] The narrowband filter is installed at the front of the camera lens or at the light output port of the structured light projector, and its center wavelength is consistent with the projection wavelength of the structured light projector.

[0023] Preferably, the image processing and analysis unit determines the actual positional offset of the connecting piece relative to the cell electrode post based on the two-dimensional image and the three-dimensional point cloud data, specifically as follows:

[0024] S401. Perform edge extraction and template matching on the two-dimensional image to locate the feature contours of the connecting piece and the battery cell terminal, and calculate the pixel coordinate offset of the center point of the connecting piece contour relative to the center point of the battery cell terminal contour. ;

[0025] S402. Using the pre-calibrated camera pixel equivalent k, convert the pixel coordinate offset into a physical plane offset. ; , ;

[0026] S403. Spatially align the three-dimensional point cloud data with the two-dimensional image to segment the point cloud of the upper surface region of the connecting piece and the point cloud of the upper surface region of the battery cell terminal. Perform plane fitting on each to obtain the plane equation of the connecting piece and the plane equation of the terminal. Calculate the distance between the two planes along the normal direction as the height offset. ;

[0027] S404, adjust the physical plane offset. With the height offset Combined into actual position offset .

[0028] Preferably, the connecting piece welding module includes a laser welding head, a two-dimensional scanning galvanometer, and an automatic focusing mechanism;

[0029] The image processing and analysis unit guides the connecting piece welding module to perform adaptive positioning compensation based on the actual position offset, specifically including:

[0030] The actual position offset In , The components are converted into coordinate offset values ​​of the two-dimensional scanning galvanometer, superimposed on the preset welding trajectory starting point coordinates, and a corrected galvanometer motion command is generated.

[0031] The The component is converted into the focus offset of the autofocus mechanism. ,in By adjusting the position of the focusing lens, the laser focus is always placed on the surface of the connecting piece;

[0032] Before the welding program starts, the corrected galvanometer motion command and the focal offset are sent to the welding controller to achieve three-dimensional adaptive positioning compensation of the welding focal point.

[0033] Preferably, the image processing and analysis unit determines whether there is an oxide layer or foreign matter on the surface of the battery cell electrode before welding based on the three-dimensional point cloud data, specifically as follows:

[0034] First, the three-dimensional point cloud data acquired before welding is segmented into pole region, and the point cloud on the pole surface is extracted.

[0035] A plane fitting is performed on the point cloud of the polar pole surface to obtain a reference plane, and the distance deviation from each point in the point cloud of the polar pole surface to the reference plane is calculated;

[0036] The region in the point cloud on the polar surface that meets the following conditions is identified as an abnormal region: the absolute value of the distance deviation exceeds a preset deviation threshold, and the projection of each point on the reference plane forms a connected region, the projected area of ​​the connected region is greater than a preset area threshold.

[0037] If the abnormal region appears as a concave region in the three-dimensional point cloud, that is, the distance deviation from each point to the reference plane is negative, then the abnormal region is mapped to the corresponding pixel region in the two-dimensional image, and the gray-scale mean of the pixel region is obtained; if the difference between the gray-scale mean of the normal pole surface and the gray-scale mean exceeds a preset gray-scale threshold, then the abnormal region is determined to be an oxide layer.

[0038] If the abnormal region appears as a raised area in the 3D point cloud, meaning that the distance deviation from each point to the reference plane is positive, then the abnormal region is mapped to the corresponding pixel region in the 2D image, and the gray-scale mean of the pixel region is obtained; if the difference between the gray-scale mean and the gray-scale mean of the normal pole surface exceeds a preset gray-scale threshold, then the abnormal region is determined to be a metallic foreign object or particle; the preset gray-scale threshold is 3 times the statistical fluctuation range of the gray-scale mean of the normal pole surface.

[0039] The average gray value of the normal pole surface is calculated by mapping the gray value of the corresponding pixel in the two-dimensional image to the normal area in the point cloud of the pole surface where the distance deviation does not exceed a preset deviation threshold.

[0040] When an oxide layer or foreign matter is detected, the image processing and analysis unit controls the connecting piece welding module or the external pole surface pretreatment device to perform local cleaning of the abnormal area, and acquires three-dimensional point cloud data again after cleaning to confirm that the surface condition is qualified.

[0041] Preferably, after welding, the image processing and analysis unit identifies the weld morphology based on the three-dimensional point cloud data to determine whether there are defects such as incomplete welding, missing welding, porosity, or weld misalignment. Specifically:

[0042] Based on the two-dimensional image, the image region of each preset solder joint is determined. The three-dimensional point cloud data is registered with the two-dimensional image through a pre-calibrated hand-eye calibration matrix, and the three-dimensional point cloud subset corresponding to each solder joint is segmented.

[0043] For each 3D point cloud subset of weld points, morphological features are extracted, including:

[0044] Weld joint height: The maximum value of the directed distance from each point within the weld joint area to the welding reference plane;

[0045] Solder joint surface depression depth: the maximum depth from the bottom of the local depression in the solder joint area to the surface of the surrounding solder joints;

[0046] Planar offset between the actual and theoretical positions of the solder joint ;

[0047] The extracted morphological features are compared with the pre-calibrated acceptable threshold range. When any feature exceeds its corresponding threshold range, the weld joint is determined to have a corresponding type of defect.

[0048] If the height of the weld joint is less than a preset minimum weld nugget height threshold, it is determined to be a cold weld; if the maximum value of the directed distance from each point in the weld joint area to the welding reference plane is less than the preset minimum weld nugget height threshold, and the difference between the directed distance from all points in the weld joint area to the welding reference plane and the thickness of the connecting piece is less than a preset surface noise threshold, it is determined to be a missed weld; if the plane offset... If the deviation exceeds the preset upper limit of weld deviation, it is judged as a weld deviation defect.

[0049] Preferably, the image processing and analysis unit is configured to determine porosity defects on the surface of the weld joint, specifically:

[0050] Local surface fitting is performed on the subset of the three-dimensional point cloud of the solder joint to obtain a smooth reference surface characterizing the macroscopic contour of the solder joint;

[0051] Calculate the vertical residual from each original point in the 3D point cloud subset of the solder joint to the smooth reference surface, and define a set of consecutive points with negative residuals and spatial adjacency as a local pit region.

[0052] Calculate the depression depth of each local pit region, and take the maximum value among the absolute values ​​of the negative residuals of all points in that region;

[0053] If any local pit area simultaneously meets the following conditions: the pit depth is greater than the preset porosity depth threshold, and the equivalent diameter of the area is within the preset porosity size range, then the weld joint is determined to have a porosity defect.

[0054] Preferably, the image processing and analysis unit is configured to classify the repairability of detected defects based on the analysis results of the post-weld 3D point cloud data, and determine the weld points that meet any of the following conditions as repairable defects:

[0055] (1) Failed soldering: The height of the solder joint is less than the preset minimum weld nugget height threshold, but the directional distance from each point in the solder joint area to the welding reference plane is greater than zero and less than the minimum weld nugget height threshold. At the same time, the standard deviation of the point cloud distribution in the solder joint area is greater than the preset surface noise threshold, that is, the point cloud distribution exhibits non-original planar characteristics.

[0056] (2) Welding deviation: Planar offset The weld deviation exceeds the preset allowable upper limit, but the weld point still partially overlaps within the effective overlap area between the connecting piece and the pole.

[0057] (3) Porosity defects: The number of pores does not exceed the preset threshold, and the sum of the projected areas of all pores on the welding reference plane does not exceed the preset total area threshold.

[0058] For solder joints determined to have repairable defects, the geometric center coordinates of the solder joint are calculated based on the three-dimensional point cloud data. , which serves as the spatial coordinates of the target point for welding repair;

[0059] The spatial coordinates of the target welding point and the defect type information are sent to the rework robot.

[0060] Preferably, the rework robot includes a multi-axis motion mechanism and an end-effector welding tool; the image processing and analysis unit guides the rework robot to perform targeted welding, specifically including:

[0061] The spatial coordinates of the target welding point are transformed from the vision system coordinate system to the robot base coordinate system to generate motion trajectory instructions;

[0062] Control the end-effector welding tool to move to a predetermined height above the target welding point and perform a fixed-point welding operation;

[0063] The repair welding parameters are set according to the defect type information as follows:

[0064] (1) For the defective weld, the repair welding power is set to 80% to 90% of the preset main welding power, and the focus is kept on the surface of the connecting piece to make up for the height of the weld nugget;

[0065] (2) For weld deviation defects, the preset main welding program is re-executed at the original preset weld point position;

[0066] (3) For porosity defects, the porosity area is remelted using the defocusing mode, and the welding power is set to 60% to 70% of the preset main welding power, so that the metal in the porosity area can flow and fill again.

[0067] After the welding repair is completed, the image processing and analysis unit controls the online welding quality detection module to collect the three-dimensional point cloud data of the area surrounding the welding repair target point again and perform a re-inspection operation:

[0068] The weld morphology features of the repair area were re-extracted and a quality assessment was performed.

[0069] If the re-inspection is successful, the weld point is marked as repaired.

[0070] If the re-inspection still fails, the repair welding and re-inspection are repeated until the re-inspection passes or the preset maximum number of reworks is reached; if the workpiece still fails after reaching the preset maximum number of reworks, it is marked as an irreparable scrap and sent to the scrap channel.

[0071] This invention provides a lithium battery assembly and testing system for power tools, which has the following advantages:

[0072] 1. By combining structured light projectors with polarizers and narrowband filters, the imaging problem of highly reflective metal surfaces such as battery cell terminals and connectors is effectively solved. The combination of the two enables clear two-dimensional images and complete three-dimensional point cloud data of the pad area to be obtained simultaneously even in production line environments with strong ambient light interference, providing a reliable data foundation for subsequent positioning, detection and judgment.

[0073] 2. By analyzing the distance deviation between the point cloud on the electrode surface and the reference plane, combined with grayscale verification of the two-dimensional image, surface defects such as oxide layers, metallic foreign objects, or particles can be accurately distinguished. After a defect is detected, the electrode surface pretreatment device is controlled to locally clean the abnormal area. After cleaning, data is collected again and re-inspected for confirmation, forming a closed-loop control of "detection-cleaning-re-inspection" to ensure the consistency of the electrode surface state before welding and reduce welding defects caused by interface contamination from the source.

[0074] 3. After welding, multi-dimensional morphological features such as weld height, surface depression depth, and actual position deviation are extracted from the three-dimensional point cloud subset of each weld point. By comparing with the pre-calibrated qualified threshold, defect types such as cold weld, missing weld, porosity, and weld deviation can be accurately identified. Compared with detection methods based on a single sensor or a single feature, the joint judgment of three-dimensional geometric information and two-dimensional grayscale information improves the accuracy of defect identification and the refinement of classification. Attached Figure Description

[0075] Figure 1 This is the overall system flowchart of the present invention;

[0076] Figure 2 This is a closed-loop flowchart of the pre-welding electrode surface defect detection and cleaning process of the present invention;

[0077] Figure 3 This is a flowchart of the post-weld weld quality inspection and closed-loop rework process of the present invention. Detailed Implementation

[0078] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0079] like Figures 1 to 3 As shown, the present invention proposes a lithium battery assembly and testing system for power tools, including a connecting piece feeding module, a connecting piece positioning and placement module, a connecting piece welding module, an online welding quality detection module, and a rework robot.

[0080] Connecting piece feeding module, used to automatically supply connecting pieces;

[0081] The connector positioning and placement module is used to place the connector on the positive and negative lead-out end faces of the battery cell assembly;

[0082] The connecting piece welding module is used to weld the connecting piece to the cell electrode, and the welding parameters of the device in the connecting piece welding module are adjustable;

[0083] The online welding quality inspection module includes an image acquisition unit and an image processing and analysis unit;

[0084] The image acquisition unit is equipped with a polarizer, a narrowband filter and a structured light projector to acquire clear two-dimensional images and three-dimensional point cloud data of the pad area before and after welding.

[0085] The image processing and analysis unit is configured as follows:

[0086] Before welding, the actual position offset of the connecting piece relative to the cell electrode is determined based on the two-dimensional image and three-dimensional point cloud data, and the connecting piece welding module is guided to perform adaptive positioning compensation.

[0087] It also uses 3D point cloud data to determine whether there is an oxide layer or foreign matter on the surface of the battery cell terminals before welding;

[0088] After welding, the weld morphology is identified based on the three-dimensional point cloud data to determine whether there are defects such as incomplete welding, missing welding, porosity, or welding deviation. When it is determined that there are repairable defects, the spatial coordinates of the defective weld point are calculated to guide the rework robot to perform fixed-point repair welding. After the repair welding is completed, the image acquisition unit is controlled to acquire three-dimensional point cloud data again and perform re-inspection. The rework robot is used to perform fixed-point repair welding on weld points with repairable defects.

[0089] The image acquisition unit acquires clear two-dimensional images and three-dimensional point cloud data of the pad area before and after welding, specifically:

[0090] S201. Control the structured light projector to project structured light with a preset coding pattern onto the pad area. The preset coding pattern is a striped pattern or a pseudo-random speckle pattern.

[0091] S202. After the structured light is projected onto the surface of the connector and the battery cell terminal, it deforms due to the change in surface height. The camera in the image acquisition unit is controlled to synchronously acquire the structured light image that has been deformed after being modulated by the surface of the pad area.

[0092] S203. The image processing and analysis unit compares the deformed structured light image with the pre-calibrated reference plane image to obtain the pixel displacement of each pixel and converts the pixel displacement into a physical offset based on the pixel size. Then, using the triangulation principle, the three-dimensional coordinates of each point in the pad area are calculated based on the physical offset to generate three-dimensional point cloud data.

[0093] Specifically, the process by which the image processing and analysis unit acquires 3D point cloud data is as follows:

[0094] First, a structured light with a known coded pattern is projected onto the pad area by a structured light projector. The coded pattern is preferably a black and white striped pattern or a pseudo-random speckle pattern with random distribution characteristics. After the structured light is projected onto the surface of the connector and the battery cell terminal, due to the height undulations in the pad area (such as the thickness of the connector, the height of the terminal boss, etc.), the originally regular coded pattern will produce a deformation corresponding to the surface morphology.

[0095] Simultaneously with structured light projection, the image processing and analysis unit controls the camera to synchronously acquire structured light images that are deformed after being modulated on the surface of the pad area. These deformed images contain three-dimensional topographic information of the pad area, in which the degree of deformation of the coded pattern is directly related to the surface height.

[0096] Then, the image processing and analysis unit compares the acquired deformed image with the pre-calibrated reference plane image. The reference plane image is the image acquired by the camera when the structured light is projected onto an ideal plane (such as a calibration plate) during the system calibration stage, and its encoded pattern has no height modulation.

[0097] For each pixel, the pixel displacement is calculated using image correlation operations. (Unit: pixels), by pixel size Convert pixel displacement to physical displacement. Finally, the height of the point relative to the reference plane is calculated using the principle of triangulation. The calculation formula is:

[0098]

[0099] Where B is the baseline distance between the optical center of the structured light projector and the optical center of the camera, and f is the focal length of the camera lens. Let B, f, and ... be the angle between the principal ray of the structured light projection and the optical axis of the camera. All parameters are known parameters obtained through system calibration.

[0100] The height values ​​of all the calculated pixels are arranged according to their spatial positions in the image to form dense three-dimensional point cloud data of the pad area. This three-dimensional point cloud and the synchronously acquired two-dimensional image have been strictly aligned in space by the system calibration matrix and can be directly used for subsequent connection piece positioning, pole surface defect detection and solder joint quality judgment.

[0101] This step-by-step acquisition process does not depend on the surface material of the connecting piece. Even in the presence of some reflective interference, the characteristics of the structured light encoded pattern can still be identified. Combined with subsequent polarization and narrowband filtering, a high-precision three-dimensional morphology can be stably obtained.

[0102] The image correlation operation is as follows: take a small square (such as a 15×15 pixel area) centered on the black dot in the reference image, and then search for the most similar small square in the deformed image. After finding it, the difference between the center positions of the two small squares is the pixel displacement of the point.

[0103] Pixel size These are inherent hardware parameters of the camera, usually provided by the manufacturer.

[0104] The image acquisition unit also includes polarizers and narrowband filters;

[0105] A polarizer is installed at the front of the camera lens, and its polarization direction is configured to be orthogonal to the main polarization direction of the light reflected by the specular surface of the battery cell electrode or connecting piece, so as to suppress brightness saturation in highly reflective areas.

[0106] Since the polarization direction of specular reflected light is relatively consistent, while the polarization state of diffuse reflected light is random, by orthogonalizing the polarizer direction with the main polarization direction of specular reflection, the intensity of specular reflected light can be greatly reduced, so that the overexposed area can be restored to the normal grayscale distribution, thereby obtaining a clear pad outline. This adjustment can be achieved by manually rotating the polarizer and observing the image histogram, or by configuring an electronically controlled rotary table to achieve adaptive adjustment.

[0107] Narrowband filters are installed at the front of the camera lens or at the light output port of a structured light projector. Their center wavelength is the same as the projection wavelength of the structured light projector, and their full width at half maximum (FWHM) is less than 20nm. They are used to allow only the structured light reflected signal to enter the camera and filter out ambient light interference.

[0108] This filter only allows reflected light of the structured light wavelength to pass through, blocking workshop lighting and other stray light, which greatly improves the signal-to-noise ratio of the structured light image. The combination of the two enables clear two-dimensional images and complete three-dimensional point clouds of highly reflective pole surfaces to be obtained simultaneously, even on production lines with strong ambient light interference.

[0109] The image processing and analysis unit determines the actual positional offset of the connecting piece relative to the cell electrode post based on the two-dimensional image and three-dimensional point cloud data, specifically:

[0110] S401. Perform edge extraction and template matching on the two-dimensional image to locate the feature contours of the connecting piece and the battery cell terminal, respectively, and calculate the pixel coordinate offset of the center point of the connecting piece contour relative to the center point of the battery cell terminal contour. ;

[0111] S402. Using the pre-calibrated camera pixel equivalent k, convert the pixel coordinate offset into a physical plane offset. ; , ;

[0112] S403. Spatially align the 3D point cloud data with the 2D image to segment the point cloud of the upper surface region of the connector plate and the point cloud of the upper surface region of the battery cell terminal. Perform plane fitting on each to obtain the plane equation of the connector plate and the plane equation of the terminal. Calculate the distance between the two planes along the normal direction as the height offset. ;

[0113] S404, Offset the physical plane With height offset Combined into actual position offset .

[0114] Specifically, the image processing and analysis unit first acquires a two-dimensional image of the pad area and the corresponding three-dimensional point cloud data. The resolution of the two-dimensional image is calibrated, and the pixel equivalent k is the physical size corresponding to each pixel; for example, k = 0.02 mm / pixel.

[0115] After the image processing and analysis unit performs Gaussian filtering on the two-dimensional image to remove noise, it uses the Canny edge detection algorithm to extract the rectangular outline of the connecting piece and the circular or rectangular outline of the battery cell terminal. The geometric center point of both outlines is then precisely located using a template matching method, and the center coordinates of the connecting piece are obtained in the image coordinate system. and pole center coordinates Thus, the pixel offset is calculated. , Using the calibration coefficient k, it can be converted into a plane offset in the world coordinate system;

[0116] Simultaneously, the 3D point cloud is aligned with the 2D image using a hand-eye calibration matrix. Utilizing the pre-defined polar contour boundaries and connecting plate contour regions in the 2D image, the 3D point cloud is segmented, and the point cloud set on the upper surface of the connecting plate is extracted. and point cloud set on the upper surface of the pole , respectively and The RANSAC (Random Sample Consensus) algorithm is used for spatial plane fitting;

[0117] Obtain the plane equation of the connecting piece And polar plane equation Since the two planes are theoretically parallel in actual assembly, the directed distance from the geometric center point of the pole plane to the connecting plate plane is taken as the height offset. ;

[0118]

[0119] in The three-dimensional coordinates of the center point of the cell electrode plane in the visual coordinate system can be taken as the geometric center of the electrode point cloud fitting plane. , , and The coefficients of the equations connecting the planar segments; the equations of the planar segments. normal vector ( , , , The direction is defined as from the pole to the connecting piece; if A value greater than 0 indicates that the plane of the connecting piece is above the plane of the pole post; otherwise, a negative value is taken.

[0120] The Canny edge detection algorithm is a well-known edge extraction algorithm in the field. It extracts contour edges from images through steps such as Gaussian filtering smoothing, gradient calculation, non-maximum suppression, and double threshold edge connection. It has the characteristics of high positioning accuracy and good edge continuity.

[0121] Template matching is a well-known image localization method in the field. It locates the target position in the image that best matches the template by calculating the normalized cross-correlation coefficient of each region of the preset template image and the image to be searched. It has the characteristics of resisting changes in illumination and strong robustness.

[0122] The hand-eye calibration matrix is ​​a well-known coordinate transformation method in the field, used to establish a spatial mapping relationship between the camera image coordinate system and the three-dimensional vision coordinate system. In this system, this matrix can be used to map the contour regions identified in the two-dimensional image to the corresponding points in the three-dimensional point cloud, or to back-project the segmentation results of the three-dimensional point cloud onto the two-dimensional image for verification, thereby achieving spatial alignment between the two-dimensional image and the three-dimensional point cloud.

[0123] The connecting piece welding module includes a laser welding head, a two-dimensional scanning galvanometer, and an automatic focusing mechanism;

[0124] The image processing and analysis unit guides the connecting piece welding module to perform adaptive positioning compensation based on the actual position offset, specifically including:

[0125] actual position offset In , The components are converted into coordinate offset values ​​of the two-dimensional scanning galvanometer, superimposed on the preset welding trajectory starting point coordinates, and a corrected galvanometer motion command is generated.

[0126] Will Components converted into focus offset of the autofocus mechanism ,in By adjusting the position of the focusing lens, the laser focus is always placed on the surface of the connecting piece;

[0127] Before the welding program starts, the corrected galvanometer motion command and focus offset are sent to the welding controller to achieve three-dimensional adaptive positioning compensation of the welding focus.

[0128] The welding controller has a pre-stored standard welding trajectory (such as a multi-point or spiral trajectory around the pole), the starting point of which corresponds to the center of the pole under ideal conditions.

[0129] Upon receiving the offset vector V(Δx,Δy,Δz), the image processing and analysis unit converts Δx and Δy into offset values ​​in the galvanometer coordinate system via a communication interface (such as EtherCAT) and sends them to the galvanometer control card. This causes the galvanometer control card to superimpose the offset value at each coordinate point along the preset path. Simultaneously, Δz is sent directly to the focusing mechanism controller as the focus offset value. The focusing mechanism moves the focusing lens in the positive or negative direction according to Δz, causing the laser focus to shift by Δz relative to the original set position. This compensates for the height deviation of the connecting piece. After compensation, the laser focus can accurately fall on the welding point of the connecting piece, avoiding welding misalignment or incomplete welding caused by misalignment.

[0130] The image processing and analysis unit determines whether there is an oxide layer or foreign matter on the surface of the battery cell electrode before welding based on the 3D point cloud data. Specifically:

[0131] First, the three-dimensional point cloud data acquired before welding is segmented into pole region, and the point cloud on the pole surface is extracted.

[0132] Plane fitting is performed on the point cloud on the polar pole surface to obtain a reference plane, and the distance deviation from each point in the point cloud on the polar pole surface to the reference plane is calculated;

[0133] The following regions in the point cloud on the polar surface are identified as abnormal regions: the absolute value of the distance deviation exceeds the preset deviation threshold, and the projection of each point on the reference plane forms a connected region, the projected area of ​​the connected region is greater than the preset area threshold.

[0134] The preset deviation threshold is determined during the system calibration phase by acquiring three-dimensional point clouds of multiple pole samples with known acceptable surface conditions, calculating the distance deviation from each point to the fitting reference plane, and statistically analyzing its standard deviation σ. The preset deviation threshold is set to 3σ, approximately 0.05 mm.

[0135] The preset area threshold is the maximum connected region projection area formed by measurement noise in the statistical calibration plane point cloud. The preset area threshold is set to 1.5 to 2 times this area, with a typical value of 0.3 mm².

[0136] If the abnormal region appears as a concave region in the 3D point cloud, that is, the distance deviation from each point to the reference plane is negative, then the abnormal region is mapped to the corresponding pixel region in the 2D image, and the gray-scale mean of the pixel region is obtained; if the difference between the gray-scale mean of the normal pole surface and the gray-scale mean exceeds the preset gray-scale threshold, then the abnormal region is determined to be an oxide layer.

[0137] The preset grayscale threshold is calculated by acquiring two-dimensional images of multiple normal pole surfaces, statistically analyzing the fluctuation range of the average grayscale value in the normal region, and setting the preset grayscale threshold to 3 times this fluctuation range.

[0138] If the abnormal region appears as a raised area in the 3D point cloud, meaning that the distance deviation from each point to the reference plane is positive, then the abnormal region is mapped to the corresponding pixel region in the 2D image, and the gray-scale mean of the pixel region is obtained; if the difference between the gray-scale mean and the gray-scale mean of the normal pole surface exceeds a preset gray-scale threshold, then the abnormal region is determined to be a metallic foreign object or particle; the preset gray-scale threshold is 3 times the statistical fluctuation range of the gray-scale mean of the normal pole surface.

[0139] Among them, the average gray value of the normal pole surface is calculated by mapping the gray value of the corresponding pixel in the two-dimensional image to the normal area in the point cloud of the pole surface where the distance deviation does not exceed the preset deviation threshold;

[0140] When an oxide layer or foreign matter is detected, the image processing and analysis unit controls the connecting piece welding module or the external pole surface pretreatment device to perform local cleaning of the abnormal area, and acquires three-dimensional point cloud data again after cleaning to confirm that the surface condition is qualified.

[0141] Specifically, before welding, the image processing and analysis unit needs to perform defect detection on the surface of the battery cell terminals to find surface defects that may affect the welding quality, such as oxide layers, metal foreign objects or particles.

[0142] Using the known pole position range after the connecting piece is placed, or the pole contour boundary obtained by matching a two-dimensional image template, the three-dimensional point cloud is cropped to segment a subset of the point cloud that belongs only to the pole surface;

[0143] A spatial plane is fitted as a reference plane using the random sampling consensus algorithm on a subset of the point cloud, and the equation is: Calculate the i-th point in the point cloud on the polar surface. Directed distance to the plane ;

[0144]

[0145] in, , , and The fitting coefficients are for the polar reference plane, which is obtained by RANSAC fitting of the point cloud of the normal region on the polar surface; since this spatial plane is an ideal plane, if In the formula, a numerator of zero indicates that the point is flush with the ideal plane. However, due to the presence of processing textures, oxide layer protrusions, or depressions on the polariton surface, most points in the actual point cloud do not exactly fall on the fitted plane. Substituting these values ​​into the formula results in a non-zero numerator. Normalizing this by dividing by the magnitude of the normal vector yields the following result. It is the perpendicular directional distance from that point to the reference plane.

[0146] This distance reflects the degree of unevenness of the point relative to an ideal flat surface. Normal pole surfaces, due to processing textures, Typically within 0.02mm, A value greater than 0 indicates a bulge. A value less than 0 indicates a depression;

[0147] Based on point cloud Values ​​generate a deviation graph, and a deviation threshold is set. =0.05mm (calibrable), mark all Points larger than the deviation threshold are considered outliers. Connectivity analysis is performed on outliers to filter out isolated noise points with an area smaller than the area threshold, such as 0.3 mm², while retaining continuous outlier regions.

[0148] Each preserved anomalous region is back-projected onto its corresponding pixel in the 2D image. If it appears as a continuous depression in the 3D point cloud... And the average gray value of the corresponding image region If the grayscale value of a normal electrode surface is lower than the threshold ΔG=30 (grayscale level 0-255), it is determined to be an oxide layer, because the oxide layer is usually dark gray or black and has an uneven surface.

[0149] If it appears as a bump Furthermore, this mean gray value minus the mean gray value of the normal pole surface If the difference exceeds the preset grayscale threshold, it is determined to be a metallic foreign object or welding slag particles;

[0150] Once any oxide layer or foreign matter is detected, the image processing and analysis unit generates a cleaning instruction and provides the center coordinates and bounding box of the abnormal area, which is then sent to the pole surface pretreatment device.

[0151] The electrode surface pretreatment device (such as a laser cleaning head) performs local scanning and cleaning based on the coordinates of the abnormal area. After cleaning, the integrated device re-acquires three-dimensional point cloud and two-dimensional image of the same electrode area. The image processing and analysis unit repeats the above steps to calculate the maximum deviation value of the cleaned area. And average gray level, if And the grayscale mean value is restored to If the surface condition is within ±ΔG, it is considered acceptable and the subsequent welding process can proceed; otherwise, cleaning will be performed again or an alarm will be issued.

[0152] After welding, the image processing and analysis unit identifies the weld morphology based on 3D point cloud data to determine whether there are defects such as incomplete welds, missing welds, porosity, or weld misalignment. Specifically:

[0153] Based on the two-dimensional image, the image region of each preset solder point is determined. The three-dimensional point cloud data is registered with the two-dimensional image through the pre-calibrated hand-eye calibration matrix, and the three-dimensional point cloud subset corresponding to each solder point is segmented.

[0154] For each 3D point cloud subset of the weld point, morphological features are extracted, including:

[0155] Solder joint height: The maximum value of the directed distance from each point in the solder joint area to the welding reference plane; the welding reference plane is a reference plane obtained by plane fitting of the point cloud of the pole post surface or connecting piece surface area around each solder joint that is not affected by welding.

[0156] Solder joint surface depression depth: the maximum depth from the bottom of the local depression in the solder joint area to the surface of the surrounding solder joints;

[0157] Planar offset between the actual and theoretical positions of the solder joint ;

[0158] The extracted morphological features are compared with the pre-calibrated acceptable threshold range. When any feature exceeds its corresponding threshold range, the weld joint is determined to have a corresponding type of defect.

[0159] Specifically, if the weld height is less than a preset minimum weld nugget height threshold, it is determined to be a cold weld; if the maximum value of the directed distance from each point in the weld area to the welding reference plane is less than the preset minimum weld nugget height threshold, and the difference between the directed distance from all points in the weld area to the welding reference plane and the thickness of the connecting piece is less than a preset surface noise threshold, it is determined to be a missed weld; if the plane offset is... If the deviation exceeds the preset upper limit of weld deviation, it is judged as a weld deviation defect.

[0160] The preset upper limit for weld deviation is determined by taking the maximum distance from the center of the overlap area to the point where the weld point can still guarantee the minimum overlap area, based on the effective overlap area size of the connecting piece and the cell electrode. The typical value is 0.3mm.

[0161] The minimum weld nugget height threshold is determined by conducting peel tests and cross-sectional metallographic analysis on weld joints under different welding parameters through process experiments. The relationship between weld joint height and welding strength is established, and the lowest weld joint height that meets the minimum peel force requirement is taken as the threshold, with a typical value of 0.2 mm.

[0162] The preset surface noise threshold is set to three times the standard deviation of the distance deviation from the plane to each point during system calibration.

[0163] The image processing and analysis unit is configured to determine porosity defects on the surface of the weld joint, specifically:

[0164] Local surface fitting is performed on a subset of the 3D point cloud of the solder joints to obtain a smooth reference surface that characterizes the macroscopic contour of the solder joints;

[0165] Calculate the vertical residual from each original point in the 3D point cloud subset of the weld joint to the smooth reference surface, and define a set of consecutive points with negative residuals and spatial adjacency as a local pit region.

[0166] Calculate the depression depth of each local pit region, and take the maximum value among the absolute values ​​of the negative residuals of all points in that region;

[0167] If any local pit area simultaneously satisfies the following conditions: the pit depth is greater than the preset porosity depth threshold, and the equivalent diameter of the area is within the preset porosity size range, then the weld joint is determined to have a porosity defect.

[0168] The preset porosity depth threshold is determined by conducting three-dimensional scanning and cross-sectional metallographic analysis on weld joints containing known porosity through process experiments, establishing the correspondence between the depth of the depression and the presence of porosity, and taking the minimum depth that can reliably distinguish between porosity and surface roughness as the threshold.

[0169] The preset pore size range is determined by conducting process tests and remelting filling tests on pores of different diameters. The maximum diameter that can be reliably filled is taken as the upper limit (typical value 0.5 mm), and the minimum diameter that has no significant impact on welding strength is taken as the lower limit (typical value 0.1 mm).

[0170] Specifically, after welding is completed, the integrated visual guidance positioning and welding quality inspection device re-acquires the three-dimensional point cloud and two-dimensional image of the solder pad area. The image processing and analysis unit then performs the following steps to determine the quality of the solder joint:

[0171] Step 1: Solder joint separation;

[0172] Using the position information of the connecting piece and pole obtained before welding, the region of interest (ROI) of each preset welding point is delineated on the two-dimensional image. The ROI is mapped to the three-dimensional point cloud space through the calibrated hand-eye matrix, and an independent point cloud subset of each welding point is extracted.

[0173] For each solder joint, a reference plane is fitted based on the point cloud of the surrounding unaffected pole or connector planar region. This plane represents the original surface before welding;

[0174] Calculate the j-th point of the weld point point cloud Directed distance to the reference plane ;

[0175]

[0176] in Let j be the original coordinates of the j-th point in the weld point point cloud. , , and The fitting coefficients are for the welding reference plane, which is obtained by RANSAC fitting of the point cloud of the normal region on the pole post surface, and the i-th point in the point cloud on the pole post surface. Directed distance to the plane Consistent;

[0177] The solder joint height is the maximum value among all positive distances within the solder joint area. ,like , A minimum weld nugget height threshold of 0.2mm (set according to the welding process) indicates that the solder has not fully melted and accumulated, and is judged as a cold weld.

[0178] If all points within the entire solder joint area All are less than the surface noise threshold (e.g., surface noise threshold = 0.03 mm) and If the value is approximately zero, it indicates that no weld nugget has formed at that location, and it is determined to be a missed weld.

[0179] For weld joints with normal height, further inspection is performed to detect local surface depressions. Moving least-squares smoothing is applied to independent subsets of the point cloud for each weld joint, and the vertical residual between the smoothed points and the original points is calculated. residual Defined as the difference between the original point coordinates and the smoothed reference surface coordinates, i.e., the height deviation along the Z-axis;

[0180] Identify residuals Less than Furthermore, a region consisting of a set of spatially adjacent continuous points is, in this case, a local pit region. A set of spatially adjacent continuous points refers to a group of points in three-dimensional space that are close to each other and can be connected to form a whole. On the surface of a solder joint, if a pit is composed of a group of points that are adjacent to each other and all lower than the surrounding surface, this group of points is a "spatially adjacent continuous point set", which completely outlines the shape and extent of the pit.

[0181] in Using the pore depth threshold, calculate the depression depth of this local pit region. , , The area of ​​the porosity defect region obtained by segmentation of a 3D point cloud or a 2D image;

[0182] like If the depth is stable within the range of 0.1mm to 0.5mm, it is determined to be a porosity, that is, a porosity defect exists.

[0183] Weld deviation determination: Calculate the geometric center coordinates of the 3D point cloud of the weld point. ;

[0184] ; ;

[0185] ( , , ( ) represents the coordinates of the actual geometric center of the weld joint. This represents the total number of point cloud concentration points at the weld joint.

[0186] The theoretical center of this weld point in the welding task is Calculate the plane offset :

[0187]

[0188] like ,like If the deviation is 0.3mm, it is determined to be a weld misalignment. At the same time, the two-dimensional image is used to confirm whether the weld point still falls within the effective overlap area between the connecting piece and the pole.

[0189] The image processing and analysis unit is configured to classify the repairability of detected defects based on the analysis results of the post-weld 3D point cloud data, and to determine the weld points that meet any of the following conditions as repairable defects:

[0190] (1) Failed soldering: The height of the solder joint is less than the preset minimum melt nugget height threshold, but there are partial melting traces in the solder joint area;

[0191] (2) Welding deviation: Planar offset The weld deviation exceeds the preset allowable upper limit, but the weld point still partially overlaps within the effective overlap area between the connecting piece and the pole.

[0192] (3) Porosity defects: The number of pores does not exceed the preset threshold, and the sum of the projected areas of all pores on the welding reference plane does not exceed the preset total area threshold.

[0193] The preset quantity threshold and preset total area threshold for pores are determined as follows: through process experiments, remelting tests are conducted on weld joints containing different quantities and areas of pores, the repair success rate is statistically analyzed, and the maximum number of pores and total area with a repair success rate higher than 95% are taken as the thresholds. Typical values ​​are no more than 3 pores and no more than 15% of the surface area of ​​the weld joint.

[0194] For weld joints determined to have repairable defects, the geometric center coordinates of the weld joint are calculated based on 3D point cloud data. , which serves as the spatial coordinates of the target point for welding repair;

[0195] Send the spatial coordinates of the target welding point and the defect type information to the rework robot:

[0196] The rework robot includes a multi-axis motion mechanism and an end-effector welding tool; the image processing and analysis unit guides the rework robot to perform targeted welding, specifically including:

[0197] Transform the spatial coordinates of the target welding point from the vision system coordinate system to the robot base coordinate system to generate motion trajectory instructions;

[0198] Control the end-effector welding tool to move to a predetermined height above the target welding point and perform a fixed-point welding operation;

[0199] Based on the defect type information, the repair welding parameters are set as follows:

[0200] (1) For the defective weld, the repair welding power is set to 80% to 90% of the preset main welding power, and the focus is kept on the surface of the connecting piece to make up for the height of the weld nugget;

[0201] (2) For weld deviation defects, the preset main welding program is re-executed at the original preset weld point position;

[0202] (3) For porosity defects, the porosity area is remelted using the defocusing mode, and the welding power is set to 60% to 70% of the preset main welding power, so that the metal in the porosity area can flow and fill again.

[0203] After the welding repair is completed, the image processing and analysis unit controls the online welding quality detection module to re-acquire the 3D point cloud data of the area surrounding the welding repair target point and performs a re-inspection operation:

[0204] The weld morphology features of the repair area were re-extracted and a quality assessment was performed.

[0205] If the re-inspection is successful, the weld point is marked as repaired.

[0206] If the re-inspection still fails, the repair welding and re-inspection are repeated until the re-inspection passes or the preset maximum number of reworks is reached; if the workpiece still fails after reaching the preset maximum number of reworks, it is marked as an irreparable scrap and sent to the scrap channel.

[0207] Specifically, after welding is completed, the image processing and analysis unit analyzes the 3D point cloud of each weld point, extracting the weld point height H and planar offset. Based on characteristic parameters such as pore depth and area, the module performs repairability classification:

[0208] Solder joints that meet any of the following conditions are considered repairable defects:

[0209] Cold solder joint: The height of the solder joint is lower than the minimum weld nugget height threshold. (e.g., 0.2mm), but there are partial melting marks in the weld area, indicating that repair welding can increase the height of the weld nugget;

[0210] Welding deviation: Planar offset Exceed (e.g., 0.3mm), but the solder joint still partially overlaps within the effective area of ​​the connecting piece and the pole post. Repositioning and welding can correct this.

[0211] Minor stomata: Stomatal depth exceeding the threshold (e.g., 0.1mm), but the number of pores does not exceed 3 and the total area of ​​pores does not exceed 15% of the surface area of ​​the weld point. It can be repaired by remelting and filling.

[0212] For solder joints determined to be repairable, the image processing and analysis unit calculates their geometric center coordinates based on a subset of the 3D point cloud of the solder joint, using these coordinates as the target point for repair welding. The calculation formula is:

[0213] ; ; ;

[0214] in The total number of point cloud concentration points at the weld joint. Let J be the three-dimensional coordinates of the j-th point in the point cloud. These coordinates reflect the actual spatial location of the weld joint and provide accurate target positioning for weld repair.

[0215] The repair robot is a six-axis articulated industrial robot with a welding tool (such as a laser welding head of the same type as the main welding device, or a micro-plasma welding gun) mounted at its end. The image processing and analysis unit is connected to the robot controller via an industrial Ethernet network to achieve real-time data interaction.

[0216] The coordinate transformation and motion planning process is as follows: First, using a pre-calibrated hand-eye calibration matrix, the spatial coordinates of the welding target point in the vision system coordinate system are transformed. Transform to coordinates in robot base coordinate system Then, the robot controller plans the motion trajectory from the current posture to a predetermined safe height (e.g., 50mm) above the target point, and then descends vertically to the focusing height above the welding target point. The hand-eye calibration matrix is ​​a 4×4 homogeneous transformation matrix obtained by simultaneously acquiring pose data in the vision system and robot base coordinate system through a calibration board.

[0217] Based on the defect type, the strategy for adjusting the welding parameters is as follows:

[0218] For defects in incomplete welds, the welding power is increased to 80% to 90% of the main weld, the focus is kept on the surface of the connecting piece, and the welding time is shortened to 50% of the main weld, so as to make up for the height of the weld nugget without damaging the surrounding area;

[0219] For weld deviation defects, repeat the complete main welding procedure at the original preset theoretical weld point position;

[0220] For porosity defects, the defocusing mode is used to remelt the porosity area, and the power is controlled at 60% to 70% of the main welding, so that the metal in the porosity area can reflow and fill.

[0221] After the welding repair operation is completed, the image processing and analysis unit returns a "welding repair complete" signal through the robot controller. The module then triggers the integrated vision-guided positioning and welding quality inspection device to re-acquire 3D point cloud and 2D images of the area around the welding target point (with a certain margin, such as 2mm).

[0222] The re-inspection process fully replicates the aforementioned post-weld defect judgment steps: segmenting the point cloud of the repair weld area, fitting a reference plane, extracting weld height, plane offset, porosity features, etc., and comparing them with the acceptable threshold. The re-inspection judgment logic is as follows:

[0223] If all characteristic parameters meet the qualification criteria ( , If no detectable porosity defects are found, the weld point is marked as successfully repaired.

[0224] If defects are still found during re-inspection, and the current number of rework attempts (n) is less than the preset maximum number of rework attempts. (like If the number of weld repairs is 2, then the weld repair and re-inspection process will be repeated.

[0225] If the number of repairs has reached... If it is still not up to standard, the workpiece is marked as an irreparable scrap by the workpiece ID reading and writing device, and sent to the scrap collection channel by the downstream sorting mechanism. At the same time, all defect data and rework history are recorded in the process database.

[0226] This closed-loop control ensures fully automated online repair and verification of welding quality, avoiding manual intervention and significantly improving the first-pass yield and product consistency of the power tool lithium battery assembly line.

[0227] In summary: Before welding, this application utilizes the spatial offset vector of a 3D point cloud to drive the laser welding head for 3D adaptive compensation of the focal point, eliminating the risk of weld misalignment and incomplete welds caused by the positional deviation of the connecting piece. Simultaneously, by analyzing the distance deviation of the point cloud on the electrode surface and combining it with 2D grayscale verification, oxide layers or foreign matter are identified, guiding local laser cleaning followed by re-inspection and confirmation, forming a pre-welding closed-loop control of "inspection-cleaning-re-inspection". After welding, multi-dimensional morphological features such as height, indentation depth, and positional deviation are extracted for each weld point to automatically identify defects such as incomplete welds, missing welds, porosity, and weld misalignment. Repairable defects are reworked using differentiated welding parameters according to type, and re-inspection is performed again after repair, forming a post-welding closed-loop control of "inspection-re-inspection". This end-to-end visual closed-loop achieves integrated management from high-precision perception, adaptive compensation, online cleaning, multi-feature judgment to automatic rework, improving welding consistency and product yield.

[0228] The threshold is set to facilitate comparison. The size of the threshold depends on the amount of sample data and the number of bases set by those skilled in the art for each set of sample data; as long as it does not affect the ratio between the parameter and the quantized value, it is acceptable.

[0229] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A lithium battery assembly and testing system for power tools, characterized in that: It includes a connecting piece feeding module, a connecting piece positioning and placement module, a connecting piece welding module, an online welding quality detection module, and a rework robot; The connector positioning and placement module is used to place the connector on the positive and negative lead-out end faces of the battery cell assembly; The connecting piece welding module is used to weld the connecting piece to the cell electrode; The online welding quality inspection module includes an image acquisition unit and an image processing and analysis unit; The image acquisition unit is equipped with a polarizer, a narrowband filter and a structured light projector to acquire clear two-dimensional images and three-dimensional point cloud data of the pad area before and after welding. The image processing and analysis unit is configured to: before welding, determine the actual positional offset of the connecting piece relative to the cell electrode based on the two-dimensional image and three-dimensional point cloud data, and guide the connecting piece welding module to perform adaptive positioning compensation; and also determine whether there is an oxide layer or foreign matter on the surface of the cell electrode before welding based on the three-dimensional point cloud data. After welding, the weld point morphology is identified based on the three-dimensional point cloud data to determine whether there are defects such as incomplete welding, missing welding, porosity, or welding deviation. When it is determined that there are repairable defects, the spatial coordinates of the defective weld point are calculated to guide the rework robot to perform fixed-point repair welding. After the repair welding is completed, the image acquisition unit is controlled to acquire three-dimensional point cloud data again and perform re-inspection. The rework robot is used to perform fixed-point repair welding on weld points with repairable defects. The image processing and analysis unit determines whether there is an oxide layer or foreign matter on the surface of the battery cell electrode before welding based on the three-dimensional point cloud data. Specifically: First, the three-dimensional point cloud data acquired before welding is segmented into pole region, and the point cloud on the pole surface is extracted. A plane fitting is performed on the point cloud of the polar pole surface to obtain a reference plane, and the distance deviation from each point in the point cloud of the polar pole surface to the reference plane is calculated; The region in the point cloud on the polar surface that meets the following conditions is identified as an abnormal region: the absolute value of the distance deviation exceeds a preset deviation threshold, and the projection of each point on the reference plane forms a connected region, the projected area of ​​the connected region is greater than a preset area threshold. If the abnormal region appears as a concave region in the three-dimensional point cloud, that is, the distance deviation of each point from the reference plane is negative, then the abnormal region is mapped to the corresponding pixel region in the two-dimensional image, and the gray average value of the pixel region is obtained. If the difference between the average gray value of a normal electrode surface and the average gray value exceeds a preset gray value threshold, the abnormal area is determined to be an oxide layer. If the abnormal region appears as a raised area in the 3D point cloud, meaning that the distance deviation from each point to the reference plane is positive, then the abnormal region is mapped to the corresponding pixel region in the 2D image, and the gray-scale mean of the pixel region is obtained; if the difference between the gray-scale mean and the gray-scale mean of the normal pole surface exceeds a preset gray-scale threshold, then the abnormal region is determined to be a metallic foreign object or particle; the preset gray-scale threshold is 3 times the statistical fluctuation range of the gray-scale mean of the normal pole surface. The average gray value of the normal pole surface is calculated by mapping the gray value of the corresponding pixel in the two-dimensional image to the normal area in the point cloud of the pole surface where the distance deviation does not exceed a preset deviation threshold. When an oxide layer or foreign matter is detected, the image processing and analysis unit controls the connecting piece welding module or the external pole surface pretreatment device to perform local cleaning of the abnormal area, and acquires three-dimensional point cloud data again after cleaning to confirm that the surface condition is qualified.

2. The lithium battery assembly and testing system for power tools according to claim 1, characterized in that: The image acquisition unit acquires clear two-dimensional images and three-dimensional point cloud data of the pad area before and after welding, specifically: S201. Control the structured light projector to project structured light with a preset coding pattern onto the pad area. The preset coding pattern is a stripe pattern or a pseudo-random speckle pattern. S202. After the structured light is projected onto the surface of the connecting piece and the battery cell electrode, it deforms due to the change in surface height. The camera in the image acquisition unit is controlled to synchronously acquire the structured light image that has been deformed after being modulated by the surface of the pad area. S203. The image processing and analysis unit compares the deformed structured light image with the pre-calibrated reference plane image to obtain the pixel displacement of each pixel and converts the pixel displacement into a physical offset using the pixel size. Then, using the triangulation principle, the three-dimensional coordinates of each point in the pad area are calculated based on the physical offset to generate the three-dimensional point cloud data.

3. The lithium battery assembly and testing system for power tools according to claim 2, characterized in that: The image acquisition unit also includes a polarizer and a narrowband filter; The polarizer is mounted on the front end of the camera lens, and its polarization direction is configured to be orthogonal to the main polarization direction of the light reflected by the mirror surface of the battery cell electrode or connecting piece, so as to suppress the brightness saturation of the highly reflective area. The narrowband filter is installed at the front of the camera lens or at the light output port of the structured light projector, and its center wavelength is consistent with the projection wavelength of the structured light projector.

4. The lithium battery assembly and testing system for power tools according to claim 3, characterized in that: The image processing and analysis unit determines the actual positional offset of the connecting piece relative to the cell electrode post based on the two-dimensional image and three-dimensional point cloud data, specifically as follows: S401. Perform edge extraction and template matching on the two-dimensional image to locate the feature contours of the connecting piece and the battery cell terminal, and calculate the pixel coordinate offset of the center point of the connecting piece contour relative to the center point of the battery cell terminal contour. ; S402. Using the pre-calibrated camera pixel equivalent k, convert the pixel coordinate offset into a physical plane offset. ; , ; S403. Spatially align the three-dimensional point cloud data with the two-dimensional image to segment the point cloud of the upper surface region of the connecting piece and the point cloud of the upper surface region of the battery cell terminal. Perform plane fitting on each to obtain the plane equation of the connecting piece and the plane equation of the terminal. Calculate the distance between the two planes along the normal direction as the height offset. ; S404, adjust the physical plane offset. With the height offset Combined into actual position offset .

5. The lithium battery assembly and testing system for power tools according to claim 4, characterized in that: The connecting piece welding module includes a laser welding head, a two-dimensional scanning galvanometer, and an automatic focusing mechanism; The image processing and analysis unit guides the connecting piece welding module to perform adaptive positioning compensation based on the actual position offset, specifically including: The actual position offset In , The components are converted into coordinate offset values ​​of the two-dimensional scanning galvanometer, superimposed on the preset welding trajectory starting point coordinates, and a corrected galvanometer motion command is generated. The The component is converted into the focus offset of the autofocus mechanism. ,in By adjusting the position of the focusing lens, the laser focus is always placed on the surface of the connecting piece; Before the welding program starts, the corrected galvanometer motion command and the focal offset are sent to the welding controller to achieve three-dimensional adaptive positioning compensation of the welding focal point.

6. The lithium battery assembly and testing system for power tools according to claim 5, characterized in that: After welding, the image processing and analysis unit identifies the weld morphology based on the three-dimensional point cloud data and determines whether there are defects such as incomplete welding, missing welding, porosity, or weld misalignment. Specifically: Based on the two-dimensional image, the image region of each preset solder joint is determined. The three-dimensional point cloud data is registered with the two-dimensional image through a pre-calibrated hand-eye calibration matrix, and the three-dimensional point cloud subset corresponding to each solder joint is segmented. For each 3D point cloud subset of weld points, morphological features are extracted, including: Weld joint height: The maximum value of the directed distance from each point within the weld joint area to the welding reference plane; Solder joint surface depression depth: the maximum depth from the bottom of the local depression in the solder joint area to the surface of the surrounding solder joints; Planar offset between the actual and theoretical positions of the solder joint ; The extracted morphological features are compared with the pre-calibrated acceptable threshold range. When any feature exceeds its corresponding threshold range, the weld joint is determined to have a corresponding type of defect. If the height of the weld joint is less than a preset minimum weld nugget height threshold, it is determined to be a cold weld; if the maximum value of the directed distance from each point in the weld joint area to the welding reference plane is less than the preset minimum weld nugget height threshold, and the difference between the directed distance from all points in the weld joint area to the welding reference plane and the thickness of the connecting piece is less than a preset surface noise threshold, it is determined to be a missed weld; if the plane offset... If the deviation exceeds the preset upper limit of weld deviation, it is judged as a weld deviation defect.

7. The lithium battery assembly and testing system for power tools according to claim 6, characterized in that: The image processing and analysis unit is configured to determine porosity defects on the surface of the weld joint, specifically: Local surface fitting is performed on the subset of the three-dimensional point cloud of the solder joint to obtain a smooth reference surface characterizing the macroscopic contour of the solder joint; Calculate the vertical residual from each original point in the 3D point cloud subset of the solder joint to the smooth reference surface, and define a set of consecutive points with negative residuals and spatial adjacency as a local pit region. Calculate the depression depth of each local pit region, and take the maximum value among the absolute values ​​of the negative residuals of all points in that region; If any local pit area simultaneously meets the following conditions: the pit depth is greater than the preset porosity depth threshold, and the equivalent diameter of the area is within the preset porosity size range, then the weld joint is determined to have a porosity defect.

8. The lithium battery assembly and testing system for power tools according to claim 7, characterized in that: The image processing and analysis unit is configured to classify the repairability of detected defects based on the analysis results of the post-weld 3D point cloud data, and to determine that weld points that meet any of the following conditions are repairable defects: (1) Failed soldering: The height of the solder joint is less than the preset minimum weld nugget height threshold, but the directional distance from each point in the solder joint area to the welding reference plane is greater than zero and less than the minimum weld nugget height threshold. At the same time, the standard deviation of the point cloud distribution in the solder joint area is greater than the preset surface noise threshold, that is, the point cloud distribution exhibits non-original planar characteristics. (2) Welding deviation: Planar offset The weld deviation exceeds the preset allowable upper limit, but the weld point still partially overlaps within the effective overlap area between the connecting piece and the pole. (3) Porosity defects: The number of pores does not exceed the preset threshold, and the sum of the projected areas of all pores on the welding reference plane does not exceed the preset total area threshold. For solder joints determined to have repairable defects, the geometric center coordinates of the solder joint are calculated based on the three-dimensional point cloud data. , which serves as the spatial coordinates of the target point for welding repair; The spatial coordinates of the target welding point and the defect type information are sent to the rework robot.

9. The lithium battery assembly and testing system for power tools according to claim 8, characterized in that: The rework robot includes a multi-axis motion mechanism and an end-effector welding tool; the image processing and analysis unit guides the rework robot to perform targeted welding, specifically including: The spatial coordinates of the target welding point are transformed from the vision system coordinate system to the robot base coordinate system to generate motion trajectory instructions; Control the end-effector welding tool to move to a predetermined height above the target welding point and perform a fixed-point welding operation; The repair welding parameters are set according to the defect type information as follows: (1) For the defective weld, the repair welding power is set to 80% to 90% of the preset main welding power, and the focus is kept on the surface of the connecting piece to make up for the height of the weld nugget; (2) For weld deviation defects, the preset main welding program is re-executed at the original preset weld point position; (3) For porosity defects, the porosity area is remelted using the defocusing mode, and the welding power is set to 60% to 70% of the preset main welding power, so that the metal in the porosity area can flow and fill again. After the welding repair is completed, the image processing and analysis unit controls the online welding quality detection module to collect the three-dimensional point cloud data of the area surrounding the welding repair target point again and perform a re-inspection operation: The weld morphology features of the repair area were re-extracted and a quality assessment was performed. If the re-inspection is successful, the weld point is marked as repaired. If the re-inspection still fails, the repair welding and re-inspection are repeated until the re-inspection passes or the preset maximum number of reworks is reached; if the workpiece still fails after reaching the preset maximum number of reworks, it is marked as an irreparable scrap and sent to the scrap channel.

Images