A safe and reliable high-rate battery manufacturing method

Abstract

The application discloses a safe and reliable high-rate battery manufacturing method, and belongs to the technical field of high-rate lithium ion battery manufacturing, and comprises the following steps: S1, adopting a pressure feedback tool table to fix the battery cell, combining binocular multispectral imaging and three-dimensional point cloud modeling, extracting the tab feature and fitting the reference line to make the length difference of each layer after trimming smaller; S2, through double-station trimming and graded negative pressure magnetic collection, the metal chip collection rate is greatly improved; S3, elastic pre-pressing ensures that the fitting gap is small, argon prevents oxidation, and the laser parameters are adaptively matched according to the thickness of the tab; S4, through joint detection of vision, a CNN model and infrared thermal imaging, combined with three-level judgment, the defect detection rate is greatly improved; S5, high-frequency inductance and high-pressure nitrogen blowing final inspection are used to eliminate metal chip residues. The application greatly improves the 10C and above high-rate discharge capacity retention rate, prolongs the cycle life, reduces the risk of thermal runaway, and balances safety and large-scale production efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of battery manufacturing, in particular to a safe and reliable high-rate battery manufacturing method. BACKGROUND

[0002] In the field of high-rate lithium ion battery manufacturing, tab processing, welding quality control and defect detection are the core links that affect the performance and safety of the battery. The current related processes still have significant technical pain points, as follows:

[0003] Current high-rate battery tab positioning mostly uses single vision imaging or manual assisted positioning. Single vision imaging is easily disturbed by the surface oxidation layer and the gap between the overlapping layers, making it difficult to accurately extract the total thickness of the tab, the length deviation of each layer and the center coordinates. The trimming link lacks a baseline fitting mechanism based on the three-dimensional features of the tab, and mostly relies on fixed templates for trimming, resulting in a common length difference of 0.08mm or more after trimming. This precision defect directly causes uneven stress on the tab during subsequent welding, and the current distribution is chaotic during the battery charging and discharging process, resulting in a significant decrease in capacity retention rate at 10C and above high-rate discharge, which cannot meet the demand for high-rate performance in scenarios such as electric tools and energy storage systems.

[0004] Secondly, in the existing tab welding process, the pre-pressing link mostly uses fixed pressure design, without dynamically adjusting the pre-pressing force according to the number of tab overlapping layers, resulting in a common gap of 0.05mm or more between the tab and the cover plate. Open nozzle is mostly used for inert gas protection, with insufficient argon purity and fixed gas output. The oxygen content in the welding area is easy to exceed 300ppm, and oxidation slag is easy to occur during tab welding. The laser welding parameters are mostly fixed values, without adaptive matching according to the tab thickness. Thin tabs are easy to burn through, and thick tabs are easy to appear virtual welding. The above welding defects result in a common welding strength of the tab being lower than 10N. During the use of the battery, local heating is easy to occur due to excessive contact resistance in the welding area, which further causes electrolyte decomposition and electrode structure aging, significantly shortening the cycle life of the battery.

[0005] In addition, current post-welding defect detection mostly relies on single vision detection or manual sampling inspection. Single vision detection can only identify appearance defects such as welding bumps and cracks, and cannot detect internal virtual welding. Manual sampling inspection is low in efficiency and subjective in judgment criteria, resulting in poor defect detection rate. The missed virtual welding cells are easy to induce thermal runaway in use. In addition, tab trimming mostly uses single station design, which needs to stop for cleaning metal chips after trimming, with a cleaning time of 20-30s per time. There is a waiting time in the connection with the welding and detection links, resulting in low overall manufacturing efficiency and difficulty in meeting the demand for large-scale production of high-rate batteries.

[0006] Therefore, the present application proposes a safe and reliable high-rate battery manufacturing method. SUMMARY

[0007] One purpose of the present application is to provide a safe and reliable high-rate battery manufacturing method. The present application can realize accurate positioning and trimming of tabs through binocular multi-spectral imaging and three-dimensional point cloud modeling, so that the length difference of each layer of tabs is small, the dimensional consistency of tabs is ensured, the metal chip collection rate is greatly improved through double-station negative pressure magnetic closed-loop collection and cleanliness final inspection, the short circuit risk caused by metal chips is completely eliminated, the welding quality is optimized through elastic pre-pressing self-adaptive adjustment, inert gas protection and laser parameter matching, oxidation slag and false welding are avoided, welding strength is improved, and post-welding defect detection rate is greatly improved through visual, infrared multi-dimensional detection and three-level judgment. At the same time, the production efficiency is improved through double-station collaborative design, and finally the capacity retention rate of high-rate battery is improved, the cycle life is prolonged, and the risk of thermal runaway is reduced. The battery safety and reliability, high-rate performance and large-scale production requirements are considered.

[0008] A safe and reliable high-rate battery manufacturing method according to an embodiment of the present application comprises the following steps:

[0009] S1, the wound or laminated lithium battery cell is clamped and fixed through an adjustable tooling table with pressure feedback, then a binocular vision camera and a multi-spectral light source are used to scan the tabs, and then a three-dimensional point cloud model of the tabs is generated through binocular disparity algorithm and multi-spectral image fusion, the total thickness, the number of overlapping layers, the length deviation and the center coordinates of each layer of tabs are extracted, and the trimming reference line is calculated so that the length difference of each layer of tabs after trimming is ≤0.03mm;

[0010] S2, the tabs are sheared through a double-station trimming mechanism, the trimming platform is provided with a chip guide groove, a negative pressure magnetic collection system sucks the metal chips into the channel of a magnetic filter screen through negative pressure, and the double-station alternating mode is used to realize the synchronous operation of trimming and reverse blowing cleaning of the collection groove;

[0011] S3, an elastic pre-pressing head is used to pre-press the trimmed tabs and the cover plate, the pre-pressing force is set to 3-5N according to the number of overlapping layers of tabs to ensure that the fitting gap is ≤0.02mm, then argon gas is output through a ring nozzle to form a local protection atmosphere, and then the laser welding parameters are adaptively matched according to the thickness of the tabs, wherein when the thickness of the tabs is 0.1-0.3mm, the laser power is 80-120W and the welding speed is 30-50mm / s, when the thickness of the tabs is 0.3-0.5mm, the laser power is 150-200W and the welding speed is 20-30mm / s, and the displacement sensor is used to correct the path in real time during the welding process;

[0012] S4, the welding area is detected by a visual detection unit and an infrared thermal imaging unit, the visual detection unit collects images by a camera and identifies appearance defects through a deep learning CNN model, the infrared thermal imaging unit identifies temperature abnormal areas by a thermal imager, and then two types of detection data are fused by an industrial computer, and three-level determination of qualified, unqualified and re-inspection is performed according to a preset threshold, and the battery cell to be re-inspected is confirmed again by a microscopic imaging device;

[0013] S5, the residual metal chips in the battery cell are scanned by a high-frequency inductive detector, if the metal chip signal is detected, the 0.8MPa high-pressure nitrogen gas is blown at an angle of 45° along the tab hole, the inductive detection is performed again after blowing, the battery cell without detecting metal chips enters the subsequent assembly link, and the battery cell still detecting metal chips is determined as unqualified and rejected.

[0014] Further, the adjustable tooling table clamp with pressure feedback in step S1 monitors the clamping force in real time through a pressure sensor, the clamping force needs to be controlled in the range of 0.5-1N, and a silica gel buffer structure is used on the contact side of the clamping plate and the battery cell to avoid damage to the battery cell shell, and the clamping action is automatically stopped when the pressure reaches the set threshold during the clamping process.

[0015] Further, when the tab is scanned by the binocular vision camera and the multi-spectral light source in step S1, the resolution of the binocular vision camera is set to 2048x1536, the vertical distance from the tab is controlled to be 300-500mm, the angle deviation is not more than 5°, the multi-spectral light source outputs blue light of 450nm±5nm, green light of 550nm±5nm and red light of 650nm±5nm in turn, and 3-5 frames of tab surface images are collected under each wavelength light source.

[0016] Further, the three-dimensional point cloud model generation and the baseline calculation of the trimming in step S1 specifically include:

[0017] Multi-spectral image fusion: the blue light B, green light G and red light R images are fused into an enhanced image I according to the weight, and the formula is:

[0018]

[0019] Wherein, The edge contour weight is equal to 0.4, β is the oxidation layer inhibition weight equal to 0.3, γ is the superposition layer recognition weight equal to 0.3, and + + =1;

[0020] Binocular disparity calculation: the disparity d of the tab surface point (x, y) is calculated by matching the left and right camera images to generate depth information Z, and the formula is:

[0021]

[0022] wherein, f is the camera focal length, b is the baseline distance of the binocular camera, is the system calibration constant;

[0023] Trimming the reference line fitting: the edge points of the polar ear The least square method is used to fit a straight line as the reference line, and the formula is:

[0024]

[0025] wherein is the slope, is the intercept, and after fitting, the length difference of each layer of polar ear after trimming is ensured to be ≤0.03mm.

[0026] Further, the preset threshold value of the three-level judgment in the step S4 is calculated as follows:

[0027] Appearance defect threshold value: assuming that the defect feature parameter set of the qualified sample is , the mean value and the standard deviation are calculated, and the defect judgment threshold value H is When the detected defect parameter > H, it is judged as an appearance defect.

[0028] Temperature difference threshold value: assuming that the temperature set of the normal welding area is , the mean value and the standard deviation are calculated, and the temperature difference threshold value is: When the temperature difference between the detected area and the normal area is , it is judged as a temperature anomaly.

[0029] The three-level judgment is divided into qualified, to be rechecked, and unqualified. When the detection result is no appearance defect and the temperature difference ≤ , it indicates qualified; when the detection result is no appearance defect and < the temperature difference ≤ , it indicates to be rechecked; when the detection result is the appearance defect parameter > H or the temperature difference > H, it indicates unqualified.

[0030] ​Further, the negative pressure value during the negative pressure magnetic collection in the step S2 is set according to the tab material, -50 to -60 kPa is adopted for an aluminum tab, -70 to -80 kPa is adopted for a copper tab, the magnetic filter screen aperture is 0.05 mm, the magnetic strength is 1000Gs, in the double-station alternating mode, the switching time of the trimming station and the cleaning station is not more than 0.5s, the back-blowing cleaning adopts nitrogen with a pressure of 50 kPa, and the duration is 0.3-0.5s.

[0031] Further, the elastic coefficient of the elastic pre-pressing head in the step S3 is 5N / mm, the pre-pressing force is set to 3N when the tab overlap layer is 2-3 layers, and is set to 4-5N when the tab overlap layer is 4-6 layers, the purity of the argon gas output by the annular nozzle is not less than 99.999%, and the gas output is controlled to be 10-15L / min, so that the oxygen content in the welding area is reduced to below 100ppm.

[0032] Further, in the laser welding parameter in the step S3, the laser welding adopts fiber laser with a wavelength of 1064nm, the spot diameter is controlled to be 0.2-0.3mm, and the sampling frequency of the displacement sensor during real-time correction of the path in the welding process is 1kHz, the welding path deviation is detected in real time, and when the deviation exceeds 0.01mm, the laser head position is automatically adjusted for correction.

[0033] Further, in the step S4, the camera resolution of the visual detection unit is 4096x3072, the annular light source with an illuminance of 5000lux is adopted when the image is collected, the deep learning CNN model is trained by not less than 100,000 samples containing welding bumps, cracks and incomplete welding defects, the recognition accuracy is ±0.01mm, the thermal imager resolution of the infrared thermal imaging unit is 640x512, the temperature measurement range is -20-500℃, the accuracy is ±0.5℃, and the sampling interval is 0.1s.

[0034] Further, in the step S5, the detection frequency of the high-frequency inductive detector is 1MHz, the sensitivity is 0.001g, the distance between the detector and the surface of the battery cell is maintained to be 2mm±0.1mm during detection, and when high-pressure nitrogen is used for blowing, the purity of the high-pressure nitrogen is not less than 99.99%, the three symmetrically arranged nozzles blow air along the 45°±2° direction of the tab hole during blowing, and the inductive detection is performed again after 1-2s interval after blowing to ensure that there is no residual metal chip.

[0035] The beneficial effects of the present application are:

[0036] 1. In this invention, the electrode positioning is achieved through binocular multispectral imaging and three-dimensional point cloud modeling, which enables accurate extraction of the total electrode thickness, center coordinates, and length deviation of each layer. Combined with the least squares method to fit the trimming baseline, the length difference of each layer of electrode after trimming is ≤0.03mm. This effectively solves the problem of uneven welding stress caused by the different lengths of electrodes in traditional trimming, thereby greatly improving the consistency of the cell electrode size and providing a uniform electrode foundation for subsequent welding. This results in a more uniform current distribution during battery charging and discharging, and effectively improves the capacity retention rate during high-rate discharge of 10C and above.

[0037] 2. In this invention, a welding method that uses elastic pre-pressure adaptive adjustment, inert gas atmosphere protection, and laser parameter adaptive matching is employed. Graded pre-pressure is set according to the number of electrode stacking layers to ensure electrode bonding gap. High-purity argon gas can effectively suppress welding oxidation. Combined with laser power and welding speed adaptive matching based on electrode thickness, the welding area is free of oxide slag and incomplete welding, and the welding strength is improved. Therefore, it can reduce the local heating problem caused by welding defects and significantly extend the battery cycle life.

[0038] 3. In this invention, a multi-dimensional defect detection system combining vision and infrared is used. The vision detection uses a high-resolution camera combined with a CNN model trained with more than 100,000 samples, which can fully cover appearance defects such as weld beads, cracks, and missing welds. Infrared thermal imaging identifies internal cold welds through a temperature difference of 5-10℃. Combined with a three-level judgment and re-inspection mechanism, the post-weld defect detection rate is greatly improved, ensuring the accuracy of rejecting defective products. At the same time, through the design of dual-station trimming, parameter adaptive welding, and detection, the overall manufacturing efficiency is greatly improved. Attached Figure Description

[0039] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0040] Figure 1 This is a flowchart of a safe and reliable high-rate battery manufacturing method proposed in this invention;

[0041] Figure 2 This is a comparison chart of metal scrap collection rates for different manufacturing methods of the safe and reliable high-rate battery manufacturing method proposed in this invention.

[0042] Figure 3 This is a comparison diagram of the welding strength of different manufacturing methods for a safe and reliable high-rate battery manufacturing method proposed in this invention.

[0043] Figure 4 A comparison chart of the high-rate discharge capacity retention rates of different manufacturing methods for the safe and reliable high-rate battery manufacturing method proposed in this invention.

[0044] Figure 5 A cycle life comparison chart of different manufacturing methods for a safe and reliable high-rate battery manufacturing method proposed by the present application. DETAILED DESCRIPTION

[0045] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0046] Embodiment one

[0047] Please refer to Figures 1-5 , the present embodiment aims at the high-rate discharge demand of electric tools, selects 18650 type cylindrical high-rate lithium ion battery as the manufacturing object, the tab of which is T2 red copper material, and a total of 800 pieces of battery cells are produced, specifically including the following steps:

[0048] S1, the wound 18650 type lithium battery is transported to the adjustable tooling table with pressure feedback one by one, the inside of the clamping plate on both sides of the tooling table is pasted with a 2mm thick silica gel buffer structure. The clamping plate is driven by a servo motor, a strain gauge type pressure sensor is integrated at the end of the clamping plate, and the clamping force signal is transmitted to the PLC controller in real time. When the clamping force reaches 0.6N, the PLC controller automatically sends a stop signal, and the servo motor stops driving, completing the fixing of the battery cell.

[0049] At 350mm above the tooling table, a binocular vision camera is fixed through an adjustable support, the perpendicularity deviation of the camera lens axis and the tab plane is adjusted to 4°; a ring-shaped multi-spectral LED light source is coaxially arranged outside the camera lens, and the light source controller sequentially outputs 450nm±5nm blue light, 550nm±5nm green light and 650nm±5nm red light according to the preset program. When each wavelength light source is lit, the camera continuously collects 4 frames of tab surface images, and the light intensity is stabilized at 2500cd / m² through the light supplementing control module during the collection process to avoid reflection interference.

[0050] The collected multi-spectral images are transmitted to an industrial computer for pretreatment;

[0051] Specifically, first, a 3x3 size Gaussian filter kernel is used to denoise the single-band image to eliminate random noise;

[0052] Then, the CLAHE algorithm is used to enhance the contrast of the image and highlight the tab edge and the interlaminar gap features.

[0053] Then a single enhanced image is generated by the multispectral image fusion formula, in which the edge definition of the polar ear is improved by 40% and the interference of the surface oxide layer is reduced by 50% compared with the single-band image.

[0054] The multispectral image fusion formula is:

[0055]

[0056] wherein is the enhanced image after fusion, , , are the gray values of the blue, green and red light images respectively, .

[0057] With the fused image as input, binocular disparity calculation is realized through Halcon software.

[0058] Specifically, first, the left and right camera images are stereo matched to obtain the disparity d of any pixel point (x, y) on the surface of the polar ear; and then the depth calculation formula is substituted into to generate the depth information Z of the polar ear surface.

[0059] Wherein the focal length of the camera after the chessboard calibration = 15 mm, = 100 mm, and the system calibration constant = 0.3, which is obtained by averaging multiple calibration experiments;

[0060] Then the depth information is combined with the two-dimensional coordinates to construct a polar ear three-dimensional point cloud model with a point cloud density of 110 points / mm².

[0061] The key feature parameters of the polar ear are extracted from the three-dimensional point cloud model, the total thickness of the polar ear is calculated as 0.75 mm by the maximum depth difference on the same vertical line, the number of superimposed layers is identified as 3 layers by the number of depth value mutations, the length deviation is calculated as 0.07 mm by the distance difference of the edge points of each layer of the polar ear to the preset reference line, and the center coordinates of the polar ear are calculated as (22.002, 16.004) mm by the point cloud gravity algorithm.

[0062] The edge point set of the polar ear is extracted , the least square method is substituted into the weight reference line fitting formula , and then k = 0.015 and b = 4.8 are obtained by solving with Matlab software. After trimming according to the reference line, the length difference of each layer of the polar ear is 0.025 mm. In the weight reference line fitting formula, k is the slope of the reference line and b is the intercept.

[0063] S2, the cell positioned in step S1 is transported to the double-station trimming mechanism by a conveyor belt. The servo motor of the trimming mechanism drives the cutter assembly to adjust the cutter spacing to 8 mm. The cutter is made of hard alloy material, and the angle between the cutter and the trimming reference line is calibrated to 0° by a laser interferometer to avoid the generation of oblique burrs during shearing;

[0064] According to the 0.25 mm thickness of the tab, the shearing speed is set to 55 mm / s. The trimming platform is made of 304 stainless steel, and a 0.4 mm wide chip guide groove is provided on the surface of the platform. The negative pressure magnetic collection system is started. Since the tab is made of copper, the negative pressure value is set to -75 kPa. A neodymium iron boron magnetic filter screen is arranged in the collection channel. The metal chips enter the channel through the guide groove under the action of negative pressure and are adsorbed by the magnetic filter screen. Non-magnetic dust is discharged to the dust collection bag with the airflow.

[0065] The double-station alternating mode is started synchronously. When the tab trimming is performed in the first station, the back blowing device in the second station is started. Nitrogen gas with a pressure of 50 kPa is sprayed in the channel in the reverse direction, and the duration is 0.4 s. The copper chips adsorbed on the magnetic filter screen are blown into the detachable recovery box;

[0066] Please refer to Figure 2 When the weight of the copper chips in the recovery box reaches 50 g, the weight sensor sends a signal to the PLC, prompting manual replacement of the recovery box. The two-station switching is realized by a servo motor driving a synchronous belt, and the switching time is controlled within 0.4 s without stopping and waiting. After the trimming of the batch of cells is completed, the metal chip collection rate reaches 99.3%, and no metal chips are found inside the cells.

[0067] S3, the trimmed cell is transported to the laser welding station. The elastic pre-pressing head pre-presses the tab and the cover plate. Since the number of tab overlapping layers is 3, the pre-pressing force can be set to 3 N. The tab and the cover plate are measured by a micrometer, and the fitting gap is 0.018 mm

[0068] The annular argon nozzle is started, and argon gas with a purity of 99.999% is output. The gas flow meter controls the gas output to be 12 L / min. An oxygen content analyzer is used to monitor the oxygen content of the welding area in real time, and the final stable value is 90 ppm, which avoids the oxidation of the tab during welding to produce oxidation slag.

[0069] The laser welding machine adaptively matches the welding parameters according to the 0.25 mm thickness of the tab:

[0070] The laser power is 100 W, the welding speed is 40 mm / s, and the spot diameter is 0.25 mm.

[0071] During the welding process, the laser head carries a laser triangulation sensor, which detects the deviation of the welding path in real time. When the deviation exceeds 0.01 mm, the sensor sends a signal to the welding machine control system, which automatically adjusts the X / Y axis position of the laser head to ensure that the welding deviation is ≤0.01 mm.

[0072] Please refer to Figure 3 After the batch of welding is completed, 50 cells are randomly selected for testing, and the tab welding strength is tested by a tensile testing machine, all ≥12N, without false welding or welding nodule phenomenon.

[0073] S4, the welded cell first enters the visual detection unit, the camera of the visual detection unit collects the image of the welding area, and a ring-shaped white LED light source is configured beside the camera to ensure that the image has no shadow and uniform brightness.

[0074] The collected image is transmitted to an industrial computer, and a deep learning CNN model is used for defect recognition. The model outputs the defect type and defect parameters, and the recognition accuracy is ±0.01 mm.

[0075] Subsequently, the cell enters the infrared thermal imaging unit, the temperature measurement range of the thermal imager is set to -20-500℃, the temperature measurement accuracy is ±0.5℃, and the sampling interval is 0.1s. The welding area is scanned by thermal imaging.

[0076] Because the contact resistance of the false welding area is large, the temperature is 5-10℃ higher than that of the normal area. The thermal imager transmits the temperature data to the industrial computer to generate a temperature distribution thermal map of the welding area.

[0077] The industrial computer fuses the visual defect data and infrared temperature data to determine the threshold value calculation criterion:

[0078] 5000 qualified samples are counted in advance, and the average value of the appearance defect parameters is , the standard deviation , the defect determination threshold

[0079] The average temperature of the normal welding area is =43℃, the standard deviation =0.7℃, and the temperature difference threshold .

[0080] According to the above threshold, three-level judgment is carried out:

[0081] Qualified: no appearance defect and the temperature difference between the welding area and the normal area is ≤ =43.7℃, 782 cells in this batch are determined to be qualified;

[0082] Unqualified: appearance defect parameter >0.065mm or temperature difference >45.1℃, 12 cells in this batch are determined to be unqualified and are automatically transferred to the isolation box by the mechanical arm.

[0083] To be rechecked: no appearance defects and 43.7℃ < temperature difference < 45.1℃, 6 pieces of battery in this batch are determined to be rechecked, and are transported to the artificial rechecking station. Through 200 times metallographic microscope, 5 pieces are finally determined to be qualified and 1 piece is determined to be unqualified.

[0084] S5, the qualified battery is transported to the cleanliness detection station. The high-frequency inductive detector is driven by the mechanical arm to keep a distance of 2mm±0.1mm from the surface of the battery, and is scanned at a uniform speed along the axis direction of the battery. The detector transmits the detection signal to the industrial computer. If the signal strength exceeds the preset threshold, it is determined that there is metal chip residue.

[0085] Please refer to Figure 4 In this batch, 3 pieces of battery are detected to have metal chip signals. The high-pressure nitrogen blowing device is started. The 3 symmetrically arranged nozzles blow air along the 45°±1° direction of the tab hole. The blowing pressure is set to 0.8MPa, and the blowing time is 2.5s. After blowing, the inductance detection is performed again after 1.5s interval. 2 pieces of battery are not detected to have metal chip signals and are determined to be qualified; 1 piece is still detected to have signals and is determined to be unqualified and is rejected.

[0086] After performance test, the performance statistics of 800 pieces of battery in this batch are shown in Table 1.

[0087] Table 1 Production and performance statistics table of Example 1 batch

[0088]

[0089] Example Two

[0090] Please refer to Figure 1 This example is aimed at the large capacity and long cycle demand of energy storage system. Square aluminum shell high rate lithium ion battery is selected as the manufacturing object. The tab is made of 1060 pure aluminum material. A total of 600 pieces of battery are produced. The specific steps are as follows:

[0091] S1, the laminated square lithium battery is transported to the adjustable tooling table. The thickness of the silicone buffer layer on the inside of the clamping plate is 1.5mm. The pressure sensor sets the clamping force to 0.9N. When the pressure reaches 0.9N, it automatically stops. The installation height of the binocular vision camera is adjusted to 450mm, the angle deviation is 3°, 3 frames of images are collected for each wavelength of the multi-spectral light source, and the light intensity is adjusted to 3000cd / m². The multi-spectral image fusion formula is consistent with that of Example 1. In the binocular disparity calculation, the camera focal length , the baseline distance , and the system calibration constant , the 3D model with a point cloud density of 105 points / mm2is generated, and the tab feature parameters are extracted: total thickness 0.75 mm, number of overlapping layers 5, length deviation 0.05 mm, and center coordinates (30.005, 22.003) mm.

[0092] The edge point set selects 60 points, and the reference line formula is consistent with that of Example 1. The solution is , and the length difference of each layer of the tab after pruning is 0.02 mm≤0.03 mm.

[0093] S2, the cutter spacing of the double-station trimming mechanism is adjusted to 12 mm, the tab thickness is 0.15 mm, and the shearing speed is set to 50 mm / s.

[0094] Please refer to Figure 2 , the negative pressure value of the negative pressure magnetic collection system for aluminum tabs is set to -55 kPa, the magnetic filter screen parameters are consistent with those of Example 1; the double-station switching time is 0.35 s, the blowback duration is 0.3 s, and the metal chip collection rate of this batch reaches 99.1%.

[0095] S3, the elastic pre-press head sets the pre-pressing force for 5-layer tabs to 5 N, and the fitting gap is 0.016 mm≤0.02 mm; the argon gas output is adjusted to 14 L / min, and the oxygen content is stabilized at 85 ppm.

[0096] Please refer to Figure 3 , the laser welding parameters: since the total thickness of the tab is 0.75 mm, the matching laser power is 180 W, the welding speed is 25 mm / s, the spot diameter is 0.3 mm, the sampling frequency of the displacement sensor is 1 kHz, the welding deviation is ≤0.01 mm, and the welding strength is all ≥18 N.

[0097] S4, the CNN model of the visual detection unit is trained with 120,000 defect samples, and the recognition accuracy is ±0.01 mm; the infrared thermal imager has a temperature measurement accuracy of ±0.4℃ and a sampling interval of 0.1 s.

[0098] In the threshold calculation, the mean value of the appearance defect parameters is , and the standard deviation is , ;

[0099] The mean value of the temperature of the normal welding area is =44℃, the standard deviation is =0.73℃, and the temperature difference threshold is .

[0100] The three-level judgment result is: 588 pieces are qualified, 8 pieces are unqualified, and 4 pieces are pending re-inspection. Among the re-inspection results, 3 pieces are qualified and 1 piece is unqualified.

[0101] S5, the detection parameters of the high-frequency inductance detector are consistent with those of embodiment 1, 4 cell cores are detected to have metal scraps, the blowing time is 3s, the re-detection is performed after an interval of 1s, 3 cell cores are qualified, 1 cell core is unqualified, and the final qualified rate is 98.2%.

[0102] Please refer to Figure 4 , after performance test: the 20C rate discharge capacity retention rate of the cell is 88%, the cycle life is 3000 times, the thermal runaway risk is reduced by 85%, the use demand of the energy storage system is met, and specific as shown in Table 2;

[0103] Table 2: Production and performance statistics of embodiment 2 batch

[0104]

[0105] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A safe and reliable high rate battery manufacturing method, characterized by, The method comprises the following steps: S1, the lithium battery cell after winding or lamination is clamped and fixed through an adjustable tooling table with pressure feedback, then a binocular vision camera and a multi-spectrum light source are used to scan the tab, three-dimensional point cloud modeling is combined to extract tab features and fit and trim the reference line, and the three-dimensional point cloud model generation and trimming reference line calculation specifically comprises: Multi-spectrum image fusion: blue light B, green light G and red light R images are fused into an enhanced image I according to weights, and the formula is: wherein, is an edge profile weight equal to 0.4, is an oxidation layer inhibition weight equal to 0.3, is a superposition layer identification weight equal to 0.3, and + + = 1. Binocular disparity calculation: the disparity d of the tab surface point (x, y) is calculated through left and right camera image matching to generate depth information Z, and the formula is: wherein is the camera focal length, is the binocular camera baseline distance, is the system calibration constant; Trimming the baseline fit: to the tab edge point set The least square method is used to fit a straight line as the baseline, the formula is: wherein is the slope, is the intercept, after fitting to ensure the length difference of each layer of tab after trimming is ≤0.03mm; S2, the tab is sheared through a double-station trimming mechanism, the trimming platform with chip guide grooves is used to cooperate with the negative pressure magnetic collection system to collect metal chips, and trimming and collection groove back blowing cleaning are synchronously performed through a double-station alternating mode; S3, the pre-pressing head is used to pre-press the trimmed tab and the cover plate, then argon is output through the annular nozzle to form a local protection atmosphere, then the laser welding parameters are adaptively matched according to the tab thickness, and the welding process is corrected in real time through the displacement sensor; S4, the welding area is detected through the visual detection unit and the infrared thermal imaging unit, then the two types of detection data are fused through the industrial computer, the three-level judgment of qualified, unqualified and re-inspection is performed according to the preset threshold, and the cell for re-inspection is confirmed again through the microscopic imaging equipment; S5, the internal residual metal chips of the cell are scanned through the high-frequency inductive detector, if the metal chip signal is detected, the cell can be obliquely blown away along the tab hole, then the inductive detection is performed again after blowing, the cell without detected metal chips enters the subsequent assembly link, and the cell with still detected metal chips is judged as unqualified and rejected.

2. The safe and reliable high rate battery manufacturing method according to claim 1, wherein The adjustable tooling table clamp in step S1 monitors the clamping force in real time through the pressure sensor, the clamping force needs to be controlled in the range of 0.5-1N, and the silicone buffer structure is used on the contact side of the clamping plate and the cell to avoid damage to the cell shell, and the clamping process is automatically stopped when the pressure reaches the set threshold.

3. The safe and reliable high rate battery manufacturing method of claim 1, wherein, The tab features in step S1 are the total thickness of the tab, the number of overlapping layers, the length deviation of each layer and the center coordinates, when the binocular vision camera and the multi-spectrum light source scan the tab, the resolution of the binocular vision camera is set to 2048x1536, the vertical distance from the tab is controlled in the range of 300-500mm, the angle deviation is not more than 5°, the multi-spectrum light source outputs 450nm±5nm blue light, 550nm±5nm green light and 650nm±5nm red light in turn, and 3-5 frames of tab surface images are collected under each wavelength light source.

4. The safe and reliable high rate battery manufacturing method of claim 1, wherein, The preset threshold calculation of the three-level judgment in step S4 is specifically: Appearance defect threshold: let the defect characteristic parameter set of qualified sample be , calculate the mean and standard deviation , and the defect determination threshold H is , and the detected defect parameter is > H. Temperature difference threshold value: let the temperature set of normal welding area be , calculate the mean and the standard deviation , and the temperature difference threshold value is: When the temperature difference between the detection area and the normal area is determined as temperature anomaly; The three-level judgment is divided into qualified, to be rechecked and unqualified. When the detection result is no appearance defect and temperature difference ≤ , it indicates qualified. When the detection result is no appearance defect and < temperature difference ≤ , it indicates to be rechecked. When the detection result is appearance defect parameter > H or temperature difference > , it indicates unqualified.

5. The safe and reliable high rate battery manufacturing method of claim 1, wherein, When the negative pressure magnetic collection is performed in step S2, the negative pressure value is set according to the tab material, -50 to -60kPa is used for aluminum tabs, and -70 to -80kPa is used for copper tabs, the magnetic filter screen aperture is 0.05mm, the magnetic strength is 1000Gs, and when the double-station alternating mode is used, the switching time of the trimming station and the cleaning station is not more than 0.5s, the back blowing cleaning adopts nitrogen with a pressure of 50kPa, and the duration is 0.3-0.5s.

6. The safe and reliable high rate battery manufacturing method of claim 1, wherein, The elastic coefficient of the elastic pre-pressing head in the step S3 is 5 N / mm, the pre-pressing force is set to 3 N when the number of overlapping layers of the tab is 2-3, and is set to 4-5 N when the number of overlapping layers of the tab is 4-6, the purity of the argon gas output by the annular nozzle is not less than 99.999%, and the gas output is controlled to be 10-15 L / min, so that the oxygen content in the welding area is reduced to below 100 ppm.

7. The safe and reliable high rate battery manufacturing method of claim 1, wherein In the step S3, the wavelength of the fiber laser used for laser welding is 1064 nm, the spot diameter is controlled to be 0.2-0.3 mm, the sampling frequency of the displacement sensor is 1 kHz during real-time correction of the path in the welding process, the deviation of the welding path is detected in real time, the position of the laser head is automatically adjusted for correction when the deviation exceeds 0.01 mm, and when the thickness of the tab is 0.1-0.3 mm, the laser power is set to 80-120 W and the welding speed is set to 30-50 mm / s, and when the thickness of the tab is 0.3-0.5 mm, the laser power is set to 150-200 W and the welding speed is set to 20-30 mm / s.

8. The safe and reliable high rate battery manufacturing method of claim 1, wherein, In the step S4, the camera resolution of the visual detection unit is 4096×3072, the annular light source with an illuminance of 5000 lux is used when collecting images, the deep learning CNN model is trained with not less than 100,000 samples containing welding bumps, cracks and lack of welding defects, and the recognition accuracy is ±0.01 mm, the thermal imager resolution of the infrared thermal imaging unit is 640×512, the temperature measurement range is -20-500℃, the accuracy is ±0.5℃, and the sampling interval is 0.1 s.

9. The safe and reliable high rate battery manufacturing method of claim 1, wherein, In the step S5, the detection frequency of the high-frequency inductive detector is 1 MHz, the sensitivity is 0.001 g, the distance between the detector and the surface of the battery cell is maintained to be 2 mm±0.1 mm during detection, the purity of the high-pressure nitrogen gas used for purging is not less than 99.99%, the three symmetrically arranged nozzles spray gas along the direction of 45°±2° of the tab hole during purging, and the inductive detection is performed again after purging for 1-2 s to ensure that there is no residual metal debris.

Images