A spaced tin plating device for battery module busbars

Abstract

The application provides a spaced type tin plating device for a battery module busbar, relates to the field of busbar processing, and comprises the steps that a target tin layer is divided into multiple sub-tin layers deposited in sequence, and a deposition strategy is dynamically selected during deposition to correct deposition deviation between layers in real time, state data of a busbar substrate or N sub-tin layers that have been deposited is collected; based on the state data collected by a sensor group, a deposition mode used for depositing an N+1 sub-tin layer is determined through a contour and temperature difference double-threshold judgment mechanism; the N+1 sub-tin layer is formed according to the determined deposition mode and a spraying compensation strategy generated based on the state data; the fusion degree and thickness compliance of a local area of the N+1 sub-tin layer are evaluated according to real-time collected deposition data; local area supplementary spraying is performed; and equipment intervention measures are output when an abnormality is detected; and the application improves the thickness uniformity of the tin layer and the interlayer fusion quality during direct spraying tin plating of the battery module busbar.

Description

Technical Field

[0001] This invention relates to the field of busbar processing technology, and in particular to an intermittent tin plating device for battery module busbars. Background Technology

[0002] Traditional battery module busbars often use intermittent tin plating to avoid material waste from large-area plating.

[0003] Related intermittent tin plating equipment typically employs direct injection of molten tin in conjunction with a mask to ensure the uniformity of the tin layer's thickness and shape.

[0004] However, the above equipment has the following shortcomings when applied to production lines for busbar processing:

[0005] 1. When the mask plate is attached to the busbar, misalignment may occur due to mechanical vibration or installation deviation, which will affect the tin plating effect.

[0006] Second, ordinary polymer masks are easily deformed by the high temperature of molten tin. High-temperature resistant metal masks are not well adapted to irregularly shaped busbars. Furthermore, different metal masks are required for different tin-plating patterns, which is not conducive to improving the tin-plating efficiency of the busbars. Summary of the Invention

[0007] The technical problem to be solved by the present invention is to provide an intermittent tin plating device for the busbar of a battery module. The device divides the target tin layer into multiple sub-tin layers, tin-plating them layer by layer, and dynamically selects the deposition strategy, dynamically controls the spraying path and the flow rate of molten tin during the tin plating process. The device identifies defects through a quality defect classification model and performs re-spraying operations to improve the uniformity of tin layer thickness and fusion quality, and reduce the defect rate and production cost.

[0008] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0009] An intermittent tin plating equipment for battery module busbars, with a spray-coated housing for maintaining ambient temperature during the tin plating process;

[0010] A sprayer is used to spray a tin layer onto the surface of the busbar.

[0011] A three-axis moving stage is used to drive the nozzle of the sprayer to move, and the nozzle of the sprayer is fixedly connected to the slide surface of the three-axis moving stage.

[0012] The sensor group is used to collect state data of the busbar substrate or the currently deposited N-layer tin sublayer. The state data includes image information, temperature distribution data and cumulative thickness data.

[0013] The controller is used to determine the deposition method used to deposit the N+1th sub-tin layer based on the state data and through a dual threshold judgment mechanism of profile and temperature difference; the deposition method is either a layer-by-layer uniform deposition method or an edge-anchor-fill deposition method; and generates a spraying compensation strategy based on the deposition method and the state data to form the N+1th sub-tin layer.

[0014] The spraying compensation strategy includes the motion path command of the sprayer nozzle, the motion path disturbance command, and the flow control command of the molten tin; during the execution of the spraying compensation strategy, the sensor group collects the deposition data of the N+1th sub-tin layer in real time.

[0015] The controller evaluates the degree of fusion and thickness conformity of a local area of ​​the N+1th sub-tin layer based on the deposition data; the controller controls the sprayer of the three-axis moving stage to perform local touch-up spraying on local areas with insufficient fusion or insufficient thickness;

[0016] The controller monitors the spraying pressure, molten tin flow rate, and nozzle trajectory deviation of the sprayer in real time. If an abnormality is detected, it outputs equipment intervention measures, including stopping spraying or issuing an equipment alarm.

[0017] Furthermore, the dual threshold judgment mechanism for contour and temperature difference includes:

[0018] The cumulative thickness distribution data of the Nth sub-tin layer acquired by the sensor group is discretized and analyzed, and the regional deviation between the current deposition thickness and the target thickness corresponding to the Nth layer is calculated using a thickness distribution difference analysis algorithm; based on the temperature distribution data, a multi-point temperature difference interpolation calculation algorithm is used to determine the temperature difference distribution in the local area of ​​the sub-tin layer.

[0019] When neither the calculated result of the regional deviation nor the calculated result of the temperature difference distribution exceeds the set threshold, the layer-by-layer uniform deposition method is selected. When either the calculated result of the thickness regional deviation or the temperature difference distribution exceeds the set threshold, the edge-anchor-fill deposition method is selected.

[0020] Furthermore, the process of forming the N+1th sub-tin layer includes:

[0021] When the layer-by-layer uniform deposition method is adopted, the controller controls the three-axis moving table to drive the nozzle of the sprayer to perform continuous spraying motion according to the motion path command;

[0022] When the edge-anchor-fill deposition method is used, the controller controls the three-axis moving stage to drive the nozzle of the sprayer to perform the following steps in sequence:

[0023] Step S301: Spray coating along a designated area on the surface of the busbar substrate to form a continuous closed tin boundary;

[0024] Step S302: Spraying a plurality of discretely distributed anchor point regions with thickening characteristics inside the closed boundary;

[0025] Step S303: Filling spraying is performed on the boundary and anchor point tin layer structure to complete the N+1th sub-tin layer;

[0026] The position, thickness, and shape of the anchor point area are adjusted based on the state data collected by the sensor group.

[0027] Furthermore, when using a layer-by-layer uniform deposition method, the motion path perturbation command is generated through the following steps:

[0028] Step S311: Obtain the cumulative thickness fluctuation data of the deposited tin sublayer;

[0029] Step S312: Perform a fast Fourier transform on the thickness fluctuation data to extract periodic feature parameters;

[0030] Step S313: Adaptively adjust the disturbance amplitude and frequency of the sprayer nozzle movement path according to the characteristic parameters, so that the new deposited layer forms a corrugated structure that enhances interlayer bonding.

[0031] Furthermore, the process of forming the N+1th sub-tin layer includes:

[0032] Based on the current state data collected by the sensor array and the preset target thickness and corrugation structure, the feature parameters of tin layer thickness deviation, corrugation structure and temperature distribution are obtained through feature vector extraction algorithm; the feature parameters are then dimensionality reduced by principal component analysis (PCA) algorithm to obtain a comprehensive feature vector including the proportion of thickness out-of-tolerance area, corrugation continuity and temperature-thickness overlap.

[0033] A multi-dimensional feature fusion algorithm is used to establish the correspondence between the comprehensive feature vector and the deposition quality index. The comprehensive feature vector is iteratively calculated multiple times through a closed-loop feedback algorithm to output motion path disturbance commands and dynamic control commands for molten tin flow rate to compensate for deposition thickness deviation and temperature fluctuation.

[0034] Furthermore, the evaluation of the fusion degree of the local region of the N+1th sub-tin layer includes:

[0035] Step S401: Perform image analysis on the interface region between the N+1th sub-tin layer and the Nth sub-tin layer or the busbar substrate.

[0036] Step S402: Use a multi-scale edge detection algorithm to identify the contour of the interface region and extract multi-dimensional geometric feature parameters;

[0037] Step S403: Based on the temperature distribution data collected in real time in the interface area, a spatial interpolation algorithm is used to reconstruct the regional temperature field, and a differential gradient algorithm is used to calculate the temperature gradient distribution characteristic parameters.

[0038] Step S404: Input the geometric feature parameters and temperature gradient distribution feature parameters into a multidimensional fusion algorithm. After feature parameter weight allocation and multiple rounds of iterative calculation, obtain a quantitative fusion degree evaluation value for accurately characterizing the fusion state of the interface region.

[0039] Furthermore, the position, thickness, and shape of the anchor point area are adjusted based on the temperature distribution data and thickness distribution data in the state data, including:

[0040] Step S302.1 Based on the real-time collected temperature distribution data of the tin layer deposition area, determine the areas with significant temperature differences within the tin layer deposition area through spatial interpolation and local temperature gradient algorithm;

[0041] Step S302.2: Combining the real-time acquired sub-tin layer deposition thickness distribution data, the thin-thickness areas are identified through a local thickness difference discrete analysis algorithm;

[0042] Step S302.3: Use a multivariate optimization search algorithm to perform comprehensive calculations on the identified areas with significant temperature differences and areas with thin thickness, and dynamically determine the precise spatial location of the anchor tin layer.

[0043] Step S302.4: The thickness of the anchor tin layer is then dynamically calculated using a thickness compensation algorithm, and the geometry of the anchor area is determined by combining the temperature distribution data, so as to achieve precise control of the anchor tin layer deposition area and stable improvement of deposition quality.

[0044] Furthermore, the criteria for judging insufficient fusion or insufficient thickness are as follows:

[0045] The fusion degree evaluation value and thickness compliance evaluation value of the local region of the N+1th sub-tin layer are input into the pre-trained quality defect classification model;

[0046] The quality defect classification model outputs a classification result indicating whether there are defects of insufficient fusion or insufficient thickness in the corresponding local area.

[0047] If the output indicates the presence of any defect, the local respraying operation is triggered.

[0048] The above-described solution of the present invention has at least the following beneficial effects:

[0049] The tin plating process on the busbars is achieved by directly spraying the tin onto the busbars through nozzles, thereby reducing the use of photomasks and increasing the flexibility of the intermittent tin plating on the busbars.

[0050] The overall tin plating layer is divided into multiple sub-tin layers and tin is plated layer by layer to ensure the uniformity of the overall tin plating layer;

[0051] By acquiring images, temperature, and thickness data of the busbar in real time, and using a dual threshold judgment mechanism of contour and temperature difference to dynamically select the deposition strategy, precise thickness and temperature control was achieved layer by layer.

[0052] By extracting feature vectors, fusing multi-dimensional features, and performing closed-loop feedback iterative calculations, nozzle paths and tin flow control commands are automatically generated, which significantly improves the uniformity of tin layer thickness and the stability of interlayer fusion.

[0053] By introducing a quality defect classification model, automatic defect identification and re-spraying are achieved, significantly reducing product defect rate and production costs. Attached Figure Description

[0054] Figure 1 This is a schematic diagram of an intermittent tin plating device for battery module busbars provided by the present invention.

[0055] Figure 2 This is a schematic diagram of a control unit for an intermittent tin plating equipment for battery module busbars provided by the present invention.

[0056] In the diagram, 101 is the spraying box; 102 is the conveying device; 103 is the three-axis moving table; 104 is the sprayer; and 105 is the busbar. Detailed Implementation

[0057] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0058] like Figures 1-2 As shown, an embodiment of the present invention proposes an intermittent tin plating device for battery module busbars, comprising: a spraying box 101 for maintaining ambient temperature during the tin plating process;

[0059] Sprayer 104 is used to spray a tin layer onto the surface of busbar 105;

[0060] The three-axis moving stage 103 is used to drive the nozzle of the sprayer 104 to move. The nozzle of the sprayer 104 is fixedly connected to the slide surface of the three-axis moving stage 103.

[0061] The sensor group is used to collect status data of the busbar 105 substrate or the currently deposited N-layer tin sublayer. The status data includes image information, temperature distribution data and cumulative thickness data.

[0062] The controller is used to determine the deposition method for depositing the N+1th sub-tin layer based on state data and through a dual threshold judgment mechanism of profile and temperature difference; the deposition method is either a layer-by-layer uniform deposition method or an edge-anchor-fill deposition method; and a spraying compensation strategy is generated based on the deposition method and state data to form the N+1th sub-tin layer.

[0063] The spraying compensation strategy includes the motion path command of the sprayer nozzle 104, the motion path disturbance command, and the flow control command of molten tin; during the execution of the spraying compensation strategy, the sensor group collects the deposition data of the N+1th sub-tin layer in real time.

[0064] Based on the deposition data, the controller evaluates the degree of fusion and thickness conformity of a local area of ​​the N+1th sub-tin layer; the controller controls the three-axis moving stage 103 and the sprayer 104 to perform local touch-up spraying on the local areas with insufficient fusion or insufficient thickness.

[0065] The controller monitors the spraying pressure, molten tin flow rate, and nozzle trajectory deviation of the sprayer 104 in real time. If an abnormality is detected, it outputs equipment intervention measures, including stopping spraying or equipment alarm.

[0066] In this embodiment of the invention, in the scenario of performing intermittent tin plating on the copper alloy busbar 105 of the battery module, the busbar 105 is conveyed to the interior of the spraying box 101 by the conveying device 102. The interior of the spraying box 101 is provided with 5 three-axis moving stages 103. The target tin layer is formed on the surface of the busbar 105 with a total thickness of 50μm, a rectangular shape of 100mm long and 80mm wide, and a quantity of 2 uniform tin layers. Each tin layer is divided into 5 sequentially deposited sub-tin layers, and the target thickness of each layer is 10μm. The sensor group includes a high-definition industrial camera, an infrared thermal imager, and a laser thickness gauge. The working principles and usage procedures of the high-definition industrial camera (not shown in the figure), the infrared thermal imager (not shown in the figure), and the laser thickness gauge (not shown in the figure) are known in the prior art and will not be described in detail here.

[0067] The sprayer 104 is equipped with a molten tin heating and heat preservation device to keep the molten tin in the storage bottle in a molten state at all times. The nozzle of the sprayer 104 is equipped with a variable flow regulating valve to realize dynamic flow control. The working principle and usage process of the sprayer 104 are known in the prior art and will not be described in detail here.

[0068] Local temperature control of the tin layer is achieved by emitting directional infrared energy to the target area through an infrared heater. The working principle and usage process of the infrared heater are well known in existing technology and will not be described in detail here.

[0069] The specific process includes:

[0070] When depositing the first sub-tin layer (N=0), a high-definition industrial camera is used to acquire surface image information of the 105 busbar substrate to determine the location of the target tin layer; an infrared thermal imager is used to acquire temperature distribution data of the substrate to understand the temperature differences in different areas of the substrate; and a laser thickness gauge is used to acquire surface flatness data of the substrate as initial thickness-related parameters. When depositing the second sub-tin layer (N=1), the corresponding state data of the first deposited sub-tin layer are acquired, including the surface image, temperature distribution, and actual cumulative thickness data of the sub-tin layer.

[0071] Based on state data, a dual threshold judgment mechanism of contour and temperature difference is used to determine the deposition method. Corresponding thresholds are set according to production accuracy requirements, such as a thickness deviation threshold of ±1μm and a temperature difference threshold of ±2℃. If the state data of the first sub-tin layer shows that its thickness distribution deviation from the target thickness is within ±1μm and the temperature difference in each region is within ±2℃, then a layer-by-layer uniform deposition method is selected when depositing the second sub-tin layer. If, before depositing the third sub-tin layer, a thickness deviation of 1.5μm is detected in the edge region of the second sub-tin layer, exceeding the set threshold, then an edge-anchor-fill deposition method is selected when depositing the third layer.

[0072] If a layer-by-layer uniform deposition method is adopted, the spraying compensation strategy generated based on the state data controls the nozzle to move at a constant speed along the surface of busbar 105 in a preset serpentine motion path. Simultaneously, based on the substrate temperature distribution, the flow rate of molten solder is appropriately reduced in slightly warmer areas and appropriately increased in slightly cooler areas. A small sinusoidal wave disturbance command is added to the motion path to induce slight fluctuations in the nozzle's smooth movement. During the spraying process, the deposition data of the third sub-tin layer is collected using real-time imaging equipment and a thickness monitoring sensor. If an edge-anchor-fill deposition method is adopted, the nozzle is controlled according to the corresponding strategy to first complete edge spraying, then form the anchor area, and finally fill the gap, while simultaneously collecting deposition data in real time.

[0073] Based on the real-time collected deposition data of the third sub-tin layer, the degree of fusion between it and the second sub-tin layer in the local area is evaluated, and the presence of obvious delamination or bubbles at the interface is observed. At the same time, the degree of conformity between the actual thickness of each local area and the target thickness is evaluated to determine whether there is any thickness deviation.

[0074] For example, if it is found that the fusion degree between a corner area of ​​the third sub-tin layer and the second sub-tin layer is insufficient, and the thickness of the area is lower than the target value (e.g., 2 μm lower), then the temperature of the area is heated to the semi-molten state of tin (approximately 240°C, 240°C in this embodiment) by a local heating device, and then the nozzle is controlled to perform local supplementary spraying on the area to improve the fusion effect and make up for the insufficient thickness.

[0075] Throughout the entire process of spraying the third tin layer, the spraying pressure (in this embodiment, the normal range is set to 0.3-0.5 MPa), molten tin flow rate (in this embodiment, the normal range is set to 5-8 ml / s), and nozzle trajectory deviation (in this embodiment, the normal range is set to ±0.1 mm) are monitored in real time. If the spraying pressure suddenly drops to 0.2 MPa, equipment intervention measures are immediately implemented, an equipment alarm is issued, and spraying is suspended until the operator troubleshoots the problem before resuming operation.

[0076] The above steps are used to deposit five sub-tin layers in sequence, and finally a spaced tin plating layer that meets the requirements is formed on the surface of busbar 105.

[0077] The dual threshold judgment mechanism based on contour and temperature difference includes:

[0078] The cumulative thickness distribution data of the Nth sub-tin layer acquired by the sensor group is discretized and analyzed, and the regional deviation between the current deposition thickness and the target thickness corresponding to the Nth layer is calculated using a thickness distribution difference analysis algorithm; based on the temperature distribution data, a multi-point temperature difference interpolation algorithm is used to determine the temperature difference distribution in the local area of ​​the sub-tin layer.

[0079] When the calculated results of regional deviation and temperature difference distribution do not exceed the set threshold, the layer-by-layer uniform deposition method is selected. When either the calculated results of thickness regional deviation or temperature difference distribution exceed the set threshold, the edge-anchor-fill deposition method is selected.

[0080] In this embodiment of the invention, before depositing the third sub-tin layer (N=2) on the battery module busbar 105, a dual threshold judgment mechanism is executed to determine the deposition method:

[0081] Thickness deviation calculation:

[0082] The surface of the second tin sublayer was divided into a 100×100 grid-like detection area (a total of 10,000 small areas). The actual thickness of each grid center was measured using a laser thickness gauge. The target thickness of the second layer was 10 μm. The thickness deviation was calculated by subtracting the target thickness of 10 μm from the actual measured thickness of each grid. The result was the thickness deviation value of that grid.

[0083] For example, if the actual measured thickness of a grid in the edge region is 11.2 μm, subtracting the target thickness of 10 μm from 11.2 μm yields a thickness deviation of +1.2 μm for that grid. Similarly, if the actual measured thickness of an adjacent grid is 11.8 μm, subtracting 10 μm from 11.8 μm also yields a thickness deviation of +1.8 μm. This subtraction calculation is performed on each of the 396 grids in the edge region to determine the thickness deviation range for each grid.

[0084] For example, during the calculation of each of the 396 grids in the edge region, it was found that the deviation of 32 grids was between +1.2μm and +1.8μm (the deviation of the remaining 364 edge grids was within ±1μm); among the 9604 grids in the middle region, one grid was measured to be 9.5μm with a deviation of -0.5μm, another grid was measured to be 10.5μm with a deviation of +0.5μm, and the deviation of all the middle grids was within -0.5μm to +0.5μm.

[0085] The maximum thickness deviation was found to be +1.8 μm (from one of the 32 out-of-tolerance grids in the edge region).

[0086] Temperature difference distribution calculation:

[0087] Eighty temperature monitoring points (30 at the edge and 50 in the center) were evenly distributed on the surface of the second tin layer. The real-time temperature of each point was measured using an infrared thermometer. The reference temperature was set to 260℃ (this temperature is the optimal value within the working temperature range of molten solder). The temperature difference was calculated by subtracting the reference temperature of 260℃ from the measured temperature of each monitoring point. The result was the temperature difference value of that monitoring point.

[0088] For areas not directly monitored, the temperature difference is estimated based on the temperature difference between adjacent monitoring points, following the principle that the temperature change in adjacent areas shows a gradual trend. For example, if the temperature difference between two adjacent monitoring points is +2.3℃ and +2.5℃ respectively, the temperature difference between the two areas is estimated to be around +2.4℃ (taking the average of the two).

[0089] For example, the measured temperatures at the five detection points in the upper left corner of the edge region ranged from 263.3℃ to 265.2℃. Based on the estimation of the temperature difference across the 15 grids in this region (including the detection points and adjacent areas), the temperature difference ranged from +3.3℃ to +5.2℃. The temperature difference for the remaining 785 edge grids and all the middle grids was calculated and estimated to range from -2.2℃ to +2.2℃. Thus, the maximum temperature difference was calculated to be +5.2℃ (from the upper left edge region).

[0090] The set thickness deviation threshold is ±1μm, and the temperature difference threshold is ±2℃. Because the maximum thickness deviation is 1.8μm (exceeding the set thickness deviation threshold) and the maximum temperature difference is 5.2℃ (exceeding the set temperature threshold), the third layer is determined to adopt the edge-anchor-fill deposition method.

[0091] Before depositing the fourth tin sublayer (N=3), the same operation was performed on the third tin sublayer:

[0092] For example, the thickness deviation of the 396 edge grids was calculated to be between -0.8μm and +0.8μm, while the deviation of the 9604 middle grids was within ±0.6μm, with a maximum deviation of +0.8μm (not exceeding the set thickness deviation threshold).

[0093] The temperature difference in all areas was calculated and estimated to be between -1.5℃ and +1.5℃ (not exceeding the set temperature threshold), so the fourth layer adopted a layer-by-layer uniform deposition method.

[0094] The process of forming the N+1th sub-tin layer includes:

[0095] When a layer-by-layer uniform deposition method is adopted, the controller controls the three-axis moving table 103 to drive the nozzle of the sprayer 104 to perform continuous spraying motion according to the motion path instruction.

[0096] When using the edge-anchor-fill deposition method, the controller controls the three-axis moving stage 103 to drive the nozzle of the sprayer 104 to perform the following steps in sequence:

[0097] Step S301: Spray coating is performed along a designated area on the surface of the busbar 105 substrate to form a continuous closed tin boundary.

[0098] Step S302: Spraying a coating inside the closed boundary to form multiple discretely distributed anchor point regions with thickening characteristics;

[0099] Step S303: Fill spraying is performed on the boundary and anchor point tin layer structure to complete the N+1th sub-tin layer;

[0100] The position, thickness, and shape of the anchor point area are adjusted based on the state data collected by the sensor group.

[0101] In this embodiment of the invention, when the third tin sublayer (N=2) is deposited using an edge-anchor-fill deposition method, the specific process is as follows:

[0102] Step S301: Spraying along the edge of the rectangular tin layer to form a continuous and closed tin boundary. The nozzle starts from the upper left corner edge of the rectangle and moves along the 100mm×80mm rectangular outline at a speed of 35mm / s and a spraying flow rate of 8.5ml / s to form a closed tin boundary with a width of 2.2mm, ensuring that the boundary is continuous and without gaps, providing a stable framework for subsequent processes.

[0103] In step S302, multiple discretely distributed anchor point regions are formed by spraying within the closed boundary. For example, based on the state data collected by the sensor group, four thin areas are found in the second sub-tin layer (located slightly to the right of the center of the rectangle, inside the upper left corner, inside the lower right corner, and inside the center of the side, with a target thickness of 1.3 μm to 1.7 μm lower) and three temperature anomaly areas (temperature 2.2℃ to 3.1℃ lower than the surrounding area). Based on this, 10 anchor point regions are set within the closed boundary, with four corresponding to the thin areas and three corresponding to the temperature anomaly areas, and three evenly distributed to enhance the structure. The anchor point regions are squares with a side length of 3.5 mm. The nozzle reciprocates in each anchor point region, with the speed reduced to 2.2 mm / s, and the molten tin spraying flow rate increased to 10.5 ml / s, so that the thickness of the anchor point region reaches 12.5 μm to 13.5 μm, which is 2.5 μm to 3.5 μm thicker than the target single layer thickness.

[0104] Step S303: Filling spraying is performed on the boundary and anchor point base. The nozzle performs scanning spraying in the boundary and anchor point area at a speed of 42 mm / s and a flow rate of 7.5 ml / s. The path avoids the anchor point area to ensure that the filled part is well integrated with the boundary and anchor point. After filling, the thickness of the area other than the anchor point reaches about 10 μm. Finally, the deposition of the third tin sublayer is completed, making the overall thickness of the third layer uniform and tightly bonded to the lower layer.

[0105] When the fourth tin sublayer (N=3) is deposited using a layer-by-layer uniform deposition method:

[0106] The target is a rectangular sub-tin layer with a length of 100mm, a width of 80mm, and a thickness of 10μm. The nozzle is controlled to continuously spray along a preset zigzag path, starting from the upper left corner of the rectangular tin layer and moving along its length. The spacing between each row is 0.4mm, the overlap between rows is 15%, the nozzle speed is 45mm / s, and the molten tin spraying flow rate is stabilized at 5.5ml / s. Throughout the process, the nozzle moves continuously without interruption. Through continuous and uniform spraying, the thickness of the fourth sub-tin layer is evenly distributed within the 100mm×80mm rectangular area, achieving the target thickness of 10μm.

[0107] When using a layer-by-layer uniform deposition method, the motion path perturbation command is generated through the following steps:

[0108] Step S311: Obtain the cumulative thickness fluctuation data of the deposited tin sublayer;

[0109] Step S312: Perform a fast Fourier transform on the thickness fluctuation data to extract periodic feature parameters;

[0110] Step S313: Adaptively adjust the disturbance amplitude and frequency of the nozzle movement path of the sprayer 104 according to the characteristic parameters, so that the new deposited layer forms a corrugated structure that enhances the interlayer bonding.

[0111] In this embodiment of the invention, when depositing the fourth sub-tin layer, a layer-by-layer uniform deposition method is adopted, and the nozzle movement path disturbance command generation process is as follows:

[0112] Step S311: The overall thickness of the first three sub-tin layers is scanned using a laser thickness gauge. Specifically, the rectangular tin layer area with a length of 100 mm and a width of 80 mm is divided into 100×100 small grids. The actual cumulative thickness at the center of each grid is measured, and the difference between the thickness of each grid and the target thickness (cumulative thickness of 30 μm) is calculated to obtain the thickness fluctuation data of the area.

[0113] For example, through thickness data analysis, it was found that the tin layer has periodic thickness fluctuations with a spatial interval of 2 mm in the length direction (the maximum fluctuation amplitude is ±0.8 μm), and a similar fluctuation feature with a spatial interval of 3.5 mm in the width direction.

[0114] Step S312: All collected thickness fluctuation data are concatenated into a continuous thickness fluctuation data sequence according to the row scanning order. Frequency analysis is performed on the thickness fluctuation data using the Fast Fourier Transform (FFT) method to determine the main periodic characteristics of the thickness fluctuations. Specifically, the frequency spectrum of the thickness fluctuation data is obtained after frequency domain transformation. Significant peaks appearing in the frequency spectrum are analyzed to determine the dominant frequency components of the thickness fluctuations, thereby identifying the spatial period corresponding to the thickness fluctuation characteristics. For example, along the length direction, there is a fluctuation with a cycle of 2 mm, and along the width direction, there is a fluctuation with a cycle of 3.5 mm. These two periodic characteristics correspond to spatial frequencies of 0.5 cycles / mm and approximately 0.29 cycles / mm, respectively.

[0115] Step S313: Based on the extracted periodic features, plan the motion disturbance command of the nozzle path to effectively reduce the existing thickness fluctuation of the first 3 sub-tin layers.

[0116] Specifically, the nozzle superimposes a periodic disturbance of a certain amplitude on the original serpentine motion path, that is:

[0117] Along the length direction, periodic fluctuations with a spatial frequency of 0.5 cycles / mm and a disturbance amplitude of 0.1 mm are superimposed;

[0118] In the width direction, there is a periodic fluctuation with a spatial frequency of approximately 0.29 cycles / mm and a perturbation amplitude of 0.08 mm.

[0119] By adjusting the nozzle path in this way, the fourth sub-tin layer deposited by the nozzle will be spatially misaligned with the existing periodic thickness fluctuations below, i.e., the distribution of peaks to troughs, thereby achieving effective complementarity of thickness fluctuations between adjacent tin layers, resulting in a significant improvement in overall thickness uniformity.

[0120] The process of forming the N+1th sub-tin layer includes:

[0121] Based on the current state data collected by the sensor array and the preset target thickness and corrugation structure, the feature parameters of tin layer thickness deviation, corrugation structure and temperature distribution are obtained through feature vector extraction algorithm; the feature parameters are then dimensionality reduced by principal component analysis (PCA) algorithm to obtain a comprehensive feature vector including the proportion of thickness out-of-tolerance area, corrugation continuity and temperature-thickness overlap.

[0122] A multi-dimensional feature fusion algorithm is used to establish the correspondence between the comprehensive feature vector and the deposition quality index. The comprehensive feature vector is iteratively calculated multiple times through a closed-loop feedback algorithm to output motion path disturbance commands and dynamic control commands for molten tin flow rate to compensate for deposition thickness deviation and temperature fluctuation.

[0123] In this embodiment of the invention, multi-dimensional feature analysis is performed based on the state data of the fourth sub-tin layer collected by the sensor group to optimize the deposition quality of the next layer. The specific analysis process is as follows:

[0124] First, a laser thickness gauge was used to divide the surface of busbar 105 into 100×100 grid points, and the actual thickness at each point was measured. The actual measured thickness at each grid point was subtracted from the target thickness (10μm) to calculate the thickness deviation value for each grid point. Then, the thickness deviation values ​​of all grid points were analyzed to determine the following key characteristic parameters: the maximum thickness deviation (e.g., +1.5μm, appearing in the upper right corner area), the arithmetic mean of the overall thickness deviation (e.g., +0.8μm), and the percentage of grid points with thickness deviations exceeding ±1μm (e.g., 30% in the upper right corner area).

[0125] Secondly, the tin layer surface is scanned using a surface profilometer to obtain a sequence of two-dimensional coordinate data of the profile, which is then unfolded along the length direction into one-dimensional height distribution data. The main wavelength (e.g., 2.2 mm) and amplitude (e.g., 0.08 mm) of the tin layer surface corrugation structure are determined using a fast Fourier transform analysis method. Furthermore, the corrugation continuity parameter is determined by the proportion of the length of the continuous corrugation region to the total measured length.

[0126] Next, the temperature field distribution of the fourth tin sublayer is captured using an infrared thermal imager to obtain temperature values ​​for each region and determine temperature characteristic parameters, including the highest temperature value (e.g., 265℃), the lowest temperature value (e.g., 257℃), and the area percentage of regions with a temperature difference exceeding 2℃ (e.g., the upper right corner area accounts for approximately 30%). Then, the spatial overlap between regions with a temperature difference exceeding 2℃ and regions with a thickness deviation exceeding ±1μm is calculated (e.g., an overlap of 75%).

[0127] All the characteristic parameters obtained from the above analysis (including thickness, ripple, and temperature-related parameters) together form a complete set of characteristic parameters, which will be used for the next step of multidimensional feature fusion analysis.

[0128] The multidimensional feature fusion algorithm is implemented using principal component analysis (PCA). The specific implementation process is as follows: First, the key feature parameters extracted from the current tin layer state data, including the maximum thickness deviation, average thickness deviation, proportion of thickness deviation areas (referring to the area ratio of grid regions where the thickness deviation exceeds a set threshold ±1μm), ripple wavelength, ripple amplitude, ripple continuity (a quantitative indicator characterizing the continuity of the ripple structure, such as the proportion of continuous ripple length to the total measured length), highest temperature, lowest temperature, proportion of temperature difference deviation areas (referring to the area ratio of grid regions where the temperature difference exceeds a set threshold ±2℃), and overlap between temperature and thickness deviation areas (referring to the area ratio of grid regions where the temperature difference deviation areas and thickness deviation areas overlap spatially), are standardized and converted into dimensionless values ​​to ensure the comparability of parameters of different dimensions and magnitudes.

[0129] Subsequently, principal component analysis (PCA) was used to reduce the dimensionality of these standardized feature parameters. During dimensionality reduction, the minimum number of principal components required for a cumulative variance contribution rate exceeding 85% (usually the first 2-3 principal components) was selected as the comprehensive feature vector to significantly reduce data complexity while retaining the main information of the original data.

[0130] Next, a mapping relationship is established between the comprehensive feature vector and the deposition quality index. The deposition quality index specifically refers to the uniformity of tin layer thickness (e.g., the proportion of grid area with thickness deviation within ±1 μm) and temperature stability (e.g., the proportion of grid area with temperature difference within ±2℃). A nonlinear regression model (random forest regression in this embodiment) is used to establish the mathematical correspondence between the aforementioned comprehensive feature vector (input) and the deposition quality index (output). The mapping relationship between the comprehensive feature vector and the deposition quality index aims to predict the impact of different spraying parameter adjustment strategies on the final deposition quality index.

[0131] In this embodiment, the construction process of the random forest regression model is as follows: First, process data from multiple batches of the actual battery module busbar tin plating production line were collected and organized, totaling approximately 1200 valid samples. Each sample recorded its characteristic parameters and corresponding deposition quality indicators. Specifically, the characteristic parameters of each sample include a comprehensive feature vector after dimensionality reduction using PCA. These samples cover a wide range of different spraying parameter combinations and corresponding deposition quality indicators to ensure the representativeness and generalization performance of the data.

[0132] The specific parameters for the random forest regression model were selected as follows: the number of decision trees was set to 120, the maximum depth of each decision tree was controlled at 8 levels, and the minimum number of samples per tree node split was limited to 8. These parameters were optimized using a 10-fold cross-validation method. Specifically, all sample data were first randomly divided into a training set (80%, i.e., 960 groups) and a validation set (20%, i.e., 240 groups); then, cross-validation optimization was performed on the number of decision trees (ranging from 80 to 150), tree depth (ranging from 5 to 12 levels), and minimum number of samples per node (ranging from 5 to 15), selecting the parameter combination that yielded the best prediction performance on the validation set. This process continued until the optimized model achieved a prediction accuracy of over 92% on the validation set (measured by the coefficient of determination between predicted and actual values).

[0133] To further clarify the influence of each feature parameter on sedimentation quality indicators, a sensitivity analysis was conducted on the input features of the random forest regression model. The specific method is as follows: First, 200 independent validation samples were randomly selected from actual production data. Each sample contained the aforementioned dimensionality-reduced comprehensive feature vector (proportion of thickness out-of-range areas, ripple continuity, temperature-thickness overlap) and the corresponding sedimentation quality indicators (thickness uniformity and temperature stability). Then, a feature-by-feature perturbation method was used for sensitivity analysis, i.e., a ±10% numerical perturbation was applied to each individual feature parameter sequentially, while the remaining feature parameters remained unchanged. Subsequently, the perturbed feature vector was re-input into the trained random forest regression model, and the magnitude of the change in the prediction results caused by the perturbation was calculated. The mean absolute percentage of the change in the prediction results (MAPE) was used as the quantitative evaluation standard for the sensitivity of this feature to the model output.

[0134] When a ±10% perturbation is applied to the proportion of areas with excessive thickness, the average change in the model-predicted deposition quality indicators is controlled within 10%; when a ±10% perturbation is applied to the ripple continuity, the average change in the model-predicted results is controlled within 7%; and when a ±10% perturbation is applied to the temperature-thickness overlap, the average change in the model-predicted results is controlled within 5%. When the sensitivity analysis shows the above results, and the model prediction errors caused by the feature perturbations are all controlled within an acceptable range, it indicates that the decision-making provided by the comprehensive feature vector and the corresponding spraying parameter optimization and control strategy proposed in this embodiment is reliable.

[0135] Finally, based on the established mapping relationship between the comprehensive feature vector and the deposition quality index (using a pre-trained random forest regression model), the nozzle motion path perturbation amplitude and the molten tin flow control coefficient are adjusted through closed-loop iterative optimization. The specific steps are as follows:

[0136] Set initial values ​​for the nozzle motion path disturbance amplitude and the molten tin flow control coefficient (e.g., initial disturbance amplitude references the previous sublayer's set value or standard value of 0.1 mm, and initial flow adjustment coefficient is set to 10%). Using the mapping relationship, predict the theoretical tin layer thickness and temperature distribution to be deposited with this set of parameters. Compare the predicted thickness and temperature distribution with target values ​​(e.g., target thickness ±1 μm, temperature fluctuation ±2 °C). If there are localized areas where the predicted thickness deviation or temperature fluctuation exceeds the allowable range, adjust the parameter combination according to preset rules (e.g., disturbance amplitude adjusted by ±0.01 mm each time, flow adjustment coefficient adjusted by ±1%). Repeat the iterative calculation process of "prediction-comparison-adjustment" until the predicted thickness and temperature distributions both meet the preset quality threshold requirements (i.e., the area ratio of grid regions with thickness deviation within ±1 μm and temperature difference within ±2 °C both reach the target value, e.g., both >95%). The final determined optimal nozzle motion path disturbance amplitude and molten tin flow control coefficient will be used for actual spraying control to compensate for the thickness deviation and temperature fluctuation of the current deposited layer.

[0137] The evaluation of the fusion degree of the local region of the N+1th sub-tin layer includes:

[0138] Step S401: Perform image analysis on the interface region between the N+1th sub-tin layer and the Nth sub-tin layer or the busbar 105 substrate.

[0139] Step S402: Use a multi-scale edge detection algorithm to identify the contour of the interface region and extract multi-dimensional geometric feature parameters;

[0140] Step S403: Based on the temperature distribution data collected in real time in the interface area, the regional temperature field is reconstructed using a spatial interpolation algorithm, and the temperature gradient distribution characteristic parameters are calculated using a differential gradient algorithm.

[0141] Step S404: Input the geometric feature parameters and temperature gradient distribution feature parameters into the multidimensional fusion algorithm. After feature parameter weight allocation and multiple rounds of iterative calculation, obtain a quantitative fusion degree evaluation value for accurately characterizing the fusion state of the interface region.

[0142] In this embodiment of the invention, the specific process for evaluating the fusion degree of the interface between the third sub-tin layer and the second sub-tin layer is as follows:

[0143] Step S401: Use a high-definition industrial camera to capture images of the interface area of ​​the two sub-tin layers. Select five feature areas, including the center and edge of the interface, and take three microscopic images of each area with a resolution of 2048×1536 pixels (magnification of 500x) to ensure that the details of the tin layer bonding at the interface are clearly presented.

[0144] Step S402: First, acquire a two-dimensional microscopic image of the tin layer interface region. After converting the image to grayscale, process it using a multi-scale edge detection method. Specifically, first, perform Gaussian filtering at different scales, i.e., blurring with standard deviations of 1.0, 1.5, and 2.0 respectively, to reduce image noise. Then, use the Canny edge detection method to extract the boundary contours of the blurred images at each scale, obtaining a series of interface contour maps. Next, fuse the different contours obtained at multiple scales to obtain a complete and continuous interface boundary contour. Based on the fused contours, extract multi-dimensional geometric feature parameters of the interface region, including the total length of the interface contour (e.g., 5.2 mm), average roughness (e.g., 0.8 μm), maximum difference in concavity and convexity height of the contour (e.g., 2.3 μm), and the number of interlocking points per unit area (e.g., 12 / mm). 2 The average interlocking depth (e.g., 1.5 μm) was used to further quantify the interface fusion state.

[0145] Step S403: Deploy 80 temperature monitoring points in the interface area using an infrared thermometer to collect real-time temperature data (e.g., range 258-264℃); process the discrete temperature points using the Kriging space interpolation algorithm to reconstruct a complete temperature field with a 100×100 grid; calculate the temperature gradient using the differential gradient algorithm, such as the maximum gradient value (1.2℃ / mm, located at the upper right edge), the average gradient (0.5℃ / mm), and the gradient direction consistency parameter (0.72), forming a set of temperature gradient distribution characteristic parameters.

[0146] In this embodiment, when using the Kriging spatial interpolation algorithm to process discrete temperature points to obtain a complete temperature field, the specific operation process is as follows: First, 80 temperature measurement points are evenly distributed within the deposition area of ​​busbar 105 using an infrared thermometer. 30 measurement points are arranged at the edge of busbar 105, and 50 measurement points are arranged in the middle area. The spacing between each measurement point is controlled between 8 and 12 mm to ensure the spatial representativeness of the measurement data. When the number of collected measurement data points reaches 80, and the spatial distribution density of the measurement points meets the spacing requirement of 8 to 12 mm, it indicates that the temperature data provided by this spatial temperature sampling strategy is reliable for the subsequent temperature field reconstruction process.

[0147] Next, the spherical variogram was chosen as the core model of the Kriging interpolation algorithm, and the nugget value was set at 0.5℃. 2 The sill value is 3.5℃. 2 Spatial interpolation of temperature data was performed under a parameter combination with a range of 15 mm. When the interpolation prediction error (root mean square error between the predicted and actual measured values) obtained under the above parameter combination is controlled within 1℃, it indicates that the determined Kriging interpolation algorithm parameters provide reliable decision-making for the complete temperature field reconstruction. Under this condition, temperature data from 80 discrete measurement points were interpolated and reconstructed into a complete continuous temperature field with a 100×100 grid and a spatial resolution of 1 mm between grid cells.

[0148] Subsequently, to calculate the spatial gradient of the temperature field, this embodiment employs a first-order central difference gradient algorithm. Specifically, for each grid point, the average temperature difference between its four adjacent grid points (up, down, left, and right) is used for gradient calculation. When the gradient calculation method is selected as the first-order central difference algorithm and the grid resolution is determined to be 1 mm, it indicates that the temperature gradient calculation results obtained using this difference method are reliable in terms of spatial resolution and accuracy. For example, in this embodiment, the maximum temperature gradient is 1.2℃ / mm, occurring near the upper right edge of the tin layer region, with an average gradient of 0.5℃ / mm.

[0149] Furthermore, this embodiment defines a gradient direction consistency parameter as an indicator characterizing the regularity of the temperature gradient distribution, specifically the proportion of grid points where the angle between the temperature gradient direction and the dominant direction is less than 15°. When the calculation results show that the gradient direction consistency parameter is between 0.70 and 0.75, it indicates that the reconstructed temperature field in this embodiment has a clear and stable gradient distribution direction, and the temperature gradient characteristic parameter set provides reliable decision-making for spraying parameter optimization and control strategies. In this embodiment, the gradient direction consistency parameter is set to 0.72, satisfying the aforementioned reliability condition.

[0150] Step S404: The multidimensional geometric feature parameters obtained in step S402 are fused and analyzed with the temperature gradient feature parameters obtained in step S403. The specific process is as follows:

[0151] First, a hierarchical evaluation structure is established, dividing all parameters into two main categories: geometric feature parameters and temperature gradient parameters. Then, based on the importance of the fusion state assessment of the tin layer interface, a pairwise comparison method is used to rank the parameters by importance, determining their relative order of importance. On this basis, a corresponding judgment matrix is ​​established. Matrix operations are performed by comparing the elements in the matrix to calculate and obtain the weight values ​​of each parameter. For example, through the above matrix calculation process, the overall weight of geometric feature parameters is determined to be approximately 60%, with the interface interlocking depth having the greatest impact on fusion degree, accounting for 25%; the overall weight of temperature gradient parameters is 40%, with the maximum temperature gradient accounting for 15%.

[0152] Subsequently, the parameters with determined weights are comprehensively evaluated: first, a preliminary fusion score is calculated using a weighted summation method; then, based on the correlation strength between the preliminary fusion score and each parameter, multiple rounds of iterative calculations are performed through gradual correction; in each iteration, the weights of key parameters in areas with lower scores are fine-tuned and optimized based on the fusion score calculated in the previous iteration, and then the next round of weighted summation calculation is performed; after multiple iterations, the calculation stops when the difference between two consecutive fusion scores is no greater than a set threshold (e.g., ±1 point); typically, stability and reliability of the fusion score can be achieved after 3 rounds of the above iterative calculations, ultimately obtaining an accurate and reliable evaluation value of the interface area fusion.

[0153] In another preferred embodiment of the present invention, when the Nth layer adopts an edge-anchor-fill deposition method and the N+1th layer adopts a layer-by-layer uniform deposition method, in order to ensure a good structural connection between the upper and lower layers, the embodiment of the present invention adopts the following corrugated connection method:

[0154] First, surface contour scanning is performed at the junction of the local thickness protrusion formed at the edge and anchor point region of the Nth layer and the filling region to extract the ripple contour data of the boundary between the anchor point region and the filling region.

[0155] Then, based on the extracted contour features, Fourier transform spectral analysis was used to determine the boundary periodic ripple structure parameters of the Nth layer, including wavelength and amplitude.

[0156] Next, the above-mentioned corrugated structure parameters of the Nth layer are input into the nozzle motion path planning of the N+1th layer, so that the corrugated disturbance structure of the N+1th layer is precisely matched with the protruding structure of the edge and anchor point area of ​​the Nth layer, so as to achieve the staggered and complementary connection between the crests and troughs, thereby ensuring the stability of the smooth transition and fusion of the tin layer structure between the two different deposition methods.

[0157] The location, thickness, and shape of the anchor point area are adjusted based on the temperature and thickness distribution data in the status data, including:

[0158] Step S302.1 Based on the real-time collected temperature distribution data of the tin layer deposition area, determine the areas with significant temperature differences within the tin layer deposition area through spatial interpolation and local temperature gradient algorithm;

[0159] Step S302.2: Combining the real-time acquired sub-tin layer deposition thickness distribution data, the thin-thickness areas are identified through a local thickness difference discrete analysis algorithm;

[0160] Step S302.3: Use a multivariate optimization search algorithm to perform comprehensive calculations on the identified areas with significant temperature differences and areas with thin thickness, and dynamically determine the precise spatial location of the anchor tin layer.

[0161] Step S302.4: The thickness of the anchor tin layer is then dynamically calculated using a thickness compensation algorithm, and the geometry of the anchor area is determined by combining the temperature distribution data, so as to achieve precise control of the anchor tin layer deposition area and stable improvement of deposition quality.

[0162] In this embodiment of the invention, the adjustment process of the anchor point region during the deposition of the third tin sublayer (target thickness 10 μm) is as follows:

[0163] Step S302.1: Use an infrared thermometer to evenly distribute 80 temperature monitoring points in the rectangular tin layer area to collect the surface temperature data of the second sub-tin layer in real time;

[0164] Using the Kriging space interpolation algorithm, the data from 80 discrete temperature measurement points are reconstructed into a continuous temperature field of 100×100 grids, with each grid point corresponding to a specific temperature value and spatial coordinates.

[0165] Then, to obtain the temperature change rate at each grid point, the spatial spacing between grid points is used as a reference (in this embodiment, the spatial distance between adjacent grid points is 1 mm, and the specific value is determined), and the following method is used for calculation:

[0166] Centered on a given grid point, observe the temperature values ​​of the four directly adjacent grid points (up, down, left, and right). Calculate the absolute value of the temperature difference between each adjacent grid point and the center point. Then, divide each temperature difference by the spatial distance (1 mm) between the two grid points to obtain the rate of temperature change in each direction. Compare these rates of temperature change in the four directions and select the maximum value as the rate of temperature change for the center grid point. Repeat this method to calculate the rate of temperature change for all grid points.

[0167] For example, in this region, the temperature of a certain grid point is 256.0℃, while the temperature of another grid point adjacent to it (at a distance of 1mm) is 257.8℃. The temperature difference between the two points is 1.8℃, so the temperature change rate of the central grid point is 1.8℃ / mm.

[0168] The gradients of other regions were calculated using the above method. For example, the gradient of the region slightly to the right of the center was 1.5℃ / mm, and the gradient of the region at the lower right edge was 1.6℃ / mm.

[0169] Step S302.2: Combining the thickness distribution data (100×100 grid) of the second sub-tin layer acquired by the laser thickness gauge, the data is processed using a local thickness difference discretization analysis algorithm: the difference between the actual thickness and the target thickness of 10μm for each grid is calculated, and then the discretization analysis is performed on the difference between adjacent grids:

[0170] First, statistical analysis was performed on a large amount of historical thickness distribution data collected in the actual tin plating process to calculate the difference distribution of grid thickness deviation and evaluate the corresponding local tin plating quality performance.

[0171] Then, based on the historical data mentioned above, the variance threshold is clearly identified by using the correspondence between dispersion (variance) and actual thickness defects. For example, when the variance of the thickness deviation difference between adjacent grids in a local area reaches or exceeds 0.3 μm², the local thickness non-uniformity is significantly enhanced, which can easily cause depositional structure defects.

[0172] Based on the above statistical analysis, to ensure the overall uniformity and quality stability of the tin plating layer, the variance threshold for the difference in thickness between adjacent grids is set to 0.3 μm. 2 (This variance threshold can be adjusted according to the actual process accuracy requirements) serves as a basis for judging whether there are obvious thin areas in local regions, thereby effectively identifying areas that need anchor point thickening deposition.

[0173] For example, four areas with thinner thickness were identified: the inner side of the upper left corner (thickness 8.2-8.8 μm, 1.2-1.8 μm lower than the target), the central area (thickness 8.5-9.0 μm, 1.0-1.5 μm lower), the inner side of the upper right corner (thickness 8.3-8.9 μm, 1.1-1.7 μm lower), and the inner side of the lower left corner (thickness 8.4-8.7 μm, 1.3-1.6 μm lower).

[0174] Step S302.3: Accurately determine the spatial location of the anchor point using a comprehensive analysis method, including the following steps:

[0175] First, identify the areas with significant temperature differences and areas with thin thickness on the surface of the second tin sublayer;

[0176] Based on the spatial distribution characteristics of the regions, the spatial overlap of the two types of regions is calculated. For example, the overlap between the region with significant temperature difference and the region with thin thickness in the upper left corner is 82%, while the overlap in the central region is 75%.

[0177] Then, each region is assigned a corresponding weight based on the degree of overlap (for example, overlapping regions are assigned a weight coefficient of 0.8, while non-overlapping regions with significant thickness or temperature differences are assigned a weight coefficient of 0.6).

[0178] Based on this, and aiming to achieve the optimal combination of tin layer thickness and temperature uniformity, a multivariate search method based on gradual adjustment of region weights is adopted:

[0179] First, select the initial position as the candidate anchor point, and calculate the comprehensive evaluation index of temperature gradient and thickness difference within a certain radius area centered on the candidate position;

[0180] Then, the anchor point position is adjusted with the goal of reducing the comprehensive index. By changing the position in turn and recalculating the comprehensive index value, the search calculation is repeated several times. After each position adjustment, the evaluation is carried out again until the difference between the comprehensive index after two consecutive adjustments is lower than the set threshold (e.g., ±1%). Then the optimal position of the anchor point can be determined.

[0181] Repeat the above optimization process to determine the spatial location of all anchor points, keeping the spacing between each anchor point between 15 and 20 mm to ensure the balance and stability of the deposition structure.

[0182] Step S302.4: After determining the spatial location of the anchor points, the optimal thickness value for each anchor point area is determined using a thickness compensation calculation method. This includes the following steps:

[0183] First, determine the base thickness compensation amount based on the actual measured thickness deviation values ​​of each thin area (e.g., 1.2-1.8μm). Then, based on previous production experience or the characteristics of molten tin deposition, set the temperature difference compensation thickness parameter (e.g., 0.2μm compensation thickness per ℃ temperature difference). Finally, add the actual measured temperature difference values ​​of the area to the base thickness compensation amount to calculate the final anchor point thickness range (e.g., 2.5-3.5μm thicker than the target single layer thickness, i.e., 12.5-13.5μm).

[0184] Next, based on the shape of the thickness difference distribution area and the temperature difference distribution area at each anchor point location, the geometry of the anchor point area is determined: for areas where the temperature difference and thickness difference highly overlap, a circular anchor point area (e.g., 3mm in diameter) is selected to ensure optimal uniformity and integration.

[0185] For narrow and elongated areas where temperature and thickness differences do not overlap or have low overlap, elliptical anchor point areas (e.g., 3mm major axis and 2mm minor axis) are used to better fit the actual area shape and reduce the fusion difference with the surrounding area. Finally, the determined position, thickness and shape parameters are input into the nozzle motion control system to precisely control the spray flow and movement trajectory of the molten tin, thereby achieving precise deposition control of the anchor point area.

[0186] The criteria for judging insufficient fusion or insufficient thickness are as follows:

[0187] The fusion degree evaluation value and thickness compliance evaluation value of the local region of the N+1th sub-tin layer are input into the pre-trained quality defect classification model;

[0188] The quality defect classification model outputs classification results that characterize whether there are defects with insufficient fusion or insufficient thickness in the corresponding local area;

[0189] If the output indicates the presence of any defect, a local respray operation is triggered.

[0190] In this embodiment of the invention, the training process of the quality defect classification model is as follows:

[0191] The training sample data consisted of 1000 sets of actual tin-plating samples collected from multiple process batches. These samples covered different tin layer thicknesses, temperature distributions, and process deviations. Of these, 800 sets were qualified samples confirmed through manual review and quality inspection, while 200 sets were unqualified samples. Unqualified samples mainly exhibited defects such as insufficient fusion or significant deviations in thickness from the target value in localized areas. Each set of samples contained three key input features:

[0192] The first feature is the fusion score, which is the result of the controller's image processing and temperature gradient assessment, used to represent the interface bonding quality between the sub-tin layer and the previous tin layer or substrate.

[0193] The second feature is the thickness compliance score, which is calculated by the ratio of the actual thickness of the local area collected by the laser thickness gauge to the target thickness;

[0194] The third characteristic is the maximum temperature difference, defined as the difference between the highest and lowest temperatures on the surface of the region during the deposition of the tin sublayer, in °C.

[0195] Before training the quality defect classification model, the sample data was preprocessed. The fusion score was linearly scaled to [0,1] from 70 to 100; the thickness compliance score was converted into the absolute value of the thickness deviation (the difference from 100%), and then the maximum deviation ±10% was used as the standard for linear scaling to the [0,1] range; the maximum temperature difference was normalized to the [0,1] range using a minimum-maximum normalization method. The three processed feature values ​​formed the input vector, and the sample label corresponded to the two categories of "qualified" or "unqualified".

[0196] The random forest model is trained using the Scikit-learn toolkit in the Python programming language. The specific training steps are as follows:

[0197] First, the 1000 samples were divided proportionally into a training set of 800 groups (640 qualified and 160 unqualified) and a test set (160 qualified and 40 unqualified).

[0198] The initial parameters for the quality defect classification model are set as follows: 100 decision trees, each with a maximum depth of 6 layers and a minimum number of samples per leaf node of 10. Ten-fold cross-validation is used to optimize the model parameters. This involves dividing the 800 training samples into 10 equal parts, with 9 parts used for training and the remaining part for validation. After 10 iterations, each part is used for validation. During cross-validation, the number of trees in the random forest (e.g., from 80 to 200), the maximum tree depth (e.g., from 5 to 10 layers), and the minimum number of leaf nodes (e.g., from 5 to 20) are optimized. The optimization goal is to simultaneously improve the model's classification accuracy and F1 score (i.e., the overall evaluation metric). The parameter optimization process ends when both the classification accuracy and F1 score obtained from cross-validation reach their maximum and stabilize without further improvement.

[0199] After training, performance is validated on an independent test set. For example, if the classification accuracy is 98.5% and the F1 score (comprehensive evaluation index) reaches 97.4% on the test set, it can be confirmed that the quality defect classification model has excellent generalization ability.

[0200] Subsequently, the feature importance calculation mechanism of the random forest algorithm itself is used to determine the contribution of each feature to the classification result. The specific implementation process is as follows:

[0201] The Random Forest algorithm calculates feature importance using the Gini index decrease. Specifically, during the training of the quality defect classification model, when each decision tree splits a feature node, it calculates the Gini index difference before and after the split. The Gini index differences contributed by each feature to all decision tree nodes are summed, averaged, and then normalized to obtain the importance ratio of each feature to the final classification result.

[0202] In this embodiment, the built-in function `feature_importances_` of the Random Forest library in Python's Scikit-learn is used to quantitatively calculate feature importance. After training is complete, calling this function will automatically output the importance ratio of each feature.

[0203] For example, the importance of the thickness conformity feature is 62%, indicating that this feature has the greatest impact on the classification results; the importance of the fusion feature is 31%, indicating that this feature has the second greatest impact; and the importance of the maximum temperature difference feature is 7%, indicating that this feature has a relatively small impact.

[0204] To determine the decision thresholds for the quality defect classification model, namely the specific decision boundaries for the fusion score and thickness compliance, the following steps are performed:

[0205] First, after the random forest classification model is trained, exhaustive search analysis is performed using performance metrics (such as classification accuracy, recall, precision, and F1 score) on the test set, targeting different combinations of thickness fit and fusion score thresholds. The specific process is as follows:

[0206] With a thickness compliance threshold range of 88% to 92%, and a step size of 0.1%;

[0207] The integration score threshold ranges from 78 to 82 points, with a step size of 0.1 points.

[0208] Within the above threshold range, threshold combination traversal calculations are performed, and each combination calculates the classification accuracy and F1 score of the classification model on the validation set.

[0209] For example, after the above traversal search process, it was found that:

[0210] When the thickness compliance threshold is 89.5% and the fusion score threshold is 79.2, the quality defect classification model achieves its highest and most stable F1 score and classification accuracy on the validation set, reaching 97.4% and 98.5% respectively. Therefore, this combination is determined to be the optimal quality judgment threshold.

[0211] Meanwhile, to ensure the reliability of the threshold selection, sensitivity analysis was conducted within ±0.5% (thickness) and ±0.5 points (fusion degree) around the threshold. It was found that the F1 score of other adjacent combinations decreased, thus confirming that this combination is the global optimum.

[0212] Ultimately, the thickness compliance of 89.5% and the fusion score of 79.2 were set as the decision boundaries for the quality defect classification model.

[0213] After the quality defect classification model is trained, it is deployed to the system control unit. In practical applications, the controller inputs the blending degree value, thickness compliance value, and maximum temperature difference value into the quality defect classification model. The quality defect classification model first outputs an overall "qualified" or "unqualified" judgment: if the prediction is qualified, it is marked as "no defect"; if the prediction is unqualified, it further analyzes the specific defect type—when the blending degree score is not higher than 79.2 points, it is judged as a defect of insufficient blending degree; when the thickness compliance is not higher than 89.5%, it is judged as a defect of insufficient thickness; if both conditions are met, it is judged as a double defect. The control system automatically triggers the corresponding local respraying repair program based on the defect judgment result.

[0214] When the quality defect classification model determines that the area has defects such as insufficient fusion or insufficient thickness, the specific calculation method for the thickness of the local re-spray is as follows:

[0215] Specifically, taking a target thickness of 10μm as a reference, the difference between the actual thickness and the target thickness in areas with insufficient thickness is first calculated. For example, when the thickness compliance is 89%, the actual thickness is 8.9μm, and the difference from the target thickness is 1.1μm. Simultaneously, the additional compensation thickness caused by insufficient fusion needs to be considered. The compensation thickness for areas with insufficient fusion is related to the degree of fusion defect; that is, for every 1-point decrease in fusion, an additional 0.1μm of supplementary spray thickness is added.

[0216] Therefore, the calculation method for the thickness of the additional spray is as follows:

[0217] When the defect is only insufficient in thickness: the thickness of the additional spray is the difference between the target thickness and the actual thickness, multiplied by the compensation coefficient, where the compensation coefficient is generally taken as 1.1 (10% redundancy compensation to ensure fusion stability).

[0218] For example, when the actual thickness is 8.9 μm, the compensation thickness is (10 μm - 8.9 μm) × 1.1 ≈ 1.2 μm, so we take the value 1.2 μm.

[0219] When the defect is only in areas with insufficient fusion: the thickness of the additional spray is the product of 80 points minus the difference obtained from the actual fusion score and 0.1μm;

[0220] For example, when the actual fusion score is 79, the thickness of the additional spray is (80-79)×0.1μm=0.1μm, and 0.1μm is used in actual production.

[0221] When the defect area has both thickness and fusion defects: the repair spray thickness is the sum of the thickness repair spray amount and the fusion repair spray amount;

[0222] For example, when the actual thickness is 8.9 μm (thickness replenishment spraying amount 1.2 μm) and the actual blending score is 77 points (blending degree replenishment spraying amount 0.3 μm),

[0223] The total thickness of the additional spray is 1.2μm + 0.3μm = 1.5μm.

[0224] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A spaced tinning apparatus for battery module busbars, characterized by, The application relates to a tin spraying device for a tin plating process. The device comprises: a spraying box for maintaining the ambient temperature during the tin plating process; a sprayer for spraying tin on the surface of a busbar; a three-axis moving table for moving the nozzle of the sprayer, the nozzle of the sprayer being fixedly connected to the surface of the sliding table of the three-axis moving table; a sensor group for collecting state data of the busbar substrate or the N layers of tin sub-layers that have been deposited, the state data including image information, temperature distribution data and accumulated thickness data; a controller for determining a deposition mode for depositing the N+1 layer of tin sub-layers based on the state data by a contour and temperature difference double-threshold judgment mechanism, the deposition mode being a layer-by-layer uniform deposition mode or an edge-anchor-filling deposition mode; and generating a spraying compensation strategy according to the deposition mode and the state data to form the N+1 layer of tin sub-layers; the spraying compensation strategy including movement path instructions, movement path disturbance instructions and flow control instructions of the nozzle of the sprayer; during the execution of the spraying compensation strategy, the sensor group collects deposition data of the N+1 layer of tin sub-layers in real time; the controller evaluates the fusion degree and thickness compliance of the local area of the N+1 layer of tin sub-layers according to the deposition data; the controller controls the three-axis moving table to drive the sprayer to locally spray the local area with insufficient fusion degree or thickness; 2. The apparatus for spacing of the busbar of the battery module according to claim 1, wherein, the controller monitors the spraying pressure, the flow of molten tin and the trajectory deviation of the nozzle of the sprayer in real time, and outputs equipment intervention measures, including stopping spraying or equipment alarm, if an abnormality is detected. The contour and temperature difference double-threshold judgment mechanism comprises: discretization analysis of the accumulated thickness distribution data of the N layer of tin sub-layers obtained by the sensor group, and calculation of the regional deviation between the current deposition thickness and the target thickness corresponding to the N layer by a thickness distribution difference analysis algorithm; determination of the temperature difference distribution of the local area of the tin sub-layer by a multi-point temperature difference interpolation calculation algorithm based on the temperature distribution data; 3. The apparatus for spacing of the busbar of the battery module according to claim 1, wherein, when the regional deviation and the temperature difference distribution calculation result do not exceed the set threshold value, the layer-by-layer uniform deposition mode is selected, and when the thickness regional deviation or the temperature difference distribution calculation result exceeds the set threshold value, the edge-anchor-filling deposition mode is selected. The process of forming the N+1 layer of tin sub-layers comprises: when the layer-by-layer uniform deposition mode is adopted, the controller controls the three-axis moving table to drive the nozzle of the sprayer to make continuous spraying movement according to the movement path instructions; when the edge-anchor-filling deposition mode is adopted, the controller controls the three-axis moving table to drive the nozzle of the sprayer to sequentially perform the following steps: step S301, spraying to form a continuous closed tin boundary along the surface of the busbar substrate; step S302, spraying to form a plurality of discrete distributed anchor point areas with thickening characteristics inside the tin boundary; step S303, filling spraying based on the tin boundary and the anchor point tin layer structure to complete the N+1 layer of tin sub-layers; wherein the position, thickness and shape of the anchor point area are adjusted based on the state data collected by the sensor group.

4. The apparatus for spacing of the busbar of the battery module according to claim 3, wherein, When a layer-by-layer uniform deposition method is adopted, the motion path perturbation instruction is generated by the following steps: In step S311, the cumulative thickness fluctuation data of the deposited sub-tin layer is obtained; In step S312, the thickness fluctuation data is subjected to fast Fourier transform to extract periodic characteristic parameters; In step S313, the perturbation amplitude and frequency of the motion path of the spray nozzle are adaptively adjusted according to the characteristic parameters, so that the newly deposited layer forms a corrugated structure that enhances the interlayer bonding.

5. The apparatus for spacing of the busbar of the battery module according to claim 3, wherein, The process of forming the N+1th sub-tin layer further includes: Based on the current state data collected by the sensor group and the preset target thickness and corrugated structure, the characteristic parameters of the tin layer thickness deviation, corrugated structure and temperature distribution are obtained; the characteristic parameters are subjected to dimension reduction processing to obtain a comprehensive characteristic vector containing the thickness out-of-tolerance area proportion, corrugated continuity and temperature-thickness coincidence degree; The corresponding relationship between the comprehensive characteristic vector and the deposition quality index is established, and the comprehensive characteristic vector is iteratively calculated multiple times to output the motion path perturbation instruction and the molten tin flow dynamic control instruction for compensating the deposition thickness deviation and temperature fluctuation.

6. The apparatus for spacing of the busbar of the battery module according to claim 1, wherein, The fusion degree of the local area of the N+1th sub-tin layer is evaluated by the following steps: In step S401, the interface area between the N+1th sub-tin layer and the Nth sub-tin layer or the busbar substrate is subjected to image analysis; In step S402, a multi-scale edge detection algorithm is used to identify the interface area profile and extract multi-dimensional geometric feature parameters; In step S403, based on the temperature distribution data collected in real time from the interface area, a spatial interpolation algorithm is used to reconstruct the regional temperature field, and a difference gradient algorithm is used to calculate the temperature gradient distribution characteristic parameters; In step S404, the geometric feature parameters and the temperature gradient distribution characteristic parameters are input into a multi-dimensional fusion algorithm, and the feature parameter weight distribution and multi-round iterative operation are performed to obtain a quantitative fusion degree evaluation value for accurately representing the fusion state of the interface area.

7. The apparatus for spacing of the busbar of the battery module according to claim 3, wherein, The position, thickness and shape of the anchor point area are adjusted according to the temperature distribution data and the thickness distribution data in the state data, including: In step S302.1, the temperature difference significant area in the tin layer deposition area is determined by a spatial interpolation and local temperature difference gradient algorithm based on the real-time collected temperature distribution data of the sub-tin layer deposition area; In step S302.2, the thickness weak area is identified by a local thickness difference discrete analysis algorithm combined with the real-time collected sub-tin layer deposition thickness distribution data; In step S302.3, a multivariate optimization search algorithm is used to comprehensively calculate the identified temperature difference significant area and thickness weak area to dynamically determine the accurate spatial position of the anchor point tin layer; In step S302.4, a thickness compensation algorithm is used to dynamically calculate the thickness of the anchor point tin layer, and the temperature distribution data is used to determine the geometric shape of the anchor point area, so as to realize accurate control of the anchor point tin layer deposition area and stable improvement of the deposition quality.

8. The apparatus for spacing of the busbar of the battery module according to claim 1, wherein, The basis for judging insufficient fusion or insufficient thickness is: The fusion degree evaluation value and the thickness compliance evaluation value of the local area of the N+1th sub-tin layer are input into a pre-trained quality defect classification model; The quality defect classification model outputs a classification result indicating whether the corresponding local area has a fusion defect or a thickness defect; If the output result indicates that there is any defect, the local re-spraying operation is triggered.

Images