Visual alignment and laser welding method and system for capacitor core

Abstract

The embodiment of the invention provides a visual alignment and laser welding method and system for a capacitor core, and the method comprises the steps: obtaining the three-dimensional point cloud data of the end face of the capacitor core and a to-be-welded lead, and carrying out the feature extraction and coordinate calculation of the three-dimensional point cloud data, and obtaining an alignment transformation parameter and an average assembly gap value; based on the three-dimensional point cloud data and the average assembly clearance value, optimal laser welding parameters are generated through a predetermined convolutional neural network model; based on the alignment transformation parameters and the optimal laser welding parameters, a dynamic focusing pulse fiber laser is driven to weld the end face of the capacitor core and the to-be-welded lead, the welding process is monitored in real time, and real-time monitoring data is obtained; and calculating a multi-dimensional quality index and quality judgment based on the real-time monitoring data to obtain a quality judgment result, and performing process data management according to the quality judgment result. According to the scheme, the positioning precision can be improved and the thermal damage risk can be reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electronics, in particular to a visual alignment and laser welding method and system for capacitor cores. BACKGROUND

[0002] In the process of capacitor manufacturing, the welding of the core (winding body) and the external lead or metal foil is a key process to achieve electrical connection, and its quality directly affects the electrical conductivity, mechanical strength and long-term reliability of the capacitor. Especially in the fields of automotive electronics, aerospace and other high reliability fields, the requirements for welding quality are extremely strict.

[0003] In current industrial production, the traditional welding method mainly faces two difficulties, one is poor positioning accuracy: relying on manual or simple mechanical clamps for alignment, it is difficult to meet the micron-level alignment requirements, and it is easy to cause offset, resulting in virtual welding or short circuit; the second is high risk of thermal damage: the energy control in the welding process is not accurate, especially for heat-sensitive thin film dielectric or metallized layer, excessive heat input is easy to cause dielectric damage, metal layer oxidation or thermal stress cracking. SUMMARY

[0004] The embodiments of the present application expect to provide a visual alignment and laser welding method and system for capacitor cores, which can improve the positioning accuracy and reduce the risk of thermal damage.

[0005] The technical solution of the present application is as follows: In a first aspect, the embodiments of the present application provide a visual alignment and laser welding method for capacitor cores, which comprises: obtaining three-dimensional point cloud data of the end face of the capacitor core and the to-be-welded lead, and performing feature extraction and coordinate calculation on the three-dimensional point cloud data to obtain alignment transformation parameters and average assembly gap values; generating optimal laser welding parameters through a pre-determined convolutional neural network model based on the three-dimensional point cloud data and the average assembly gap values; driving a dynamic focusing pulsed fiber laser to weld the end face of the capacitor core and the to-be-welded lead based on the alignment transformation parameters and the optimal laser welding parameters, and performing real-time monitoring on the welding process to obtain real-time monitoring data; calculating multi-dimensional quality indicators and quality judgments based on the real-time monitoring data to obtain quality judgment results, and performing process data management according to the quality judgment results.

[0006] In the above-mentioned scheme, the three-dimensional point cloud data of the end face of the capacitor core and the to-be-welded lead is obtained, and the three-dimensional point cloud data is subjected to feature extraction and coordinate calculation to obtain alignment transformation parameters and average assembly gap values, which comprises: The capacitor core and the to-be-welded lead are globally positioned by a binocular stereo vision system, and a target region containing the capacitor core end face and the to-be-welded lead is determined; The target region is scanned to obtain three-dimensional point cloud data, wherein the three-dimensional point cloud data includes three-dimensional coordinates and reflection intensity information of each data point in the target region; Based on the three-dimensional coordinates of the three-dimensional point cloud data, the center point coordinates and the normal vector of the capacitor core end face and the to-be-welded surface of the to-be-welded lead are calculated respectively; Based on the center point coordinates and the normal vector, the alignment transformation parameters are calculated, and the spatial relationship between the capacitor core and the lead point cloud is calculated according to the alignment transformation parameters to obtain the average assembly gap value.

[0007] In the above scheme, the optimal laser welding parameters are generated based on the three-dimensional point cloud data and the average assembly gap value through a pre-determined convolutional neural network model, comprising: Based on the three-dimensional point cloud data and the average assembly gap value, a numerical feature vector is calculated; The numerical feature vector is processed by feature extraction, feature fusion and parameter regression through the convolutional neural network model to obtain the optimal laser welding parameters; wherein the optimal laser welding parameters include at least one of laser peak power, pulse width, pulse frequency, welding scanning speed and spot compensation diameter.

[0008] In the above scheme, the numerical feature vector is calculated based on the three-dimensional point cloud data and the average assembly gap value, comprising: Based on the reflection intensity information in the three-dimensional point cloud data, a low value region of intensity anomaly is identified to obtain a cleanliness index value; Based on the three-dimensional coordinates in the three-dimensional point cloud data, the height root mean square error of the point cloud of the to-be-welded surface and the best fitting plane is calculated to obtain a flatness index value; Based on the cleanliness index value, the flatness index value and the average assembly gap value, the numerical feature vector is determined.

[0009] In the above scheme, the numerical feature vector is processed by feature extraction, feature fusion and parameter regression through the convolutional neural network model to obtain the optimal laser welding parameters, comprising: One-dimensional feature extraction is performed on the numerical feature vector through the convolutional neural network model to obtain a one-dimensional feature map; The one-dimensional feature map is flattened to obtain a one-dimensional vector, and the one-dimensional vector is nonlinearly fused to obtain a fused feature; The fusion features are subjected to welding parameter regression processing to obtain the optimal laser welding parameters.

[0010] In the scheme, the end face of the capacitor core and the lead to be welded are welded by driving the dynamic focusing pulsed fiber laser based on the alignment transformation parameters and the optimal laser welding parameters, and the welding process is monitored in real time to obtain real-time monitoring data, which includes: Based on the alignment transformation parameters, the mobile capacitor core and the lead to be welded are moved to the target welding position; Based on the optimal laser welding parameters, the dynamic focusing pulsed fiber laser is driven to weld at the target welding position, the welding process is monitored in real time, and the dynamic images and temperature distribution information of the welding pool are obtained; The dynamic images and the temperature distribution information are subjected to feature extraction to obtain pool dynamic features and thermal cycle features, and the pool dynamic features and the thermal cycle features are determined as the real-time monitoring data.

[0011] In the scheme, based on the real-time monitoring data, multi-dimensional quality indicators and quality judgments are calculated to obtain quality judgment results, and process data management is performed according to the quality judgment results, which includes: Based on the pool dynamic features and the thermal cycle features in the real-time monitoring data, multi-dimensional quality indicators of the welding points are calculated to obtain forming quality indicator values, heat affected zone index indicator values, and connectivity prediction indicator values; The forming quality indicator values, the heat affected zone index indicator values, and the connectivity prediction indicator values are compared with respective corresponding process standard threshold values to obtain quality judgment results; Based on the quality judgment results, process data management is performed; wherein the process data management includes one of qualified release, online compensation, and rejection.

[0012] In a second aspect, the embodiments of the present application provide a visual alignment and laser welding system for capacitor cores, which includes an acquisition module, a generation module, a monitoring module, and a calculation module, wherein, The acquisition module is configured to acquire three-dimensional point cloud data of the end face of the capacitor core and the lead to be welded, and to perform feature extraction and coordinate calculation on the three-dimensional point cloud data to obtain alignment transformation parameters and average assembly gap values. The generation module is configured to generate optimal laser welding parameters based on the three-dimensional point cloud data and the average assembly gap values through a pre-determined convolutional neural network model. The monitoring module is configured to drive the dynamic focus pulsed fiber laser to weld the end face of the capacitor core and the lead to be welded based on the alignment transformation parameter and the optimal laser welding parameter, and to monitor the welding process in real time to obtain real-time monitoring data. The computing module is configured to calculate a multi-dimensional quality index and quality determination based on the real-time monitoring data, to obtain a quality determination result, and to perform process data management according to the quality determination result.

[0013] In a third aspect, an embodiment of the present application provides a visual alignment and laser welding device for a capacitor core, comprising a processor and a memory, wherein The memory is configured to store a computer program. The processor is configured to call and run the computer program from the memory to execute the method of the first aspect.

[0014] In a fourth aspect, an embodiment of the present application provides a computer readable storage medium storing executable instructions for causing a processor to execute the method of the first aspect.

[0015] The embodiment of the present application provides a visual alignment and laser welding method and system for a capacitor core. The method comprises the following steps: obtaining three-dimensional point cloud data of the end face of the capacitor core and the lead to be welded, and performing feature extraction and coordinate calculation on the three-dimensional point cloud data to obtain alignment transformation parameters and an average assembly gap value; generating optimal laser welding parameters through a pre-determined convolutional neural network model based on the three-dimensional point cloud data and the average assembly gap value; driving a dynamic focus pulsed fiber laser to weld the end face of the capacitor core and the lead to be welded based on the alignment transformation parameters and the optimal laser welding parameters, and monitoring the welding process in real time to obtain real-time monitoring data; calculating a multi-dimensional quality index and quality determination based on the real-time monitoring data to obtain a quality determination result, and performing process data management according to the quality determination result. In the above-mentioned solution, the three-dimensional point cloud data of the end face of the capacitor core and the lead to be welded is obtained, and the alignment transformation parameters and the average assembly gap value are obtained by performing feature extraction and coordinate calculation on the three-dimensional point cloud data; the optimal welding parameters are automatically generated based on the three-dimensional point cloud data and the average assembly gap value, such as the gap, flatness and cleanliness, which realizes the leap from “fixed parameter welding” to “perception-decision-adaptive welding”, ensures the accuracy of the welding joint position, and improves the positioning accuracy. At the same time, the accuracy of energy input is ensured, and the risk of thermal damage is reduced. BRIEF DESCRIPTION OF DRAWINGS

[0016] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the specification, serve to explain the technical solutions of this application. Obviously, the drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0017] Figure 1 This is an optional flowchart illustrating a method for visual alignment and laser welding of capacitor cores, provided as an embodiment of this application. Figure 2 A schematic diagram of a visual alignment and laser welding system for capacitor cores provided in this application embodiment; Figure 3 This is a schematic diagram of a device for visual alignment and laser welding of capacitor cores, provided as an embodiment of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the specific technical solutions of this application will be further described in detail below with reference to the accompanying drawings of the embodiments of this application. The following embodiments are used to illustrate this application, but are not intended to limit the scope of this application.

[0019] Unless otherwise defined, all technical and scientific terms used in this application have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. The terminology used in this application is for the purpose of describing embodiments of this application only and is not intended to be limiting of this application.

[0020] In the following description, references to "some embodiments," "this embodiment," "this application embodiment," and examples, etc., describe a subset of all possible embodiments. However, it is understood that "some embodiments" may be the same subset or different subset of all possible embodiments and may be combined with each other without conflict.

[0021] If the application documents contain similar descriptions such as "first / second", the following explanation shall be added: In the following description, the terms "first / second / third" are used only to distinguish similar objects and do not represent a specific order of objects. It is understood that "first / second / third" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0022] This application provides a method for visual alignment and laser welding of capacitor cores. Figure 1This is an optional flowchart illustrating a visual alignment and laser welding method for capacitor cores provided in an embodiment of this application, which will be combined with... Figure 1 The steps shown are explained.

[0023] S101. Obtain three-dimensional point cloud data of the capacitor core end face and the lead to be soldered, and perform feature extraction and coordinate calculation on the three-dimensional point cloud data to obtain the alignment transformation parameters and average assembly gap value.

[0024] In some embodiments of this application, the three-dimensional point cloud data is obtained through a global positioning module and a local precision measurement module.

[0025] In some embodiments of this application, a visual alignment and laser welding method for capacitor cores is adapted to laser welding scenarios for capacitor cores.

[0026] In some embodiments of this application, a visual alignment and laser welding method for capacitor cores is adapted to a visual alignment and laser welding system for capacitor cores.

[0027] In some embodiments of this application, a binocular stereo vision system is used to globally locate the capacitor core and the lead to be soldered, determining a target area containing the end face of the capacitor core and the lead to be soldered; the target area is scanned to obtain three-dimensional point cloud data; wherein, the three-dimensional point cloud data includes the three-dimensional coordinates and reflection intensity information of each data point in the target area; based on the three-dimensional coordinates of the three-dimensional point cloud data, the center point coordinates and normal vectors of the end face of the capacitor core and the lead to be soldered are calculated respectively; based on the center point coordinates and normal vectors, alignment transformation parameters are calculated; and, according to the alignment transformation parameters, the spatial relationship between the point clouds of the core and the lead is calculated to obtain the average assembly gap value.

[0028] It should be noted that the capacitor core is also called the capacitor element; the lead wire to be soldered is also called the lead terminal.

[0029] S102. Based on three-dimensional point cloud data and average assembly gap value, the optimal laser welding parameters are generated through a pre-determined convolutional neural network model.

[0030] In some embodiments of this application, a numerical feature vector is calculated based on three-dimensional point cloud data and average assembly gap value; through a convolutional neural network model, feature extraction, feature fusion and parameter regression are performed on the numerical feature vector to obtain the optimal laser welding parameters; wherein, the optimal laser welding parameters include at least one of laser peak power, pulse width, pulse frequency, welding scanning speed and spot compensation diameter.

[0031] In some embodiments of this application, the convolutional neural network model is pre-trained. Specifically, this includes: Loss function: Huber Loss is used. This loss function is less sensitive to outliers than mean squared error (MSE), is more robust to regression problems, and can mitigate the impact of mislabeled data. Optimizer: Adam optimizer is used, with an initial learning rate set to 0.001. Training data: "Process conditions-welding parameters-welding results" triplet data from a historical database is used; where the process parameters corresponding to welding results verified as "good" are used as training labels. Training process: The dataset is divided into training, validation, and test sets in a 70:15:15 ratio. Early stopping is used during training to monitor the validation set loss and avoid overfitting.

[0032] S103. Based on the alignment transformation parameters and the optimal laser welding parameters, drive the dynamic focusing pulsed fiber laser to weld the end face of the capacitor core to the lead to be welded, and monitor the welding process in real time to obtain real-time monitoring data.

[0033] In some embodiments of this application, based on the alignment transformation parameters, the movable capacitor core and the lead to be welded are moved to the target welding position; based on the optimal laser welding parameters, a dynamic focusing pulsed fiber laser is driven to perform welding at the target welding position, the welding process is monitored in real time, and dynamic images and temperature distribution information of the weld pool are obtained; features are extracted from the dynamic images and temperature distribution information to obtain dynamic features and thermal cycling features of the weld pool, and the dynamic features and thermal cycling features of the weld pool are determined as real-time monitoring data.

[0034] S104. Based on real-time monitoring data, calculate multi-dimensional quality indicators and quality judgments to obtain quality judgment results, and perform process data management based on the quality judgment results.

[0035] In some embodiments of this application, based on the dynamic characteristics of the molten pool and the thermal cycling characteristics in real-time monitoring data, multi-dimensional quality indicators of the weld joint are calculated to obtain forming quality index value, heat-affected zone index value, and connectivity prediction index value; the forming quality index value, heat-affected zone index value, and connectivity prediction index value are compared with their respective corresponding process standard thresholds to obtain quality judgment results; based on the quality judgment results, process data management is performed; wherein, process data management includes one of qualified release, online compensation, and rejection.

[0036] For example, the online quality assessment and closed-loop control process is as follows: 1. Objective: To determine the quality of weld joints in real time and to promptly address or adjust parameters for any non-conforming situations.

[0037] 2. Multi-feature quality judgment: Based on real-time monitoring of the dynamic characteristics of the molten pool and thermal cycle characteristics, the quality indicators of the solder joint are calculated in real time and compared with the preset process standard thresholds to achieve millisecond-level initial quality judgment.

[0038] 3. Judgment Logic and Execution: The system performs parallel calculations on the following three core quality indicators: Forming quality indicators: Based on post-weld images captured by coaxial vision, image analysis algorithms are used to detect the presence of macroscopic defects such as spatter, dents, and surface cracks. Example of a judgment threshold: Number of spatter particles < 3, and no continuous cracks.

[0039] Heat Affected Zone Index (HAZ): Based on infrared thermography data, this index calculates the ratio of the area of ​​the heat-affected zone (the area where the temperature exceeds the material's tolerance threshold, such as 150°C) to the nominal weld area. Example threshold: This ratio ≤ 2.0.

[0040] Connectivity prediction metrics: Combining molten pool stability (fluctuation frequency below the set value) and total process energy input (within the theoretical requirement range), a lightweight model is used to predict the reliability of metallurgical bonding. Outputs are categorized as "strong," "critical," or "weak."

[0041] Example 1: In a scenario where the anode foil and leads of a certain type of aluminum electrolytic capacitor are being welded, the following criteria are used to determine if the heat-affected zone exceeds the limit: Monitoring data: Infrared thermal imaging shows that the area of ​​the heat-affected zone with a temperature exceeding 180°C is 2.5 times the nominal solder joint area.

[0042] Judgment process: The system calculates the "thermal impact index" to be 2.5.

[0043] Judgment result: This indicator is greater than the threshold of 2.0, and is judged as "abnormal" on its own. The system immediately marks this solder joint status as "pending processing".

[0044] Real-time decision-making and execution: Objective: Based on the quality judgment results, automatically execute the corresponding closed-loop control actions within the production line cycle time (usually <1 second) and manage the process data.

[0045] Decision-making logic and execution path: Based on the judgment result, the system automatically enters one of the following three paths: Path 1: Release if qualified.

[0046] Triggering condition: All three core quality indicators are better than the threshold.

[0047] Action performed: The system marks the weld point as qualified. Simultaneously, it triggers a data archiving procedure, binding the "full-process process parameter spectrum" corresponding to the weld point with the capacitor's unique code, storing it in the quality database, and generating a traceable "quality ID." The workpiece then proceeds to the next process step.

[0048] Path 2: Minor anomalies, online compensation.

[0049] Triggering condition: One or a few indicators are slightly out of tolerance, but the defect is considered to be appropriate to be corrected in situ.

[0050] Action Execution: The system immediately triggers the "micro-compensation welding" procedure. For example, if the "molten pool area is too small" indicator suggests the connection may be insufficient, the system controls the laser to precisely apply one or more low-energy, short-pulse-width compensation pulses at the original weld point location to perform local remelting and strengthening. After compensation, the system immediately performs a secondary judgment. If it passes, it proceeds to Path 1; if it still fails, it proceeds to Path 3.

[0051] Path 3: Seriously unqualified, eliminated and trained.

[0052] Triggering conditions: Key indicators (such as macroscopic cracks) are severely out of tolerance, or are still unqualified after compensation.

[0053] Perform the following actions: a. Immediate rejection: The system controls the rejection mechanism (such as a cylinder or robotic arm) to remove the defective product from the main line and place it in a special waste box.

[0054] b. Data Storage and Learning: The complete data of this welding operation (including process parameters and monitoring characteristics that led to the defect) is automatically saved as an "anomaly case." This case will be transmitted to a cloud-based or server-side process optimization platform for periodic "offline" updates to the parameter decision model, thereby enabling the system to avoid similar defects.

[0055] In a further example, when the weld joint in Example 1 is marked due to exceeding the "heat-affected zone index," the process of online micro-compensation and model learning is as follows: Decision-making process: The system determines that the defect is a "minor anomaly" caused by slightly higher energy and automatically enters path two (online compensation).

[0056] Perform the following actions: Compensation welding: A series of rapid, short pulses with 30% of the original energy and double the frequency are applied to the welded area using a laser. The purpose is to improve the microstructure of the heat-affected zone through rapid annealing, rather than remelting.

[0057] Secondary assessment: After compensation, the infrared thermal image shows that the high temperature range of the heat-affected zone has shrunk, and the "heat impact index" has been recalculated to 1.8, which meets the threshold.

[0058] Final processing: The system updates the status of the weld point to "qualified after compensation" and archives the data according to path one. Simultaneously, the complete data pair showing that "initial parameters led to a large thermal impact, but this can be effectively corrected by low-energy fast-frequency pulses" is sent to the model training server as a valuable sample for optimizing the boundary conditions of the decision-making model.

[0059] Understandably, by acquiring 3D point cloud data of the capacitor core end face and the leads to be welded, and by performing feature extraction and coordinate calculation on the 3D point cloud data, alignment transformation parameters and average assembly gap values ​​are obtained. Based on the 3D point cloud data and the average assembly gap value, optimal welding parameters are automatically generated based on conditions such as gap, flatness, and cleanliness. This achieves a leap from "fixed parameter welding" to "perception-decision-adaptive welding," ensuring the accuracy of the weld joint position and improving positioning precision. Simultaneously, it ensures the accuracy of energy input and reduces the risk of thermal damage.

[0060] In some embodiments of this application, S101 can be implemented by S201-S204, as follows: S201. Using a binocular stereo vision system, the capacitor core and the leads to be soldered are globally located to determine the target area including the end face of the capacitor core and the leads to be soldered.

[0061] S202. Scan the target area to obtain three-dimensional point cloud data; wherein, the three-dimensional point cloud data includes the three-dimensional coordinates and reflection intensity information of each data point in the target area.

[0062] S203. Based on the three-dimensional coordinates of the three-dimensional point cloud data, calculate the center point coordinates and normal vector of the capacitor core end face and the soldering surface of the lead to be soldered, respectively.

[0063] S204. Based on the center point coordinates and normal vector, calculate the alignment transformation parameters, and based on the alignment transformation parameters, calculate the spatial relationship between the core and the lead point cloud to obtain the average assembly gap value.

[0064] For example, obtaining the precise three-dimensional spatial coordinates and relative positional relationship between the capacitor core end face and the lead to be soldered includes: S11, Multi-sensor imaging: The system uses two sets of vision modules.

[0065] Global positioning module: Composed of two high-resolution area array CCD cameras, this binocular stereo vision system quickly and coarsely locates the capacitor cores and leads entering the workstation, identifying their approximate position and orientation. The purpose is to provide visual guidance for the next level of precision measurement, confining the high-precision scanning range to the target's vicinity (i.e., the target area), thereby significantly improving the overall detection speed of the system.

[0066] Local precision measurement module: Using a line laser scanner or structured light projector, the core gold-plated end face and the lead wire to be soldered area are scanned after coarse positioning to obtain high-precision three-dimensional point cloud data. Each data point in the point cloud data contains three-dimensional coordinates (X, Y, Z) and reflection intensity information (I), which fully characterizes the micro-geometry and surface optical properties of the measured area.

[0067] S12, Feature extraction and coordinate calculation.

[0068] Geometric feature extraction: Based on the three-dimensional coordinates (X, Y, Z) of the point cloud, the coordinates (O_c) and normal vector (N_c) of the end face of the gold-plated core layer, as well as the coordinates (O_l) and normal vector (N_l) of the center point of the lead wire to be soldered, are extracted through plane fitting and edge detection algorithms.

[0069] Alignment transformation solution: Based on the process principle of "minimizing assembly clearance and ensuring surface normal alignment", an optimal alignment transformation parameter (rotation matrix R and translation vector T) is calculated. This transformation will be used to control the actuator to ensure that the lead wire is precisely aligned with the core end face.

[0070] State parameter calculation: After applying the alignment transformation parameters (R, T), the spatial relationship between the core and the lead point cloud is recalculated, the average assembly gap value (Gap_avg) is output, and the theoretical solder joint coordinate set is generated in this state.

[0071] In some embodiments of this application, S102 can be implemented by S301-S302, as follows: S301. Calculate the numerical feature vector based on 3D point cloud data and average assembly gap value.

[0072] In some embodiments of this application, based on the reflection intensity information in the three-dimensional point cloud data, low-value areas with abnormal intensity are identified to obtain a cleanliness index value; based on the three-dimensional coordinates in the three-dimensional point cloud data, the root mean square error of the height between the point cloud of the surface to be welded and the best-fit plane is calculated to obtain a flatness index value; based on the cleanliness index value, the flatness index value, and the average assembly gap value, a numerical feature vector is determined.

[0073] For example, based on the above-output three-dimensional point cloud data and its reflection intensity information (I), the three key conditions affecting welding quality are quantitatively evaluated, and a computable feature vector is generated: 1. Cleanliness Index: By statistically analyzing the distribution (mean, standard deviation) of reflection intensity information (I) in the point cloud of the area to be welded, and identifying low-value areas with abnormal intensity, a cleanliness index value is obtained to indirectly assess the degree of surface oxidation and contaminant adhesion. Poor uniformity of reflection intensity or the presence of local low-value areas indicates poor cleanliness.

[0074] 2. Flatness index: Based on the three-dimensional coordinates (X, Y, Z) of the point cloud, the root mean square error (RMS) of the height of the point cloud of the surface to be welded relative to its best-fit plane is calculated to obtain the flatness index value, thereby quantifying the micro-undulations and unevenness of the surface.

[0075] 3. Assembly Gap Index: The calculated average assembly gap value (Gap_avg) serves as a key input. This value characterizes the average distance between the core and the lead wire's contact surface.

[0076] Numerical feature vectors are determined based on cleanliness index values, flatness index values, and average assembly gap values.

[0077] S302. Using a convolutional neural network model, feature extraction, feature fusion, and parameter regression are performed on the numerical feature vector to obtain the optimal laser welding parameters. The optimal laser welding parameters include at least one of the following: laser peak power, pulse width, pulse frequency, welding scanning speed, and spot compensation diameter.

[0078] In some embodiments of this application, a one-dimensional feature map is obtained by extracting one-dimensional features from the numerical feature vector using a convolutional neural network model; the one-dimensional feature map is flattened to obtain a one-dimensional vector; and the one-dimensional vector is nonlinearly fused to obtain fused features; the fused features are then subjected to welding parameter regression processing to obtain the optimal laser welding parameters.

[0079] For example, the numerical feature vector is input into a pre-trained lightweight neural network decision model. This model is trained on a database of "process conditions-welding parameters-welding quality" from historical successful welding cases. Based on the current specific condition features (numerical feature vector), the model maps and outputs a set of optimized laser welding parameters, including: Laser peak power (P): Adjust the energy input according to cleanliness and gap.

[0080] Pulse width (τ): controls the single-point heat input time, affecting the melt depth and heat-affected zone.

[0081] Pulse frequency (f): Works in conjunction with scanning speed to control solder joint overlap rate.

[0082] Welding scanning speed (v): determines the overall processing cycle time and linear heat input.

[0083] Light spot compensation diameter (d): The light spot size is dynamically fine-tuned according to the assembly gap to maintain stable energy density.

[0084] The input to the pre-trained lightweight neural network decision model is the process condition feature vector F_condition, with dimensions [1, 6]. This vector is composed of cleanliness index, flatness index, assembly gap index, and their derived statistics. The output is the optimized laser welding parameter vector P_weld, with dimensions [1, 5], corresponding to: laser peak power (P), pulse width (τ), pulse frequency (f), welding scanning speed (v), and spot compensation diameter (d).

[0085] The network structure of the convolutional neural network model is explained in detail below: One-dimensional feature convolution extraction: A one-dimensional convolutional layer (Conv1D) is used to perform initial transformation and interaction on the input features. Specific parameters: number of filters 16, kernel size 3, padding method 'same', activation function 'ReLU'. Subsequently, a batch normalization layer is applied to standardize the feature distribution, accelerating training and improving stability. Finally, a one-dimensional max pooling layer (MaxPooling1D) with a pooling size of 2 is used for downsampling, compressing the sequence length, and extracting the main feature patterns. Output feature map size: [1, 16].

[0086] Feature Flattening and Fusion: The one-dimensional feature map is flattened into a one-dimensional vector of length 16. The flattened vector is then fed into a fully connected (Dense) layer containing 32 neurons, using the 'ReLU' activation function for non-linear fusion of high-level features. To prevent overfitting, a dropout layer is applied after the fusion layer to output the fused features.

[0087] Welding parameter regression output: The fused features are input into the final fully connected output layer (Dense). This layer contains 5 neurons, each corresponding to one of the 5 welding parameters to be predicted. A linear activation function is used for numerical regression to obtain the optimal laser welding parameters.

[0088] In some embodiments of this application, S103 can be implemented by S401-S403, as follows: S401. Based on the alignment transformation parameters, move the movable capacitor core and the lead to be soldered to the target soldering position.

[0089] S402. Based on the optimal laser welding parameters, drive the dynamic focusing pulsed fiber laser to perform welding at the target welding position, monitor the welding process in real time, and obtain dynamic images and temperature distribution information of the weld pool.

[0090] S403. Extract features from dynamic images and temperature distribution information to obtain dynamic features of the molten pool and thermal cycling features, and determine the dynamic features of the molten pool and thermal cycling features as real-time monitoring data.

[0091] For example, laser welding execution includes precision motion control and dynamically focused laser processing, as detailed below: Precision motion control: The control system drives a six-axis robot or a precision XYθ platform to move and stabilize the core and lead wire at the theoretical welding position based on the alignment transformation parameters (R, T).

[0092] Dynamic focusing laser processing: The pulsed fiber laser starts working based on the laser welding parameter set (P, τ, f, v, d). The laser beam, through a galvanometer system and a dynamic focusing module, controls its focal point to scan the welding area along a preset path (such as a circle or spiral). The spot compensation diameter (d) is applied in real time to adjust the focal spot size to fit the actual assembly gap.

[0093] Real-time monitoring of the coaxial process: During laser welding, the welding process is monitored in situ using sensors integrated into the optical path. Melt pool visual monitoring: Real-time dynamic images of the weld pool are captured using a high-speed CMOS camera integrated coaxially with the laser beam. Key morphological features (melt pool dynamic features) such as area, aspect ratio, and contour fluctuation frequency of the weld pool are extracted from the image sequence.

[0094] Infrared thermal imaging monitoring: Using an infrared thermal imager configured coaxially or off-axis, the temperature field distribution of the solder joint center and heat-affected zone is monitored in real time to obtain thermal cycle characteristics such as maximum temperature, heating rate and cooling rate.

[0095] The core of this embodiment lies in building an interconnected, data-driven intelligent welding control system.

[0096] Visual precision perception and positioning layer: Adopting a master-slave collaborative strategy of "global binocular vision for rapid coarse positioning + local line laser precision scanning", it efficiently acquires the precise three-dimensional shape and surface optical properties of the welding target, providing high-precision spatial coordinates and process status input for subsequent steps.

[0097] Intelligent process decision and adaptation layer: Based on the geometric features (assembly gap, flatness) and surface state features (cleanliness) extracted by the perception layer, a set of optimal laser welding parameters are mapped and output in real time through a lightweight pre-trained one-dimensional convolutional neural network model.

[0098] Dynamic laser execution and process monitoring layer: Based on decision parameters, it drives a dynamic focusing pulsed fiber laser to perform welding, and through coaxial integrated high-speed vision and infrared thermal imaging sensors, it performs millisecond-level in-situ synchronous monitoring of the molten pool dynamics and temperature field, making the welding process monitorable.

[0099] Online quality assessment and closed-loop control layer: Based on real-time monitoring data, the system calculates multi-dimensional quality indicators online and makes immediate judgments. The system has three real-time response paths: data archiving and traceability of qualified parts; in-situ micro-compensation welding for repairable defects; and immediate removal of severe defects and feedback of cases to the decision-making model, forming a closed loop of "monitoring-judgment-execution-learning" to drive continuous self-optimization of the process.

[0100] Understandably, the master-slave 3D vision system, employing "binocular vision coarse positioning + line laser precision measurement," enables fully automated, non-contact, micron-level spatial positioning and attitude measurement of capacitor cores and leads, achieving significantly superior alignment accuracy compared to traditional fixtures or manual methods. Combined with an intelligent parameter decision model based on a one-dimensional convolutional neural network, it automatically generates optimal welding parameters based on real-time detected gaps, flatness, cleanliness, and other conditions, achieving a leap from "fixed parameter welding" to "perception-decision-adaptive welding." This ensures the dual accuracy of weld joint position and energy input, greatly improving product consistency and reliability. Integrating coaxial vision and infrared thermal imaging monitoring into the welding optical path enables millisecond-level in-situ monitoring of the molten pool dynamics and temperature field. The system can determine welding quality online in real time and perform in-situ low-energy micro-compensation repair on "minor anomalies" weld points, while immediately and automatically rejecting "severely defective" parts. This online closed-loop control system of "processing-monitoring-judgment-compensation / rejection" moves the quality control node from "after the fact" to "during the process," effectively preventing irreversible defects such as damage to the medium layer caused by improper heat input, and significantly reducing scrap rate and rework costs.

[0101] Based on the above embodiments of the visual alignment and laser welding method for capacitor cores, this application also provides a visual alignment and laser welding system for capacitor cores, such as... Figure 2 As shown, Figure 2 This application provides a schematic diagram of a visual alignment and laser welding system for capacitor cores. The system includes: an acquisition module 201, a generation module 202, a monitoring module 203, and a calculation module 204. The acquisition module 201 is used to acquire three-dimensional point cloud data of the capacitor core end face and the lead to be soldered, and to perform feature extraction and coordinate calculation on the three-dimensional point cloud data to obtain alignment transformation parameters and average assembly gap value. The generation module 202 is used to generate optimal laser welding parameters based on the three-dimensional point cloud data and the average assembly gap value through a pre-determined convolutional neural network model. The monitoring module 203 is used to drive a dynamic focusing pulsed fiber laser to weld the end face of the capacitor core to the lead to be welded based on the alignment transformation parameters and the optimal laser welding parameters, and to monitor the welding process in real time to obtain real-time monitoring data. The calculation module 204 is used to calculate multi-dimensional quality indicators and quality judgments based on the real-time monitoring data, obtain quality judgment results, and perform process data management based on the quality judgment results.

[0102] In some embodiments of this application, the acquisition module 201 is further configured to perform global positioning of the capacitor core and the lead to be soldered using a binocular stereo vision system, and determine a target area containing the end face of the capacitor core and the lead to be soldered; scan the target area to obtain the three-dimensional point cloud data; wherein the three-dimensional point cloud data includes the three-dimensional coordinates and reflection intensity information of each data point in the target area; calculate the center point coordinates and normal vector of the end face of the capacitor core and the lead to be soldered based on the three-dimensional coordinates of the three-dimensional point cloud data; calculate the alignment transformation parameters based on the center point coordinates and the normal vector; and calculate the spatial relationship between the point cloud of the core and the lead according to the alignment transformation parameters to obtain the average assembly gap value.

[0103] In some embodiments of this application, the generation module 202 is further configured to calculate a numerical feature vector based on the three-dimensional point cloud data and the average assembly gap value; and to perform feature extraction, feature fusion and parameter regression processing on the numerical feature vector through the convolutional neural network model to obtain the optimal laser welding parameters; wherein the optimal laser welding parameters include at least one of laser peak power, pulse width, pulse frequency, welding scanning speed and spot compensation diameter.

[0104] In some embodiments of this application, the calculation module 204 is further configured to identify low-value areas with abnormal intensity based on the reflection intensity information in the three-dimensional point cloud data, and obtain a cleanliness index value; calculate the root mean square error of the height between the point cloud of the surface to be welded and the best-fit plane based on the three-dimensional coordinates in the three-dimensional point cloud data, and obtain a flatness index value; and determine the numerical feature vector based on the cleanliness index value, the flatness index value, and the average assembly gap value.

[0105] In some embodiments of this application, the generation module 202 is further configured to: extract one-dimensional features from the numerical feature vector using the convolutional neural network model to obtain a one-dimensional feature map; flatten the one-dimensional feature map to obtain a one-dimensional vector; and perform nonlinear fusion on the one-dimensional vector to obtain fused features; and perform welding parameter regression processing on the fused features to obtain the optimal laser welding parameters.

[0106] In some embodiments of this application, the monitoring module 203 is used to move the movable capacitor core and the lead to be welded to the target welding position based on the alignment transformation parameters; drive the dynamic focusing pulsed fiber laser to perform welding at the target welding position based on the optimal laser welding parameters; monitor the welding process in real time to obtain dynamic images and temperature distribution information of the weld pool; extract features from the dynamic images and temperature distribution information to obtain dynamic features and thermal cycling features of the weld pool; and determine the dynamic features and thermal cycling features of the weld pool as the real-time monitoring data.

[0107] In some embodiments of this application, the calculation module 204 is further configured to calculate multi-dimensional quality indicators of the weld joint based on the dynamic characteristics of the molten pool and the thermal cycling characteristics in the real-time monitoring data, and obtain forming quality index value, heat-affected zone index value, and connectivity prediction index value; compare the forming quality index value, the heat-affected zone index value, and the connectivity prediction index value with their respective corresponding process standard thresholds to obtain quality judgment results; and perform process data management based on the quality judgment results; wherein, the process data management includes one of qualified release, online compensation, and rejection.

[0108] Based on the above embodiments of the method for visual alignment and laser welding of capacitor cores, this application also provides a device for visual alignment and laser welding of capacitor cores, such as... Figure 3 As shown, Figure 3 This is a schematic diagram of a visual alignment and laser welding device for capacitor cores provided in an embodiment of this application. The device 3 includes a processor 301 and a memory 302. The memory 302 stores a computer program; the processor 301 retrieves and runs the computer program from the memory to execute a visual alignment and laser welding method for capacitor cores as described in the above embodiment.

[0109] In the embodiments of this application, the processor 301 described above can be at least one of the following: Application-Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), Digital Signal Processing Device (DSPD), Programmable Logic Device (PLD), Field Programmable Gate Array (FPGA), Central Processing Unit (CPU), Controller, Microcontroller, and Microprocessor. It is understood that for different devices, the electronic device used to implement the above processor function can also be other types, and the embodiments of this application do not specifically limit it.

[0110] This application provides a computer-readable storage medium storing a computer program for implementing, when executed by a processor, a method for visual alignment and laser welding of capacitor cores as described in any of the above embodiments.

[0111] For example, the program instructions corresponding to the visual alignment and laser welding method for capacitor cores in this embodiment can be stored on storage media such as optical discs, hard disks, and USB flash drives. When the program instructions corresponding to the visual alignment and laser welding method for capacitor cores in the storage media are read or executed by an electronic device, the visual alignment and laser welding method for capacitor cores as described in any of the above embodiments can be implemented.

[0112] Furthermore, in the embodiments of this application, the functional modules can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional module.

[0113] If the integrated unit is implemented as a software functional module and is not sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this embodiment, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the method of this embodiment. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0114] It should be understood that the phrases "one embodiment," "an embodiment," or "some embodiments" mentioned throughout the specification mean that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "in one embodiment," "in one embodiment," or "in some embodiments" appearing throughout the specification do not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. It should be understood that in the various embodiments of this application, the sequence numbers of the above-described processes do not imply a sequential order of execution; the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application. The sequence numbers of the embodiments in this application are merely for descriptive purposes and do not represent the superiority or inferiority of the embodiments. The descriptions of the various embodiments above tend to emphasize the differences between the various embodiments; their similarities or commonalities can be referred to mutually, and for the sake of brevity, these will not be repeated here.

[0115] The modules described above as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules. They may be located in one place or distributed across multiple network units. Some or all of the modules may be selected to achieve the purpose of this embodiment according to actual needs.

[0116] In addition, each functional module in the various embodiments of this application can be integrated into one processing unit, or each module can be a separate unit, or two or more modules can be integrated into one unit; the integrated modules can be implemented in hardware or in the form of hardware plus software functional units.

[0117] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media that can store program code, such as mobile storage devices, read-only memory (ROM), magnetic disks, or optical disks.

[0118] The methods disclosed in the several method embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments.

[0119] The features disclosed in the several product embodiments provided in this application can be arbitrarily combined without conflict to obtain new product embodiments.

[0120] The features disclosed in the several method or device embodiments provided in this application can be arbitrarily combined without conflict to obtain new method or device embodiments.

[0121] The above description is merely an implementation method of the embodiments of this application, but the protection scope of the embodiments of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the protection scope of this application.

Claims

1. A method for visual alignment and laser welding of capacitor cores, characterized in that, The method includes: Acquire three-dimensional point cloud data of the capacitor core end face and the lead to be soldered, and perform feature extraction and coordinate calculation on the three-dimensional point cloud data to obtain alignment transformation parameters and average assembly gap value. Based on the three-dimensional point cloud data and the average assembly gap value, the optimal laser welding parameters are generated through a pre-determined convolutional neural network model. Based on the alignment transformation parameters and the optimal laser welding parameters, a dynamic focusing pulsed fiber laser is driven to weld the end face of the capacitor core to the lead to be welded, and the welding process is monitored in real time to obtain real-time monitoring data. Based on the real-time monitoring data, multi-dimensional quality indicators and quality judgments are calculated to obtain quality judgment results, and process data management is performed based on the quality judgment results.

2. The method according to claim 1, characterized in that, The acquisition of three-dimensional point cloud data of the capacitor core end face and the lead to be soldered, and the feature extraction and coordinate calculation of the three-dimensional point cloud data to obtain alignment transformation parameters and average assembly gap values, include: Using a binocular stereo vision system, the capacitor core and the lead to be soldered are globally located to determine the target area including the end face of the capacitor core and the lead to be soldered. The target area is scanned to obtain the three-dimensional point cloud data; wherein, the three-dimensional point cloud data includes the three-dimensional coordinates and reflection intensity information of each data point in the target area; Based on the three-dimensional coordinates of the three-dimensional point cloud data, calculate the center point coordinates and normal vector of the capacitor core end face and the soldering surface of the lead to be soldered, respectively; Based on the center point coordinates and the normal vector, the alignment transformation parameters are calculated, and based on the alignment transformation parameters, the spatial relationship between the core and the lead point cloud is calculated to obtain the average assembly gap value.

3. The method according to claim 1, characterized in that, The process of generating optimal laser welding parameters based on the 3D point cloud data and the average assembly gap value using a pre-determined convolutional neural network model includes: Based on the three-dimensional point cloud data and the average assembly gap value, a numerical feature vector is calculated; The optimal laser welding parameters are obtained by performing feature extraction, feature fusion, and parameter regression on the numerical feature vector using the convolutional neural network model; wherein the optimal laser welding parameters include at least one of laser peak power, pulse width, pulse frequency, welding scanning speed, and spot compensation diameter.

4. The method according to claim 3, characterized in that, The calculation of the numerical feature vector based on the three-dimensional point cloud data and the average assembly gap value includes: Based on the reflection intensity information in the three-dimensional point cloud data, low-value areas with abnormal intensity are identified to obtain cleanliness index values. Based on the three-dimensional coordinates in the three-dimensional point cloud data, the root mean square error of the height between the point cloud of the surface to be welded and the best-fit plane is calculated to obtain the flatness index value. The numerical feature vector is determined based on the cleanliness index value, the flatness index value, and the average assembly gap value.

5. The method according to claim 3, characterized in that, The process of extracting features, fusing features, and regressing parameters from the numerical feature vector using the convolutional neural network model to obtain the optimal laser welding parameters includes: One-dimensional feature extraction is performed on the numerical feature vector using the convolutional neural network model to obtain a one-dimensional feature map; The one-dimensional feature map is flattened to obtain a one-dimensional vector, and the one-dimensional vector is nonlinearly fused to obtain a fused feature. The optimal laser welding parameters are obtained by performing welding parameter regression processing on the fusion characteristics.

6. The method according to claim 1, characterized in that, The method involves driving a dynamically focused pulsed fiber laser to weld the end face of the capacitor core to the lead to be welded, based on the alignment transformation parameters and the optimal laser welding parameters. The welding process is monitored in real time to obtain real-time monitoring data, including: Based on the aforementioned alignment transformation parameters, the movable capacitor core and the lead to be soldered are moved to the target soldering position; Based on the optimal laser welding parameters, the dynamic focusing pulsed fiber laser is driven to perform welding at the target welding position, and the welding process is monitored in real time to obtain dynamic images and temperature distribution information of the weld pool. Feature extraction is performed on the dynamic image and the temperature distribution information to obtain the dynamic features of the molten pool and the thermal cycle features, and the dynamic features of the molten pool and the thermal cycle features are determined as the real-time monitoring data.

7. The method according to claim 1, characterized in that, The process of calculating multi-dimensional quality indicators and quality judgments based on the real-time monitoring data to obtain quality judgment results, and managing process data based on the quality judgment results, includes: Based on the dynamic characteristics of the molten pool and thermal cycling characteristics in the real-time monitoring data, multi-dimensional quality indicators of the weld joint are calculated to obtain the forming quality index value, thermal influence index value and connectivity prediction index value. The forming quality index value, the heat-affected index value, and the connectivity prediction index value are compared with their respective corresponding process standard thresholds to obtain the quality judgment result. Based on the quality assessment results, process data management is performed; wherein, the process data management includes one of qualified release, online compensation, and rejection.

8. A visual alignment and laser welding system for capacitor cores, characterized in that, The visual alignment and laser welding system for capacitor cores includes: an acquisition module, a generation module, a monitoring module, and a calculation module, wherein... The acquisition module is used to acquire three-dimensional point cloud data of the capacitor core end face and the lead to be soldered, and to perform feature extraction and coordinate calculation on the three-dimensional point cloud data to obtain alignment transformation parameters and average assembly gap value. The generation module is used to generate optimal laser welding parameters based on the three-dimensional point cloud data and the average assembly gap value through a pre-determined convolutional neural network model. The monitoring module is used to drive a dynamic focusing pulsed fiber laser to weld the end face of the capacitor core to the lead to be welded based on the alignment transformation parameters and the optimal laser welding parameters, and to monitor the welding process in real time to obtain real-time monitoring data. The calculation module is used to calculate multi-dimensional quality indicators and quality judgments based on the real-time monitoring data, obtain quality judgment results, and perform process data management based on the quality judgment results.

9. A device for visual alignment and laser welding of capacitor cores, characterized in that, include: Processor and memory, of which, The memory is used to store computer programs; The processor is configured to call and run the computer program from the memory to perform the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores executable instructions for causing a processor to execute, thereby implementing the method of any one of claims 1 to 7.

Images