Capacitor electrical performance intelligent detection method and system based on visual detection

Abstract

The invention discloses a capacitor electrical performance intelligent detection method and system based on visual detection, after a capacitor enters a detection area, image acquisition is carried out, a visual system completes preliminary appearance screening, and position and attitude positioning is synchronously carried out; a test pin mechanism of the test machine dynamically adjusts the pressing position and angle according to the positioning information, and photographs the contact point of the test pin and the capacitor before and after the test pin is pressed; the water flows through a secondary visual inspection station, and whether newly added appearance defects are generated due to testing or not is rechecked through visual inspection; meanwhile, the system generates a unique ID for each capacitor; according to the electrical performance parameter grading and the appearance condition, warehouse division is realized; meanwhile, the equipment state and the product abnormal index are monitored, and an early warning or maintenance program is triggered. According to the invention, artificial fatigue and subjectivity can be eliminated, the detection consistency is higher, the manual maintenance cost is reduced, the stability of the production process is improved, material blocking is reduced, the data query and processing speed is improved, and the batch quality risk is avoided in advance.

Description

Technical Field

[0001] This invention relates to a capacitor electrical performance testing method, and more specifically, to a visual detection-based intelligent method and system for testing capacitor electrical performance. Background Technology

[0002] Multilayer capacitors (MLCCs) are core components in electronic devices, and their quality directly affects the reliability and stability of end devices. Therefore, after production, they must undergo rigorous electrical performance testing and visual inspection before leaving the factory. Electrical performance testing must cover key parameters such as withstand voltage, capacitance, loss, and insulation resistance, and is mainly completed using equipment such as dedicated withstand voltage testing machines and four-parameter testing machines. Visual inspection needs to check for various defects such as blackened terminals, scratches, tinning of the ceramic body, chipping, and fragmentation. Current technologies mostly use six-sided visual sorting machines or manual microscope sampling inspection. In current testing technologies, while the electrical performance testing equipment is highly automated, its accuracy and stability heavily rely on manual maintenance. The cleanliness and flatness of the test probes need to be manually checked every two hours, and the test discs need to be manually wiped every three hours. The timeliness of this maintenance directly determines the testing quality. Delayed maintenance can easily lead to misjudgments due to prong oxidation, contamination, or test disc positioning errors. This not only results in defective products flowing downstream but can also cause scratches or even breakage, leading to unnecessary production losses. Furthermore, in the appearance inspection stage, the high misjudgment rate of existing six-sided appearance sorting machines due to the diverse types of defects and complex judgment criteria necessitates a secondary inspection of acceptable products using a microscope, increasing labor costs and reducing efficiency. More importantly, current technology treats electrical performance testing and appearance inspection as independent processes, lacking a data correlation mechanism. In reality, there is a significant inherent connection between the two, such as... The lack of silver in the probe directly affects the product's voltage resistance, and tin contamination of the ceramic body may cause arcing during testing. However, this correlation has not been effectively explored, making it impossible to trace the root cause of quality problems and hindering targeted improvements at the process level. In addition, existing testing methods suffer from several common defects: First, they are highly dependent on manual labor. Coordinating process flows, monitoring equipment status, and sorting out defective products all require extensive manual intervention, which is not only inefficient but also prone to human errors such as mixing materials and missing inspections. Second, data from each process is isolated, with electrical performance parameters and appearance defect information stored separately, making comprehensive analysis impossible and hindering the discovery of deeper process hazards. Third, process quality control is lagging behind, relying mainly on post-incident sampling or retrospective analysis after problems occur, failing to achieve real-time monitoring and risk warning of the testing process, leading to the escalation of quality problems. Fourth, equipment utilization is low, with frequent downtime caused by test probe malfunctions, material jams, and frequent parameter adjustments, severely impacting production continuity.

[0003] Therefore, those skilled in the art are dedicated to providing a visual inspection-based intelligent detection method and system for capacitor electrical performance that can effectively solve the above-mentioned technical problems. Summary of the Invention

[0004] To achieve the above objectives, the present invention provides a visual inspection-based intelligent detection method for capacitor electrical performance, comprising the following steps: S1: After the capacitor enters the detection area, it is image acquired and the vision system completes the initial appearance screening and simultaneously performs position and orientation localization. S2: The electrical performance testing machine receives positioning information from the vision system. The test probe mechanism of the testing machine dynamically adjusts the pressing position and angle according to the positioning information to ensure that the test probe is aligned with the capacitor electrode, thus completely solving the problems of test defects, product scratches or breakage caused by position deviation. At the same time, before and after the test probe is pressed down, the contact point between the test probe and the capacitor is photographed. The image analysis is used to determine whether there are any abnormal states on the surface of the test probe that affect the contact performance. It also detects whether any new defects appear on the surface of the capacitor electrode after the test due to the test operation, and monitors whether the pressing depth and flatness of the test probe are consistent. S3: After the capacitor completes the electrical performance test, it flows through the secondary visual inspection station to re-inspect whether any new appearance defects have been caused by the test. At the same time, the system generates a unique ID for each capacitor, binds all the electrical performance parameters of the capacitor with the appearance image data before and after the test, and stores them in a unified database. S4: Accurately divide the equipment into compartments based on electrical performance parameters and appearance; conduct root cause analysis of defects by mining the associated database through background machine learning algorithms to locate potential problems in the production process, provide direction for process improvement, and monitor equipment status and abnormal product indicators to trigger early warning or maintenance procedures, thereby achieving process optimization and predictive maintenance.

[0005] Furthermore, step S1, which involves feeding and initial visual positioning detection, specifically includes: The capacitors enter the inspection area via a vibratory feeder or conveyor belt, and a high-resolution industrial camera first acquires images of the six surfaces of each capacitor. The preliminary appearance screening specifically includes: judging in real time whether there are foreign objects such as fragments, chipped edges, severe dirt, steel balls, etc. in the capacitor according to the predefined appearance standard card digital model. If they are found, they are immediately sorted to the waste bin to avoid them entering the electrical testing stage and causing equipment jamming or contamination. The position and attitude positioning specifically includes: accurately locating the position and center coordinates of the capacitor's electrode tips, providing high-precision positioning guidance signals for the robotic arm or test probe of the subsequent electrical performance testing equipment; The appearance standard card digital model is a predefined digital standard model used for appearance inspection, and the high-precision positioning guidance signal ensures that the robotic arm or test probe of the subsequent electrical performance testing equipment can accurately connect to the capacitor electrode.

[0006] Furthermore, the abnormal state in S2 specifically includes oxidation, pitting, and dirt; the new defects in S2 specifically include scratches, indentations, or solder adhesion. The newly added appearance defects in S3 specifically include breakdown points, arcing marks, and excessively deep test indentations; the specific electrical performance parameters include withstand voltage, capacity, loss, and insulation resistance.

[0007] Furthermore, the number of high-resolution industrial cameras is three, which are fixedly arranged in a ring around the circumference of the capacitor. The angle between the optical axis of an adjacent high-resolution industrial camera and the line connecting the center of the capacitor is 120°, and the optical axes of the three high-resolution industrial cameras are all perpendicular to the cross-section of the capacitor's conveying direction to avoid viewing angle coverage deviation caused by dynamic changes in the conveying path. The high-resolution industrial cameras are arranged in a ring along the vertical direction of the capacitor's conveying path to ensure that the three circumferential sides of the capacitor are acquired synchronously without distortion. The three high-resolution industrial cameras respectively acquire images of the three circumferential sides of the capacitor. At the same time, adjustable-focus industrial cameras are set above and below the conveying path to synchronously acquire images of the upper and lower end faces of the capacitor. During the acquisition process, the system automatically calls the preset camera parameter configuration according to the capacitor specifications and dynamically adjusts the light intensity and angle through a combination of ring and bar light sources. The bar light source is turned on to suppress reflections in the reflective areas of the capacitor's ceramic body, while the ring light source is turned on to highlight details in the metal end areas, ensuring that the six-sided images are free of shadows and overexposure. At the same time, the image acquisition and capacitor conveying speed are synchronized through a servo motor. When the capacitor reaches the preset acquisition station, the servo motor triggers the camera shutter to avoid image blurring caused by conveying displacement. The system simultaneously acquires capacitor images using near-infrared, visible light, and ultraviolet multispectral industrial cameras, and uses image fusion algorithms to overlay defect features from different spectra. After acquisition, the system automatically stitches and calibrates the coordinates of the six images to generate a three-dimensional image dataset with the center of the capacitor as the origin. This dataset includes metadata such as acquisition time, camera number, and light source parameters.

[0008] Furthermore, after the image acquisition is completed, the following steps are also included: S1a: Layered compression is applied to the 3D image dataset. The bottom layer retains 16-bit grayscale details of the original image, with grayscale values ​​ranging from 0 to 65535, to meet the requirements of subsequent high-precision defect identification. The compression ratio of the middle layer adaptive entropy coding ranges from 10:1 to 30:1, dynamically adjusted according to the ceramic material of the capacitor. The compression ratio of the middle layer adaptive entropy coding is... The expression is: In the formula This indicates the compression ratio of the middle-layer adaptive entropy coding; This indicates the density of the ceramic material in the current capacitor. This indicates the minimum density value of common ceramic materials; This indicates the maximum density value of common ceramic materials; The upper layer generates a 256×256 pixel thumbnail for quick preview; S1b: Assign an index tag to each frame of image, which includes the acquisition station number, timestamp, unique pre-ID of capacitor, light source parameters, and camera calibration value. This tag is embedded in the header of the compressed file in XML format and does not occupy additional storage space. When storing, the system synchronously writes the layered compressed image data and the index tag into the distributed database. S1c: The database adopts dynamic storage with hot data caching and cold data archiving. Image data collected within a preset time period is stored in the SSD cache area. In the dynamic storage strategy, the preset time of the SSD cache area can be dynamically adjusted according to the production cycle. When the detection task is at its peak, the cache time is automatically extended to 72 hours and shortened to 24 hours during off-peak periods to optimize the efficiency of storage resource allocation. Image data that exceeds the preset time and has no abnormal markers is automatically archived to the distributed hard disk array.

[0009] Furthermore, the machine learning algorithm includes a defect correlation analysis model and an equipment status prediction model. The defect correlation analysis model generates a multi-dimensional association rule base by mining the mapping relationship between capacitor appearance defects, electrical performance parameters, and production process data. The equipment status prediction model is based on historical data of test probe usage time, contact frequency, and degree of contamination, combined with real-time detected abnormal status parameters. The remaining maintenance cycle for the test pin is expressed as: In the formula Indicates the remaining maintenance cycle of the test probe; Indicates the rated maintenance cycle of the test probe; Indicates the number of times the test probe has been used; Indicates the rated number of uses for the test probe; Indicates the current contaminated area of ​​the test probe; Indicates the threshold for the contaminated area of ​​the test probe; This indicates that the number of times the item is used affects the weight. The degree of soiling indicates how much weight is affected; Establish a degradation prediction model to predict the maintenance cycle of test pins in advance, thereby shortening the maintenance response time by more than 30% and reducing the product defect rate caused by test pin failure.

[0010] Furthermore, it also includes a dynamic calibration step for electrical performance parameters. The system collects test data of laboratory standard capacitors every hour, compares it with the test results of the production line testing machine, and calculates the deviation coefficient. The deviation compensation model is optimized using the random forest algorithm, and automatically adjusts the test parameters of the testing machine by combining the capacitor specifications, ambient temperature and humidity, and testing machine running time data to ensure that the test error of electrical performance parameters is controlled within ±0.5% under different operating conditions. The calibration data is bound to the unique ID of the capacitor and archived. The compensation value for electrical performance parameter testing is expressed as follows: In the formula This indicates the compensation value for electrical performance parameter testing; This represents the difference between the actual ambient temperature and the standard temperature. This represents the difference between the actual ambient humidity and the standard humidity. This indicates the difference between the actual output voltage of the tester and the standard voltage. , , These represent the deviation compensation coefficients for temperature, humidity, and voltage, respectively. It also includes an environmental interference adaptive adjustment module, which monitors the dust concentration, airflow speed, and electromagnetic interference intensity in the detection area in real time. When the dust concentration exceeds the preset threshold, the high-voltage electrostatic dust removal device is automatically activated. When the airflow speed affects the stability of the capacitor's posture, the servo motor is linked to adjust the conveying buffer distance. When electromagnetic interference causes test signal fluctuations, the anti-interference filtering algorithm of the test machine is triggered to dynamically optimize the signal acquisition parameters and ensure the stability of test accuracy under environmental fluctuations.

[0011] Furthermore, the secondary visual inspection station adopts an AI defect classification model, which is trained through deep learning and can subdivide newly added appearance defects into three levels: fatal defects, serious defects, and minor defects. In the AI ​​defect classification model, fatal defects include penetration points and through cracks; severe defects include large-area arcing marks; and minor defects include tiny indentations. Fatal defects directly trigger sorting to the waste bin, while severe defects are marked and enter the re-inspection channel. Minor defects are classified into bins according to customer-customized standards. At the same time, the model supports online self-learning, feeding back the defect judgment results of manual review and correction to the model to continuously optimize the classification accuracy.

[0012] Furthermore, it also includes a multi-station collaborative scheduling mechanism, in which the system synchronizes the equipment load and testing progress data of each testing station in real time through the industrial internet platform; The threshold for workstation congestion warning is expressed as follows: In the formula Indicates the workstation congestion warning threshold; Indicates the rated load capacity of the workstation; Indicates the upstream conveying speed; Indicates the upstream rated conveying speed; This indicates the current operating efficiency of the equipment at the workstation; Indicates the rated operating efficiency of the workstation equipment; Indicates the influence coefficient of conveying speed; Indicates the influence coefficient on equipment efficiency; When a congestion occurs at a certain workstation, the conveying speed of the upstream vibratory feeder or conveyor belt is automatically adjusted, and some products to be tested are diverted to a backup testing workstation. When a camera or test probe malfunctions, a workstation switching command is immediately triggered to ensure that the overall testing process is uninterrupted and the overall utilization rate of the equipment is increased to over 95%. The distributed database supports data interaction with enterprise MES and ERP systems, synchronizing capacitor testing data to the MES system in real time for production progress tracking and process parameter adjustment; and synchronizing warehouse data, qualified product quantity, and scrap statistics to the ERP system to provide data support for inventory management and order delivery.

[0013] A visual inspection-based intelligent capacitor electrical performance testing system includes: The image acquisition and positioning module receives capacitors transported to the detection area via a vibratory feeder or conveyor belt. It acquires images of all six sides of the capacitors using multiple industrial cameras and performs position and orientation positioning. The industrial cameras include three high-resolution cameras arranged in a ring around the capacitor's circumference, and adjustable-focus industrial cameras positioned above and below the transport path. The angle between the optical axis of adjacent high-resolution cameras and the line connecting the capacitor's center is 120°. A ring-shaped strip light source is provided to dynamically adjust the light intensity and angle. A servo motor synchronizes the image acquisition with the capacitor transport speed. After acquisition, the six-sided images are automatically stitched and their coordinates calibrated to generate a three-dimensional image dataset. The image processing and storage module is used to perform layered compression processing on the 3D image dataset. The bottom layer retains 16-bit grayscale details, the middle layer dynamically adjusts the adaptive entropy coding compression ratio according to the ceramic material of the capacitor, and the top layer generates thumbnails. Each frame of the image is assigned an index tag containing information such as the acquisition station number and timestamp, which is embedded in the header of the compressed file. The processed image data and index tags are synchronously written to a distributed database that adopts a dynamic storage strategy. The electrical performance testing and monitoring module receives positioning information output from the image acquisition and positioning module, and controls the test probe mechanism to dynamically adjust the pressing position and angle. Before and after the test probe is pressed down, the image acquisition unit takes pictures of the contact point and the surface of the capacitor electrode to analyze and determine the abnormal state of the test probe and new defects in the electrode. At the same time, the multi-view vision system monitors the pressing depth and flatness of the test probe. The module also performs electrical performance parameter testing and dynamic calibration of electrical performance parameters. The environmental adaptive adjustment module is used to monitor the dust concentration, airflow speed and electromagnetic interference intensity in the detection area in real time. When the monitoring data exceeds the preset threshold, it triggers the high-voltage electrostatic dust removal device, the servo motor transmission parameter adjustment and the test machine anti-interference filtering algorithm to ensure the stability of the detection environment. The secondary visual re-inspection and compartmentation module is used to receive capacitors that have completed electrical performance testing. Through an AI defect classification model trained by deep learning, it re-inspects and classifies whether there are any new appearance defects in the capacitors. It generates a unique ID for each capacitor, binds and archives the electrical performance parameters with the appearance image data before and after the test, and achieves accurate compartmentation based on the electrical performance classification and appearance condition. The intelligent analysis and early warning module is used to mine database data through machine learning algorithms, build defect correlation analysis models and equipment status prediction models, conduct defect root cause analysis and equipment status prediction; monitor abnormal product indicators and equipment status in real time, and trigger corresponding early warning or maintenance procedures; at the same time, it establishes a multi-workstation collaborative scheduling mechanism to achieve dynamic balance of load and progress of each inspection workstation, and supports data interaction with enterprise MES system and ERP system.

[0014] The present invention has the following beneficial effects: 1. This invention precisely aligns the electrodes, avoiding test misjudgments, product scratches or breakage caused by positional deviations. Automated visual inspection replaces manual inspection, covering six sides of the appearance and test needle status, eliminating human fatigue and subjectivity, and resulting in higher test consistency. Multispectral image acquisition, reflection suppression and dynamic calibration technology reduce image distortion and false negative rate, and control the electrical performance test error within ±0.5%. 2. This invention eliminates the need for manual intervention throughout the entire process, from capacitor feeding, initial screening, and electrical performance testing to secondary inspection and sorting. It automatically monitors the oxidation and contamination of test probes, replacing manual inspections every 2 hours, and simultaneously enables predictive maintenance of equipment, reducing manual maintenance costs. It also avoids errors such as mixing and missed inspections caused by manual operation, thus improving the stability of the production process. 3. This invention assigns a unique ID to each capacitor, binds electrical performance parameters with appearance image data before and after testing, establishes a one-item-one-file full-dimensional database, and uses machine learning to mine data correlations to locate potential problems in the production process, which facilitates process improvement. 4. Automatic diversion of congested products and switching of faulty workstations increases the overall utilization rate of equipment to over 95%; early warning of equipment failure and product abnormalities reduces downtime caused by material jams and parameter adjustments, optimizes production continuity, and uses layered compression and dynamic storage strategies to balance data accuracy and storage efficiency, thereby improving data query and processing speed. 5. Real-time monitoring of environmental factors such as dust, temperature, humidity, and electromagnetic interference, and automatic activation of adaptive adjustment mechanisms to ensure detection stability; at the same time, the AI ​​defect classification model accurately grades defects and quickly sorts out defective products, transforming traditional post-event sampling inspection into in-process control plus pre-event early warning, and avoiding batch quality risks in advance. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the intelligent detection method for capacitor electrical performance based on visual inspection in this invention. Figure 2 This is a schematic block diagram of the intelligent capacitor electrical performance detection system based on vision detection in this invention; Figure 3 This is a detailed flowchart illustrating the intelligent capacitor detection process in this invention. Figure 4 This is a schematic block diagram of the image acquisition and positioning module in this invention; Figure 5 This is a schematic block diagram of the data processing and storage process in this invention. Detailed Implementation

[0016] The present invention will be further described below with reference to the accompanying drawings and embodiments: In the description of this invention, it should be noted that the terms "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0017] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "setting," and "connection" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0018] like Figures 1 to 5 As shown, a visual inspection-based intelligent detection method for capacitor electrical performance includes the following steps: S1: After the capacitor enters the detection area, it is image acquired and the vision system completes the initial appearance screening and simultaneously performs position and orientation localization. S2: The electrical performance testing machine (withstand voltage, four-parameter testing machine) receives positioning information from the vision system. The test probe mechanism of the testing machine dynamically adjusts the pressing position and angle based on this positioning information to ensure the test probe is aligned with the capacitor electrode, completely solving the problems of test defects, product scratches, or breakage caused by positional deviations. Simultaneously, before and after the test probe is pressed down, the contact point between the test probe and the capacitor is photographed. Image analysis is used to determine if there are any abnormal conditions on the test probe surface that affect contact performance (such as oxidation, pits, or dirt, replacing the original manual finger touch inspection). After testing, it detects whether any new defects appear on the capacitor electrode surface due to the testing operation (such as scratches, indentations, or solder adhesion, achieving real-time monitoring of damage during the testing process). It also monitors whether the test probe pressing depth and flatness are consistent. The test probe pressing depth and flatness are measured using multi-view vision. The multi-view vision system acquires at least three sets of images from different perspectives during the test probe pressing process, calculates the pressing depth based on the triangulation principle, and determines whether the flatness meets the standard through image grayscale value distribution uniformity analysis. S3: After the capacitor completes the electrical performance test, it flows through the secondary visual inspection station to re-inspect whether any new appearance defects have been caused by the test. At the same time, the system generates a unique ID for each capacitor, binds all the electrical performance parameters of the capacitor with the appearance image data before and after the test, and stores them in a unified database. S4: Accurately categorize products based on electrical performance parameters (e.g., J, K, M grades) and appearance (e.g., products with good electrical performance but minor scratches, or products with good electrical performance and perfect appearance); utilize backend machine learning algorithms to mine related databases and conduct defect root cause analysis to pinpoint potential problems in the production process, providing direction for process improvement (e.g., frequent micro-arc traces on electrode edges after withstand voltage testing of a batch of capacitors, indicating problems with raw materials or sintering processes; products with poor insulation resistance generally have micro-cracks in the ceramic body or terminal contamination); simultaneously monitor equipment status (e.g., test probe contamination) and abnormal product indicators (e.g., product solder loss rate exceeding a threshold), triggering warnings or maintenance procedures (e.g., automatic prompts or triggering of cleaning procedures, and recording the event), achieving process optimization and predictive maintenance.

[0019] Step S1 consists of material loading and initial visual positioning detection, specifically including: The capacitors enter the inspection area via a vibratory feeder or conveyor belt, and a high-resolution industrial camera first acquires images of the six surfaces of each capacitor. The preliminary appearance screening specifically includes: judging in real time whether there are foreign objects such as fragments, chipped edges, severe dirt, steel balls, etc. in the capacitor according to the predefined appearance standard card digital model. If they are found, they are immediately sorted to the waste bin to avoid them entering the electrical testing stage and causing equipment jamming or contamination. The position and attitude positioning specifically includes: accurately locating the position and center coordinates of the capacitor's electrode tips, providing high-precision positioning guidance signals for the robotic arm or test probe of the subsequent electrical performance testing equipment; The appearance standard card digital model is a predefined digital standard model used for appearance inspection, and the high-precision positioning guidance signal ensures that the robotic arm or test probe of the subsequent electrical performance testing equipment can accurately connect to the capacitor electrode.

[0020] The abnormal states in S2 specifically include oxidation, pitting, and dirt (replacing the original manual finger touch inspection); the new defects in S2 specifically include scratches, indentations, or solder adhesion (achieving real-time monitoring of damage during the testing process); The newly added appearance defects in S3 specifically include breakdown points, arcing marks, and excessively deep test indentations; the specific electrical performance parameters include withstand voltage, capacity, loss, and insulation resistance.

[0021] The system comprises three high-resolution industrial cameras, which are fixedly arranged in a ring around the circumference of the capacitor. The angle between the optical axis of an adjacent high-resolution industrial camera and the line connecting the center of the capacitor is 120°, and the optical axes of all three high-resolution industrial cameras are perpendicular to the cross-section of the capacitor's conveying direction to avoid viewing angle coverage deviations caused by dynamic changes in the conveying path. The high-resolution industrial cameras are arranged in a ring along the vertical direction of the capacitor's conveying path to ensure that distorted images of the three circumferential sides of the capacitor are acquired synchronously. The three high-resolution industrial cameras acquire images of the three circumferential sides of the capacitor, and adjustable-focus industrial cameras are set above and below the conveying path to synchronously acquire images of the upper and lower end faces of the capacitor. During the data acquisition process, the system automatically calls the preset camera parameter configuration (including exposure time, ISO, and lens focal length) according to the capacitor specifications and dynamically adjusts the light intensity and angle through a combination of ring and bar light sources. Bar light sources are activated to suppress reflections in the reflective areas of the capacitor's ceramic body, while ring light sources are activated to highlight details in the metal end areas, ensuring that the six-sided images are free of shadows and overexposure. Simultaneously, image acquisition and capacitor transport speed are synchronized via a servo motor. When the capacitor reaches the preset acquisition station, the servo motor triggers the camera shutter, preventing image blurring caused by transport displacement. Near-infrared, visible light, and ultraviolet multispectral industrial cameras are used to simultaneously acquire capacitor images. Image fusion algorithms are used to overlay defect features from different spectra, enhancing the detection capabilities for defects that conventional vision cannot identify, such as internal cracks in the ceramic body and latent oxidation of the electrodes. The fused data is then bound to a 3D image dataset to improve the dimensionality and accuracy of defect identification. After acquisition, the system automatically stitches and calibrates the coordinates of the six images to generate a three-dimensional image dataset with the center of the capacitor as the origin. This dataset includes metadata such as acquisition time, camera number, and light source parameters.

[0022] The technical solution of this invention can solve the problems in the prior art, such as missed images of the end face due to single-view acquisition, reflection caused by fixed light source, image distortion caused by shadow, and image blurring caused by asynchronous acquisition and transmission. Compared with the operation method of relying on manual adjustment of camera angle, fixed light source parameters, and manual screening of clear images afterward, it can significantly reduce the image acquisition misjudgment rate and reduce the subsequent detection deviation caused by image quality problems.

[0023] After the image acquisition is completed, the following steps are also included: S1a: Layered compression is applied to the 3D image dataset. The bottom layer retains 16-bit grayscale details of the original image, with grayscale values ​​ranging from 0 to 65535, to meet the requirements of subsequent high-precision defect identification. The compression ratio of the middle layer adaptive entropy coding ranges from 10:1 to 30:1, dynamically adjusted according to the ceramic material of the capacitor (such as alumina or niobium oxide). The compression ratio of the middle layer adaptive entropy coding is... The expression is: In the formula This indicates the compression ratio of the middle-layer adaptive entropy coding; This indicates the density of the ceramic material in the current capacitor. This indicates the minimum density value of common ceramic materials; This indicates the maximum density value of common ceramic materials; The upper layer generates a 256×256 pixel thumbnail for quick preview; S1b: Assign an index tag to each frame of image, which includes the acquisition station number, timestamp, unique pre-ID of capacitor, light source parameters, and camera calibration value. This tag is embedded in the header of the compressed file in XML format and does not occupy additional storage space. When storing, the system synchronously writes the layered compressed image data and the index tag into the distributed database. S1c: The database adopts dynamic storage with hot data caching and cold data archiving. Image data collected within a preset time period is stored in the SSD cache. In the dynamic storage strategy, the preset time of the SSD cache can be dynamically adjusted according to the production cycle (adjustment range is 24-72 hours). When the detection task is at its peak, the cache time is automatically extended to 72 hours and shortened to 24 hours during off-peak periods to optimize storage resource allocation efficiency. Image data that exceeds the preset time and has no abnormal markers is automatically archived to the distributed hard disk array.

[0024] The machine learning algorithm includes a defect correlation analysis model and an equipment status prediction model. The defect correlation analysis model generates a multi-dimensional association rule base by mining the mapping relationship between capacitor appearance defects, electrical performance parameters, and production process data. The equipment status prediction model is based on historical data of test probe usage time, contact frequency, and degree of contamination, combined with real-time detected abnormal status parameters. The remaining maintenance cycle for the test pin is expressed as: In the formula Indicates the remaining maintenance cycle of the test probe; Indicates the rated maintenance cycle of the test probe; Indicates the number of times the test probe has been used; Indicates the rated number of uses for the test probe; Indicates the current contaminated area of ​​the test probe; Indicates the threshold for the contaminated area of ​​the test probe; This indicates that the number of times the item is used affects the weight. The degree of soiling indicates how much weight is affected; Establish a degradation prediction model to anticipate test probe maintenance cycles in advance. This reduces maintenance response time by more than 30% and lowers product defect rates caused by test probe failure.

[0025] This invention also includes a dynamic calibration step for electrical performance parameters. In this step, the system collects test data of a laboratory standard capacitor every hour, compares it with the test results of the production line testing machine, and calculates the deviation coefficient. The deviation compensation model is optimized using a random forest algorithm, and automatically adjusts the test parameters of the testing machine by combining the capacitor specifications, ambient temperature and humidity, and testing machine runtime data to ensure that the test error of electrical performance parameters is controlled within ±0.5% under different operating conditions. The calibration data is bound to the capacitor's unique ID and archived. The compensation value for electrical performance parameter testing is expressed as follows: In the formula This indicates the compensation value for electrical performance parameter testing; This represents the difference between the actual ambient temperature and the standard temperature. This represents the difference between the actual ambient humidity and the standard humidity. This indicates the difference between the actual output voltage of the tester and the standard voltage. , , These represent the deviation compensation coefficients for temperature, humidity, and voltage, respectively. It also includes an environmental interference adaptive adjustment module, which monitors dust concentration, airflow speed, and electromagnetic interference intensity in the detection area in real time. When the dust concentration exceeds a preset threshold, it automatically activates the high-voltage electrostatic dust removal device; when the airflow speed affects the stability of the capacitor's posture, it links the servo motor to adjust the conveying buffer distance; when electromagnetic interference causes fluctuations in the test signal, it triggers the anti-interference filtering algorithm of the test machine to dynamically optimize the signal acquisition parameters and ensure the stability of test accuracy under fluctuating detection environment. The secondary vision inspection station adopts an AI defect classification model, which, through deep learning training, can subdivide newly added appearance defects into three levels: fatal defects, serious defects, and minor defects. Fatal defects (such as penetration points and through cracks) directly trigger sorting to the waste bin, while serious defects (such as large-area arc marks) are marked and enter the re-inspection channel; minor defects (such as tiny indentations) are sorted into bins according to customer-customized standards; at the same time, the model supports online self-learning, feeding back the defect judgment results of manual review and correction to the model to continuously optimize the classification accuracy.

[0026] This invention also includes a multi-station collaborative scheduling mechanism, in which the system synchronizes the equipment load and testing progress data of each testing station in real time through an industrial internet platform; The threshold for workstation congestion warning is expressed as follows: In the formula Indicates the workstation congestion warning threshold; Indicates the rated load capacity of the workstation; Indicates the upstream conveying speed; Indicates the upstream rated conveying speed; This indicates the current operating efficiency of the equipment at the workstation; Indicates the rated operating efficiency of the workstation equipment; Indicates the influence coefficient of conveying speed; Indicates the influence coefficient on equipment efficiency; When a workstation becomes congested (e.g., the backlog of products to be tested on the electrical performance testing machine exceeds a preset threshold), the conveying speed of the upstream vibratory feeder or conveyor belt is automatically adjusted, and some products to be tested are diverted to a backup testing workstation. When a camera or test probe malfunctions, a workstation switching command is immediately triggered to ensure uninterrupted testing process and improve equipment utilization to over 95%. The distributed database supports data interaction with the enterprise's MES and ERP systems, synchronizing capacitor testing data (unique ID, electrical performance parameters, appearance defect information, testing time) to the MES system in real time for production progress tracking and process parameter adjustment. Warehouse data, the number of qualified products, and waste statistics are synchronized to the ERP system to provide data support for inventory management and order delivery.

[0027] A visual inspection-based intelligent capacitor electrical performance testing system includes: S100: Image acquisition and positioning module, used to receive capacitors transported to the detection area via a vibratory feeder or conveyor belt, and to acquire images of all six sides of the capacitors using multiple sets of industrial cameras, achieving position and orientation positioning; the industrial cameras include three high-resolution industrial cameras arranged in a ring around the circumference of the capacitor, and adjustable-focus industrial cameras positioned above and below the transport path, with the angle between the optical axis of adjacent high-resolution industrial cameras and the line connecting the center of the capacitor being 120°; a ring-shaped strip light source is provided to dynamically adjust the light intensity and angle, and a servo motor is used to synchronize image acquisition with the capacitor transport speed; after acquisition, the six-sided images are automatically stitched and coordinate calibrated to generate a three-dimensional image dataset; S200: Image processing and storage module, used to perform layered compression processing on 3D image datasets. The bottom layer retains 16-bit grayscale details, the middle layer dynamically adjusts the adaptive entropy coding compression ratio according to the ceramic material of the capacitor, and the top layer generates thumbnails. It assigns an index tag containing information such as the acquisition station number and timestamp to each frame of image, embeds it in the header of the compressed file, and synchronously writes the processed image data and index tags to a distributed database that adopts a dynamic storage strategy. S300: Electrical performance testing and monitoring module, used to receive positioning information output by the image acquisition and positioning module, control the test probe mechanism to dynamically adjust the pressing position and angle; before and after the test probe is pressed down, the image acquisition unit takes pictures and analyzes the contact point and the surface of the capacitor electrode to determine the abnormal state of the test probe and the new defects of the electrode, and at the same time monitors the pressing depth and flatness of the test probe through the multi-view vision system; simultaneously completes the electrical performance parameter test and performs dynamic calibration of the electrical performance parameters; S400: Environmental adaptive adjustment module, used to monitor dust concentration, airflow speed and electromagnetic interference intensity in the detection area in real time. When the monitoring data exceeds the preset threshold, it will trigger the high-voltage electrostatic dust removal device, the servo motor transmission parameter adjustment and the test machine anti-interference filtering algorithm to ensure the stability of the detection environment. S500: Secondary visual re-inspection and compartmentation module, used to receive capacitors that have completed electrical performance testing, and re-inspect and classify the capacitors for new appearance defects through an AI defect classification model trained by deep learning; generate a unique ID for each capacitor, bind and archive the electrical performance parameters with the appearance image data before and after the test, and achieve accurate compartmentation based on electrical performance classification and appearance condition. S600: Intelligent Analysis and Early Warning Module, used to mine database data through machine learning algorithms, build defect correlation analysis models and equipment status prediction models, conduct defect root cause analysis and equipment status prediction; monitor product abnormal indicators and equipment status in real time, trigger corresponding early warning or maintenance procedures; at the same time, establish a multi-station collaborative scheduling mechanism to achieve dynamic balance of load and progress of each inspection station, and support data interaction with enterprise MES system and ERP system.

[0028] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A visual inspection-based intelligent detection method for capacitor electrical performance, characterized in that, Includes the following steps: S1: After the capacitor enters the detection area, it is image acquired and the vision system completes the initial appearance screening and simultaneously performs position and orientation localization. S2: The electrical performance testing machine receives positioning information from the vision system. The test probe mechanism of the testing machine dynamically adjusts the pressing position and angle according to the positioning information to ensure that the test probe is aligned with the capacitor electrode. At the same time, before and after the test probe is pressed down, the contact point between the test probe and the capacitor is photographed. The image analysis is used to determine whether there are any abnormal states on the surface of the test probe that affect the contact performance. The test probe also detects whether any new defects appear on the surface of the capacitor electrode after the test due to the test operation, and monitors whether the pressing depth and flatness of the test probe are consistent. S3: After the capacitor completes the electrical performance test, it flows through the secondary visual inspection station to re-inspect whether any new appearance defects have been caused by the test. At the same time, the system generates a unique ID for each capacitor, binds all electrical performance parameters of the capacitor with the appearance image data before and after the test, and stores them in a unified database. S4: Based on electrical performance parameters and appearance, compartments are divided; through background machine learning algorithms, related databases are mined to conduct defect root cause analysis to locate potential problems in the production process, while monitoring equipment status and product abnormal indicators to trigger early warnings or maintenance procedures.

2. The intelligent detection method for capacitor electrical performance based on vision detection as described in claim 1, characterized in that, Step S1 consists of material loading and initial visual positioning detection, specifically including: The capacitors enter the inspection area via a vibratory feeder or conveyor belt, and a high-resolution industrial camera first acquires images of the six surfaces of each capacitor. The preliminary appearance screening specifically includes: judging in real time whether there are foreign objects such as fragments, chipped edges, severe dirt, steel balls, etc. in the capacitor according to the predefined appearance standard card digital model. If they are found, they are immediately sorted to the waste bin to avoid them entering the electrical testing stage and causing equipment jamming or contamination. The position and attitude positioning specifically includes: accurately locating the position and center coordinates of the capacitor's electrode tips, providing high-precision positioning guidance signals for the robotic arm or test probe of the subsequent electrical performance testing equipment; The appearance standard card digital model is a predefined digital standard model used for appearance inspection, and the high-precision positioning guidance signal ensures that the robotic arm or test probe of the subsequent electrical performance testing equipment can accurately connect to the capacitor electrode.

3. The intelligent detection method for capacitor electrical performance based on vision detection as described in claim 2, characterized in that, The abnormal states in S2 specifically include oxidation, pitting, and contamination; the new defects in S2 specifically include scratches, indentations, or solder adhesion. The newly added appearance defects in S3 specifically include breakdown points, arcing marks, and excessively deep test indentations; the specific electrical performance parameters include withstand voltage, capacity, loss, and insulation resistance.

4. The intelligent detection method for capacitor electrical performance based on vision detection as described in claim 3, characterized in that, The system comprises three high-resolution industrial cameras, which are fixedly arranged in a ring around the circumference of the capacitor. The angle between the optical axis of an adjacent high-resolution industrial camera and the line connecting the center of the capacitor is 120°, and the optical axes of all three high-resolution industrial cameras are perpendicular to the cross-section of the capacitor's conveying direction to avoid viewing angle coverage deviations caused by dynamic changes in the conveying path. The high-resolution industrial cameras are arranged in a ring along the vertical direction of the capacitor's conveying path to ensure that distorted images of the three circumferential sides of the capacitor are acquired synchronously. The three high-resolution industrial cameras acquire images of the three circumferential sides of the capacitor, and adjustable-focus industrial cameras are set above and below the conveying path to synchronously acquire images of the upper and lower end faces of the capacitor. During the acquisition process, the system automatically calls the preset camera parameter configuration according to the capacitor specifications and dynamically adjusts the light intensity and angle through a combination of ring and bar light sources. The bar light source is turned on to suppress reflections in the reflective areas of the capacitor's ceramic body, while the ring light source is turned on to highlight details in the metal end areas, ensuring that the six-sided images are free of shadows and overexposure. At the same time, the image acquisition and capacitor conveying speed are synchronized through a servo motor. When the capacitor reaches the preset acquisition station, the servo motor triggers the camera shutter. The system simultaneously acquires capacitor images through near-infrared, visible light, and ultraviolet multispectral industrial cameras, and uses image fusion algorithms to overlay defect features under different spectra. After the data acquisition is completed, the system automatically stitches and calibrates the coordinates of the six images to generate a three-dimensional image dataset with the center of the capacitor as the origin.

5. The intelligent detection method for capacitor electrical performance based on vision detection as described in claim 4, characterized in that, After the image acquisition is completed, the following steps are also included: S1a: Layered compression is applied to the 3D image dataset. The bottom layer retains the 16-bit grayscale details of the original image, with grayscale values ​​ranging from 0 to 65535. The compression ratio of the middle layer's adaptive entropy coding ranges from 10:1 to 30:1, dynamically adjusted according to the ceramic material of the capacitor. The compression ratio of the middle layer's adaptive entropy coding is... The expression is: In the formula This indicates the compression ratio of the middle-layer adaptive entropy coding; This indicates the density of the ceramic material in the capacitor. This indicates the minimum density value of common ceramic materials; This indicates the maximum density value of common ceramic materials; The upper layer generates a 256×256 pixel thumbnail; S1b: Assign an index tag to each frame of image, which includes the acquisition station number, timestamp, unique pre-ID of capacitor, light source parameters, and camera calibration value. This tag is embedded in the header of the compressed file in XML format and does not occupy additional storage space. When storing, the system synchronously writes the layered compressed image data and the index tag into the distributed database. S1c: The database adopts dynamic storage with hot data caching and cold data archiving. Image data collected within a preset time period is stored in the SSD cache area. In the dynamic storage strategy, the preset time of the SSD cache area can be dynamically adjusted according to the production cycle. When the detection task is at its peak, the cache time is automatically extended to 72 hours and shortened to 24 hours during off-peak periods. Image data that exceeds the preset time and has no abnormal markers is automatically archived to the distributed hard disk array.

6. The intelligent detection method for capacitor electrical performance based on vision detection as described in claim 5, characterized in that, The machine learning algorithm includes a defect correlation analysis model and an equipment status prediction model. The defect correlation analysis model generates a multi-dimensional association rule base by mining the mapping relationship between capacitor appearance defects, electrical performance parameters, and production process data. The equipment status prediction model is based on historical data of test probe usage time, contact frequency, and degree of contamination, combined with real-time detected abnormal status parameters. The remaining maintenance cycle for the test pin is expressed as: In the formula Indicates the remaining maintenance cycle of the test probe; Indicates the rated maintenance cycle of the test probe; Indicates the number of times the test probe has been used; Indicates the rated number of uses for the test probe; Indicates the current contaminated area of ​​the test probe; Indicates the threshold for the contaminated area of ​​the test probe; This indicates that the number of times the item is used affects the weight. The degree of soiling indicates how much weight is affected; Establish a recession prediction model to predict the maintenance cycle of test needles in advance.

7. The intelligent detection method for capacitor electrical performance based on vision detection as described in claim 6, characterized in that, It also includes a dynamic calibration step for electrical performance parameters. The system collects test data of laboratory standard capacitors every hour, compares it with the test results of the production line testing machine, and calculates the deviation coefficient. The deviation compensation model is optimized using the random forest algorithm. Combined with capacitor specifications, ambient temperature and humidity, and testing machine running time data, the test parameters of the testing machine are automatically adjusted to ensure that the test error of electrical performance parameters is controlled within ±0.5% under different operating conditions. The calibration data is bound to the unique ID of the capacitor and archived. The compensation value for electrical performance parameter testing is expressed as follows: In the formula This indicates the compensation value for electrical performance parameter testing; This represents the difference between the actual ambient temperature and the standard temperature. This represents the difference between the actual ambient humidity and the standard humidity. This indicates the difference between the actual output voltage of the tester and the standard voltage. , , These represent the deviation compensation coefficients for temperature, humidity, and voltage, respectively. It also includes an environmental interference adaptive adjustment module, which monitors the dust concentration, airflow speed, and electromagnetic interference intensity in the detection area in real time. When the dust concentration exceeds the preset threshold, the high-voltage electrostatic dust removal device is automatically activated. When the airflow speed affects the stability of the capacitor's posture, the servo motor is linked to adjust the conveying buffer distance. When electromagnetic interference causes test signal fluctuations, the anti-interference filtering algorithm of the test machine is triggered to dynamically optimize the signal acquisition parameters and ensure the stability of test accuracy under environmental fluctuations.

8. The intelligent detection method for capacitor electrical performance based on vision detection as described in claim 7, characterized in that, The secondary visual inspection station uses an AI defect classification model, which is trained through deep learning and can classify new appearance defects into three levels: fatal defects, serious defects, and minor defects. In the AI ​​defect classification model, fatal defects include penetration points and through cracks; serious defects include large-area arcing marks; minor defects include tiny indentations. Fatal defects directly trigger sorting to the waste bin, while serious defects are marked and enter the re-inspection channel. Minor defects are classified into different warehouses based on customized standards; meanwhile, the model supports online self-learning, feeding back the defect judgment results of manual review and correction to the model.

9. The intelligent detection method for capacitor electrical performance based on vision detection as described in claim 8, characterized in that, It also includes a multi-station collaborative scheduling mechanism, in which the system synchronizes the equipment load and testing progress data of each testing station in real time through the industrial internet platform. The threshold for workstation congestion warning is expressed as follows: In the formula Indicates the workstation congestion warning threshold; Indicates the rated load capacity of the workstation; Indicates the upstream conveying speed; Indicates the upstream rated conveying speed; This indicates the current operating efficiency of the equipment at the workstation; Indicates the rated operating efficiency of the workstation equipment; Indicates the influence coefficient of conveying speed; Indicates the influence coefficient on equipment efficiency; When a congestion occurs at a certain workstation, the conveying speed of the upstream vibratory feeder or conveyor belt is automatically adjusted, and some of the products to be tested are diverted to a backup testing workstation; when a camera or test probe malfunctions, a workstation switching command is immediately triggered. The distributed database supports data interaction with enterprise MES and ERP systems, synchronizing capacitor testing data to the MES system in real time; and synchronizing warehouse data, qualified product quantity, and waste statistics to the ERP system.

10. A visual inspection-based intelligent detection system for capacitor electrical performance, characterized in that, include: The image acquisition and positioning module receives capacitors transported to the detection area via a vibratory feeder or conveyor belt. It acquires images of all six sides of the capacitors using multiple industrial cameras and performs position and orientation positioning. The industrial cameras include three high-resolution cameras arranged in a ring around the capacitor's circumference, and adjustable-focus industrial cameras positioned above and below the transport path. The angle between the optical axis of adjacent high-resolution cameras and the line connecting the capacitor's center is 120°. A ring-shaped strip light source is provided to dynamically adjust the light intensity and angle. A servo motor synchronizes the image acquisition with the capacitor transport speed. After acquisition, the six-sided images are automatically stitched and their coordinates calibrated to generate a three-dimensional image dataset. The image processing and storage module is used to perform layered compression processing on the 3D image dataset. The bottom layer retains 16-bit grayscale details, the middle layer dynamically adjusts the adaptive entropy coding compression ratio according to the ceramic material of the capacitor, and the top layer generates thumbnails. Each frame of image is assigned an index tag containing information such as the acquisition station number and timestamp, which is embedded in the header of the compressed file. The processed image data and the index tag are synchronously written to a distributed database that adopts a dynamic storage strategy. The electrical performance testing and monitoring module is used to receive the positioning information output by the image acquisition and positioning module, and control the test probe mechanism to dynamically adjust the pressing position and angle. Before and after the test probe is pressed down, the image acquisition unit takes pictures and analyzes the contact point and the surface of the capacitor electrode to determine the abnormal state of the test probe and the new defects of the electrode. At the same time, the multi-view vision system monitors the pressing depth and flatness of the test probe. Simultaneously complete electrical performance parameter testing and perform dynamic calibration of electrical performance parameters; The environmental adaptive adjustment module is used to monitor the dust concentration, airflow speed and electromagnetic interference intensity in the detection area in real time. When the monitoring data exceeds the preset threshold, it triggers the high-voltage electrostatic dust removal device, the servo motor transmission parameter adjustment and the test machine anti-interference filtering algorithm to ensure the stability of the detection environment. The secondary visual re-inspection and compartmentation module is used to receive capacitors that have completed electrical performance testing. Through an AI defect classification model trained by deep learning, it re-inspects and classifies whether there are any new appearance defects in the capacitors. It generates a unique ID for each capacitor, binds and archives the electrical performance parameters with the appearance image data before and after the test, and achieves accurate compartmentation based on the electrical performance classification and appearance condition. The intelligent analysis and early warning module is used to mine database data through machine learning algorithms, build defect correlation analysis models and equipment status prediction models, conduct defect root cause analysis and equipment status prediction; monitor abnormal product indicators and equipment status in real time, and trigger corresponding early warning or maintenance procedures; at the same time, it establishes a multi-workstation collaborative scheduling mechanism to achieve dynamic balance between the load and progress of each inspection workstation.

Images