Automated Detection Method and Device for Capacitor Pins

Abstract

The present application provides an automated detection method and device for capacitor pins, which relates to the technical field of capacitor detection. Samples of capacitors are collected from a batch of capacitors; the detection items are sorted according to the feasibility of fault repair to obtain a sequence of items to be detected; a detection device is scheduled to perform detection data collection to obtain a sample detection data set; pin defects are identified based on the sample detection data set to obtain sample defect information; the sample defect information is aggregated to obtain a sample detection defect distribution; the quality of the sample capacitors is evaluated based on the sample defect information to obtain a batch quality distribution; the sample detection defect distribution and the batch quality distribution are comprehensively analyzed to output a quality detection result. The present application solves the technical problems in the prior art that due to the fixed detection process and the lack of consideration of the feasibility of fault repair, the detection efficiency is low and the accuracy of batch quality processing is insufficient, improves the quality detection efficiency and enhances the accuracy of batch capacitor quality control.

Description

Technical Field

[0001] This application relates to the technical field of capacitor detection, and specifically to an automated detection method and device for capacitor pins. Background Art

[0002] As a common component in electronic devices, the quality of the pins of a capacitor has an important impact on the product performance and service life. The current capacitor pin detection methods mainly include manual detection, fixed-process automatic detection, and statistical sampling detection. Manual detection relies on the operator's observation and tool assistance. Although it has high flexibility, it is inefficient, highly subjective, and difficult to meet the requirements of large-scale detection. Fixed-process automatic detection uses automated equipment to detect each characteristic of the capacitor pins one by one. Although the detection speed is increased, the detection process is fixed, and the feasibility of defect repair is not considered, which easily leads to waste of resources. Statistical sampling detection infers the quality distribution of batches through the detection of a small number of samples and is suitable for large-scale detection. However, its defect analysis accuracy is limited, it cannot deeply analyze the defect distribution, and it is difficult to comprehensively reflect the batch quality. These existing methods cannot fully meet the requirements of modern manufacturing for intelligent and efficient detection. Summary of the Invention

[0003] This application provides an automated detection method and device for capacitor pins, which solves the technical problems of low detection efficiency and insufficient accuracy in batch capacitor quality processing in the prior art due to the fixed detection process and the lack of consideration of the feasibility of fault repair, and achieves the technical effects of improving the quality detection efficiency of capacitor pins and enhancing the accuracy of batch capacitor quality control.

[0004] In view of the above problems, on the one hand, this application provides an automated detection method for capacitor pins. The method includes: collecting detection samples from a batch of capacitors to obtain K sample capacitors, where the K sample capacitors and the batch of capacitors satisfy a preset sampling ratio; sorting the detection items according to the feasibility of fault repair to obtain a sequence of items to be detected; using the sequence of items to be detected as a constraint, scheduling a detection device to perform detection data collection on K groups of capacitor pins of the K sample capacitors to obtain K sample detection data sets; identifying pin defects based on the K sample detection data sets to obtain K sample defect information, where the sample defect information includes a sample defect label sequence and a sample defect level sequence; aggregating the K sample defect information to obtain a sample detection defect distribution; evaluating the quality of the K sample capacitors based on the K sample defect information to obtain a batch quality distribution; comprehensively analyzing the sample detection defect distribution and the batch quality distribution, and outputting the quality detection result of the batch of capacitors.

[0005] On the other hand, this application also provides an automated detection device for capacitor pins. The device includes:

[0006] A sample collection module for collecting detection samples of a batch of capacitors to obtain K sample capacitors, where the K sample capacitors and the batch of capacitors satisfy a preset sampling ratio; a detection item sorting module for sorting detection items according to the feasibility of fault repair to obtain a sequence of items to be detected; a detection data set collection module for scheduling a detection device to perform detection data set collection on K groups of capacitor pins of the K sample capacitors with the sequence of items to be detected as a constraint to obtain K sample detection data sets; a pin defect identification module for identifying pin defects based on the K sample detection data sets to obtain K sample defect information, where the sample defect information includes a sample defect label sequence and a sample defect grade sequence; a defect information aggregation module for aggregating the K sample defect information to obtain a sample detection defect distribution; a quality evaluation module for evaluating the quality of the K sample capacitors based on the K sample defect information to obtain a batch quality distribution; a detection result output module for comprehensively analyzing the sample detection defect distribution and the batch quality distribution and outputting a quality detection result of the batch of capacitors.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0008] By means of a preset sampling ratio, detection samples of a batch of capacitors are collected to obtain K sample capacitors, ensuring that the collected samples are representative and can reflect the quality status of the entire batch of capacitors, providing basic data for subsequent detection. Sort the detection items according to the feasibility of fault repair to optimize the detection process, giving priority to dealing with defects that are easy to repair and have a greater impact, ensuring the importance and practicality of the detection items. According to the optimized sequence of detection items, schedule the detection device to collect data on the pins of the sample capacitors, ensuring the accuracy and integrity of the detection data and providing data support for subsequent defect identification. Through the analysis of the detection data, identify the defects of the pins and label and grade the defects, realizing the refined management of the defects and providing detailed information for quality evaluation. Aggregate the defect information of multiple samples to form a sample detection defect distribution, providing macroscopic data for batch quality evaluation and facilitating the discovery of common problems in mass production. Based on the sample defect information, evaluate the quality of the sample capacitors to form a batch quality distribution, providing comprehensive quality information and facilitating the timely discovery and solution of quality problems. Comprehensively analyze the sample detection defect distribution and the batch quality distribution, and output the final quality detection result, providing an intuitive and comprehensive quality assessment and avoiding the one-sidedness of single-index judgment.

[0009] In summary, through the dynamic detection item ranking based on the feasibility of fault repair, the present application realizes the accurate identification and hierarchical screening of capacitor pin defects; through sample collection and defect information aggregation, it accurately evaluates the quality distribution of batch capacitors; and in combination with the intelligent decision output mechanism, it optimizes the processing method of batch capacitors, significantly improving the efficiency of capacitor pin detection and the accuracy of quality evaluation, and ensuring the rationality of the processing of batch capacitor pins.

[0010] The above description is only an overview of the technical solution of the present application. In order to be able to understand the technical means of the present application more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features and advantages of the present application more obvious and understandable, the specific embodiments of the present application are specifically exemplified below. Brief Description of the Drawings

[0011] Figure 1 It is a schematic flow chart of the automatic detection method for capacitor pins provided by an embodiment of the present application.

[0012] Figure 2 It is a schematic flow chart of obtaining the sequence of items to be detected in the automatic detection method for capacitor pins provided by an embodiment of the present application.

[0013] Figure 3 It is a schematic flow chart of obtaining the defect information of K samples in the automatic detection method for capacitor pins provided by an embodiment of the present application.

[0014] Figure 4 It is a schematic structural diagram of the automatic detection device for capacitor pins provided by an embodiment of the present application.

[0015] Description of the reference numerals: The sample collection module 10, the detection item ranking module 20, the detection data set collection module 30, the pin defect identification module 40, the defect information aggregation module 50, the quality evaluation module 60, the detection result output module 70. Detailed Embodiments

[0016] By providing an automatic detection method and device for capacitor pins, the embodiment of the present application solves the technical problems of low detection efficiency and insufficient accuracy in batch quality processing in the prior art due to the fixed detection process and the lack of consideration of the feasibility of fault repair, and achieves the technical effects of improving the efficiency of capacitor pin quality detection and enhancing the accuracy of batch capacitor quality control.

[0017] Embodiment 1, as Figure 1 shown, the embodiment of the present application provides an automatic detection method for capacitor pins, and the method includes:

[0018] Step S1: Collect detection samples from a batch of capacitors to obtain K sample capacitors, where the K sample capacitors and the batch of capacitors meet a preset sampling ratio.

[0019] Specifically, a batch of capacitors refers to a large number of capacitors that need to be detected and is the overall object of the detection task. According to a preset sampling ratio, a certain number of capacitors are randomly selected from the batch of capacitors as sample capacitors to obtain K sample capacitors. Here, K is a positive integer representing the total number of selected sample capacitors. For example, if the preset sampling ratio is 10% and there are 1000 capacitors in the batch, then 100 capacitors need to be randomly selected as sample capacitors (i.e., K = 100). Use a simple numbering system to number each capacitor, and then use a random number generator to randomly determine 100 numbers between 1 and 1000, and extract the capacitors corresponding to these numbers as sample capacitors.

[0020] Through the above sampling, while ensuring the representativeness of the samples, the detection workload can be reduced, the detection efficiency can be improved, and the waste of resources caused by detecting all capacitors can be avoided.

[0021] Step S2: Sort the detection items according to the feasibility of fault repair to obtain a sequence of items to be detected.

[0022] Specifically, the feasibility of fault repair refers to the possibility of the difficulty and cost of repairing a capacitor pin after a fault occurs. Classify the possible faults of the capacitor pins, and define the repair feasibility for each type of defect according to historical data or engineering experience. Sort the detection items corresponding to various faults according to the level of repair feasibility to determine the sequence of items to be detected. The detection device adjusts the action process according to this sequence. For example, it gives priority to detecting breaks and analyzes the surface shape or cracks of the pins through an optical microscope. Optimizing the detection process through the sorting of detection items can improve the detection efficiency.

[0023] Step S3: Using the sequence of items to be detected as a constraint, schedule the detection device to perform detection data collection on the K groups of capacitor pins of the K sample capacitors to obtain K sample detection data sets.

[0024] Specifically, according to the sequence of items to be detected obtained in Step S2, use the corresponding detection device to detect the K groups of capacitor pins of the K sample capacitors in turn, collect the corresponding detection data, and summarize to generate K sample detection data sets. Among them, each sample capacitor corresponds to a group of capacitor pins, that is, a sample detection data set. The sample detection data set contains all the detection data records generated by each sample capacitor, such as feature data of the pin shape, size, surface state, etc.

[0025] By scheduling the detection device to collect data in a specific order, the validity and accuracy of the data are ensured, providing a reliable data source for subsequent defect identification and helping to accurately judge the status of the capacitor pins.

[0026] Step S4: Perform pin defect identification based on the K sample detection data sets to obtain K sample defect information, where the sample defect information includes a sample defect label sequence and a sample defect grade sequence.

[0027] Specifically, use data analysis software or algorithms to analyze the K sample detection data sets to determine whether there are defects in the pins, as well as the type and degree of the defects. For example, for the data of the pin appearance, if the bending angle of the pin exceeds a certain threshold, it is marked with a "bent" defect label, and the defect grade is determined as mild, moderate or severe according to the bending degree. Generate K sample defect information based on the analysis and judgment results of the K sample detection data sets, and each sample defect information includes a sample defect label sequence and a sample defect grade sequence. Among them, the sample defect label sequence is used to mark the defect type of each sample capacitor pin; the sample defect grade sequence is used to mark the severity of the defects of each sample capacitor pin.

[0028] By accurately identifying the defect conditions of each sample capacitor pin, detailed basic information is provided for subsequent overall analysis, which helps to accurately evaluate the quality of the sample capacitors.

[0029] Step S5: Aggregate the K sample defect information to obtain the sample detection defect distribution.

[0030] Specifically, use statistical software to statistically analyze the data in the defect label sequence and defect grade sequence of each sample, calculate the proportion of each defect type in all samples, as well as the proportion distribution of different defect grades, etc., to obtain the sample detection defect distribution. The sample detection defect distribution describes the distribution of all sample capacitor pin defects in the overall batch of capacitors, including the proportions of different defect types and defect grades.

[0031] By aggregating the K sample defect information to obtain the sample detection defect distribution, the defect conditions of the sample capacitor pins can be grasped as a whole, and information such as the concentration trend of defects, common defect types and severity distribution can be found, providing a reference for evaluating the quality of the entire batch.

[0032] Step S6: Perform quality evaluation on the K sample capacitors according to the K sample defect information to obtain the batch quality distribution.

[0033] Specifically, according to the preset quality standards, quality evaluation is carried out by combining the defect information of each sample capacitor. For example, if the defect level of a sample capacitor is mild and does not affect the overall function, it may be judged as qualified; if the defect level is severe and difficult to repair, it is judged as scrapped; those in between are judged as to be repaired. Then, based on the quality evaluation results of the sample capacitors, the quality distribution of the entire batch of capacitors is deduced through statistical methods to obtain the batch quality distribution, including the proportions of different quality states such as qualified, to be repaired, and scrapped. For example, methods of probability theory and mathematical statistics can be used for the deduction.

[0034] Obtaining the batch quality distribution through quality evaluation can reasonably generalize the quality situation of the samples to the entire batch, provide a basis for comprehensively evaluating the quality of the entire batch of capacitors, and help make correct processing decisions.

[0035] Step S7: Comprehensively analyze the detected defect distribution of the samples and the batch quality distribution, and output the quality inspection results of the batch of capacitors.

[0036] Specifically, analyze the data in the detected defect distribution of the samples and the batch quality distribution through comparison, weighing, etc. According to the situations of these two distributions, determine the final quality inspection results. For example, whether it is qualified and what kind of processing is required (such as all rework, partial repair, etc.). By comprehensively considering various aspects of information, the one-sidedness of single-index judgment is avoided, and the quality inspection results of the batch of capacitors can be accurately output to ensure that the processing decisions for the batch of capacitors are scientific and reasonable.

[0037] Further, as Figure 2 shown, step S2 includes:

[0038] Step S21: Use the capacitor model of the batch of capacitors as the retrieval instruction to perform network data call to obtain multiple repair feasibility evaluation indicators for multiple sample defect types, where the repair feasibility evaluation indicators include repair difficulty, repair cost, repair life loss, repair performance loss, and repair time consumption.

[0039] Step S22: Synchronize the multiple repair feasibility evaluation indicators to the repair feasibility evaluation function to obtain multiple defect repair feasibility coefficients.

[0040] Step S23: Match and obtain multiple sample detection devices for the multiple sample defect types.

[0041] Step S24: Aggregate the multiple sample detection devices based on device consistency to obtain M types of sample detection devices.

[0042] Step S25: Based on the M types of sample detection devices and multiple defect repair feasibility coefficients, perform parallel detection priority ranking on the multiple sample defect types, and output the sequence of items to be detected.

[0043] Specifically, determine the model of the batch of capacitors to be detected, generate a retrieval instruction according to the capacitor model, call historical data through an online database (such as a production record database or an industry fault analysis platform), and extract the common defects of this type of capacitor and the corresponding repair feasibility evaluation indicators for each defect. These repair feasibility evaluation indicators are key parameters for measuring the possibility of defect repair, including the following: repair difficulty, that is, the complexity of the repair operation; repair cost, that is, the material cost and labor cost required for repair; repair life loss, that is, the degree of shortening of the capacitor life after repair; repair performance loss, that is, the degree of decline in the capacitor performance after repair; repair time consumption, that is, the time required to complete the repair.

[0044] The repair feasibility evaluation function is a mathematical model or algorithm used to comprehensively evaluate multiple indicators (such as difficulty, cost, time consumption) and output a repair feasibility coefficient. Exemplarily, a multi-index comprehensive evaluation model (such as the weighted average method or the fuzzy comprehensive evaluation method) can be used to construct the repair feasibility evaluation function, and convert multiple repair feasibility evaluation indicators into a defect repair feasibility coefficient. This defect repair feasibility coefficient is a numerical result used to represent the feasibility degree of a certain sample defect type repair. The higher the value, the higher the repair feasibility. Integrating multiple repair feasibility evaluation indicators into one coefficient simplifies the evaluation of defect repair feasibility, enables direct comparison of the repair feasibility between different sample defect types, and provides a unified quantitative standard for subsequent ranking.

[0045] According to different sample defect types, match the corresponding detection devices from the pre-established detection device database. This database stores the mapping relationships between various sample defect types and the corresponding detection devices. For example, the database records relationships such as "pin breakage - X-ray detection device", "capacitance value deviation - capacitance tester", etc. Accurately matching the detection devices corresponding to the sample defect types ensures the accuracy and effectiveness of subsequent detections, and provides support at the hardware device level for parallel detection priority ranking.

[0046] Analyze the multiple sample detection devices that are matched, and aggregate them according to the similarity or correlation in terms of functions, principles, detection ranges, etc. of the detection devices to obtain M types of sample detection devices. Here, M is a positive integer, representing the number of sample detection devices after aggregation. Exemplarily, a device consistency evaluation matrix can be established, where the rows and columns respectively represent different sample detection devices, and the matrix elements represent the degree of consistency between two devices (which can be a value between 0 and 1, 0 indicating complete inconsistency, and 1 indicating complete consistency). Then, through a clustering algorithm (such as the K-means clustering algorithm), the sample detection devices with a high degree of consistency are aggregated together to obtain M types of sample detection devices. For example, if the optical microscope and the electron microscope have a high degree of consistency, they can be aggregated into one type of sample detection device for detecting microscopic defects on the surface of pins. By aggregating the sample detection devices, the types of detection devices are reduced, the subsequent detection arrangements are simplified, and at the same time, it is beneficial to the optimal allocation of detection resources and the improvement of detection efficiency.

[0047] Comprehensively consider the availability, efficiency of the M types of sample detection devices, and multiple defect repair feasibility coefficients, conduct parallel detection priority ranking for multiple sample defect types, determine the order of detection for multiple sample defect types. When ranking, it is necessary to consider the situation where multiple detections can be carried out simultaneously (parallel detection), and output the determined order as the sequence of items to be detected. Determining a reasonable sequence of items to be detected can, in the case of parallel detection, preferentially detect those sample defect types with high repair feasibility and convenient detection, improve the overall detection efficiency, and reduce the waste of detection time and resources.

[0048] Further, step S25 includes:

[0049] Step S251: Aggregate the multiple defect repair feasibility coefficients according to the M types of sample detection devices to obtain M groups of defect repair feasibility coefficients.

[0050] Step S252: Obtain M comprehensive repair feasibility coefficients by summing up the M groups of defect repair feasibility coefficients within the group.

[0051] Step S253: Serialize the M types of sample detection devices according to the M comprehensive repair feasibility coefficients to obtain the detection device scheduling priority.

[0052] Step S254: Conduct parallel detection priority ranking for the multiple sample defect types according to the M groups of defect repair feasibility coefficients to obtain M parallel detection priorities.

[0053] Step S255: Associatively store the detection device scheduling priority and the M parallel detection priorities to generate the sequence of items to be detected.

[0054] Specifically, first establish a mapping relationship to associate each sample detection device with the defect repair feasibility coefficient of the corresponding sample defect type. For example, if there are 3 sample detection devices (M = 3), namely a microscope for detecting the appearance of pins, a multimeter for detecting electrical performance, and an X-ray machine for detecting internal structures. For each detection device, find the defect repair feasibility coefficient of the sample defect type associated with it. Suppose the sample defect types corresponding to the microscope are pin bending and pin surface scratches, then group the defect repair feasibility coefficients of these two sample defect types together, and so on. Eventually, M groups of defect repair feasibility coefficients are obtained. Through this aggregation method, a connection is established between the defect repair feasibility coefficient and the sample detection device, providing a grouped data basis for subsequent prioritization by comprehensively considering the detection device and repair feasibility.

[0055] For the obtained M groups of defect repair feasibility coefficients, perform the in-group summation operation respectively to obtain M comprehensive repair feasibility coefficients corresponding to the M sample detection devices. Among them, each sample detection device corresponds to a comprehensive repair feasibility coefficient, which represents the overall degree of the repair feasibility of the sample defect types related to this sample detection device. For example, for a group of defect repair feasibility coefficients (a1, a2, a3) corresponding to a certain sample detection device, the comprehensive repair feasibility coefficient A = a1 + a2 + a3. Calculating the comprehensive repair feasibility coefficient simplifies the representation of the repair feasibility related to each sample detection device, facilitating comparison and prioritization among different sample detection devices.

[0056] Sort the M sample detection devices according to the magnitudes of the M comprehensive repair feasibility coefficients. For example, if a larger comprehensive repair feasibility coefficient indicates a higher priority for detection, then arrange the M sample detection devices in descending (or ascending, according to the specific priority definition) order of the comprehensive repair feasibility coefficient to obtain the detection device scheduling priority for arranging the scheduling order of the detection devices. A sorting algorithm (such as the bubble sort algorithm) can be used to implement this sorting process.

[0057] Based on the M groups of defect repair feasibility coefficients, prioritize the sample defect types within each group. For example, for the sample defect types corresponding to a certain group of defect repair feasibility coefficients, determine their order in parallel detection according to the coefficient magnitudes. A sorting method similar to step S253 can be adopted to sort the sample defect types according to specific priority rules (such as the larger the coefficient, the earlier the detection) to obtain M parallel detection priorities. According to these parallel detection priorities, when performing parallel detection, it is possible to proceed in a reasonable order, giving priority to detecting those sample defect types with higher repair feasibility and improving the overall detection efficiency.

[0058] The associated storage detection device schedules priorities and M parallel detection priorities to generate a sequence of items to be detected. A data structure (such as a linked list or a dictionary) can be used to implement the associated storage. For example, in a dictionary, the detection device scheduling priority is used as the key, and the corresponding M parallel detection priorities are used as the values. In this way, the two can be associated to form a complete sequence of items to be detected.

[0059] By integrating the order information of the detection device scheduling and the parallel detection of the sample defect types, a sequence of items to be detected is generated, providing a clear operation guide for the subsequent detection work, ensuring that the detection process proceeds in a reasonable order, and improving the accuracy and efficiency of the detection.

[0060] Further, step S3 includes:

[0061] Step S31: Guided by the detection device scheduling priority in the sequence of items to be detected, schedule the first detection device entity of the first sample detection device, where the first detection device entity is an image acquisition device.

[0062] Step S32: Analyze the first parallel detection priority to obtain a sequence of image acquisition requirements, where the first parallel detection priority includes H sample defect types.

[0063] Step S33: Constrained by the sequence of image acquisition requirements, run the first detection device entity to perform automated acquisition of detection data on the first set of capacitor pins of the first sample capacitor, and obtain a first acquisition detection data set, where the first acquisition detection data set includes a first detection image sequence corresponding to the sequence of image acquisition requirements.

[0064] Step S34: And so on, analyze the M - 1 parallel detection priorities, and according to the analysis results, run the M - 1 detection device entities of the M - 1 sample detection devices to perform automated acquisition of detection data on the first set of capacitor pins, and obtain M - 1 acquisition detection data sets, where the first acquisition detection data set and the M - 1 acquisition detection data sets constitute the first sample detection data set.

[0065] Step S35: And so on, perform acquisition of detection data sets on the K sets of capacitor pins to obtain K sample detection data sets.

[0066] Specifically, obtain the detection device scheduling priority from the sequence of items to be detected, and determine the first detection device entity of the first sample detection device to be scheduled according to this priority. In actual operation, the scheduling process can be implemented through an automated control system. For example, in a detection equipment management system, the rules for the detection device scheduling priority are preset in advance. When the detection starts, the system automatically selects and activates the first detection device entity according to this rule, that is, the image acquisition device, such as an industrial camera or other devices used to collect images of capacitor pins.

[0067] The first parallel detection priority is the parallel detection priority corresponding to the sample defect type of the first sample detection device, which includes the priority order of H sample defect types. Among them, H is a positive integer less than or equal to M. Analyze the first parallel detection priority, extract the information related to image acquisition from it, and arrange this information in the priority order to form an image acquisition requirement sequence. This image acquisition requirement sequence is the order of image acquisition requirements determined according to the first parallel detection priority, which clarifies the image acquisition order for different sample defect types. For example, if the first parallel detection priority indicates that the appearance bending of the pins should be detected first, and then the scratches on the surface of the pins, then in the image acquisition requirement sequence, the image for detecting the bending of the pins will be arranged first, and then the image for detecting the scratches on the pins.

[0068] Randomly select a sample capacitor for detection from K sample capacitors, denoted as the first sample capacitor, and the first sample capacitor has the first group of capacitor pins. Take the image acquisition requirement sequence as a constraint condition and input it into the first detection device entity (image acquisition device). The image acquisition device acquires images of the first group of capacitor pins of the first sample capacitor according to this requirement sequence. For example, in the control program of an industrial camera, according to the instructions in the image acquisition requirement sequence, adjust parameters such as the shooting angle and focal length of the camera, and sequentially acquire images under different requirements to form a first detection image sequence, and these images together constitute the first acquisition detection data set. Among them, each image in the first detection image sequence is applied to a defect detection item.

[0069] Similarly, analyze the remaining M - 1 parallel detection priorities, and according to the analysis results, run the M - 1 detection device entities of the corresponding M - 1 sample detection devices to perform automated acquisition of detection data for the first group of capacitor pins, and obtain M - 1 acquisition detection data sets. Summarize the first acquisition detection data set and the M - 1 acquisition detection data sets to form the first sample detection data set. For example, after scheduling the first sample detection device to collect data for the pins of the first sample capacitor, schedule the remaining sample detection devices one by one according to the detection device scheduling priority to collect detection data for the pins of the first sample capacitor.

[0070] In the same way as collecting the detection data set of the first sample described above, the pins of each sample capacitor are sequentially detected to collect the data set, and the corresponding K sample detection data sets are obtained. For example, after the first sample detection device finishes collecting image data of the pins of the remaining sample capacitors, the next sample detection device is scheduled according to the detection device scheduling priority to collect detection data of the pins of each sample capacitor, and this process is repeated until the detection data sets of the K groups of capacitor pins of the K sample capacitors are all collected, and finally K sample detection data sets are obtained.

[0071] Through the above steps, the detection data of the pins of the K sample capacitors are collected in an orderly and comprehensive manner, providing a complete data basis for subsequent operations such as pin defect identification and quality evaluation, and ensuring the accuracy and effectiveness of the entire detection process.

[0072] Further, as Figure 3 shown, step S4 includes:

[0073] Step S41: Construct H pin defect recognition networks according to the H sample defect types, where the input data of the pin defect recognition network is the sample detection image, and the output data is the sample defect label and the sample defect level.

[0074] Step S42: Set the model scheduling priority of the H pin defect recognition networks with the first parallel detection priority as the constraint.

[0075] Step S43: After synchronizing the first detection image sequence to the H pin defect recognition networks, start the H pin defect recognition networks one by one for pin appearance defect recognition with the model scheduling priority as the constraint, and output the first sample defect label sequence and the first sample defect level sequence.

[0076] Step S44: Associatively store the first sample defect label sequence and the first sample defect level sequence to obtain the first sample defect information.

[0077] Step S45: And so on, perform pin defect recognition according to the K sample detection data sets to obtain the K sample defect information.

[0078] Specifically, a pin defect recognition network is constructed for each type of sample defect. During the construction process, deep learning frameworks such as TensorFlow or PyTorch can be used. First, determine the network structure, for example, adopting a combination of multiple convolutional layers, pooling layers, and fully connected layers. Then, use a large number of labeled sample detection images (including known sample defect labels and sample defect grades) to train the network. These labeled data can be obtained through manual labeling or from an existing capacitor pin defect database. By constructing pin defect recognition networks for different types of sample defects, the defect types and grades of the pins can be identified more accurately, improving the accuracy and specificity of defect recognition.

[0079] According to the order of the H types of sample defects in the first parallel detection priority, set the model scheduling priorities for the corresponding H pin defect recognition networks to reasonably arrange the startup order of each network. For example, if the detection order of a certain type of sample defect is ranked ahead in the first parallel detection priority, then the model scheduling priority of the corresponding pin defect recognition network is higher. These priority relationships can be stored and managed through a priority list or a configuration file. Determine the running order of the pin defect recognition networks so that when performing pin apparent defect recognition, the networks can be called in a reasonable order, improving the recognition efficiency and matching the previous detection order.

[0080] Input the first detection image sequence into the H pin defect recognition networks and start these networks one by one according to the model scheduling priority. During the operation of the network, the network analyzes and calculates based on the input image data and outputs the corresponding sample defect labels and sample defect grades, that is, the first sample defect label sequence and the first sample defect grade sequence. For example, in a network constructed using the TensorFlow framework, the session mechanism can be used to start the network and obtain the output results.

[0081] Use a data structure (such as a dictionary or a data table) to associate and store the first sample defect label sequence and the first sample defect grade sequence. For example, a dictionary can be created with the sample defect label as the key and the corresponding sample defect grade as the value to obtain the first sample defect information.

[0082] Perform pin defect recognition on the K sample detection data sets respectively in a similar way to obtaining the first sample defect information. For each sample detection data set, construct the corresponding pin defect recognition network, set the model scheduling priority, perform pin apparent defect recognition, associate and store the defect label and grade sequences, and finally obtain K sample defect information.

[0083] Through the construction of a defect recognition network, priority scheduling, and batch detection, the above steps achieve the automated recognition and storage of sample pin defects. The construction of a dedicated network enables the precise recognition of different defect types; through priority scheduling, computing resources are reasonably allocated, the detection efficiency is improved, and the detection results are uniformly stored, providing complete data support for subsequent analysis.

[0084] Furthermore, the repair feasibility evaluation function is as follows:

[0085] ; where R is the defect repair feasibility coefficient, D is the repair difficulty, C is the repair cost, L is the repair life loss, P is the repair performance loss, and T is the repair time consumption.

[0086] Specifically, the repair feasibility function consists of and two terms. For this term, is the product of the repair difficulty and the repair cost, reflecting the repair cost from the comprehensive perspectives of difficulty and cost. The larger the value of this term, the more resources and technical support are required for the repair, so the repair may be more challenging and costly.

[0087] In the term, the numerator is the product of the repair life loss and the repair performance loss, considering the impact of the repair on the capacitor life and performance; the denominator makes the impact of the repair time consumption on the result non-linear. As

[0088] increases, this part of the value will decrease rapidly. Then, the square root is taken after multiplying these two parts to obtain an intermediate result that comprehensively considers various factors. The longer the repair time consumption, the greater the impact of the loss of life and performance after the repair on the repair feasibility. A longer repair time may result in poorer performance and shorter life of the repaired device.

[0089] This repair feasibility evaluation function quantifies multiple important factors in the repair process, including repair difficulty, cost, impact on life and performance, time consumption, and the variation relationship between repair cost and repair difficulty. The defect repair feasibility coefficient R calculated through this function can quantitatively evaluate the repair feasibility of different sample defect types, providing a scientific and reasonable basis for subsequent operations such as determining the priority of detection items.

[0090] Furthermore, step S5 further includes:

[0091] Step S51: Using the capacitor model of the batch capacitors as a retrieval instruction, perform network data calls to obtain historical quality inspection data, where the historical quality inspection data includes multiple historical defect detection sequences of multiple historical defective pins.

[0092] Step S52: Through defect correlation identification of the multiple historical defect detection sequences, obtain a defect correlation relationship graph.

[0093] Step S53: Generate defect detection normalization rules based on the defect correlation relationship graph.

[0094] Step S54: Use the defect detection normalization rules to traverse the K sample defect information, perform redundant defect detection result elimination, and obtain K updated defect information.

[0095] Step S55: According to the multiple sample defect types, aggregate the K updated defect information to obtain the sample detection defect distribution, where the sample detection defect distribution includes multiple groups of sample defect level percentages of the multiple sample defect types.

[0096] Specifically, generate a retrieval instruction according to the capacitor model, and call the historical quality inspection data related to this type of capacitor through a network database (such as a production record database or an industry fault analysis platform). These historical quality inspection data include multiple historical defect detection sequences of multiple historical defective pins, that is, the detection result sequences recorded for multiple historical defective pins, such as information on detected defect types, levels, etc. By obtaining historical quality inspection data, it provides more reference basis for subsequent defect correlation analysis and quality assessment, and can use historical experience to optimize the current detection results.

[0097] Perform defect correlation identification on the defect types in multiple historical defect detection sequences, find the mutual relationships between these defect types, and then construct a defect correlation relationship graph based on the analyzed mutual relationships. This graph is a data structure that graphically represents the association relationships between defect types, where nodes represent defect types and edges represent the association relationships between defect types. Exemplarily, data mining and graph algorithms can be used to achieve defect correlation identification. First, extract and count the defect types in multiple historical defect detection sequences. Then, analyze information such as the frequency of co-occurrence of different defect types. For example, if defect type A and defect type B often co-occur in the same historical defect detection sequence, then it can be considered that there is an association relationship between them. Use a graph construction tool (such as the NetworkX library in Python) to construct a defect correlation relationship graph, with defect types as nodes and association relationships as edges. The defect correlation relationship graph intuitively shows the relationships between different defect types, helps to deeply understand the internal connections of capacitor pin defects, and provides a basis for formulating defect detection normalization rules.

[0098] Generate defect detection normalization rules based on the relationships between nodes and edges in the defect correlation relationship graph. This normalization rule is used to standardize and unify defect detection results. For example, if it is found in the graph that defect type A and defect type B have a strong association relationship and can be regarded as different manifestations of the same defect under certain conditions, then in the defect detection normalization rule, it can be stipulated that when A and B are detected simultaneously, they are merged or processed in a unified manner according to a certain method. The defect detection normalization rule can improve the accuracy and consistency of defect detection results and avoid result differences caused by different detection angles or standards.

[0099] Traverse each sample defect information in the K sample defect information. For each defect type and level among them, make a judgment according to the defect detection normalization rule. If it is found that there are situations that need to be merged or eliminated, corresponding operations are performed. For example, if according to the rule, two defect types of a certain sample are regarded as the same defect, then they are merged into one defect, and relevant information such as the sample defect level is updated. For example, if bending and abnormal electrical parameters are detected simultaneously, retain the sample defect level corresponding to bending. By eliminating redundant defect detection results, more accurate and refined information about the capacitor pin defects of K samples, that is, K updated defect information, can be obtained, improving the accuracy of sample defect information, reducing the interference caused by redundant information, and making the subsequent defect distribution analysis more reliable.

[0100] For K updated defect information, count the occurrence times of each sample defect type in different samples and the corresponding defect level conditions. Then calculate the defect level percentage of each sample defect type in all samples to generate the sample detection defect distribution. For example, for the sample defect type A, it is counted that there are n samples with A defect among K samples, and classified and counted according to different defect levels, and the percentage of each level is calculated. Data statistics and analysis tools (such as the Pandas library in Python) can be used to implement these calculations. The finally obtained sample detection defect distribution can comprehensively reflect the distribution and severity of various sample defect types in the samples, providing an important basis for the quality assessment of capacitors, production process improvement, etc.

[0101] Further, step S7 includes:

[0102] Step S71: Configure defect interference weights for the multiple sample defect types to obtain multiple sample defect interference weights.

[0103] Step S72: Configure defect level weights for the multiple sample defect types to obtain multiple groups of sample defect level weights.

[0104] Step S73: Use the multiple sample defect interference weights and multiple groups of sample defect level weights to perform weighted processing on the multiple groups of sample defect level percentages to obtain a defect comprehensive interference coefficient.

[0105] Step S74: Extract the sampling qualified percentage from the batch quality distribution, where the batch quality distribution also includes the sampling scrap percentage and the sampling repair percentage.

[0106] Step S75: Preset a sampling qualified threshold and a defect comprehensive interference threshold.

[0107] Step S76: When and only when the defect comprehensive interference coefficient and the sampling qualified percentage respectively meet the sampling qualified threshold and the defect comprehensive interference threshold, the quality detection result is qualified.

[0108] Specifically, configure defect interference weights for multiple sample defect types to obtain multiple sample defect interference weights. These sample defect interference weights are used to measure the interference degree of each sample defect type on the overall performance or quality of the capacitor. The interference weights of each sample defect type can be determined according to empirical data, industry standards or through experiments. For example, if the pin break has a very large impact on the capacitor performance, it may be given a higher interference weight, such as 0.8; while the slight scratch on the pin surface has a relatively small impact on the performance, it may be given a lower interference weight, such as 0.2. The sample defect interference weights can more accurately reflect the impact degree of different defect types on the capacitor quality, providing important parameters for the subsequent comprehensive evaluation of the capacitor quality.

[0109] Similarly, defect level weights are configured for multiple sample defect types to obtain multiple groups of sample defect level weights, which measure the impact degree of the same sample defect type at different levels on the overall performance or quality of the capacitor. For example, for the defect type of pin bending, slight bending (level 1) may be assigned a relatively low level weight, such as 0.3; moderate bending (level 2) may be assigned 0.6; severe bending (level 3) may be assigned 0.9. The sample defect level weights help to refine the impact assessment of different levels of defects, making the quality assessment more detailed and accurate.

[0110] Determine the actual proportion of each sample defect type in the sample according to multiple groups of sample defect level percentages. Then, for each sample defect type, multiply its sample defect interference weight by the corresponding sample defect level weight, and then multiply by the sample defect level percentage of this defect type. Finally, add up the calculation results of all sample defect types to obtain the comprehensive defect interference coefficient. For example, there are two sample defect types A and B. The sample defect level percentage of A is p A , the interference weight is w A , and the level weight is w A−level ; the sample defect level percentage of B is p B , the interference weight is w B , and the level weight is w B−level . Then the comprehensive defect interference coefficient C = w A ×w A−level ×p A +w B ×w B−level ×p B . The comprehensive defect interference coefficient can comprehensively evaluate the impact of defects on the quality of the capacitor considering multiple factors, providing a comprehensive quantitative index for judging whether the quality of the capacitor is qualified.

[0111] The batch quality distribution includes information such as the sampling pass percentage, sampling scrap percentage, and sampling repair percentage. Extract the sampling pass percentage from the batch quality distribution data.

[0112] Two thresholds, namely the sampling qualification threshold and the comprehensive defect interference threshold, are preset according to product quality requirements or industry standards. Among them, the sampling qualification threshold is used to judge whether the sampling qualification percentage is qualified. The comprehensive defect interference threshold is used to judge whether the comprehensive defect interference coefficient is within the acceptable range. For example, the sampling qualification threshold can be set to 90%, indicating that the proportion of qualified products in the sampled products should reach at least 90%; the comprehensive defect interference threshold can be set to 0.3, indicating that the comprehensive defect interference coefficient should not exceed 0.3. These thresholds can be stored in the configuration file or system parameters, providing a clear standard for judging whether the quality inspection result of the capacitor is qualified, making the quality assessment objective and operable.

[0113] Compare the comprehensive defect interference coefficient with the comprehensive defect interference threshold, and at the same time compare the sampling qualification percentage with the sampling qualification threshold. If the comprehensive defect interference coefficient is less than or equal to the comprehensive defect interference threshold, and the sampling qualification percentage is greater than or equal to the sampling qualification threshold, the quality inspection result is qualified; otherwise, it is unqualified. When the quality inspection result is unqualified, the batch of capacitors is collectively scrapped.

[0114] The above steps comprehensively analyze the defect interference weight, grade weight and batch quality distribution, and use the double-threshold rule to quantitatively evaluate the quality inspection result of the batch of capacitors. Considering the influence of defect types and grades, a more scientific quality assessment is achieved, providing a data basis and operation guidance for product quality control and further optimization.

[0115] In summary, the automatic detection method for capacitor pins provided by the embodiments of the present application has the following technical effects:

[0116] The embodiments of the present application realize the efficient and accurate detection of the pins of a batch of capacitors through an automatic detection device and a refined defect recognition and quality evaluation method. First, representative samples are collected by presetting the sampling ratio to ensure the representativeness of the detection data. Secondly, the detection items are sorted according to the feasibility of fault repair to optimize the detection process and improve the detection efficiency. Then, the detection device is scheduled to collect data on the pins of the sample capacitors to ensure the accuracy and integrity of the detection data. Next, through the analysis of the detection data, the defects of the pins are identified, and the defects are labeled and graded to realize the refined management of the defects. Finally, by comprehensively analyzing the defect distribution of the sample detection and the batch quality distribution, the final quality inspection result is output, providing comprehensive and intuitive quality information to facilitate the timely discovery and solution of quality problems. Generally speaking, the embodiments of the present application not only improve the efficiency of capacitor pin quality inspection, but also improve the accuracy of quality inspection, thereby enhancing the accuracy of batch capacitor quality control and meeting the requirements of precise quality control.

[0117] Embodiment 2, as Figure 4As shown in the figure, based on the same inventive concept as in the foregoing Embodiment 1, the embodiment of the present application provides an automatic detection device for capacitor pins, and the device includes:

[0118] A sample collection module 10, configured to collect detection samples for a batch of capacitors to obtain K sample capacitors, where the K sample capacitors and the batch of capacitors satisfy a preset sampling ratio.

[0119] A detection item sorting module 20, configured to sort detection items according to the feasibility of fault repair to obtain a sequence of items to be detected.

[0120] A detection data set collection module 30, configured to use the sequence of items to be detected as a constraint to schedule a detection device to perform detection data set collection on K groups of capacitor pins of the K sample capacitors to obtain K sample detection data sets.

[0121] A pin defect identification module 40, configured to identify pin defects according to the K sample detection data sets to obtain K sample defect information, where the sample defect information includes a sample defect label sequence and a sample defect level sequence.

[0122] A defect information aggregation module 50, configured to aggregate the K sample defect information to obtain a sample detection defect distribution.

[0123] A quality evaluation module 60, configured to evaluate the quality of the K sample capacitors according to the K sample defect information to obtain a batch quality distribution.

[0124] A detection result output module 70, configured to comprehensively analyze the sample detection defect distribution and the batch quality distribution and output the quality detection result of the batch of capacitors.

[0125] Further, the detection item sorting module 20 of the embodiment of the present application is further configured to perform the following steps:

[0126] Use the capacitor model of the batch of capacitors as a retrieval instruction to perform online data call to obtain multiple repair feasibility evaluation indicators for multiple sample defect types, where the repair feasibility evaluation indicators include repair difficulty, repair cost, repair life loss, repair performance loss, and repair time; synchronize the multiple repair feasibility evaluation indicators to a repair feasibility evaluation function to obtain multiple defect repair feasibility coefficients; match and obtain multiple sample detection devices for the multiple sample defect types; aggregate the multiple sample detection devices based on device consistency to obtain M sample detection devices; and perform parallel detection priority sorting on the multiple sample defect types according to the M sample detection devices and the multiple defect repair feasibility coefficients, and output the sequence of items to be detected.

[0127] Further, the detection item sorting module 20 of the embodiment of the present application is further configured to perform the following steps:

[0128] Aggregate the multiple defect repair feasibility coefficients according to the M sample detection devices to obtain M groups of defect repair feasibility coefficients; obtain M comprehensive repair feasibility coefficients by adding the M groups of defect repair feasibility coefficients within the group; serialize the M sample detection devices according to the M comprehensive repair feasibility coefficients to obtain the detection device scheduling priority; perform parallel detection priority sorting on the multiple sample defect types according to the M groups of defect repair feasibility coefficients to obtain M parallel detection priorities; associate and store the detection device scheduling priority and the M parallel detection priorities to generate the sequence of items to be detected.

[0129] Further, the detection data set collection module 30 of the embodiment of the present application is further configured to perform the following steps:

[0130] Guided by the detection device scheduling priority in the sequence of items to be detected, schedule the first detection device entity of the first sample detection device, where the first detection device entity is an image acquisition device; parse the first parallel detection priority to obtain an image acquisition requirement sequence, where the first parallel detection priority includes H sample defect types; with the image acquisition requirement sequence as a constraint, run the first detection device entity to perform automatic acquisition of detection data on the first group of capacitor pins of the first sample capacitor to obtain a first acquisition detection data set, where the first acquisition detection data set includes a first detection image sequence corresponding to the image acquisition requirement sequence; and so on, parse the M - 1 parallel detection priorities, and according to the parsing results, run the M - 1 detection device entities of the M - 1 sample detection devices to perform automatic acquisition of detection data on the first group of capacitor pins to obtain M - 1 acquisition detection data sets, where the first acquisition detection data set and the M - 1 acquisition detection data sets constitute the first sample detection data set; and so on, perform detection data set collection on the K groups of capacitor pins to obtain K sample detection data sets.

[0131] Further, the pin defect recognition module 40 of the embodiment of the present application is further configured to perform the following steps:

[0132] Construct H pin defect recognition networks according to the H types of sample defect types. Among them, the input data of the pin defect recognition network is the sample detection image, and the output data is the sample defect label and the sample defect level; Set the model scheduling priority of the H pin defect recognition networks with the first parallel detection priority as the constraint; After synchronizing the first detection image sequence to the H pin defect recognition networks, start the H pin defect recognition networks one by one for pin appearance defect recognition with the model scheduling priority as the constraint, and output the first sample defect label sequence and the first sample defect level sequence; Associatively store the first sample defect label sequence and the first sample defect level sequence to obtain the first sample defect information; By analogy, perform pin defect recognition according to the K sample detection data sets to obtain the K sample defect information.

[0133] Further, the repair feasibility evaluation function is as follows:

[0134] ; where R is the defect repair feasibility coefficient, D is the repair difficulty, C is the repair cost, L is the repair life loss, P is the repair performance loss, and T is the repair time.

[0135] Further, the defect information aggregation module 50 in the embodiment of the present application is further configured to perform the following steps:

[0136] Use the capacitor model of the batch capacitor as the retrieval instruction to call network data and obtain historical quality detection data, where the historical quality detection data includes multiple historical defect detection sequences of multiple historical defective pins; Obtain the defect association relationship map by performing defect association recognition on the multiple historical defect detection sequences; Generate defect detection normalization rules based on the defect association relationship map; Use the defect detection normalization rules to traverse the K sample defect information to eliminate redundant defect detection results and obtain K updated defect information; Aggregate the K updated defect information according to the multiple sample defect types to obtain the sample detection defect distribution, where the sample detection defect distribution includes multiple groups of sample defect level percentages of the multiple sample defect types.

[0137] Further, the detection result output module 70 in the embodiment of the present application is further configured to perform the following steps:

[0138] Configure defect interference weights for the multiple sample defect types to obtain multiple sample defect interference weights; configure defect level weights for the multiple sample defect types to obtain multiple groups of sample defect level weights; use the multiple sample defect interference weights and multiple groups of sample defect level weights to perform weighted processing on the multiple groups of sample defect level percentages to obtain a comprehensive defect interference coefficient; extract the sampling pass percentage from the batch quality distribution, where the batch quality distribution further includes a sampling scrap percentage and a sampling repair percentage; preset a sampling pass threshold and a comprehensive defect interference threshold; and the quality inspection result is qualified if and only if the comprehensive defect interference coefficient and the sampling pass percentage respectively meet the sampling pass threshold and the comprehensive defect interference threshold.

[0139] Through the foregoing detailed description of the automated detection method for capacitor pins in this specification, those skilled in the art can clearly know the automated detection device for capacitor pins in this embodiment. For the device disclosed in Embodiment 2, since it corresponds to the method disclosed in Embodiment 1, it has corresponding functional modules and beneficial effects. For the relevant parts, refer to the description in the method section.

[0140] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to these embodiments shown herein, but will be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An automated detection method for capacitor pins, characterized in that, The method includes: Collecting detection samples from a batch of capacitors to obtain K sample capacitors, where the K sample capacitors and the batch of capacitors satisfy a preset sampling ratio; Sorting detection items according to the feasibility of fault repair to obtain a sequence of items to be detected; Constrained by the sequence of items to be detected, scheduling a detection device to perform detection data collection on K groups of capacitor pins of the K sample capacitors to obtain K sample detection data sets; Identifying pin defects based on the K sample detection data sets to obtain K sample defect information, where the sample defect information includes a sample defect label sequence and a sample defect level sequence; Aggregating the K sample defect information to obtain a sample detection defect distribution; Evaluating the quality of the K sample capacitors based on the K sample defect information to obtain a batch quality distribution; Comprehensively analyzing the sample detection defect distribution and the batch quality distribution, and outputting the quality detection result of the batch of capacitors; The sorting the detection items according to the feasibility of fault repair to obtain a sequence of items to be detected includes: Using the capacitor model of the batch of capacitors as a retrieval instruction to perform network data call to obtain multiple repair feasibility evaluation indicators for multiple sample defect types, where the repair feasibility evaluation indicators include repair difficulty, repair cost, repair life loss, repair performance loss, and repair time consumption; Synchronizing the multiple repair feasibility evaluation indicators to a repair feasibility evaluation function to obtain multiple defect repair feasibility coefficients; Matching to obtain multiple sample detection devices for the multiple sample defect types; Aggregating the multiple sample detection devices based on device consistency to obtain M sample detection devices; According to the M sample detection devices and the multiple defect repair feasibility coefficients, performing parallel detection priority sorting on the multiple sample defect types, and outputting the sequence of items to be detected; The repair feasibility evaluation function is as follows: ； Where R is the defect repair feasibility coefficient, D is the repair difficulty, C is the repair cost, L is the repair life loss, P is the repair performance loss, and T is the repair time consumption.

2. The automated detection method for capacitor pins according to claim 1, wherein, According to the M sample detection devices and the multiple defect repair feasibility coefficients, performing parallel detection priority sorting on the multiple sample defect types, and outputting the sequence of items to be detected, the method includes: Aggregating the multiple defect repair feasibility coefficients according to the M sample detection devices to obtain M groups of defect repair feasibility coefficients; Obtaining M comprehensive repair feasibility coefficients by summing the M groups of defect repair feasibility coefficients within the group; Serializing the M sample detection devices according to the M comprehensive repair feasibility coefficients to obtain a detection device scheduling priority; Performing parallel detection priority sorting on the multiple sample defect types according to the M groups of defect repair feasibility coefficients to obtain M parallel detection priorities; Associatively storing the detection device scheduling priority and the M parallel detection priorities to generate the sequence of items to be detected.

3. The automated detection method for capacitor pins according to claim 2, wherein, Constrained by the sequence of items to be detected, schedule the detection device to perform detection data acquisition on the K groups of capacitor pins of the K sample capacitors, and obtain K sample detection data sets. The method includes: Guided by the scheduling priority of the detection device in the sequence of items to be detected, schedule the first detection device entity of the first sample detection device, where the first detection device entity is an image acquisition device; Parse the first parallel detection priority to obtain an image acquisition requirement sequence, where the first parallel detection priority includes H types of sample defect types; Constrained by the image acquisition requirement sequence, run the first detection device entity to perform automated acquisition of detection data on the first group of capacitor pins of the first sample capacitor, and obtain a first acquisition detection data set, where the first acquisition detection data set includes a first detection image sequence corresponding to the image acquisition requirement sequence; And so on, parse the M-1 parallel detection priorities, and according to the parsing results, run the M-1 detection device entities of the M-1 sample detection devices to perform automated acquisition of detection data on the first group of capacitor pins, and obtain M-1 acquisition detection data sets, where the first acquisition detection data set and the M-1 acquisition detection data sets constitute the first sample detection data set; And so on, perform detection data acquisition on the K groups of capacitor pins to obtain K sample detection data sets.

4. The automated detection method for capacitor pins according to claim 3, wherein Perform pin defect identification based on the K sample detection data sets to obtain K sample defect information, where the sample defect information includes a sample defect label sequence and a sample defect grade sequence. The method includes: Construct H pin defect identification networks according to the H types of sample defect types, where the input data of the pin defect identification network is the sample detection image, and the output data is the sample defect label and the sample defect grade; Constrained by the first parallel detection priority, set the model scheduling priority of the H pin defect identification networks; After synchronizing the first detection image sequence to the H pin defect identification networks, constrained by the model scheduling priority, start the H pin defect identification networks one by one to perform pin apparent defect identification, and output a first sample defect label sequence and a first sample defect grade sequence; Associate and store the first sample defect label sequence and the first sample defect grade sequence to obtain first sample defect information; And so on, perform pin defect identification based on the K sample detection data sets to obtain the K sample defect information.

5. The automated detection method for capacitor pins according to claim 4, wherein The method further includes: Using the capacitor model of the batch of capacitors as a retrieval instruction, perform network data call to obtain historical quality detection data, where the historical quality detection data includes multiple historical defect detection sequences of multiple historical defective pins; Obtain a defect association relationship graph by performing defect association identification on the multiple historical defect detection sequences; Generate a defect detection normalization rule based on the defect association relationship graph; Use the defect detection normalization rule to traverse the K sample defect information and eliminate redundant defect detection results to obtain K updated defect information; Aggregate the K updated defect information according to the multiple sample defect types to obtain the sample detection defect distribution, where the sample detection defect distribution includes multiple groups of sample defect level percentages of the multiple sample defect types.

6. The automated detection method for capacitor pins according to claim 5, characterized in that, Comprehensively analyze the sample detection defect distribution and the batch quality distribution, and output the quality detection result of the batch of capacitors. The method includes: Configure defect interference weights for the multiple sample defect types to obtain multiple sample defect interference weights; Configure defect level weights for the multiple sample defect types to obtain multiple groups of sample defect level weights; Use the multiple sample defect interference weights and the multiple groups of sample defect level weights to perform weighted processing on the multiple groups of sample defect level percentages to obtain a defect comprehensive interference coefficient; Extract the sampling qualified percentage from the batch quality distribution, where the batch quality distribution further includes a sampling scrap percentage and a sampling repair percentage; Preset a sampling qualified threshold and a defect comprehensive interference threshold; When and only when the defect comprehensive interference coefficient and the sampling qualified percentage respectively meet the sampling qualified threshold and the defect comprehensive interference threshold, the quality detection result is qualified.

7. An automated detection device for capacitor pins, characterized in that, The device is used to execute the automatic detection method for capacitor pins according to any one of claims 1-6, and includes: A sample collection module, configured to collect detection samples for a batch of capacitors to obtain K sample capacitors, where the K sample capacitors and the batch of capacitors meet a preset sampling ratio; A detection item sorting module, configured to sort detection items according to the feasibility of fault repair to obtain a sequence of items to be detected; A detection data set collection module, configured to use the sequence of items to be detected as a constraint, schedule a detection device to perform detection data set collection on K groups of capacitor pins of the K sample capacitors to obtain K sample detection data sets; A pin defect identification module, configured to identify pin defects according to the K sample detection data sets to obtain K sample defect information, where the sample defect information includes a sample defect label sequence and a sample defect level sequence; A defect information aggregation module, configured to aggregate the K sample defect information to obtain a sample detection defect distribution; A quality evaluation module, configured to perform quality evaluation on the K sample capacitors according to the K sample defect information to obtain a batch quality distribution; A detection result output module, configured to comprehensively analyze the sample detection defect distribution and the batch quality distribution, and output the quality detection result of the batch of capacitors.

Images