An automated inductive device detection and classification system and method

Abstract

The application discloses an automatic detection and classification system for inductance devices, which comprises an automatic feeding module, a conveying track, an appearance detection module, an LCR parameter measurement module, a saturation current scanning module, a temperature rise test module, an automatic sorting module and a fusion grading module, and forms a multi-station flow line structure. The system sequentially completes appearance detection, LCR parameter measurement, saturation current scanning and temperature rise test, the fusion grading module performs hard constraint screening through a first-level rule engine, a second-level neural network determination unit performs grade division, and the automatic sorting module sorts devices according to the determination result, so that the automatic detection and classification of inductance devices are realized.

Description

Technical Field

[0001] This application belongs to the field of industrial product quality inspection technology, specifically relating to an automated inspection and classification system and method for inductor devices. Background Technology

[0002] As a core passive component in electronic circuits that performs functions such as energy storage, filtering, oscillation, and impedance matching, the electrical performance parameters and appearance quality of inductors directly affect the reliability and consistency of the entire system. With the electronics manufacturing industry moving towards mass production and high consistency, the incoming inspection and outgoing grading of inductor devices place higher demands on testing efficiency, testing dimensions, and the objectivity of grading.

[0003] Currently, the testing of inductors mainly employs discrete manual or semi-automatic methods. For visual inspection, surface defects are typically screened manually using visual inspection or simple photoelectric sensors. For electrical parameter measurement, inductance, quality factor, and DC resistance are measured offline using an independent LCR digital bridge. For saturation current testing, an external DC bias source and current source are required, with operators increasing the current in steps and manually recording the inductance decay point. For temperature rise testing, the inductor is placed in a constant temperature chamber with the rated current applied, and temperature changes are manually read. Each step is independent, and the selection of testing dimensions, the execution of the testing process, and the grading determination all heavily rely on the manual judgment of on-site operators, resulting in the following shortcomings:

[0004] (1) In the existing scheme, basic electrical parameters (inductance, quality factor, DC resistance), power parameters (saturation current, temperature rise characteristics) and appearance quality can theoretically be measured. However, which dimensions are actually inspected and which test items are performed depends on the operator's on-site judgment based on experience, the urgency of the task, or the availability of equipment. For example, during incoming material inspection, the operator may skip the saturation current scan and only measure the inductance due to time constraints; during factory grading, the temperature rise test may be omitted because it takes a long time; and the appearance inspection may reduce the sampling ratio or relax the judgment criteria due to human eye fatigue.

[0005] (2) In the existing scheme, the testing equipment is arranged in an island layout, and the devices need to be manually transferred one by one between multiple workstations. The progress of the testing process, the connection between each workstation, and the control of the testing cycle all depend on the on-site coordination of the operators. When a workstation is interrupted due to equipment failure or personnel absence, the entire testing process stops;

[0006] (3) Although there are unified grading standards in the existing schemes for different workstations, the actual judgment requires the testing personnel to manually compare each item against the standard. Especially for devices at the grade boundary, different testing personnel have different understandings and implementation standards of the standard clauses, resulting in poor consistency of grading results between different personnel or different batches. When the testing cycle is mismatched, the devices accumulate or wait between workstations, resulting in low testing efficiency. Summary of the Invention

[0007] This application provides an automated detection and classification system for inductor devices, which aims to solve the problems of low detection efficiency and poor classification results in existing technologies.

[0008] In a first aspect, an automated detection and classification system for inductor devices is provided, the system comprising: a feeding module, a conveying module, a detection module, a sorting actuator, and a control module;

[0009] The detection module includes an appearance inspection module, an electrical parameter measurement module, a saturation current testing module, and a temperature rise testing module arranged sequentially along the conveying module to form a multi-station assembly line structure.

[0010] The control module is communicatively connected to the feeding module, conveying module, each detection module and the sorting execution mechanism, and is used to coordinate the timing and data interaction of each module.

[0011] The feeding module is used to feed the inductors to be tested one by one to the input end of the conveying module;

[0012] The control module includes a timing control unit, a grading unit, and a sorting unit. The timing control unit assigns a number to each inductor to be tested that is sequentially fed by the feeding module, and records each number and its corresponding feeding time to form a mapping table. The timing control unit also controls the conveying module to convey the devices at a predetermined pace, so that each inductor to be tested passes sequentially through the positions of each detection module. The grading unit receives the detection data uploaded by each detection module, associates the detection data of the same device with the number to form a detection data set, performs a grading judgment on the detection data set according to the grading judgment rules, and outputs the judgment result. The sorting unit generates a sorting instruction based on the judgment result and sends it to the sorting execution mechanism.

[0013] Each monitoring module of the conveying module is equipped with a function to record the arrival time of a device and report it to the timing control unit when the device is detected to arrive. The timing control unit queries the mapping table based on the arrival time and determines the number corresponding to the loading time that matches the arrival time as the number of the current inductor to be tested.

[0014] The appearance inspection module is used to acquire image data of the inductor to be tested and perform appearance inspection on the device based on image processing. When an appearance defect is detected, it is uploaded to the control module so that the control module controls the rejection mechanism to remove the inductor with the appearance defect from the conveying module. The rejection mechanism is located between the appearance inspection module and the electrical parameter measurement module.

[0015] The electrical parameter measurement module is used to send an AC test signal to the inductor under test to measure the actual inductance, the equivalent series resistance of the actual inductor device, and the actual DC resistance, and to calculate the inductance deviation rate, quality factor deviation, and DC resistance deviation.

[0016] The saturation current scanning module is configured to apply an increasing DC bias current to the inductor to be detected, monitor the change in the inductance, and determine the saturation current based on the decay characteristics of the inductance with the current.

[0017] The temperature rise test module is used to apply a working current to the inductor to obtain the temperature rise value of the device.

[0018] The sorting execution mechanism is used to sort the devices into the corresponding collection channels according to the execution instructions sent by the control module.

[0019] Optionally, in the above scheme, the conveying module further includes a gripping unit;

[0020] The control module controls the grasping unit at the corresponding position to grasp the inductor to be detected based on the arrival signal of the inductor detected by the position detection unit, and controls the corresponding detection module to start detection; after the detection is completed, the control module controls the grasping unit to release the inductor to be detected.

[0021] Optionally, in the above scheme, the appearance inspection module includes a CCD vision sensor, a four-zone ring light source, and an image processing unit. The image processing unit is configured to detect the size of the inductor in the image of the inductor to be inspected acquired by the CCD vision sensor using an edge detection algorithm, and to detect the size defects and surface defects of the inductor in the image of the inductor to be inspected acquired by the CCD vision sensor.

[0022] In the above scheme, optionally, the inductance deviation rate is calculated using the following formula:

[0023] δ_L = |L_meas - L_nom| / L_nom * 100%;

[0024] Where L_meas is the actual inductance and L_nom is the nominal inductance;

[0025] The quality factor deviation is first calculated using the actual quality factor through the equivalent series resistance of the actual inductor, and then the quality factor deviation is calculated using the following formula:

[0026] δ_Q = |Q_meas - Q_std| / Q_std * 100%;

[0027] Where Q_meas is the actual quality factor and Q_std is the standard quality factor;

[0028] The DC resistance deviation is calculated using the following formula:

[0029] δ_DCR = |DCR_meas - DCR_nom| / DCR_nom * 100%;

[0030] Where DCR_meas is the measured DC resistance and DCR_nom is the nominal DC resistance.

[0031] Optionally, in the above scheme, the system further includes: a self-resonant frequency measurement module: configured to detect the actual self-resonant frequency point and compare the actual self-resonant frequency point with the nominal self-resonant frequency. When the actual self-resonant frequency point is greater than or equal to the nominal self-resonant frequency, it is determined to meet the design specifications; otherwise, it is determined to be unqualified.

[0032] Optionally, in the above scheme, the classification determination rule is a two-level classification rule:

[0033] In the first level, the grading unit determines whether each detection parameter in the detection data set meets the qualification conditions based on a preset hard constraint threshold. If any parameter does not meet the conditions, it is determined to be unqualified.

[0034] In the second stage, for the inductor under test that has passed the first stage's determination, the grading unit uses the detection parameters as input features and outputs the determination result of the device's grade through a neural network model.

[0035] In the above scheme, optionally, the hard constraint threshold of the first level includes: the inductance deviation rate is less than 10%, the quality factor deviation rate is less than 15%, the DC resistance deviation rate is less than 10%, the temperature rise is less than 40℃, and the saturation current margin is greater than 120%; the saturation current margin is the ratio of the saturation current to the rated operating current.

[0036] In the above scheme, optionally, the neural network is a BP neural network, which takes the five parameters of inductance deviation rate, quality factor deviation rate, DC resistance deviation rate, saturation current margin and temperature rise value as input features, outputs the probability distribution of the device belonging to the general-purpose level, high-efficiency level and power level, and takes the one with the highest probability as the final level.

[0037] Optionally, in the above scheme, the sorting execution mechanism includes multiple sorting points arranged sequentially along the output end of the conveying module, each sorting point corresponding to a collection channel, and each sorting point is equipped with a position detector and an actuator;

[0038] The sorting unit binds the judgment result with the number of the inductor to be tested, and sends the corresponding sorting instruction to each sorting point;

[0039] When the position detector at any sorting point detects the arrival of the inductor to be detected, the sorting point determines, according to the sorting instruction, whether the collection channel corresponding to the judgment result of the currently arriving inductor to be detected is consistent with the collection channel corresponding to the sorting point:

[0040] If they match, the sorting point controls the actuator to push the device into the corresponding collection channel;

[0041] If there is a discrepancy, the sorting point remains inactive, and the detection inductor continues to the next sorting point.

[0042] Secondly, an automated detection and classification method for inductor devices, executed by a control module in the system described in any one of the first aspects above, includes the following steps:

[0043] The control feeding module feeds the inductors to be tested one by one to the input terminal of the conveying module, assigns a number to each device fed in sequence, records each number and its corresponding feeding time, and forms a mapping table;

[0044] The control conveying module sequentially conveys the inductor to be tested to the location of each detection module at a predetermined pace; when the position detection unit of the receiving conveying module detects the arrival of the inductor, it records the arrival time, queries the mapping table according to the arrival time, and determines the tag number corresponding to the feeding time that matches the arrival time as the number of the currently arriving inductor to be tested.

[0045] The appearance inspection module receives the appearance inspection results. If the appearance inspection results indicate an appearance defect, the rejection mechanism will remove the corresponding inductor to be inspected from the conveying module.

[0046] Receive inductance deviation rate, quality factor deviation rate, and DC resistance deviation rate reported by the electrical parameter measurement module;

[0047] Receive the saturation current reported by the saturation current test module;

[0048] Receive the temperature rise value reported by the temperature rise test module;

[0049] The detection data of the same device are associated with the number to form a detection data set, and the detection data set is graded according to the grading judgment rule, and the judgment result is output.

[0050] Based on the determination result, a sorting instruction is generated and sent to the sorting execution mechanism to control the sorting execution mechanism to sort the devices to the corresponding collection channel.

[0051] Compared with the prior art, this application has at least the following beneficial effects:

[0052] A multi-station production line structure is formed by a feeding module, a conveying module, and sequentially arranged appearance inspection modules, electrical parameter measurement modules, saturation current testing modules, and temperature rise testing modules along the conveying module. This, combined with a control module to coordinate the timing and data interaction of each module, achieves fully automated continuous operation of the inductor testing process. This overcomes the problems of low testing efficiency and asynchronous cycle times caused by the independent operation of each testing device and the need for manual transfer of each device in existing technologies. Simultaneously, the control module assigns a number to each inductor to be tested and establishes a mapping table. This allows each testing unit on the conveying module to only report the arrival time of the inductor, and the control module can uniquely determine the number of the inductor by querying the mapping table. This achieves automatic association between the testing data of each module and the same inductor without relying on physical identifiers (such as barcodes or RFID). This solves the industry problem of inductors being too small to be labeled or engraved with physical identifiers, and avoids the influence of physical identifiers on inductor characteristics. This approach addresses potential interference and overcomes the problems of fragmented data across testing stages and the inability to establish complete performance profiles for individual inductors in existing technologies. Building upon this, the grading unit of the control module uses a two-level grading rule to determine the various test data of the same device. First, it uses hard constraint thresholds to quickly eliminate non-conforming products. Then, it uses a neural network model to perform multi-parameter nonlinear fusion and grade probability determination on devices that pass the initial screening. This replaces the existing grading method that relies on manual comparison by testing personnel against standards, effectively eliminating inconsistencies in grading results caused by differences in standard understanding and implementation between different personnel or batches. This significantly improves the objectivity, consistency, and accuracy of the grading results. The sorting unit automatically generates sorting instructions based on the judgment results and controls the sorting execution mechanism to sort the devices one by one to the corresponding collection channels, forming a complete automated closed loop from material supply, testing, grading to sorting. This comprehensively improves the efficiency, data traceability, and grading reliability of inductor device testing and classification. Attached Figure Description

[0053] Figure 1 This is a structural diagram of an automated detection and classification system for inductors provided in one embodiment of this application.

[0054] Figure 2 This is a schematic diagram of the mechanical structure of the conveyor track, photoelectric sensor and positioning gripper provided in one embodiment of this application.

[0055] Figure 3 This is a schematic diagram of the mechanical structure and sorting logic of an automatic sorting module provided in one embodiment of this application. Detailed Implementation

[0056] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments.

[0057] In one embodiment, such as Figure 1 and Figure 2 As shown, an automated testing and sorting system for inductor devices is provided, including: a feeding module, a conveying module, a testing module, a sorting actuator, and a control module;

[0058] The testing module includes an appearance inspection module, an electrical parameter measurement module, a saturation current testing module, and a temperature rise testing module arranged sequentially along the conveying module to form a multi-station assembly line structure.

[0059] The control module is communicatively connected to the feeding module, conveying module, each detection module and the sorting execution mechanism, and is used to coordinate the timing and data interaction of each module.

[0060] The feeding module is used to feed the inductors to be tested one by one to the input terminal of the conveying module;

[0061] The control module includes a timing control unit, a grading unit, and a sorting unit. The timing control unit assigns a number to each inductor to be tested that is sequentially fed by the feeding module, and records each number and its corresponding feeding time to form a mapping table. The timing control unit also controls the conveying module to convey the devices at a predetermined pace, so that each inductor to be tested passes through the location of each detection module in sequence. The grading unit receives the detection data uploaded by each detection module, associates the detection data of the same device with numbers to form a detection data set, performs a grading judgment on the detection data set according to the grading judgment rules, and outputs the judgment result. The sorting unit generates a sorting instruction based on the judgment result and sends it to the sorting execution mechanism.

[0062] Each monitoring module of the conveying module is equipped with a position detection unit, which is used to record the arrival time when the device is detected to arrive and report it to the timing control unit. The timing control unit queries the mapping table according to the arrival time and determines the number corresponding to the loading time that matches the arrival time as the number of the current inductor to be tested.

[0063] The appearance inspection module is used to acquire image data of the inductor to be tested and perform appearance inspection on the device based on image processing. When an appearance defect is detected, it is uploaded to the control module so that the control module controls the rejection mechanism to remove the inductor with the appearance defect from the conveying module. The rejection mechanism is located between the appearance inspection module and the electrical parameter measurement module.

[0064] The electrical parameter measurement module is used to send an AC test signal to the inductor under test to measure the actual inductance, the equivalent series resistance of the actual inductor device, and the actual DC resistance, and to calculate the inductance deviation rate, quality factor deviation, and DC resistance deviation.

[0065] The saturation current scanning module is configured to apply an increasing DC bias current to the inductor under test, monitor the change in the inductance under test, and determine the saturation current based on the decay characteristics of the inductance with the current.

[0066] The temperature rise test module is used to apply operating current to inductor devices to obtain the temperature rise value of the devices;

[0067] The sorting actuator is used to sort devices into the corresponding collection channels according to the execution instructions sent by the control module.

[0068] In this embodiment, the conveying module's conveying track is equipped with a drive mechanism, a position detection unit consisting of photoelectric sensors, and a gripping module consisting of positioning claws. The drive mechanism is used to convey the device at a fixed interval. The photoelectric sensors are used to detect whether the device has reached the detection station, and when the device is detected to have arrived, they trigger the positioning claws at that detection station to position and clamp the device. Multiple photoelectric sensors are installed one-to-one on the side of the conveying track at each detection station, with their sensing ends facing the device's passing position on the track. Multiple positioning claws are also installed one-to-one at each detection station and are driven by a cylinder to open and close perpendicular to the conveying direction. When any photoelectric sensor detects that a device has reached the corresponding detection station, it outputs a trigger signal, causing the drive mechanism to pause conveying, and the positioning claws at that station to clamp the device for the corresponding detection module to perform detection. After the detection is completed, the positioning claws release, and the drive mechanism resumes conveying.

[0069] When the appearance inspection module determines that there are size defects and surface defects, it classifies the inductor under test as an appearance defective product and writes a rejection mark on the device number. A rejection mechanism is set between the appearance inspection module and the LCR parameter measurement module. The rejection mechanism includes a pneumatic push rod and an NG collection channel that is laterally connected to the conveyor track. When the device with the rejection mark is transferred to the rejection mechanism along the conveyor track, the pneumatic push rod pushes it into the NG collection channel so that it does not enter the subsequent inspection station. After the inductor under test is rejected, the original cycle position it occupied becomes an empty inspection position. When the photoelectric sensors of the subsequent inspection stations detect the inspection position, they do not trigger the inspection.

[0070] In the temperature rise testing module, after the saturation current scan is completed, the module applies the rated operating current to the device and collects the inductor surface temperature in real time using an infrared temperature sensor, recording the temperature rise value ΔT. Specifically, during the application of the rated operating current, the infrared temperature sensor continuously collects the device surface temperature according to a preset sampling period; when the device surface temperature reaches thermal equilibrium, i.e., the rate of temperature change is less than a preset rate of change threshold (e.g., within a preset duration of time, such as 5 minutes), the temperature rise test is completed. When the temperature rise stabilizes, the rated operating current is stopped; if the power-on time reaches the preset maximum test time (e.g., 60 minutes), the rated operating current is also stopped to balance the test cycle; the temperature rise value The difference between the highest surface temperature T_max reached by the device at the time of termination and the initial temperature T_0 of the device before the test begins is... .

[0071] In the grading module, the detection data of the same device are associated by number: the automatic feeding module assigns a unique position number to each device fed, and the photoelectric sensors at each detection station record the arrival time of the device and bind it with the number. The detection data uploaded by each detection module is associated with the same device by the number as the primary key, so that the detection data of multiple inductor devices under test transported on the production line at the same time correspond one by one and are not confused with each other.

[0072] When the control module controls the rejection mechanism to remove the device from the conveying module, the original position occupied by the device becomes a detection vacancy. The detection vacancy and its bound logical position number continue to flow with the conveying module. When the arrival time reported by the position detection unit of the subsequent detection module corresponds to the detection vacancy, the timing control unit controls the corresponding detection module to skip the detection.

[0073] In the aforementioned automated testing and classification system for inductor devices, a multi-station assembly line structure is formed by a feeding module, a conveying module, and sequentially arranged appearance inspection modules, electrical parameter measurement modules, saturation current testing modules, and temperature rise testing modules along the conveying module. This, combined with a control module, coordinates the timing and data interaction of each module, achieving fully automated continuous operation of the inductor device testing process. This overcomes the problems of low testing efficiency and asynchronous cycle times caused by the independent operation of each testing device and the need for manual transfer of each device in existing technologies. Simultaneously, the control module assigns a number to each inductor to be tested and establishes a mapping table. This allows each testing unit on the conveying module to only report the arrival time of the inductor to be tested, and the control module can uniquely determine the number of the inductor by querying the mapping table. This achieves automatic association between the testing data of each module and the same inductor to be tested without relying on physical identifiers (such as barcodes or RFID). This solves the industry problem of inductor devices being too small to be physically identified by affixed or engraved, and avoids the need for physical identification. The system addresses the potential interference of identification marks on inductor characteristics and overcomes the problems of fragmented data across testing stages and the inability to establish complete performance profiles for individual inductors in existing technologies. Furthermore, the grading unit of the control module uses a two-level grading rule to determine the various test data of the same device: first, it quickly eliminates defective products using hard constraint thresholds; then, it uses a neural network model to perform multi-parameter nonlinear fusion and grade probability determination on devices that pass the initial screening. This replaces the existing grading method that relies on manual comparison by testing personnel against standards, effectively eliminating inconsistencies in grading results caused by differences in standard understanding and implementation between different personnel or batches, significantly improving the objectivity, consistency, and accuracy of the grading results. The sorting unit automatically generates sorting instructions based on the judgment results and controls the sorting execution mechanism to sort the devices one by one to the corresponding collection channels, forming a complete automated closed loop from material supply, testing, grading to sorting, thereby comprehensively improving the efficiency, data traceability, and grading reliability of inductor device testing and classification.

[0074] In one embodiment, the conveying module further includes a gripping unit; the control module controls the gripping unit at the corresponding position to grip the inductor to be detected based on the arrival signal of the inductor to be detected detected by the position detection unit, and controls the corresponding detection module to start detection; after the detection is completed, the control module controls the gripping unit to release the inductor to be detected.

[0075] In one embodiment, the appearance inspection module includes a CCD vision sensor, a four-zone ring light source, and an image processing unit. The image processing unit is configured to detect the size of the inductor in the image of the inductor to be inspected acquired by the CCD vision sensor using an edge detection algorithm, and to detect size defects and surface defects of the inductor in the image of the inductor to be inspected acquired by the CCD vision sensor.

[0076] In this embodiment, visual inspection and preliminary screening are performed: the inductive devices are visually inspected using a CCD vision sensor located above the inspection track.

[0077] (1) Size inspection: The edge detection algorithm (Canny) is used to extract the outline of the device, calculate the length, width and height of the device, and compare them with the standard size library. If the deviation exceeds ±5%, it is judged as an appearance defect.

[0078] (2) Surface defect detection: Template matching algorithm (normalized cross-correlation coefficient NCC) is used to detect surface scratches, device offset, pin deformation and other defects. The NCC threshold is set to 0.85. If it is lower than the threshold, it is judged as a surface defect.

[0079] (3) Preliminary screening: Devices that pass the appearance inspection will proceed to the next step of electrical parameter inspection. Unqualified devices will be judged as appearance defects and their location numbers will be written into the rejection mark. A rejection mechanism is set between the appearance inspection station and the LCR parameter measurement station. The rejection mechanism includes a pneumatic push rod and an NG collection channel that is laterally connected to the conveyor rail. When the device with the rejection mark is transferred to the first rejection mechanism along the conveyor rail, it will be pushed into the NG collection channel (NG collection box) by the pneumatic push rod so that it will not enter the subsequent inspection station. After the device is rejected, the original cycle position it occupied becomes an empty inspection position. When the photoelectric sensors of the subsequent inspection stations detect the inspection position, they will not trigger the inspection. The inspection position will be directly transferred to the automatic sorting module along the conveyor rail.

[0080] In one embodiment, the inductance deviation rate is calculated using the following formula:

[0081] ;

[0082] Where L_meas is the actual inductance and L_nom is the nominal inductance;

[0083] The quality factor deviation is first calculated using the actual quality factor through the equivalent series resistance of the actual inductor, and then the quality factor deviation is calculated using the following formula:

[0084] ;

[0085] Where Q_meas is the actual quality factor and Q_std is the standard quality factor;

[0086] The DC resistance deviation is calculated using the following formula:

[0087] ;

[0088] Where DCR_meas is the measured DC resistance and DCR_nom is the nominal DC resistance.

[0089] In this embodiment, electrical parameters are automatically detected: inductors that pass the appearance inspection enter the electrical parameter detection station, where the electrical parameters are automatically measured by an LCR digital bridge.

[0090] (1) Inductance L measurement: An AC test signal is applied to the inductor under test by an LCR digital bridge. The test frequency is set to 1kHz and the test level is 1Vrms. The bridge automatically reads the inductance value. (unit: The deviation rate is calculated by comparing it with the nominal value L_nom. ;

[0091] (2) Quality factor Q measurement: While applying an AC test signal to measure the inductance, the bridge also measures the equivalent series resistance Rs of the inductor and calculates the value according to the formula. (in The quality factor is calculated from f (where f is the test frequency). , and standard quality factor Compare and calculate the Q-bias. ; The measured inductance;

[0092] (3) DC resistance (DCR) measurement: The DC resistance between the pins is measured using the four-wire method. (Unit: mΩ), compare with the nominal value DCR_nom to calculate the deviation rate. ;

[0093] In one embodiment, the system further includes: a self-resonant frequency measurement module: configured to detect the actual self-resonant frequency point and compare the actual self-resonant frequency point with the nominal self-resonant frequency; if the actual self-resonant frequency point is greater than or equal to the nominal self-resonant frequency, it is determined to meet the design specifications; otherwise, it is determined to be unqualified.

[0094] In this embodiment, self-resonant frequency (SRF) measurement (optional): for high-frequency inductive devices, the scanning frequency range is... Detecting the self-resonant frequency point With respect to the nominal self-resonant frequency When a comparison is performed, If the inductor meets the design specifications, it is deemed to meet the requirements; otherwise, it is deemed NG-SRF. Inductors that fail the test do not require further testing and are directly sent to the non-conforming collection channel via the sorting actuator.

[0095] In one embodiment, the classification rule is a two-level classification rule:

[0096] In the first level, the grading unit determines whether each detection parameter in the detection data set meets the qualification conditions based on a preset hard constraint threshold. If any parameter does not meet the conditions, it is determined to be unqualified.

[0097] In the second stage, for the inductor under test that has passed the first stage judgment, the grading unit uses the detection parameters as input features and outputs the judgment result of the device's level through a neural network model.

[0098] In this embodiment, the hard constraints of the first-level rule engine include: the inductance deviation rate is not greater than... The quality factor deviation rate is not greater than The DC resistance deviation rate is not greater than The temperature rise value is not greater than The saturation current margin is not less than If any one of the five parameters fails to meet the corresponding hard constraint, the test is deemed unqualified. The saturation current margin and the temperature rise value are two independent hard constraint parameters, and are related by an OR condition, meaning the saturation current margin must be less than the specified value. Or the temperature rise value is greater than If any of the following conditions are met, the application will be deemed unqualified.

[0099] The second-level neural network uses inductance deviation rate Q-value deviation rate DCR deviation rate Saturation current margin Temperature rise The five parameters are input features. The output probability distribution of the device after training the BP neural network model is classified as general-purpose, high-efficiency, and power-level. The one with the highest probability is taken as the final level.

[0100] The definitions of each level are as follows: General Level — , , Saturation current margin Temperature rise High-efficiency level — And the Q value is greater than 90% of the nominal value. Saturation current margin 150%-180%, temperature rise Power stage — , Saturation current margin greater than Temperature rise less than ;

[0101] The parameter ranges for each of the above levels are used to provide a reference standard when labeling historical samples with level labels during the training phase. The neural network learns from the labeled samples to establish a non-linear mapping relationship between the combination of five parameters and the level. The output is the probability distribution of the device belonging to each level, rather than checking whether each parameter falls within the above range. Therefore, for devices with parameter intersections near the level boundary, the neural network makes a judgment by combining all five parameters through the learned non-linear mapping relationship. The final level is based on the probability distribution output by the network.

[0102] The method for constructing the BP neural network model is as follows: historical detection data is collected to form a sample set, and each sample contains the inductance deviation rate. Q-value deviation rate DCR deviation rate δ_DCR, saturation current margin Isat_margin, temperature rise value Five parameters and corresponding level labels are defined. The level labels are labeled by technical experts based on the parameter ranges of each level and the results of whole-machine application verification. The five input features are normalized to eliminate dimensional differences. The sample set is divided into training set and validation set according to a preset ratio. The Adam optimizer is used to iteratively train the network weights and biases with the cross-entropy loss function as the objective function. Training stops when the validation set loss no longer decreases for a preset number of consecutive rounds. The trained network parameters are saved for online hierarchical access.

[0103] In one embodiment, such as Figure 3 As shown, the sorting execution mechanism includes multiple sorting points arranged sequentially along the output end of the conveying module. Each sorting point corresponds to a collection channel, and each sorting point is equipped with a position detector and an actuator. The sorting unit binds the judgment result with the number of the inductor to be tested and sends the corresponding sorting instruction to each sorting point. When the position detector of any sorting point detects the arrival of the inductor to be tested, the sorting point determines, according to the sorting instruction, whether the collection channel corresponding to the judgment result of the currently arriving inductor to be tested is consistent with the collection channel corresponding to the sorting point. If they are consistent, the sorting point controls the actuator to push the device into the corresponding collection channel. If they are inconsistent, the sorting point remains inactive, and the inductor to be tested continues to flow to the next sorting point.

[0104] In this embodiment, the automatic sorting module includes a controller and multiple sorting points arranged sequentially along the conveying direction at the end of the conveying track. Each sorting point corresponds to a collection channel (NG collection channel, general-purpose collection channel, high-efficiency collection channel, and power collection channel). Each sorting point is equipped with a position sensor and an actuator (solenoid valve and pneumatic push rod driven by the solenoid valve). Each collection channel is connected to the conveying track via a lateral slide. The sorting logic is as follows: the controller's classification module outputs a judgment result and binds the judgment result with the position number of the inductor. When the position sensor of a sorting point detects the arrival of a device, the controller queries the judgment result bound to that position number. If the sorting point matches the collection channel corresponding to the judgment result, the solenoid valve of that sorting point is triggered, causing the pneumatic push rod to move and push the device into the corresponding collection channel via the lateral slide. At the same time, the solenoid valves of the other sorting points remain inactive. If they do not match, the sorting point remains inactive, and the device continues to flow with the conveying track to the next sorting point until it reaches the sorting point corresponding to its judgment result. This achieves one-to-one correspondence between the device and the collection channel, avoiding missorting.

[0105] In one embodiment, an automated detection and classification method for inductor devices is provided, executed by the control module of the system described above, and includes the following steps:

[0106] The control feeding module feeds the inductors to be tested one by one to the input terminal of the conveying module, assigns a number to each device fed in sequence, records each number and its corresponding feeding time, and forms a mapping table;

[0107] The control conveying module sequentially conveys the inductor to be tested to the location of each detection module at a predetermined pace; when the position detection unit of the receiving conveying module detects the arrival of the inductor, it records the arrival time, queries the mapping table according to the arrival time, and determines the tag number corresponding to the feeding time that matches the arrival time as the number of the currently arriving inductor to be tested.

[0108] The appearance inspection module receives the appearance inspection results. If the appearance inspection results indicate an appearance defect, the rejection mechanism will remove the corresponding inductor to be inspected from the conveying module.

[0109] Receive inductance deviation rate, quality factor deviation rate, and DC resistance deviation rate reported by the electrical parameter measurement module;

[0110] Receive the saturation current reported by the saturation current test module;

[0111] Receive the temperature rise value reported by the temperature rise test module;

[0112] The detection data of the same device are associated with the number to form a detection data set, and the detection data set is graded according to the grading judgment rule, and the judgment result is output.

[0113] Based on the determination result, a sorting instruction is generated and sent to the sorting execution mechanism to control the sorting execution mechanism to sort the devices to the corresponding collection channel.

[0114] Example 1: Automated detection and grading of power inductors:

[0115] This embodiment uses a TDK SPM series power inductor (6.8uH, rated current 3.2A, DCR 28m ohms) as an example. Step S1: The material is fed by a vibratory feeder and transferred to the inspection table by a vacuum robot. Step S2: Six-sided images are captured using a four-zone light source, and the rejection rate is calculated. Step S3: Inductance 6.82uH (+2.9%), Q value , Ohms (94.6%). Step S4: , margin Step S5: Temperature rises by 27.5 degrees Celsius. Step S6: The rule engine passes all tests; the neural network determines the power level probability to be 0.72. Step S7: The pneumatic robotic arm sorts the components to the power level channel.

[0116] Example 2: Detection and grading of high-frequency RF inductors:

[0117] This embodiment uses a Murata LQW series wire-wound inductor (220nH, The system is used as the target. Impedance characteristics are scanned in the 1MHz to 3GHz frequency band. Systems with an SRF greater than 2.45GHz and a peak Q value greater than 80 are classified as high-efficiency and suitable for 2.4GHz WiFi front-end circuits.

[0118] Example 3: Application of batch incoming material inspection:

[0119] This example uses the incoming inspection of 50,000 Sumida power inductors (10uH, rated current 2.5A) as a scenario. The system completed all inspections in 4.2 hours: 380 inductors with appearance defects were rejected. 420 failed the hard constraint test. ), General grade 15600 ( ), 22,000 high-efficiency units ( ), power stage 11600 ( The detection data is automatically uploaded to the MES system.

[0120] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. An automated testing and classification system for inductor devices, characterized in that, The system includes: a feeding module, a conveying module, a detection module, a sorting actuator, and a control module; The detection module includes an appearance inspection module, an electrical parameter measurement module, a saturation current testing module, and a temperature rise testing module arranged sequentially along the conveying module to form a multi-station assembly line structure. The control module is communicatively connected to the feeding module, conveying module, each detection module and the sorting execution mechanism, and is used to coordinate the timing and data interaction of each module. The feeding module is used to feed the inductors to be tested one by one to the input end of the conveying module; The control module includes a timing control unit, a grading unit, and a sorting unit. The timing control unit assigns a number to each inductor to be tested that is sequentially fed by the feeding module, and records each number and its corresponding feeding time to form a mapping table. The timing control unit also controls the conveying module to convey the devices at a predetermined pace, so that each inductor to be tested passes sequentially through the positions of each detection module. The grading unit receives the detection data uploaded by each detection module, associates the detection data of the same device with the number to form a detection data set, performs a grading judgment on the detection data set according to the grading judgment rules, and outputs the judgment result. The sorting unit generates a sorting instruction based on the judgment result and sends it to the sorting execution mechanism. Each monitoring module of the conveying module is equipped with a function to record the arrival time of a device and report it to the timing control unit when the device is detected to arrive. The timing control unit queries the mapping table based on the arrival time and determines the number corresponding to the loading time that matches the arrival time as the number of the current inductor to be tested. The appearance inspection module is used to acquire image data of the inductor to be tested and perform appearance inspection on the device based on image processing. When an appearance defect is detected, it is uploaded to the control module so that the control module controls the rejection mechanism to remove the inductor with the appearance defect from the conveying module. The rejection mechanism is located between the appearance inspection module and the electrical parameter measurement module. The electrical parameter measurement module is used to send an AC test signal to the inductor under test to measure the actual inductance, the equivalent series resistance of the actual inductor device, and the actual DC resistance, and to calculate the inductance deviation rate, quality factor deviation, and DC resistance deviation. The saturation current scanning module is configured to apply an increasing DC bias current to the inductor to be detected, monitor the change in the inductance, and determine the saturation current based on the decay characteristics of the inductance with the current. The temperature rise test module is used to apply a working current to the inductor to obtain the temperature rise value of the device. The sorting execution mechanism is used to sort the devices into the corresponding collection channels according to the execution instructions sent by the control module.

2. The automated detection and classification system for inductor devices according to claim 1, characterized in that, The conveying module also includes a gripping unit; The control module controls the grasping unit at the corresponding position to grasp the inductor to be detected based on the arrival signal of the inductor detected by the position detection unit, and controls the corresponding detection module to start detection; after the detection is completed, the control module controls the grasping unit to release the inductor to be detected.

3. The automated detection and classification system for inductor devices according to claim 1, characterized in that, The appearance inspection module includes a CCD vision sensor, a four-zone ring light source, and an image processing unit. The image processing unit is configured to detect the size of the inductor in the image of the inductor to be inspected acquired by the CCD vision sensor using an edge detection algorithm, and to detect the size defects and surface defects of the inductor in the image of the inductor to be inspected acquired by the CCD vision sensor.

4. The automated detection and classification system for inductor devices according to claim 1, characterized in that, The inductance deviation rate is calculated using the following formula: ； Where L_meas is the actual inductance and L_nom is the nominal inductance; The quality factor deviation is first calculated using the actual quality factor through the equivalent series resistance of the actual inductor, and then the quality factor deviation is calculated using the following formula: ； Where Q_meas is the actual quality factor and Q_std is the standard quality factor; The DC resistance deviation is calculated using the following formula: ； Where DCR_meas is the measured DC resistance and DCR_nom is the nominal DC resistance.

5. The automated detection and classification system for inductor devices according to claim 1, characterized in that, The system also includes: a self-resonant frequency measurement module: configured to detect the actual self-resonant frequency point and compare the actual self-resonant frequency point with the nominal self-resonant frequency. When the actual self-resonant frequency point is greater than or equal to the nominal self-resonant frequency, it is determined to meet the design specifications; otherwise, it is determined to be unqualified.

6. The automated detection and classification system for inductor devices according to claim 1, characterized in that, The classification determination rule is a two-level classification rule: In the first level, the grading unit determines whether each detection parameter in the detection data set meets the qualification conditions based on a preset hard constraint threshold. If any parameter does not meet the conditions, it is determined to be unqualified. In the second stage, for the inductor under test that has passed the first stage's determination, the grading unit uses the detection parameters as input features and outputs the determination result of the device's grade through a neural network model.

7. The automated detection and classification system for inductor devices according to claim 6, characterized in that, The first-level hard constraint thresholds include: the inductance deviation rate is less than 10%, the quality factor deviation rate is less than 15%, the DC resistance deviation rate is less than 10%, the temperature rise is less than 40℃, and the saturation current margin is greater than 120%; the saturation current margin is the ratio of the saturation current to the rated operating current.

8. The automated detection and classification system for inductor devices according to claim 6, characterized in that, The neural network is a BP neural network, which takes the five parameters of inductance deviation rate, quality factor deviation rate, DC resistance deviation rate, saturation current margin and temperature rise value as input features, and outputs the probability distribution of the device belonging to the general-purpose level, high-efficiency level and power level, and takes the one with the highest probability as the final level.

9. The automated detection and classification system for inductor devices according to claim 1, characterized in that, The sorting actuator includes multiple sorting points arranged sequentially along the output end of the conveying module. Each sorting point corresponds to a collection channel, and each sorting point is equipped with a position detector and an actuator. The sorting unit binds the judgment result with the number of the inductor to be tested, and sends the corresponding sorting instruction to each sorting point; When the position detector at any sorting point detects the arrival of the inductor to be detected, the sorting point determines, according to the sorting instruction, whether the collection channel corresponding to the judgment result of the currently arriving inductor to be detected is consistent with the collection channel corresponding to the sorting point: If they match, the sorting point controls the actuator to push the device into the corresponding collection channel; If there is a discrepancy, the sorting point remains inactive, and the inductor to be tested continues to the next sorting point.

10. An automated detection and classification method for inductive devices, characterized in that, Performed by the control module in the system according to any one of claims 1 to 9, the method includes the following steps: The control feeding module feeds the inductors to be tested one by one to the input terminal of the conveying module, assigns a number to each device fed in sequence, records each number and its corresponding feeding time, and forms a mapping table; The control conveying module sequentially conveys the inductor to be tested to the location of each detection module at a predetermined pace; when the position detection unit of the receiving conveying module detects the arrival of the inductor, it records the arrival time, queries the mapping table according to the arrival time, and determines the tag number corresponding to the feeding time that matches the arrival time as the number of the currently arriving inductor to be tested. The appearance inspection module receives the appearance inspection results. If the appearance inspection results indicate an appearance defect, the rejection mechanism will remove the corresponding inductor to be inspected from the conveying module. Receive inductance deviation rate, quality factor deviation rate, and DC resistance deviation rate reported by the electrical parameter measurement module; Receive the saturation current reported by the saturation current test module; Receive the temperature rise value reported by the temperature rise test module; The detection data of the same device are associated with the number to form a detection data set, and the detection data set is graded according to the grading judgment rule, and the judgment result is output. Based on the determination result, a sorting instruction is generated and sent to the sorting execution mechanism to control the sorting execution mechanism to sort the devices to the corresponding collection channel.

Images