Electronic component processing system

Abstract

This invention provides an electronic component processing system that can more effectively estimate the cause of abnormalities in the processing mechanism that handles electronic components. [Solution] An electronic component processing system 1 for processing electronic components comprises a processing mechanism 100 for processing electronic components, a control device 110 for controlling the processing mechanism 100, a storage device 30 that stores an inference model 32 for inferring the cause of an abnormality in the processing mechanism 100, a diagnostic device 20 that diagnoses the cause of an abnormality using the inference model 32, and a display device 40 that displays the cause of an abnormality based on the diagnosis by the diagnostic device 20. The inference model 32 is trained to infer the cause of an abnormality based on training data, and the training data includes the cause of an abnormality and feature data indicating the feature quantities corresponding to the cause of the abnormality. The control device 110 outputs feature data based on information obtained from the processing mechanism 100 to the diagnostic device 20, and the diagnostic device 20 obtains the cause of an abnormality from the inference model by inputting the feature data to the inference model.

Description

Technical Field

[0001] The present disclosure relates to an electronic component processing system for processing electronic components.

Background Art

[0002] Electronic components are manufactured through various processing steps. The various processing steps may include, for example, a step of mounting an electronic component at a predetermined location and a step of inspecting the mounting result. For example, Japanese Patent Application Laid-Open No. 2003-110288 (Patent Document 1) describes a system including a mounting device for mounting an electronic component and an inspection device for inspecting the mounting state of the electronic component, and configured to estimate the cause of an abnormality in the mounting device based on the inspection result by the inspection device.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] The system disclosed in Japanese Patent Application Laid-Open No. 2003-110288 (Patent Document 1) estimates the cause of an abnormality when the inspection criteria are not met according to the inspection criteria preset by the designer. Therefore, in this type of conventional system, there is a problem that the cause of an abnormality that may occur when the inspection criteria are met cannot be sufficiently estimated.

[0005] The present disclosure has been made to solve such problems, and an object thereof is to provide an electronic component processing system capable of more effectively estimating the cause of an abnormality in a processing mechanism for processing electronic components.

Means for Solving the Problems

[0006] The electronic component processing system relating to this disclosure is an electronic component processing system for processing electronic components, comprising: a processing mechanism for processing electronic components; a control device for controlling the processing mechanism; a storage device for storing an inference model for inferring the cause of an abnormality in the processing mechanism; a diagnostic device for diagnosing the cause of an abnormality using the inference model; and a display device for displaying the cause of an abnormality based on the diagnosis by the diagnostic device. The inference model is trained to infer the cause of an abnormality based on training data, and the training data includes the cause of an abnormality and feature data indicating the feature quantities corresponding to the cause of the abnormality. The control device outputs feature data based on information obtained from the processing mechanism to the diagnostic device, and the diagnostic device obtains the cause of an abnormality from the inference model by inputting the feature data into the inference model. [Effects of the Invention]

[0007] According to this disclosure, it is possible to provide an electronic component processing system that can more effectively estimate the cause of abnormalities in a processing mechanism that processes electronic components. [Brief explanation of the drawing]

[0008] [Figure 1] This is a block diagram showing the configuration of the electronic component processing system according to this embodiment. [Figure 2] This diagram shows an overview of the configuration of the visual inspection mechanism. [Figure 3] This is a magnified view of the area around the alignment guide of the visual inspection mechanism. [Figure 4] This is a block diagram used to explain machine learning inference models. [Figure 5] This is a flowchart showing the processing procedure of the inspection device. [Figure 6] This is a flowchart showing the processing procedure of the diagnostic device. [Figure 7] This figure shows an example of a diagnostic result displayed on a display device. [Figure 8] This diagram shows an overview of the configuration of the electrical characteristics testing mechanism related to the modified example. [Figure 9] This is a block diagram used to explain the machine learning of the inference model involved in the modified example. [Modes for carrying out the invention]

[0009] Each embodiment will be described in detail below with reference to the drawings. The same or corresponding parts in the drawings are denoted by the same reference numerals, and their descriptions will not be repeated.

[0010] [Electronic component processing system 1] Figure 1 is a block diagram showing the configuration of the electronic component processing system 1 according to this embodiment. The configuration and processing overview of the electronic component processing system 1 will be described below using Figure 1.

[0011] The electronic component processing system 1 includes an inspection device 10, a diagnostic device 20, a storage device 30, a display device 40, and a tablet terminal 50.

[0012] The display device 40 may be a liquid crystal display device or an organic EL display device. The tablet terminal 50 is carried by a service technician who maintains the inspection device 10. A notebook computer or a desktop computer may be used instead of the tablet terminal 50.

[0013] The inspection device 10 is, for example, a visual inspection device for inspecting the appearance of electronic components. The inspection device 10 includes a visual inspection mechanism 100, a control device 110, a drive mechanism 120, and an adjustment mechanism 130.

[0014] The control device 110 outputs a command to the drive mechanism 120 to drive the visual inspection mechanism 100. The visual inspection mechanism 100 is equipped with a camera 104 for photographing electronic components. The visual inspection mechanism 100 is driven by the drive mechanism 120. The visual inspection mechanism 100 performs visual inspection of a large number of electronic components using the camera 104. The control device 110 acquires information related to the visual inspection from the visual inspection mechanism 100. The information related to the visual inspection includes image data acquired by the camera 104.

[0015] The control device 110 outputs a command for adjusting the state of the appearance inspection mechanism 100 to the adjustment mechanism 130. The adjustment mechanism 130 adjusts the state of the appearance inspection mechanism 100 according to the command. When the state of the appearance inspection mechanism 100 is adjusted by the adjustment mechanism 130, the abnormality of the appearance inspection mechanism 100 is eliminated.

[0016] The diagnostic device 20 is configured to access the storage device 30. More specifically, the diagnostic device 20 is configured to access the database 31 and the inference model 32. The diagnostic device 20 is not limited to software processing, and may execute various processes using dedicated hardware (electronic circuits). The diagnostic device 20 may include the storage device 30.

[0017] The storage device 30 may be a hard disk drive or a solid state drive or the like. The storage device 30 stores the database 31 and the inference model 32. The database 31 includes the history related to the maintenance of the appearance inspection mechanism 100. The inference model 32 is a model trained by machine learning. The storage device storing the database 31 and the storage device storing the inference model 32 may be different.

[0018] When an abnormality occurs in the appearance inspection mechanism 100, the diagnostic device 20 diagnoses the cause of the abnormality of the appearance inspection mechanism 100. The diagnostic device 20 acquires the data necessary for diagnosis from the inspection device 10. The data necessary for diagnosis includes, for example, the image data acquired by the camera 104. The diagnostic device 20 executes an inference process related to the cause of the abnormality using the inference model 32. The diagnostic device 20 executes the inference process according to the inference program stored in the memory 22. More specifically, the diagnostic device 20 inputs the data necessary for diagnosis into the inference model 32 and acquires the inference result from the inference model 32.

[0019] The diagnostic device 20 transmits the inference result to the inspection device 10 as a diagnostic result. The diagnostic device 20 displays the inference result on the display device 40 as a diagnostic result. The diagnostic result includes the cause of the abnormality of the appearance inspection mechanism 100.

[0020] The control device 110 included in the inspection device 10 receives the diagnostic result. Based on the diagnostic result, the control device 110 identifies the cause of the abnormality in the visual inspection mechanism 100. The control device 110 outputs the necessary commands to the adjustment mechanism 130 to resolve the abnormality. This should resolve the abnormality in the visual inspection mechanism 100.

[0021] If the abnormality in the visual inspection mechanism 100 is not resolved, the diagnostic result may be inaccurate. In this case, the control device 110 sends a message requesting inspection to the tablet terminal 50. The control device 110 is configured to communicate with the tablet terminal 50 via the internet 90.

[0022] The service technician goes to the site where the problematic inspection device 10 is located and adjusts the visual inspection mechanism 100 appropriately. This resolves the abnormality in the visual inspection mechanism 100. The service technician uses a tablet terminal 50 to create a maintenance report. The maintenance report includes the cause of the abnormality and image data of the visual inspection mechanism 100 that characterizes the cause of the abnormality. The tablet terminal 50 accesses the database 31 and updates the maintenance history in the database 31 with the maintenance report. The maintenance history includes the cause of the abnormality and the corresponding feature data.

[0023] The control device 110 comprises a processor 111, a memory 112, and an interface 113. The control device 110 is, for example, a microcomputer. The memory 112 includes an area where a program executed by the processor 111 is stored, and an area where the processor 111 temporarily stores program code, work memory, etc.

[0024] The diagnostic device 20 comprises a processor 21, memory 22, and interface 23. The diagnostic device 20 is, for example, a microcomputer. The memory 22 includes an area where programs executed by the processor 21 are stored, and an area where the processor 21 temporarily stores program code, work memory, etc.

[0025] The processors 21,111 are typically composed of a CPU (Central Processing Unit) or an MPU (Multi-Processing Unit), etc. The processors 21,111 are an example of an arithmetic unit. The memory 22,112 includes volatile memory such as DRAM (dynamic random access memory) and SRAM (static random access memory), ROM (Read Only Memory), and non-volatile memory such as flash memory. The memory 22,112 may also be an SSD (solid state drive) or an HDD (hard disk drive), etc.

[0026] The diagnostic device 20 and the storage device 30 may be housed in a single enclosure. The diagnostic device 20 and the storage device 30 may be configured to communicate via a network such as the Internet.

[0027] [Visual inspection mechanism 100] Figure 2 is a diagram showing an overview of the configuration of the visual inspection mechanism 100. Figure 3 is a magnified view of the area around the alignment guide 103 of the visual inspection mechanism 100.

[0028] The visual inspection mechanism 100 photographs the appearance of the workpiece W to be inspected and checks whether there are any problems with the appearance of the workpiece W. The visual inspection mechanism 100 includes a transport table 101, a parts supply unit 102, an alignment guide 103, a camera 104, a discharge mechanism 105, collection boxes 106A and 106B, and a rotor 108. The visual inspection mechanism 100 is an example of a processing mechanism for handling electronic components.

[0029] The transport table 101 transports the workpieces W to be inspected. The transport table 101 is made of a transparent glass material. The rotor 108 rotates the transport table 101 at a constant speed in the direction of the arrow shown in Figure 2. The parts supply unit 102 is a linear feeder. The parts supply unit 102 has an inclined surface that descends in the direction of the transport table 101. Multiple workpieces W are arranged in a line on the inclined surface of the parts supply unit 102. The parts supply unit 102 supplies the workpieces W to the transport table 101 by vibrating the workpieces W.

[0030] The workpiece W is, for example, a rectangular electronic component with six faces. As shown in Figure 3, a number of workpieces W arranged in a line are sequentially supplied from the component supply unit 102 to the transport table 101. The alignment guide 103 aligns the workpieces W supplied from the component supply unit 102 to the transport table 101 in a line along the transport line of the transport table 101.

[0031] The rotational speed of the transport table 101 is faster than the speed at which the parts supply unit 102 supplies workpieces W to the transport table 101. Therefore, the spacing between workpieces W on the transport table 101 is longer than the spacing between workpieces W on the parts supply unit 102. The spacing between workpieces W on the transport table 101 is determined by the relationship between the speed at which the parts supply unit 102 supplies workpieces W to the transport table 101 and the rotational speed of the transport table 101. Multiple workpieces W supplied to the transport table 101 are arranged in a line at a constant interval and rotate together with the transport table 101 around the rotor 108.

[0032] The transport table 101 is equipped with six cameras 104 to capture images of the six sides of the workpiece W. Cameras 104 are image sensing cameras and process images on a pixel-by-pixel basis. Cameras 104 perform image binarization. However, cameras 104 may perform other processing methods instead of binarization. Cameras 104 have a function to automatically adjust focus. Cameras 104 are mounted on brackets. Cameras 104 have a function to automatically adjust their shooting position within a range of, for example, about 20 millimeters.

[0033] The camera 104 includes a lens 104A directed at the subject and a lighting unit 104B that illuminates the area around the subject. The lighting unit 104B has an automatic dimming function and includes an LED. The camera 104 transmits the acquired image data to the control device 110 (see Figure 1). The control device 110 stores the acquired image data. Based on the image data, the control device 110 inspects the appearance of the workpiece W and determines whether the workpiece is good or defective.

[0034] The discharge mechanism 105 drops the inspected workpiece W into either the recovery box 106A or 106B. The discharge mechanism 105 guides the workpiece W to either the recovery box 106A or 106B by spraying air onto it. The control device 110 controls the discharge mechanism 105 so that good products are guided to recovery box 106A and defective products are guided to recovery box 106B.

[0035] An adjustment mechanism 130 is located near the visual inspection mechanism 100. The control device 110 controls the adjustment mechanism 130 to adjust the visual inspection mechanism 100 if an abnormality occurs in the visual inspection mechanism 100. The adjustment mechanism 130 includes, for example, a glass cleaning unit. The glass cleaning unit is provided on the route where the workpiece W is placed. The glass cleaning unit is fitted with a cloth for cleaning the glass surface of the transport table 101. The cloth may have detergent attached to it.

[0036] The control device 110, for example, controls the glass cleaning unit to clean any dirt adhering to the glass surface of the transport table 101. The control device 110 stops supplying workpieces W to the transport table 101 before cleaning the glass surface of the transport table 101. After confirming that there are no workpieces W on the transport table 101, the control device 110 cleans the glass surface of the transport table 101.

[0037] In addition to the glass cleaning unit, the adjustment mechanism 130 includes an illumination adjustment unit for adjusting the illumination of the lighting unit 104B, a camera adjustment mechanism for adjusting focus problems or misalignment of the camera 104, a parts supply unit adjustment mechanism for adjusting misalignment of the parts supply unit 102, and an alignment guide adjustment mechanism for adjusting abnormalities in the alignment guide 103.

[0038] [Machine learning with inference model 32] Figure 4 is a block diagram illustrating the machine learning of the inference model 32. As shown in Figure 4, the inference model 32 is trained by supervised learning in the training unit 61 of the computer 60. For training, for example, a deep learning method is used. In general, deep learning provides an input layer, an intermediate layer, and an output layer in the inference model. Training data is input to the input layer. It is preferable to have multiple intermediate layers. For example, it is preferable to have three or more intermediate layers. The output layer is a layer that outputs the results from the intermediate layers. In this embodiment, the output layer outputs the estimated results regarding the cause of the anomaly. In this embodiment, for example, the error is calculated by a loss function, and the inference result is optimized by gradient descent.

[0039] The training unit 61 consists of the hardware and software of the computer 60. The training unit 61 accepts feature data and ground truth data as training data used for machine learning of the inference model 32. The feature data is, for example, image data acquired by the camera 104. The ground truth data is the cause of anomalies in the visual inspection mechanism 100 that corresponds to the feature data.

[0040] The correct data, as shown in Figure 4, includes (a) dirt (scratches, foreign matter) on lens 104A, (b) poor focus, (c) misalignment of camera 104, (d) poor illumination of lighting unit 104B, (e) dirt (scratches, foreign matter) on transport table 101, (f) misalignment of parts supply unit 102, and (g) abnormality of alignment guide 103 (misalignment of imaging position).

[0041] For example, if dirt adheres to the lens 104A, it is expected that noise corresponding to the dirt will appear in the image acquired by the camera 104 at a specific location. If a focus problem occurs, the image acquired by the camera 104 will be blurred.

[0042] The training unit 61 receives a large dataset consisting of the training data shown in Figure 4 and trains the inference model 32. The large dataset includes sets of the above-mentioned ground truth data (a) and the corresponding feature data for the ground truth data (a), sets of the above-mentioned ground truth data (b) and the corresponding feature data for the ground truth data (b), sets of the above-mentioned ground truth data (c) and the corresponding feature data for the ground truth data (c), sets of the above-mentioned ground truth data (d) and the corresponding feature data for the ground truth data (d), sets of the above-mentioned ground truth data (e) and the corresponding feature data for the ground truth data (e), sets of the above-mentioned ground truth data (f) and the corresponding feature data for the ground truth data (f), and sets of the above-mentioned ground truth data (g) and the corresponding feature data for the ground truth data (g). The trained inference model 32 is stored in the memory device 30.

[0043] "The above set of ground truth data (a) and the corresponding feature data for ground truth data (a)" is an example of the first training data. "The above set of ground truth data (c) and the corresponding feature data for ground truth data (c)" is an example of the second training data. "The above set of ground truth data (d) and the corresponding feature data for ground truth data (d)" is an example of the third training data.

[0044] The storage device 30 stores the database 31. The database 31 contains the maintenance history of the visual inspection mechanism 100. When the amount of maintenance history updates in the database 31 reaches a predetermined amount, the diagnostic device 20 uses the updated maintenance history as training data to further train the inference model 32.

[0045] Thus, the inference model 32 is trained to infer the cause of anomalies in the visual inspection mechanism 100 based on the training data. More specifically, the inference model 32 is trained to infer the cause of anomalies in the visual inspection mechanism 100 based on training data that includes feature data (image data) and ground truth data.

[0046] [Processing by inspection device 10] Figure 5 is a flowchart showing the processing procedure of the inspection device 10. The contents of the inspection device 10 will be explained below according to the flowchart. Note that the processing based on this flowchart is mainly performed by the control device 110.

[0047] First, the inspection device 10 performs an inspection process (step S11). The inspection process includes taking a photograph of the workpiece W with the camera 104, inspecting the six sides of the workpiece W based on the image data acquired by the photograph, determining whether the workpiece W is a good product or a defective product, saving the determination result to the processing database in the control device 110, and guiding the workpiece W to either the collection box 106A or the collection box 106B according to the determination result. Users often perform inspections of workpiece W in lot units. One lot contains multiple workpieces W that conform to product standards. Typically, the product standards for workpiece W differ from lot to lot.

[0048] Next, the inspection device 10 determines whether the defect rate of the workpiece W exceeds a threshold (step S12). The inspection device 10 calculates the defect rate by calculating the number of defective items relative to the number of workpieces W inspected. The threshold for the defect rate is set by the designer. The inspection device 10 may accept user input to change the threshold for the defect rate.

[0049] If the defect rate of the workpiece W does not exceed a threshold, the inspection device 10 continues the inspection process. If the defect rate of the workpiece W exceeds a threshold, the inspection device 10 stops the inspection process (step S13). Next, the inspection device 10 instructs the diagnostic device 20 to diagnose the cause of the abnormality (step S14). The diagnostic instruction includes image data acquired by the six cameras 104 when the defect rate of the workpiece W exceeds a threshold. The diagnostic device 20 diagnoses the cause of the abnormality using the image data. The inspection device 10 may also include image data acquired before the point in time when the defect rate of the workpiece W exceeded a threshold in the diagnostic instruction.

[0050] The image data transmitted in step S14 is an example of feature data based on information acquired from the visual inspection mechanism. The image data is data based on images captured by six cameras 104 mounted on the visual inspection mechanism 100. The inspection device 10 may transmit the images captured by the six cameras 104 themselves as image data to the diagnostic device 20. Alternatively, the inspection device 10 may transmit image data to the diagnostic device 20 after applying a predetermined process to each of the images captured by the six cameras 104. The predetermined process may be a process to sharpen the image or a process to enhance the edges of the image.

[0051] Upon receiving the command, the diagnostic device 20 estimates the cause of the abnormality using image data and transmits the cause of the abnormality with the highest confidence score to the inspection device 10. The inspection device 10 receives the diagnostic result (cause of the abnormality) (step S15). The inspection device 10 has a function to identify the abnormal location based on the diagnostic result and to adjust the identified abnormal location. Before adjusting the abnormal location, the inspection device 10 determines whether the previous diagnostic result and the current diagnostic result are the same (step S16).

[0052] If the previous diagnostic result and the current diagnostic result are not the same, the inspection device 10 adjusts the abnormal part using the adjustment mechanism 130 (step S17). For example, if the cause of the abnormality is dirt on the transport table 101, the inspection device 10 controls a glass cleaning unit, which is an example of the adjustment mechanism 130, to clean the dirt attached to the transport table 101. Alternatively, if the cause of the abnormality is dirt on the lens 104A, the inspection device 10 controls a lens cleaning unit, which is an example of the adjustment mechanism 130, to clean the dirt attached to the lens 104A.

[0053] If the cause of the malfunction is insufficient illumination, the inspection device 10 controls the illumination adjustment unit, which is an example of the adjustment mechanism 130, to adjust the illumination of the lighting unit 104B, which has inappropriate illumination. If the cause of the malfunction is poor focus or misalignment of the camera 104, the inspection device 10 controls the camera adjustment mechanism, which is an example of the adjustment mechanism 130, to adjust the state of the camera 104, which is experiencing poor focus or misalignment, to an appropriate state.

[0054] In this manner, the inspection device 10 adjusts the abnormal area using the adjustment mechanism 130 according to the diagnostic result (cause of the abnormality). Next, the inspection device 10 resumes the inspection process (step S18).

[0055] If the current diagnostic result is the same as the previous one, it means that the abnormality has not been resolved despite the adjustments made to the faulty parts according to the previous diagnostic result. Therefore, if the current diagnostic result is the same as the previous one, the inspection device 10 requests an inspection from a service technician (step S19). More specifically, the inspection device 10 sends an inspection request message to the tablet terminal 50 carried by the service technician, and completes the process according to this flowchart.

[0056] The service technician goes to the site where the malfunctioning inspection device 10 is located and adjusts the visual inspection mechanism 100 appropriately. The service technician creates a maintenance report using a tablet terminal 50 (see Figure 1). The tablet terminal 50 accesses the database 31 (see Figure 1) and updates the maintenance history in the database 31 with the maintenance report.

[0057] [Processing by diagnostic device 20] Figure 6 is a flowchart showing the processing procedure of the diagnostic device 20. The contents of the diagnostic device 20 will be explained below according to the flowchart. First, the diagnostic device 20 determines whether or not it has received a diagnostic command from the inspection device 10 (step S21). The diagnostic command includes image data.

[0058] When the diagnostic device 20 receives a diagnostic command from the inspection device 10, it inputs the image data included in the command into the inference model 32 (step S22). Based on the image data, the inference model 32 infers the cause of the abnormality in the visual inspection mechanism 100. The inference model 32 outputs the inference result.

[0059] The inference model 32 is trained to infer multiple anomaly causes, as shown in Figure 4. The inference results output by the inference model 32 include confidence scores for each anomaly cause.

[0060] Next, the diagnostic device 20 obtains inference results from the inference model 32 (step S23). The diagnostic device 20 determines the confidence score corresponding to each cause of abnormality (step S24). Next, the diagnostic device 20 transmits the cause of abnormality with the highest confidence score to the inspection device 10 (step S25).

[0061] Next, the diagnostic device 20 displays the cause of the abnormality where the confidence score exceeds the threshold on the display device 40 along with the confidence score (step S26), and completes the process based on this flowchart. The threshold for the confidence score is set by the designer. The diagnostic device 20 may accept user input to change the threshold for the confidence score.

[0062] In this example, we have described how to display anomaly causes whose confidence score exceeds a threshold, along with their confidence scores, on the display device 40. However, the diagnostic device 20 may also display all anomaly causes that are being inferred in the inference model 32, along with their confidence scores, on the display device 40.

[0063] If the diagnostic device 20 has not received a diagnostic command from the inspection device 10, it determines whether the database 31 has been updated (step S27). The database 31 is updated by a tablet terminal 50 carried by a service technician. When the database 31 is updated, new maintenance history is added to the database 31. The maintenance history includes the cause of the abnormality and image data of the visual inspection mechanism 100 that characterizes the cause of the abnormality.

[0064] If database 31 has not been updated, the process based on this flowchart ends. If database 31 has been updated, the diagnostic device 20 determines whether the amount of updates has reached a specified amount (step S28). If the amount of updates has reached a specified amount, the diagnostic device 20 trains the inference model 32 using the maintenance history as training data (step S29).

[0065] In this way, the diagnostic device 20 trains the inference model 32 using the maintenance history so that the inference model 32 can infer the cause of the anomaly. The maintenance history may contain information about the cause of the anomaly that is not included in the inference target of the inference model 32. Such anomalies are examples of third-order anomalies. If the maintenance history contains a third-order anomaly and feature data corresponding to the third-order anomaly, the diagnostic device 20 trains the inference model 32 using the maintenance history so that the inference model 32 can infer the third-order anomaly. This improves the diagnostic accuracy of the diagnostic device 20.

[0066] [Example of display on display device 40] Figure 7 shows an example of the diagnostic results displayed on the display device 40. As shown in Figure 7, the display device 40 displays the cause of the abnormality in the inspection device 10 along with a confidence score. In particular, Figure 7 shows an example in which the causes of abnormality are displayed from top to bottom of the screen in the order of the cause with the highest confidence score, the second highest confidence score, and the third highest confidence score. By looking at the display device 40, the user can identify the most likely cause of the abnormality.

[0067] Figure 7 shows "lens contamination" as the cause of the anomaly with the highest confidence score, and "poor focus" as the cause of the anomaly with the second highest confidence score. The cause of the anomaly with the highest confidence score is an example of the first cause of the anomaly, and the cause of the anomaly with the second highest confidence score is an example of the second cause of the anomaly.

[0068] Alternatively, instead of displaying the numerical value of the confidence score on the display device 40, a symbol indicating the level of the confidence score may be displayed on the display device 40.

[0069] [Differentiation] Figure 8 shows an overview of the configuration of the electrical characteristics inspection mechanism 100A related to a modified example. In this embodiment, the electrical characteristics inspection mechanism 100A shown in Figure 8 may be used as the inspection device 10 instead of the visual inspection mechanism 100.

[0070] The electrical characteristics inspection mechanism 100A inspects the electrical characteristics of the workpiece Wa to be inspected. The electrical characteristics inspection mechanism 100A comprises a transport table 101A, a component supply unit 102A, measuring devices 1011-1013, a storage section 1015, a discharge mechanism 105A, recovery boxes 106A and 106B, and a rotor 108A. The electrical characteristics inspection mechanism 100A is an example of a processing mechanism for processing electronic components.

[0071] The rotor 108A rotates the transport table 101A at a constant speed in the direction of the arrow shown in Figure 8. The component supply unit 102A supplies workpieces Wa to the transport table 101A. Workpieces Wa are, for example, capacitors, which are an example of electronic components. A large number of workpieces Wa arranged in a row are sequentially supplied from the component supply unit 102A to the transport table 101. The workpieces Wa supplied to the transport table 101 are stored in the storage compartments 1015. One workpiece Wa is stored in each storage compartment 1015.

[0072] The workpiece Wa stored in the storage section 1015 is guided to measuring devices 1011, 1012, and 1013 by the rotation of the transport table 101A. Each of the measuring devices 1011 to 1013 is equipped with measuring terminals T1 and T2. Each of the measuring devices 1011 to 1013 measures the electrical characteristics of the workpiece Wa by bringing measuring terminal T1 into contact with the positive electrode of the workpiece Wa, and measuring terminal T2 into contact with the negative electrode of the workpiece Wa.

[0073] Measuring device 1011 measures the capacitance of workpiece Wa. Measuring devices 1012 and 1013 are, for example, voltage application units. Measuring device 1012 measures the resistance value to determine whether or not there is a short circuit fault in workpiece Wa. Measuring device 1013 measures the insulation resistance value of workpiece Wa.

[0074] The control device 110 determines whether the capacitance of workpiece Wa meets the standard based on the measurement value from measuring device 1011. The control device 110 determines whether there is a short circuit fault in workpiece Wa based on the measurement value from measuring device 1012. The control device 110 determines whether the insulation resistance value of workpiece Wa meets the product standard based on the measurement value from measuring device 1013.

[0075] The control device 110 determines a workpiece Wa is a good product if its capacitance meets the product standard, no short-circuit faults have occurred, and its insulation resistance value meets the product standard. The control device 110 determines a workpiece Wa is a defective product if its capacitance does not meet the product standard, a short-circuit fault has occurred, or its insulation resistance value does not meet the product standard. The control device 110 controls the discharge mechanism 105 so that good products are guided to the collection box 106A and defective products are guided to the collection box 106B.

[0076] The control device 110 controls the discharge mechanism 105A so that good products are guided to the collection box 106A and defective products are guided to the collection box 106B.

[0077] Near the electrical characteristics inspection mechanism 100A related to the modified configuration, an adjustment mechanism 130A corresponding to the adjustment mechanism 130 is located. If an abnormality occurs in the electrical characteristics inspection mechanism 100A, the control device 110 controls the adjustment mechanism 130A and performs control to resolve the cause of the abnormality. The adjustment mechanism 130A may include, for example, a cleaning unit for cleaning the measurement terminals T1 and T2 of the measuring devices 1011 to 1013.

[0078] [Machine learning of inference model 32A related to the modified example] Figure 9 is a block diagram illustrating the machine learning of the inference model 32A related to the modified example. When the electrical characteristic inspection mechanism 100A related to the modified example is applied to the inspection device 10, the inference model 32A is installed in the storage device 30 of the electronic component processing system 1 instead of the inference model 32.

[0079] As shown in Figure 9, the inference model 32A is trained by supervised learning in the training unit 61 of the computer 60. The training unit 61 accepts feature data and ground truth data as training data to be used for machine learning of the inference model 32A.

[0080] Feature data may include, for example, the unmeasured rate, the filling rate, and the capacitance distribution. The unmeasured rate indicates the ratio of the number of workpieces Wa whose insulation resistance could not be measured to the total number of workpieces Wa targeted for measurement by the measuring device 1013. The capacitance distribution indicates the distribution of capacitance of workpieces Wa (capacitors) obtained when inspecting workpieces Wa in lot units.

[0081] The filling rate indicates the ratio of the number of workpieces Wa transported from the parts supply unit 102A to the transport table 101A to the number of workpieces Wa transported to the transport table 101A to the storage unit 1015. If the filling rate is not 100%, it means that some of the workpieces Wa transported to the transport table 101A were not stored in the storage unit 1015 due to some factor. For example, if the storage unit 1015 is damaged or clogged with debris, the workpieces Wa will not be stored in the storage unit 1015. Alternatively, if the transport table 101A experiences significant vibration, the workpieces Wa may not be stored in the storage unit 1015.

[0082] The inspection device 10 has a function to calculate the unmeasured rate, filling rate, and capacitance distribution based on the information output from the electrical characteristics inspection mechanism 100A when the workpiece Wa is being inspected by the electrical characteristics inspection mechanism 100A.

[0083] The correct data represents the cause of the malfunction in the 100A electrical characteristics inspection mechanism, corresponding to the feature data.

[0084] The correct data, as shown in Figure 9, includes (a1) contamination of the measurement terminals T1 and T2 (including wear and damage, etc.), (b1) misalignment of the transport position (including damage to a specific position of the rotor), (c1) abnormalities in the parts supply unit 102A, and (d1) lot contamination.

[0085] Normally, measuring devices 1011-1013 contact the external electrodes at both ends of the workpiece Wa with measuring terminals T1 and T2. When measuring devices 1011-1013 process a large number of workpieces, dirt may accumulate on the measuring terminals T1 and T2, or the measuring terminals T1 and T2 may wear down.

[0086] When the visual inspection mechanism 100 is the target, the feature data corresponding to each of the ground truth data (a) to (g) is image data. In contrast, when the electrical characteristic inspection mechanism 100A is the target, the feature data corresponding to each of the ground truth data (a1) to (d1) is not uniform.

[0087] The feature data corresponding to the ground truth data (a1) is the rate of unmeasured data. The ground truth data (a1) is applied, for example, when workpieces Wa are supplied to the transport table 101 in a single column, as shown in Figure 8.

[0088] In the modified example, the transport table 101 may be configured to supply workpieces Wa in two rows. In this case, one storage section 1015 accommodates both the first row of workpieces Wa and the second row of workpieces Wa. The measuring device 1013 measures both the first row of workpieces Wa and the second row of workpieces Wa simultaneously. If the positions of the measuring terminals T1 and T2 of the measuring device 1013 and the terminals of the two workpieces Wa housed in the storage section 1015 are misaligned from the appropriate positions, neither the first row of workpieces Wa nor the second row of workpieces Wa can be measured. In such a case, "(e1) misalignment of measuring terminals T1 and T2" may be further adopted as correct data.

[0089] In such cases, the inference model 32 may be trained using a large number of training datasets, each consisting of a set of ground truth data (e1) and corresponding feature data. The feature data corresponding to the ground truth data (e1) is the unmeasured rate, similar to the case of ground truth data (a1).

[0090] The feature data corresponding to the correct data (b1) and (c1) are the packing rates. The feature data corresponding to the correct data (d1) is the capacitance distribution. In the inspection device 10, a large number of capacitors are inspected by lot. Normally, the capacitance of capacitors differs from lot to lot. Therefore, the capacitance of capacitors included in a single lot is generally consistent. For example, lot A may contain 10 picofarad capacitors, and lot B may contain 50 picofarad capacitors. However, due to some factor, a 50 picofarad capacitor may be mixed into lot A. This situation is called "lot mixing".

[0091] The training unit 61 receives a large dataset consisting of the training data shown in Figure 9 and trains the inference model 32. The large dataset includes sets of the above-mentioned ground truth data (a1) and the corresponding feature data, sets of the above-mentioned ground truth data (b1) and the corresponding feature data, sets of the above-mentioned ground truth data (c1) and the corresponding feature data, and sets of the above-mentioned ground truth data (d1) and the corresponding feature data. The large dataset may further include sets of the above-mentioned ground truth data (e1) and the corresponding feature data. The trained inference model 32A is stored in the memory device 30.

[0092] "The above set of ground truth data (a1) and the corresponding feature data for ground truth data (a1)" is an example of the fourth training data. "The above set of ground truth data (e1) and the corresponding feature data for ground truth data (e1)" is an example of the fifth training data.

[0093] Thus, the inference model 32A is trained to infer the cause of an anomaly in the electrical characteristic inspection mechanism 100A based on the training data. More specifically, the inference model 32A is trained to infer the cause of an anomaly in the electrical characteristic inspection mechanism 100A based on training data that includes feature data and ground truth data.

[0094] In this embodiment, it is assumed that the electrical characteristics inspection mechanism 100A, which is a modified example, is used instead of the visual inspection mechanism 100 of the electronic component processing system 1 shown in Figure 1. In other words, in this embodiment, it is assumed that the system configuration described using Figures 1 to 7 will also apply when the electrical characteristics inspection mechanism 100A is used instead of the visual inspection mechanism 100.

[0095] [Aspect] Those skilled in the art will understand that each of the embodiments described above is a specific example of the following embodiments.

[0096] (Section 1) An electronic component processing system according to one embodiment is an electronic component processing system for processing electronic components, comprising: a processing mechanism for processing electronic components; a control device for controlling the processing mechanism; a storage device for storing an inference model for inferring the cause of an abnormality in the processing mechanism; a diagnostic device for diagnosing the cause of an abnormality using the inference model; and a display device for displaying the cause of an abnormality based on the diagnosis by the diagnostic device, wherein the inference model is trained to infer the cause of an abnormality based on training data, the training data includes the cause of an abnormality and feature data indicating the feature quantities corresponding to the cause of the abnormality, the control device outputs feature data based on information obtained from the processing mechanism to the diagnostic device, and the diagnostic device obtains the cause of an abnormality from the inference model by inputting the feature data to the inference model.

[0097] (Article 2) The electronic component processing system referred to in Article 1, wherein the processing mechanism is an appearance inspection mechanism for inspecting the appearance of electronic components, or an electrical characteristics inspection mechanism for inspecting the electrical characteristics of electronic components.

[0098] (Article 3) The electronic component processing system of Article 2, wherein the processing mechanism is an appearance inspection mechanism, the appearance inspection mechanism includes a camera for photographing electronic components, the camera has a lens and an illumination device, the feature data includes image data acquired by the camera, the training data includes first training data, second training data, and third training data, the first training data includes first data indicating that the cause of the abnormality of the appearance inspection mechanism is lens contamination and image data corresponding to the first data, the second training data includes second data indicating that the cause of the abnormality of the appearance inspection mechanism is camera misalignment and image data corresponding to the second data, and the third training data includes third data indicating that the cause of the abnormality of the appearance inspection mechanism is illumination device and image data corresponding to the third data.

[0099] (Clause 4) The electronic component processing system of paragraph 2, wherein the processing mechanism is an electrical characteristic inspection mechanism, the electrical characteristic inspection mechanism includes a measuring device for measuring the electrical characteristic values ​​of electronic components, the measuring device has measuring terminals that are in contact with the electronic components, the feature data includes unmeasured rate data indicating the rate at which the measuring device could not measure the electrical characteristic values ​​of electronic components, the training data includes fourth training data and fifth training data, the fourth training data includes fourth data indicating that the cause of the abnormality of the electrical characteristic inspection mechanism is wear or dirt on the measuring terminals and unmeasured rate data corresponding to the fourth data, and the fifth training data includes fifth data indicating that the cause of the abnormality of the electrical characteristic inspection mechanism is misalignment of the measuring terminals and unmeasured rate data corresponding to the fifth data.

[0100] (Article 5) An electronic component processing system according to any one of Articles 1 to 4, wherein the cause of an abnormality includes a first cause of abnormality and a second cause of abnormality, the diagnostic device obtains a first confidence level from an inference model in which the cause of abnormality is estimated to be the first cause of abnormality and a second confidence level in which the cause of abnormality is estimated to be the second cause of abnormality, and the display device displays the first cause of abnormality and the second cause of abnormality so that the user can identify the difference between the first confidence level and the second confidence level when the diagnostic result of the diagnostic device includes both the first cause of abnormality and the second cause of abnormality.

[0101] (Article 6) An electronic component processing system according to any one of Articles 1 to 5, further comprising an adjustment mechanism for adjusting the processing mechanism, wherein the diagnostic device outputs the cause of the abnormality to the control device, the control device outputs a command to the adjustment mechanism to resolve the abnormality in the processing mechanism corresponding to the cause of the abnormality, and the adjustment mechanism adjusts the processing mechanism based on the command.

[0102] (Clause 7) In the electronic component processing system described in paragraph 6, if the abnormality in the processing mechanism corresponding to the cause of the abnormality is not resolved, the control device shall notify a terminal device carried by a service technician.

[0103] (Clause 8) The electronic component processing system of paragraph 7, wherein the diagnostic device is configured to access a database in which maintenance history entered by a service technician is stored, and the maintenance history includes the cause of an anomaly and characteristic data corresponding to the cause of the anomaly.

[0104] (Clause 9) The electronic component processing system of Clause 8, wherein the diagnostic device trains the inference model using the maintenance history so that the inference model infers the cause of an anomaly.

[0105] (Clause 10) In the electronic component processing system of Clause 9, if the cause of an anomaly includes a third cause of an anomaly that is not included in the inference target of the inference model, and the maintenance history includes the third cause of an anomaly and feature data corresponding to the third cause of an anomaly, the diagnostic device trains the inference model using the maintenance history so that the inference model infers the third cause of an anomaly.

[0106] The embodiments disclosed herein should be considered in all respects to be illustrative and not restrictive. The scope of this disclosure is indicated by the claims rather than by the description of the embodiments above, and all modifications within the meaning and scope of the claims are intended to be included. [Explanation of Symbols]

[0107] 1 Electronic component processing system, 10 Inspection device, 20 Diagnostic device, 21, 111 Processor, 22, 112 Memory, 23, 113 Interface, 30 Storage device, 31 Database, 32, 32A Inference model, 40 Display device, 50 Tablet terminal, 60 Computer, 61 Training unit, 90 Internet, 100 Visual inspection mechanism, 100A Electrical characteristics inspection mechanism, 101, 101A Transport table, 102, 102A Parts supply unit, 103 Alignment guide, 104 Camera, 104A Lens, 104B Lighting unit, 105, 105A Discharge mechanism, 106A, 106B Recovery box, 108 Rotor, 110 Control device, 120 Drive mechanism, 130 Adjustment mechanism, 1011-1013 Measuring device, 1015 Storage unit, T1, T2 Measurement terminals, W, Wa workpiece.

Claims

1. An electronic component processing system for processing electronic components, A processing mechanism for processing the aforementioned electronic components, A control device for controlling the processing mechanism, A storage device that stores an inference model for inferring the cause of an abnormality in the processing mechanism, A diagnostic device that diagnoses the cause of the abnormality using the aforementioned inference model, The system includes a display device that displays the cause of the abnormality based on the diagnosis made by the diagnostic device, The inference model is trained to infer the cause of the anomaly based on the training data. The training data includes the cause of the anomaly and feature data indicating the feature corresponding to the cause of the anomaly. The control device outputs the feature data based on the information obtained from the processing mechanism to the diagnostic device. The diagnostic device is an electronic component processing system that obtains the cause of the abnormality from the inference model by inputting the feature data into the inference model.

2. The electronic component processing system according to claim 1, wherein the processing mechanism is an appearance inspection mechanism for inspecting the appearance of the electronic component, or an electrical characteristics inspection mechanism for inspecting the electrical characteristics of the electronic component.

3. The processing mechanism is the visual inspection mechanism, The aforementioned visual inspection mechanism includes a camera for photographing the electronic components, The aforementioned camera has a lens and an illumination device, The feature data includes image data acquired by the camera, The aforementioned training data includes first training data, second training data, and third training data. The first training data includes first data indicating that the cause of the abnormality in the visual inspection mechanism is dirt on the lens, and image data corresponding to the first data. The second training data includes second data indicating that the cause of the abnormality in the visual inspection mechanism is a misalignment of the camera, and the image data corresponding to the second data. The electronic component processing system according to claim 2, wherein the third training data includes third data indicating that the cause of the abnormality in the visual inspection mechanism is the lighting device and the image data corresponding to the third data.

4. The processing mechanism is the electrical characteristic inspection mechanism, The electrical characteristic testing mechanism includes a measuring device for measuring the electrical characteristic values ​​of the electronic component. The measuring device has measuring terminals that make contact with the electronic component, The feature data includes unmeasured rate data indicating the rate at which the measuring device could not measure the electrical characteristic value of the electronic component. The aforementioned training data includes the fourth training data and the fifth training data. The fourth training data includes fourth data indicating that the cause of the abnormality in the electrical characteristics inspection mechanism is wear or dirt on the measurement terminals, and the unmeasured rate data corresponding to the fourth data. The electronic component processing system according to claim 2, wherein the fifth training data includes fifth data indicating that the cause of the abnormality in the electrical characteristic inspection mechanism is a misalignment of the measurement terminals, and the unmeasured rate data corresponding to the fifth data.

5. The aforementioned cause of abnormality includes the first cause of abnormality and the second cause of abnormality, The diagnostic device obtains a first confidence level from the inference model that the cause of the abnormality is estimated to be the first cause of the abnormality, and a second confidence level that the cause of the abnormality is estimated to be the second cause of the abnormality. The electronic component processing system according to any one of claims 1 to 4, wherein the display device displays the first abnormal cause and the second abnormal cause so that the user can identify the difference between the first reliability and the second reliability when the diagnostic result of the diagnostic device includes the first abnormal cause and the second abnormal cause.

6. The processing mechanism further comprises an adjustment mechanism for adjusting the aforementioned processing mechanism, The diagnostic device outputs the cause of the abnormality to the control device. The control device outputs a command to the adjustment mechanism to resolve the abnormality in the processing mechanism corresponding to the cause of the abnormality, The electronic component processing system according to any one of claims 1 to 4, wherein the adjustment mechanism adjusts the processing mechanism based on the command.

7. The electronic component processing system according to claim 6, wherein the control device notifies a terminal device carried by a service technician if the abnormality in the processing mechanism corresponding to the cause of the abnormality is not resolved.

8. The diagnostic device is configured to access a database that stores maintenance history entered by the service technician. The electronic component processing system according to claim 7, wherein the maintenance history includes the cause of the abnormality and the feature data corresponding to the cause of the abnormality.

9. The electronic component processing system according to claim 8, wherein the diagnostic device trains the inference model using the maintenance history so that the inference model infers the cause of the abnormality.

10. The aforementioned cause of an anomaly includes a third cause of an anomaly that is excluded from the inference target of the inference model, If the maintenance history includes the third cause of an anomaly and the feature data corresponding to the third cause of an anomaly, the diagnostic device trains the inference model using the maintenance history so that the inference model infers the third cause of an anomaly, according to claim 9.

Images