Inspection management system, inspection management device, teacher data generation method, and program

Abstract

To provide a technique capable of generating teacher data that is able to improve the accuracy of an inspection model used for intermediate inspection in a manufacturing line of a product having a plurality of manufacturing processes.SOLUTION: An inspection management system is used in a manufacturing line having a plurality of manufacturing processes related to manufacture of a product and also having a plurality of manufacturing devices corresponding to the plurality of processes and an inspection device that performs an inspection based on a captured image of the product. The inspection management system includes teacher data generation means for generating teacher data used for an inspection model corresponding to each of intermediate inspections, by using images used in the respective intermediate inspections related to one or more intermediate manufacturing processes in the manufacturing processes, and a final inspection result that is a result of a final inspection related to a final manufacturing process of the product.SELECTED DRAWING: Figure 2

Description

Technical Field

[0001] The present invention relates to an inspection management system, an inspection management device, a teacher data generation method, and a program.

Background Art

[0002] In a product manufacturing line, an inspection device for the product is arranged in the middle process or the final process of the line, and inspections such as defect detection and sorting of defective products are performed using images taken of the product. For example, in a manufacturing line of a component mounting substrate, generally, a process of printing solder paste on a printed wiring board (printing process), a process of mounting components on the substrate on which the solder paste is printed (mounting process), and a process of heating the substrate after component mounting to solder the components to the substrate (reflow process) are included, and it is known to perform image inspection after each process (for example, Patent Documents 1 to 3, etc.).

[0003] Also, when performing inspection after each of the above processes (especially the intermediate process), it is also conceivable to perform automatic inspection using an inspection model (including not only a learned inference model but also a rule-based pass / fail determination program, etc. The same applies hereinafter). However, if the accuracy of the inspection model is low, an oversight may occur in which a product that is actually a good product is regarded as a defective product in the inspection, or a miss may occur in which a product that is actually a defective product is regarded as a good product. Therefore, it is necessary to perform learning of the inspection model (rule creation in the case of rule-based, the same applies hereinafter) so that appropriate pass / fail determination can be made in each process.

[0004] In this regard, conventionally, for an image of a product determined to be a good product in an automatic inspection in an intermediate process (for example, either the printing process or the mounting process in a manufacturing line of a component mounting substrate), a correct label of a good product is assigned, and for an image of a product determined to be defective in the automatic inspection, the result of a visual reinspection by an operator is assigned as the correct label, thereby creating teacher data used for an inspection model for the target intermediate inspection. Hereinafter, an image of a product used in the intermediate inspection is also referred to as an intermediate inspection image, a sample image, or the like.

[0005] Originally, products that were judged to be non-defective in the intermediate inspection (including visual inspection) but were actually defective products in the final inspection (also judged to be defective in visual inspection) should have been judged to be defective in the intermediate inspection. However, according to the conventional method, the sample images of such products were regarded as teacher data of non-defective products in the intermediate process. Similarly, products that were judged to be defective in the intermediate inspection but became non-defective products in the final inspection should have been judged to be non-defective in the intermediate inspection. However, according to the conventional method, the sample images of such products were regarded as teacher data of defective products in the intermediate process.

[0006] That is, in the conventional method, there was a problem that inappropriate data with incorrect correct labels was mixed in the teacher data used for generating the inspection model, making it difficult to generate an inspection model with high accuracy.

Prior Art Documents

Patent Documents

[0007]

Patent Document 1

Patent Document 2

Patent Document 3

Summary of the Invention

Problems to be Solved by the Invention

[0008] The present invention has been made in view of the above circumstances, and an object thereof is to provide a technique capable of generating teacher data that can improve the accuracy of an inspection model used for intermediate inspection in a production line of products having a plurality of manufacturing processes.

Means for Solving the Problems

[0009] In order to achieve the above object, the present invention adopts the following configuration. That is, An inspection management system used in a production line having a plurality of manufacturing processes related to the production of a product, a plurality of manufacturing apparatuses corresponding to the plurality of processes, and an inspection apparatus that performs inspection based on an image obtained by photographing the product, teacher data generation means for generating teacher data used for an inspection model corresponding to each of the intermediate inspections, using the images used in each of the intermediate inspections related to one or more intermediate manufacturing processes in the manufacturing process and the final inspection result which is the result of the final inspection related to the final manufacturing process of the product; An inspection management system having

[0010] In this specification, "intermediate manufacturing process" means a manufacturing process other than the final manufacturing process, including the first manufacturing process. Also, "intermediate inspection" is an inspection of intermediate products less than the final product, including the first inspection. Also, "inspection" includes not only automatic inspection by a program but also visual inspection by an operator. Also, in this specification, when simply referred to as "product", it refers to products at all stages of the manufacturing process, including both final products and intermediate products. Also, "inspection model" includes not only a learned inference model (so-called AI) by machine learning and manual learning, but also a rule-based pass / fail judgment program and the like. That is, "used for an inspection model" includes both being used as a learning dataset for an inference model and being used as reference data for rule-based rule construction.

[0011] With such a configuration, teacher data reflecting the inspection result in the final process can be created, and by using such teacher data, the accuracy of the inspection model used for intermediate inspection can be improved.

[0012] Further, when the final inspection result is a pass judgment, the teacher data generation means may assign a correct label of good to the image used in the intermediate inspection that is the target of the teacher data generation, and when the final inspection result is a fail judgment, assign a correct label of defective to the image used in the intermediate inspection that is the target of the teacher data generation.

[0013] According to such a configuration, regardless of the result of the intermediate inspection, the result of the final inspection can be used as the correct label, that is, appropriate teacher data in which omissions and oversights are corrected can be automatically created. Note that when learning (or creating) the inspection model, it is also possible to use only either teacher data with a correct label for good products or teacher data with a correct label for defective products.

[0014] Further, the teacher data generation means acquires repair presence / absence information indicating whether or not the product has been repaired when the result of the intermediate inspection is defective, and when there is information indicating that the product after the intermediate inspection, which is the target of the teacher data generation, has been repaired, it may be determined not to use the image of the product for which the repair has been performed in the teacher data.

[0015] If a product determined to be defective as a result of the intermediate inspection is repaired and then sent to the subsequent process, the intermediate inspection image used during the intermediate inspection, which is the target of teacher data generation, will not be associated with the product on which the final inspection was performed. Therefore, it is not appropriate to assign the final inspection result as correct data to the intermediate inspection image. For this reason, by not using the inspection image of the product repaired after the intermediate inspection as teacher data as in the above configuration, it is possible to prevent inappropriate teacher data from being used during the learning of the inspection model, and thereby reduce the factors that degrade the accuracy of the inspection model. Also, the inspection management system may further include display means for at least displaying a first user interface for the user to confirm and correct the teacher data generated by the teacher data generation means, and input means for receiving user operation inputs.

[0016]

[0017] ​When there is a problem in processes other than the intermediate process (other intermediate processes and the final process) that are the targets of teacher data generation and the final inspection result is defective, or when the final inspection result of a product that has been repaired (when the system has not obtained information indicating that the repair has been performed) is found to be non-defective after a defective determination in the intermediate inspection, assigning the final inspection result as the correct label to the intermediate inspection image would result in inappropriate teacher data. Therefore, by providing a configuration that allows the user to confirm and correct the teacher data generated by the teacher data generation means as described above, inappropriate teacher data can be excluded or corrected. This can reduce the factors that lower the accuracy of the inspection model.

[0018] Also, the first user interface may include a display that lists all the images used in the inspections for each manufacturing process of one of the products. According to such a configuration, since the images of the products taken in each of the multiple manufacturing processes can be compared side by side, it is possible to easily determine which manufacturing process caused the final inspection result to be defective when confirming the teacher data for a product with a defective final inspection result. Also, it is possible to easily determine whether a repair has been performed in any of the processes (and if so, in which process).

[0019] Furthermore, the inspection management system may further include a classification means that classifies the teacher data to assist the user in confirming and correcting the teacher data.

[0020] Even if the user can confirm and correct the teacher data, it is difficult to extract the teacher data that should be confirmed and corrected from a large number of teacher data. Therefore, for example, by classifying the teacher data according to the combination patterns of information such as the results of automatic intermediate inspection by a program, the results of visual intermediate inspection by visual inspection, the results of final inspection, and the presence or absence of repair (the presence or absence of the repair information itself), the handling of the teacher data can be facilitated.

[0021] Further, the inspection management system may include display means for at least displaying a second user interface for searching the generated teacher data. Note that the display means for displaying the first user interface and the display means for displaying the second user interface may be configured by the same hardware or different hardware. According to such a configuration, it is possible to easily extract data of patterns that the user wants to confirm, and it is possible to efficiently confirm and correct teacher data.

[0022] Further, the second user interface may include a graph display that shows, for each inspection program to which at least the inspection model is applied, the quantity of the teacher data that can be used in the inspection model so that the quantity for each classification is known. Further, the graph display can narrow down and display the teacher data according to at least the type of defect specified in the inspection program, and the second user interface may include an area for displaying the image related to the teacher data displayed after narrowing down.

[0023] According to such a configuration, regarding the inspection program in the target intermediate process, the user can search for the teacher data created (and classified) based on not only the result of the intermediate inspection but also the final inspection result, and can refer to the inspection image. As a result, the final inspection result can be linked to the manufacturing result (product state) of a certain intermediate process, so that more appropriate process analysis can be performed and more effective process improvement can be considered.

[0024] Further, the intermediate inspection includes automatic inspection by a program and visual inspection in which an inspector visually determines the image, and the classification means assigns a preset classification label to the teacher data to be classified according to a combination of the result of the automatic inspection and / or the result of the visual inspection performed using the image related to the teacher data to be classified and the corresponding result of the final inspection.

[0025] Also, the classification means When both the result of the automatic inspection performed using the image related to the teacher data to be classified and the corresponding final inspection result are judged as non-defective products, a classification label of "non-defective product" is attached to the teacher data. When the result of the automatic inspection performed using the image related to the teacher data to be classified is judged as a non-defective product, and the corresponding final inspection result is judged as a defective product, a classification label of "missed defect, need to confirm" is attached to the teacher data. When the result of the automatic inspection performed using the image related to the teacher data to be classified is judged as a defective product, the result of the visual inspection is judged as a non-defective product, and the corresponding final inspection result is judged as a non-defective product, a classification label of "overlooked defect" is attached to the teacher data. When the result of the automatic inspection performed using the image related to the teacher data to be classified is judged as a defective product, the result of the visual inspection is judged as a non-defective product, and the corresponding final inspection result is judged as a defective product, a classification label of "visual inspection result, need to confirm" is attached to the teacher data. When the result of the automatic inspection performed using the image related to the teacher data to be classified is judged as a defective product, the result of the visual inspection is judged as a defective product, and the corresponding final inspection result is judged as a non-defective product, a classification label of "overlooked defect, need to confirm" is attached to the teacher data. When the result of the automatic inspection performed using the image related to the teacher data to be classified is judged as a defective product, the result of the visual inspection is judged as a defective product, and the corresponding final inspection result is judged as a defective product, a classification label of "actual defect" may be attached to the teacher data.

[0026] Note that in this specification, "actual defect" means that the result of the automatic inspection of the target process is judged as a defective product and the result of the visual inspection is also judged as a defective product. According to such a configuration, the user can arbitrarily select the target for checking and correcting the teacher data according to the classification set in detail according to the pattern.

[0027] Further, the intermediate inspection includes a formal inspection that treats the determination result as a formal inspection result, and a provisional inspection that makes a determination using the inspection model but does not reflect it as an inspection result. The inspection model evaluation means may further be provided to evaluate the accuracy of the inspection model by comparing the inspection result of the formal inspection with the determination result in the provisional inspection. According to such a configuration, it is possible to easily confirm the accuracy of the inspection model while operating the production line.

[0028] Further, the inspection management system may further include a report creation means for creating a report on the accuracy of the inspection model using the aggregation of the intermediate inspection results, which are the results of each of the intermediate inspections for each predetermined period, and the aggregation of the final inspection results, which are the results of the final inspection, and an output means for outputting the report.

[0029] According to such a configuration, the user can grasp the accuracy of each inspection model using the data of the intermediate inspection results and the final inspection results for each predetermined period. Further, it becomes possible to review each inspection model on a daily, weekly, or monthly basis.

[0030] Further, the report may include a third user interface that shows, for each of a plurality of items of a first level of abstraction related to the intermediate inspection, a misjudgment graph sorted by one of the misjudgment number or the misjudgment rate for each of the items of a second level of abstraction included in the item of the first level of abstraction that was misjudged during the intermediate inspection, and preferentially shows the misjudgment graph for the item name of the first level of abstraction and the items of the second level of abstraction related to the item for which one of the misjudgment number or the misjudgment rate in the items of the second level of abstraction is larger.

[0031] Further, the misjudgment graph may sort either the number of misjudgments or the misjudgment rate in descending order, and highlight the items of the second abstraction level for which the other of the number of misjudgments or the misjudgment rate exceeds a threshold value. Here, the "misjudgment" refers to a case where the result of the automatic inspection in the intermediate process is different from the final pass / fail judgment result after the final inspection. With such a configuration, the user can visually and easily grasp the inspection models (inspection criteria, part numbers, etc.) that should be corrected with particular priority.

[0032] Further, the report may include a diagram showing the plotting of the labels identifying each inspection model used in the intermediate inspection on two-dimensional coordinates with the number of cases where the intermediate inspection result is defective and the final inspection result is non-defective, and the number of cases where the intermediate inspection result is non-defective and the final inspection result is defective, as axes.

[0033] By referring to such a graph, the user can determine whether the inspection criteria should be relaxed or tightened while observing the relationship between different types of misjudgments, namely, "defective in intermediate inspection and non-defective in final inspection" and "non-defective in intermediate inspection and actually defective in final inspection".

[0034] Further, the inspection management system may further include an evaluation index calculation means for calculating a predetermined index related to the accuracy evaluation of the inspection model using a plurality of pieces of information on the intermediate inspection results, which are the results of each intermediate inspection, and a plurality of pieces of information on the final inspection results, which are the results of the final inspection, respectively, and a second inspection model evaluation means for evaluating the accuracy of the inspection model based on the index.

[0035] Further, the intermediate inspection and the final inspection include automatic inspection by a program and visual inspection by an inspector's visual judgment, and the index may include a value based on the value obtained by dividing the number of inspection targets where the result of the automatic inspection in the intermediate inspection is defective and the result of the visual inspection in the final inspection is defective by the number of inspection targets where the result of the visual inspection in the final inspection is defective.

[0036] In addition, the intermediate inspection and the final inspection include automatic inspection by a program and visual inspection by an inspector's visual judgment. As the index, a value based on a value obtained by dividing the number of inspection targets for which the result of the automatic inspection in the intermediate inspection is defective and the result of the visual inspection in the final inspection is defective by the number of inspection targets for which the result of the automatic inspection in the intermediate inspection is defective may be included.

[0037] In addition, the intermediate inspection and the final inspection include automatic inspection by a program and visual inspection by an inspector's visual judgment. As the index, a value based on a value obtained by dividing the number of inspection targets for which the result of the automatic inspection in the intermediate inspection is defective and the result of the final inspection is non-defective by the number of inspection targets for which the result of the automatic inspection in the intermediate inspection is defective may be included. by the number of inspection targets for which the result of the automatic inspection in the intermediate inspection is defective may be included.

[0038] In addition, the intermediate inspection and the final inspection include automatic inspection by a program and visual inspection by an inspector's visual judgment. As the index, a value based on a value obtained by dividing the number of inspection targets for which the result of the automatic inspection in the intermediate inspection is non-defective and the result of the visual inspection in the final inspection is defective by the number of inspection targets for which the result of the visual inspection in the final inspection is defective may be included.

[0039] In this way, by calculating the evaluation index and using it as an index for evaluating the accuracy of the inspection model, it becomes possible to objectively evaluate the accuracy of a plurality of inspection models using an absolute evaluation index.

[0040] In addition, the product may be a component mounting substrate, the final manufacturing process may be a reflow process, and the intermediate manufacturing process may include a printing process and a mounting process. The present invention can be effectively applied to such an inspection management system for a manufacturing line. In this case, the final inspection is any inspection after the reflow process, and may be an inspection by a plurality of inspection devices such as AOI and AXI, or may include a visual inspection.

[0041] In addition, the present invention can also be regarded as the following inspection management device. That is, An inspection management device used in a manufacturing line having a plurality of manufacturing processes related to the manufacture of a product, a plurality of manufacturing apparatuses corresponding to the plurality of processes, and an inspection device that performs inspection based on an image obtained by photographing the product, comprising: teacher data generation means for generating teacher data to be used for an inspection model corresponding to each of the intermediate inspections, using the images used in each of the intermediate inspections related to one or more intermediate manufacturing processes in the manufacturing process and the final inspection result which is the result of the final inspection related to the final manufacturing process of the product.

[0042] Further, the present invention can also be regarded as a method for generating the following teacher data. That is, In a manufacturing line having a plurality of manufacturing processes related to the manufacture of a product, a plurality of manufacturing apparatuses corresponding to the plurality of processes, and an inspection device that performs inspection based on an image obtained by photographing the product, a method for generating teacher data for an inspection model used in each of the intermediate inspections related to one or more intermediate manufacturing processes in the manufacturing process, comprising: an image acquisition step of acquiring the images used for inspection in each of the intermediate inspections; a final result acquisition step of acquiring a final inspection result which is the result of the final inspection related to the final manufacturing process; and a generation step of generating the teacher data using the images and the final inspection result.

[0043] In the generation step, when the final inspection result is a pass determination, a correct label of pass may be assigned to the images used for the intermediate inspection that is the target of the teacher data generation, and when the final inspection result is a fail determination, a correct label of fail may be assigned to the images used for the intermediate inspection that is the target of the teacher data generation.

[0044] Further, the present invention can also be regarded as a program for causing a computer to execute the above method, and a computer-readable recording medium on which such a program is non-temporarily recorded.

[0045] Also, each of the above configurations and processes can be combined with each other to constitute the present invention as long as no technical contradiction occurs.

Advantages of the Invention

[0046] According to the present invention, it is possible to provide a technique capable of generating teaching data that can increase the accuracy of an inspection model used for intermediate inspection in a production line of a product having a plurality of manufacturing processes.

Brief Description of the Drawings

[0047]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

Figure 17

Figure 18

BEST MODE FOR CARRYING OUT THE INVENTION

[0048] <APPLICATION EXAMPLE> The present invention can be applied, for example, as an inspection management apparatus 1 used in an inspection management system 100 as shown in FIG. 1. FIG. 1 is a schematic diagram of an inspection management system 100 on a component mounting line of a printed circuit board according to this application example. As shown in FIG. 1, in the mounting line according to this application example, a solder printing apparatus X1, a post-solder printing inspection apparatus Y1, a mounter X2, a post-mounting inspection apparatus Y2, a reflow furnace X3, and a post-reflow inspection apparatus Y3 are provided in order from the upstream side.

[0049] The solder printing apparatus X1 is an apparatus for printing solder on the electrode portions of a printed circuit board, the mounter X2 is an apparatus for placing electronic components to be mounted on the substrate on the solder paste, and the reflow furnace X3 is a heating apparatus for solder-joining the electronic components to the substrate.

[0050] Also, each inspection device Y1, Y2, Y3 inspects the state of the substrate at the exit of each process and determines whether there is a defect or a risk of a defect. Hereinafter, the inspection by the post-printing inspection device Y1 is called "post-printing inspection", the inspection by the post-mounting inspection device Y2 is called "post-mounting inspection", and the inspection by the post-reflow inspection device Y3 is called "post-reflow inspection" or "final inspection". Also, when explaining the "post-printing inspection" and the "post-mounting inspection" without particularly distinguishing them, it may be referred to as "intermediate inspection".

[0051] Each of the above-described manufacturing devices (X1, X2, X3) and each inspection device (Y1, Y2, Y3) are connected to the inspection management device 1 via a network such as a LAN (Local Area Network). The inspection management device 1 is composed of a general-purpose computer system including a processor such as a CPU (Central Processing Unit), a main storage device such as a ROM and a RAM, an auxiliary storage device (such as a storage device), an input device (such as a keyboard, a mouse, a controller, a touch panel), an output device (such as a display, a printer, a speaker), and the like. Note that the functions of the inspection management device 1 described later may be realized by the processor reading and executing a program stored in the auxiliary storage device.

[0052] The inspection management device 1 generates teacher data for learning an inspection model for automatic inspection performed by the inspection devices (Y1, Y2) in the middle of the process, by the functions of each functional unit described later. FIG. 2 is a block diagram showing an outline of the functional configuration of the inspection management device 1 according to this application example. As shown in FIG. 2, the inspection management device 1 has a control unit 10, an output unit 20, a communication unit 30, a storage unit 40, and an input unit 50. Further, the control unit 10 includes, as functional modules, a post-printing inspection information acquisition unit 101, a post-mounting inspection information acquisition unit 102, a post-reflow inspection information acquisition unit 103, a repair information acquisition unit 104, and a teacher data generation unit 105.

[0053] The post-printing inspection information acquisition unit 101 acquires post-printing inspection information, i.e., inspection data during inspection by the post-solder printing inspection device Y1, specifically, data including image data used in the inspection and the judgment result therefor ("good" or "bad"), via the communication unit 30, etc. Also, the post-mounting inspection information acquisition unit 102 acquires post-mounting inspection information, i.e., inspection data during inspection by the post-mounting inspection device Y2, and the post-reflow inspection information acquisition unit 103 acquires post-reflow inspection information, i.e., inspection data during final inspection by the post-reflow inspection device Y3. Note that in each intermediate inspection and final inspection, if a visual inspection by an inspector is performed after an automatic inspection by an inspection program is performed, the judgment result of the visual inspection is also included in the inspection data acquired here.

[0054] When a repair is performed on a substrate at any stage of the manufacturing process, the repair information acquisition unit 104 acquires "repair information" that includes at least information identifying the substrate on which the repair was performed and information associated with the substrate on which the repair was performed.

[0055] The teacher data generating unit 105 generates teacher data used for an automatic inspection model implemented in the post-solder-printing inspection device Y1 or the post-mounting inspection device Y2. Specifically, based on the inspection image data and inspection results acquired during intermediate inspection of a certain board and the inspection results acquired during final inspection of the board, teacher data is generated in which the same judgment content as the final inspection result is assigned as a correct answer label to the inspection image data.

[0056] To explain in more detail, for example, when generating training data for an inspection model to be implemented in the solder printing inspection device Y1, if the final inspection result of the board indicated by the image data used in the inspection by the solder printing inspection device Y1 running under the current inspection program is a good product, a correct answer label is attached to the image data, and if it is a defective product, a defective correct answer label is attached to the image data. Label the solution.

[0057] According to the inspection management apparatus 1 as described above, regardless of the result of the intermediate inspection, the result of the final inspection is used as the correct label, that is, it becomes possible to automatically create appropriate teacher data in which omissions and oversights are corrected.

[0058] <Embodiment 1> Subsequently, with reference to the drawings (including the drawings once described in the above application example) in sequence, an example of a mode for carrying out the present invention will be described in more detail. However, the specific configurations described in the embodiments are not intended to limit the scope of the present invention only to those, unless otherwise specified.

[0059] (System Configuration) FIG. 1 is a diagram schematically showing a configuration example of a component mounting line which is an inspection management system 100 according to Embodiment 1. As described above, the component mounting line mainly consists of three processes: solder printing on a substrate, mounting of components, and reflow (soldering of solder).

[0060] The solder printing apparatus X1 is an apparatus that prints paste-like solder on the electrode portions (pads) on the printed circuit board by screen printing. The mounter X2 is an apparatus for picking up electronic components to be mounted on the substrate and placing the components on the solder paste at the corresponding location, and is also called a chip mounter. The reflow furnace X3 is a heating apparatus for heating and melting the solder paste and then cooling it to solder-join the electronic components to the substrate. When there are a large number and variety of electronic components to be mounted on the substrate, a plurality of mounters X2 may be provided on the component mounting line.

[0061] In addition, on the component mounting line, an inspection apparatus is installed to inspect the state of the substrate at the outlet of each process of solder printing, component mounting, and reflow, and to determine whether there is a defect or a risk of a defect.

[0062] The post-solder printing inspection device Y1 is a device for inspecting the printing state of solder paste on a substrate carried out from the solder printing device X1. In the post-solder printing inspection device Y1, an inspection image including the solder paste printed on the substrate is acquired, and determination is made as to whether it is appropriate for various inspection items. Examples of the inspection items include the volume, area, height, displacement, and shape of the solder. Note that the inspection device may be provided with means for two-dimensionally or three-dimensionally measuring the substrate (particularly the solder paste) in addition to a camera.

[0063] The post-mounting inspection device Y2 is a device for inspecting the arrangement state of electronic components on a substrate carried out from the mounter X2. In the post-mounting inspection device Y2, an inspection image including components (which may be a part of the component such as a component body or an electrode) placed on the solder paste is acquired, and determination is made as to whether it is appropriate for various inspection items. Examples of the inspection items include component displacement, angular (rotation) displacement, missing components (the component is not arranged), wrong components (different components are arranged), wrong polarity (the polarities of the electrodes on the component side and the substrate side are different), front-back inversion (the component is arranged backward), and component height. Note that the fact that it may be provided with measuring means other than a camera is the same as that of the post-solder printing inspection device Y1.

[0064] The post-reflow inspection device Y3 is a device for inspecting the quality of soldering on a substrate carried out from the reflow oven X3. In the post-reflow inspection device Y3, an inspection image of the substrate including the soldered part after reflow is acquired, and determination is made as to whether it is appropriate for various inspection items. Note that the post-reflow inspection device Y3 may have a function related to X-ray image inspection in addition to the configuration for visual inspection.

[0065] As inspection items by visual inspection, in addition to the same items as those after mounting, the quality of the solder fillet shape and the like are also included. For measuring the shape of solder, for example, the phase shift method, the color highlight method, etc. can be used. As inspection items for X-ray image inspection, for example, there are component misalignment, solder height, solder volume, solder ball diameter, length of back fillet, quality of solder joint, etc. Note that as the X-ray image, an X-ray transmission image or a CT (Computed Tomography) image may be used. Note that even in the case including X-ray image inspection, in the following description, the inspection device after the final process is referred to as the post-reflow inspection device Y3, and the inspection after the final process is referred to as the "final inspection" or the "post-reflow inspection".

[0066] Also, each inspection device Y1, Y2, Y3 according to this embodiment is provided with a display device and an input means for visually checking the substrate to be inspected. When visual inspection is performed by an inspector in each process, the determination content by visual inspection is taken as the inspection result in each process. Note that the display device and input means for visual inspection may be provided as a terminal separate from each inspection device.

[0067] In this embodiment, the substrate processed by the solder printing device X1 and the mounter X2 is a semi-finished product, and the substrate carried out from the reflow furnace X3 becomes a finished product. Also, the inspections performed by the post-solder printing inspection device Y1 and the post-mounting inspection device Y2 are intermediate inspections, and the inspection by the post-reflow inspection device Y3 is the final inspection.

[0068] (Inspection Management Device) Each of the above-described manufacturing apparatuses (X1, X2, X3) and each inspection apparatus (Y1, Y2, Y3) are connected to the inspection management apparatus 1 via communication means. As described above, the inspection management apparatus 1 is constituted by a general-purpose computer system. Note that the inspection management apparatus 1 may be constituted by a single computer or may be constituted by a plurality of computers. Alternatively, it is also possible to implement all or part of the functions of the inspection management apparatus 1 in the computer incorporated in any one of the manufacturing apparatuses (X1, X2, X3) or the inspection apparatuses (Y1, Y2, Y3). Alternatively, part of the functions of the inspection management apparatus 1 may be realized by a server (such as a cloud server) on a wide area network.

[0069] As shown in FIG. 2, the inspection management apparatus 1 includes a control unit 10, an output unit 20, a communication unit 30, a storage unit 40, and an input unit 50. Further, as functional modules, the control unit 10 includes each functional unit of a post-print inspection information acquisition unit 101, a post-mount inspection information acquisition unit 102, a post-reflow inspection information acquisition unit 103, a repair information acquisition unit 104, a teacher data generation unit 105, a data classification unit 106, a UI (User Interface) creation unit 107, and a teacher data update unit 108.

[0070] The output unit 20 is a means for outputting various information such as a user interface (UI) screen described later, and is configured to include a display device such as a liquid crystal display.

[0071] The communication unit 30 communicates with each of the manufacturing apparatuses (X1, X2, X3) and each inspection apparatus (Y1, Y2, Y3) by a wired or wireless method and executes transmission and reception of data. The communication unit 30 is configured to include, for example, an integrated circuit such as a communication IC (Integrated Circuits).

[0072] The storage unit 40 includes a main storage device, an auxiliary storage device, etc., and stores various types of information such as printed inspection information, post-mount inspection information, post-reflow inspection information, repair information, etc., in addition to the teacher data as described later. Further, it may store an inspection program including an inspection model. Note that the storage unit 40 may be configured to include an external storage device such as a server or the like.

[0073] The input unit 50 is a means for receiving the input of various types of information to the inspection management device 1, and is configured to include an input device such as a keyboard, a mouse, a controller, a touch panel, etc.

[0074] Subsequently, each functional block included in the control unit 10 will be described. Note that since each functional unit of the printed inspection information acquisition unit 101, the post-mount inspection information acquisition unit 102, the post-reflow inspection information acquisition unit 103, and the repair information acquisition unit 104 is as described in the above application example, a new description will be omitted.

[0075] (Regarding the teacher data generation unit) As described in the above application example, the teacher data generation unit 105 generates teacher data used for learning an automatic inspection model mounted on the post-solder printing inspection device Y1 or the post-mount inspection device Y2. The teacher data generated by the teacher data generation unit 105 may be stored in the storage unit 40. Note that as the learning of the automatic inspection model using the teacher data, for example, it is assumed to obtain a feature amount suitable for estimating a substrate that becomes a real defect after the reflow process from an image at the intermediate inspection from a large number of teacher data, and to determine a determination threshold value of the feature amount with the least omission and overlooking.

[0076] As described in the description of the application example, the teacher data generation unit 105 generates teacher data by attaching the same determination content as the final inspection result to the inspection image data at the intermediate inspection as the correct label. Further, based on the repair information acquired by the repair information acquisition unit 104, when the board corresponding to the target inspection image data has been repaired in any of the intermediate processes, the teacher data generation unit 105 determines not to use the inspection image data as teacher data. Since the intermediate processes include the printing process and the mounting process, when explaining without specifying the intermediate process that is the target of the teacher data, terms such as "target process" are used. Also, determining not to use it as teacher data for the inspection model is hereinafter also expressed as "excluding" from the teacher data.

[0077] Figure 3 is a table showing the relationship between the correct labels automatically assigned by the teacher data generation unit 105 to the inspection images of the target process according to the combination patterns of the automatic inspection results, visual inspection results, and subsequent post-reflow inspection results in the target process. As shown in Figure 3, regardless of the results of the automatic inspection or visual inspection in the target process, the teacher data generation unit 105 assigns a correct label of "OK" (meaning good product; the same applies hereinafter) if the result of the subsequent post-reflow inspection is a good product, and a correct label of "NG" (meaning defective product; the same applies hereinafter) if it is an actual defect. In addition, when repair has been performed in any of the intermediate processes (repair available), the data is excluded from the teacher data.

[0078] If a board determined to be defective during the intermediate inspection is repaired and then sent to the subsequent process, the intermediate inspection image used during the intermediate inspection that is the target of teacher data generation will not be linked to the board on which the final inspection was performed. Therefore, it is not appropriate to assign the final inspection result as the correct data to the intermediate inspection image. For this reason, by not using the inspection images of the repaired boards as teacher data, it is possible to prevent the use of inappropriate teacher data for the learning of the inspection model, and it is possible to reduce the factors that reduce the accuracy of the inspection model.

[0079] However, even when there is no repair information (the presence or absence of repair is unknown), the teacher data generation unit 105 automatically assigns a correct label to the inspection image of the target process. FIG. 4 shows the relationship between the automatic inspection results and visual inspection results in the printing process and the mounting process, respectively, the subsequent post-reflow inspection, and the correct label automatically assigned by the teacher data generation unit 105 to the inspection image of the target process, according to the combination pattern. As shown in FIG. 4, regardless of the inspection results of the intermediate processes other than the target process, the teacher data generation unit 105 automatically assigns a correct label with the same content as the final inspection result to the inspection image of the target process.

[0080] (Regarding the data classification unit) The data classification unit 106 classifies the teacher data generated by the teacher data generation unit 105 to assist the user in confirmation and correction. Specifically, classification labels are assigned according to the combination pattern of the inspection results of automatic inspection and visual inspection in each process of the product related to the teacher data and the final inspection result. Note that the number of combination patterns and the number of classification labels do not necessarily have to match (that is, a plurality of combination patterns may be aggregated into one classification label). Then, the teacher data to which the classification label is assigned by the data classification unit 106 is stored in the storage unit 40.

[0081] FIG. 5 is a table showing the relationship between the classification labels automatically assigned by the data classification unit 106 to the teacher data according to the combination pattern of the automatic inspection result and the visual inspection result in the target process, and the subsequent post-reflow inspection. Regarding the repair information, since it is excluded from the teacher data when there is repair, only the case where the presence or absence of repair is unknown (or there is information indicating that it has not been repaired) is the target for assigning the classification label.

[0082] As shown in FIG. 5, when both the result of the automatic inspection performed using the inspection image related to the teacher data to be classified and the corresponding final inspection result are judged as good products, the data classification unit 106 assigns a "good product" classification label to the teacher data.

[0083] In addition, when the result of the automatic inspection performed using the inspection image related to the teacher data to be classified is a pass determination, and the corresponding final inspection result is a fail determination, the data classification unit 106 attaches a classification label of "missed, need confirmation" to the teacher data.

[0084] In addition, when the result of the automatic inspection performed using the inspection image related to the teacher data to be classified is a fail determination, the result of the visual inspection is a pass determination, and the corresponding final inspection result is a pass determination, the data classification unit 106 attaches a classification label of "overlooked" to the teacher data.

[0085] In addition, when the result of the automatic inspection performed using the inspection image related to the teacher data to be classified is a fail determination, the result of the visual inspection is a pass determination, and the corresponding final inspection result is a fail determination, the data classification unit 106 attaches a classification label of "visual result needs confirmation" to the teacher data.

[0086] In addition, when the results of both the automatic inspection and the visual inspection performed using the inspection image related to the teacher data to be classified are fail determinations, and the corresponding final inspection result is a pass determination, the data classification unit 106 attaches a classification label of "overlooked, need confirmation" to the teacher data.

[0087] In addition, when the results of both the automatic inspection and the visual inspection performed using the inspection image related to the teacher data to be classified are fail determinations, and the corresponding final inspection result is a fail determination, the data classification unit 106 attaches a classification label of "actual defect" to the teacher data.

[0088] When the user confirms and corrects the teacher data, it is difficult to extract the teacher data to be confirmed and corrected from a large number of teacher data. In this regard, by classifying the teacher data in detail according to the pattern of information combination as described above, the handling of the teacher data can be facilitated.

[0089] (UI Creation Unit) The UI creation unit 107 creates a user interface (UI) for the user to check and correct the teacher data generated and stored in the storage unit 40. Note that the created UI is output to the display device included in the output unit 20. Examples of the UI created by the UI creation unit 107 include a UI for searching (or narrowing down) the teacher data stored in the storage unit 40 based on the classification given by the data classification unit 106. In addition, the UI creation unit 107 also creates a UI for listing inspection images of each target substrate up to the final process of the teacher data that is likely to need correction, and comparing the states of the substrates in each process.

[0090] FIG. 6 shows an example of a UI for searching teacher data created by the UI creation unit 107. In the example of the UI shown in FIG. 6, for the inspection program of the target intermediate inspection, a search item area S for narrowing down teacher data according to the classification is arranged at the upper part of the screen. In the search item area S, a checkbox is provided for each classification label except "good product", and the label of the selected classification can narrow down the teacher data. In the example of FIG. 6, the classifications of "overlooked", "actual defect", and "missed and need to confirm" are selected.

[0091] Also, in the middle part of the screen, a graph display area P is arranged to show the quantity of teacher data generated within a specified period and labeled with a specified classification by inspection program. From left to right, the target is specifically narrowed down and displayed in the order of inspection program, component part number group included in the program, component part number included in the component part number group, type of defect, etc. Specifically, by selecting any item of the graph related to the upper item, the items of the graph related to the lower items included in the item can be displayed.

[0092] In addition, at the lower part of the screen, an inspection image display area G for displaying inspection images related to the teacher data for the finally narrowed-down parts is arranged. By selecting the displayed image or the "confirmation" button, it is possible to transition to a UI for confirming and correcting the individual teacher data related to the displayed image. Note that the UI illustrated in FIG. 6 in the present embodiment corresponds to the second user interface according to the present invention.

[0093] According to the UI as shown in FIG. 6 like this, regarding the inspection program (part number group, part number, defect type included) in the target intermediate process, the user can search for the teacher data created (and classified) based on not only the result of the intermediate inspection but also the final inspection result, and can refer to the inspection image. For this reason, regarding the manufacturing result (product state) of a certain intermediate process, the manufacturing result can be grasped based on the final inspection result, so that a more appropriate process analysis can be performed, and a more effective process improvement can be considered.

[0094] FIG. 7 is an example of a UI for confirming and correcting individual teacher data that is transitioned from the UI of FIG. 6. As shown in FIG. 7, images after printing, after mounting, and after reflow for the area including the inspection target parts of a certain substrate are displayed in a list, and for the substrate (more specifically, the parts) for which the final inspection result is defective, it is possible to easily determine which manufacturing process caused the final inspection result to be defective. Also, it is possible to easily determine whether or not repair has been performed in any of the processes (and if so, in which process). Note that the UI illustrated in FIG. 7 in the present embodiment corresponds to the first user interface according to the present invention.

[0095] In addition, in the UI shown in FIG. 7, the current classification label assigned to the teacher data to be confirmed is in the form of a pull-down menu and is displayed together with the inspection image. In the pull-down menu In addition, there are other items such as "OK", "NG", and "Exclude". The user can select "OK", "NG", or "Exclude" from the pull-down menu to modify the correct data currently assigned to the target teacher data or determine to exclude it from the teacher data. Also, if the items in the pull-down menu are not changed, the correct data and classification labels will be maintained in their current state.

[0096] The user refers to the list image as shown in FIG. 7 and, according to the classification assigned to the target teacher data, considers whether the teacher data should be corrected (or excluded) in accordance with the combination pattern of the automatic inspection results and visual inspection results in each of the printing process and the mounting process, and the subsequent post-reflow inspection. As described above, since the teacher data generation unit 105 automatically assigns correct data even when there is no repair information (the presence or absence of repair is unknown), the user also determines whether repair has been performed in any of the intermediate processes based on the repair information (or inspection images, etc.) not input to the system.

[0097] The tables shown in FIGS. 8 to 10 show the relationship between the classification assigned to the teacher data, the combination pattern of the automatic inspection results and visual inspection results in each of the printing process and the mounting process, the post-reflow inspection, the user's judgment result, and the content of the change in the correct label.

[0098] FIG. 8 shows the combination of user correction patterns when the target process is the printing process and the inspection result in the mounting process is a good product (automatic inspection "OK" or visual inspection "OK"). FIG. 9 shows the combination of user correction patterns when the target process is the printing process and the inspection result in the mounting process is defective (automatic inspection "NG" and visual inspection "NG"). FIG. 10 shows the combination of user correction patterns when the target process is the mounting process.

[0099] For example, in Fig. 8 (i.e., when the target process is the printing process), for the training data with the classification label "Check for omission", if the automatic inspection result of the printing process is "OK" and the post-reflow inspection result is actually defective, the current correct label is "NG". In contrast, the user determines whether the cause of the defect lies in the printing process or in subsequent processes (the mounting process and the reflow process) based on the list image as illustrated in Fig. 7.

[0100] Here, if it is determined that the cause lies in the printing process, the correct label of the target training data does not need to be corrected (it is also possible to end the correction without making any selection, or to select "NG" again from the pull-down menu). On the other hand, if it is determined that the cause of the defect lies in the mounting process or the reflow process, the correct label of the training data is corrected from "NG" to "OK". This is because there is no problem with the substrate at the stage of the printing process, so the training data with the correct label of "NG" attached to such inspection images is inappropriate data, and if such training data is mixed, the accuracy of the inspection model will decrease.

[0101] Also, in Fig. 9, for the training data with the classification label "Check for oversight", if the inspection result of the printing process is "NG" including visual inspection and the result of the post-reflow inspection is a good product, the current correct label is "OK". In this case, the user first determines whether there is repair. If there is repair and the final inspection result is a good product, the target training data is excluded. On the other hand, if there is no repair, based on the list image as illustrated in Fig. 7, it is determined whether there is a cause judged to be defective in each process.

[0102] And if it is determined that there are signs of defects in the inspection image of the printing process, the correct label is corrected from "OK" to "NG". On the other hand, if it is determined that there are no signs of defects in the inspection image of the printing process but there are signs of defects in the image of the mounting process (or the reflow process), there is no problem with "OK" as the correct label in the printing process, so no correction is required. However, in this case It is necessary to consider correcting the correct labels of the teacher data in the mounting process, or reviewing the inspection results of the reflow process, etc.

[0103] In addition, the same applies to the case of correcting the teacher data with the mounting process shown in FIG. 10 as the target process. In this case, however, it is not affected by the inspection results of the printing process.

[0104] (Teacher data update unit) As described above, the user determines and decides whether to correct (exclude) the teacher data via the UI (and the input unit 50) created by the UI creation unit 107. The teacher data update unit 108 reflects the determined content in the teacher data stored in the storage unit 40.

[0105] Next, based on FIG. 11, the flow of processing performed by the inspection management device 1 will be described. FIG. 11 is a flowchart showing an example of the processing executed by the inspection management device 1. As shown in FIG. 11, the post-printing inspection information acquisition unit 101 of the inspection management device 1 acquires post-printing inspection information (S101). Next, the post-mounting inspection information acquisition unit 102 acquires post-mounting inspection information (S102). Further, the post-reflow inspection information acquisition unit 103 acquires post-reflow inspection information (S103). Subsequently, the repair information acquisition unit 104 acquires repair information (S104). Note that the description of the content of each piece of information has already been given, so the description here is omitted.

[0106] Next, the teacher data generation unit 105 generates teacher data based on the information acquired in steps S101 to S104 (S105). Subsequently, the data classification unit 106 assigns classification labels to the teacher data generated in step S105 and classifies it (S106). Then, the inspection management device 1 stores the teacher data classified in step S106 in the storage unit 40.

[0107] Next, the inspection management device 1 determines whether there is an instruction to display the UI screen from the user (S108). If it is determined that there is no instruction for the UI screen, the process returns to step S101 and the subsequent processes are repeated. On the other hand, if it is determined in step S108 that there is an instruction to display the UI screen from the user, the UI creation unit 107 creates a UI for teacher data extraction (narrowing down), and displays this on the output unit 20 (S109). Note that the UI for teacher data extraction can be a screen as exemplified in FIG. 6 as described above.

[0108] Then, the inspection management device 1 receives an input from the user via the input unit 50 (S110). In response to this, the UI creation unit 107 further creates a UI for teacher data correction, and displays this on the output unit 20 (S111). Note that the UI for teacher data correction can be a screen as exemplified in FIG. 7 as described above. The inspection management device 1 receives an input from the user via the input unit 50 (S112). In response to this, the teacher data update unit 108 updates the correct label of the teacher data to be corrected stored in the storage unit 40, or deletes the teacher data to be excluded (S113). Thereafter, the inspection management device 1 determines whether a predetermined end condition is satisfied (S114). If it is determined that the condition is satisfied, the series of processes is terminated. On the other hand, if it is determined in step S114 that the end condition is not satisfied, the process returns to step S108 and the series of processes is repeated.

[0109] According to the inspection management system 100 according to the present embodiment as described above, it is possible to automatically and appropriately generate teacher data used for an inspection model used at the time of intermediate inspection of a substrate. Furthermore, by classifying the generated teacher data according to the progress of the inspection of the target substrate, it is possible to appropriately and efficiently extract the teacher data. Also, by creating and displaying a UI suitable for the extraction and correction of teacher data, the user can easily confirm and correct the generated teacher data.

[0110] <Embodiment 2> In the above-described inspection management apparatus 1, it was only configured to generate and update the teacher data used for the inspection model. However, the present invention is not limited to such an aspect, and the inspection management apparatus may further include additional functional units. FIG. 12 is a functional block diagram of an inspection management apparatus 2 according to such a second embodiment.

[0111] Regarding the inspection management apparatus 2 according to the present embodiment, the same components as those of the inspection management apparatus 1 of the above-described first embodiment are denoted by the same reference numerals, and the description thereof will not be repeated (the same applies to the following description of each modification example and embodiment). As shown in FIG. 12, the inspection management apparatus 2 according to the present embodiment has the same configuration as the inspection management apparatus 1 of the first embodiment, except that a provisional inspection result acquisition unit 109 and an inspection model evaluation unit 110 are newly added as functional units of the control unit 12.

[0112] The inspection apparatuses for intermediate inspection (post-solder printing inspection apparatus Y1 and post-mounting inspection apparatus Y2) connected to the inspection management apparatus 2 according to the present embodiment are configured to perform a main inspection by the main inspection program that treats the determination result as the official inspection result, and a provisional inspection that performs determination by the inspection model being tested but does not reflect the result as the inspection result. The result of the provisional inspection is held separately from the result of the main inspection as the provisional inspection result.

[0113] The provisional inspection result acquisition unit 109 of the inspection management apparatus 2 acquires data of the provisional inspection result from the post-solder printing inspection apparatus Y1 and the post-mounting inspection apparatus Y2 via the communication unit 30. Further, the inspection model evaluation unit 110 evaluates the accuracy of the inspection model by comparing the acquired provisional inspection result (i.e., the determination result of the inspection model being tested) with the result of the main inspection (the result appropriately judged as good or bad including visual inspection). Specifically, the accuracy can be evaluated based on how well the provisional inspection result matches the main inspection result.

[0114] By evaluating the accuracy of the inspection model learned with the generated teacher data in this way, it is possible to confirm whether the inspection model at the current stage has an accuracy that can withstand implementation.

[0115] (Modified Example) Note that the inspection model evaluation unit 110 in the above inspection management device 2 has been described as evaluating the accuracy of the inspection model by comparing the results of the preliminary inspection and the results of the main inspection. However, it is also possible to absolutely evaluate a plurality of inspection models using specific evaluation indicators. FIG. 13 is a functional block diagram of an inspection management device 3 according to such a modified example. As shown in FIG. 13, the inspection management device 3 according to this modified example includes an evaluation index calculation unit 131 as a functional unit of the control unit 13, and has the same configuration as the inspection management device 2 except that the inspection model evaluation unit 132 evaluates the accuracy of the inspection model by a method different from that of the inspection management device 2.

[0116] The evaluation index calculation unit 131 calculates evaluation indicators related to the accuracy (performance) of the inspection model using a plurality of results of each intermediate inspection and the corresponding results of the final inspection respectively. Specifically, for example, as an index for evaluating the inspection model for post-mount inspection, the number of inspection targets for which the result of the automatic inspection in the post-mount inspection is defective and the result of the post-reflow inspection is actually defective is divided by the number of actually defective products in the post-reflow inspection (the number of all inspection targets that are actually defective in the post-reflow inspection regardless of the pass / fail determination in the post-mount inspection), and the value obtained by expressing the quotient as a percentage (the actual defective detection rate after intermediate reflow) can be used as an index.

[0117] In addition to this, as an index, it is also possible to use a value obtained by dividing the number of inspection targets for which the result of the automatic inspection after mounting is defective and the result of the post-reflow inspection is actually defective by the number of inspection targets for which the result of the automatic inspection after mounting is defective (the total number determined to be defective during the automatic inspection after mounting), and expressing the quotient as a percentage (the correct rate of NG determination at the intermediate stage).

[0118] It is also possible to use a value obtained by dividing the number of inspection targets for which the result of the automatic inspection after mounting is defective and the result of the post-reflow inspection is non-defective by the number of inspection targets for which the result of the automatic inspection after mounting is defective, and expressing the quotient as a percentage (the incorrect rate of NG determination at the intermediate stage).

[0119] Furthermore, it is also possible to use the value (intermediate reflow actual defect miss rate) obtained by dividing the number of inspection targets that passed the automatic inspection after mounting and became actual defects in the post-reflow inspection by the number of actual defects in the post-reflow inspection, expressed as a percentage. This index indicates the proportion of products that become actual defects after reflow but are missed during the intermediate inspection. However, the "actual defects after reflow" here includes actual defects after reflow caused by other intermediate processes (printing process) or the final process (reflow process), so it cannot be judged as the miss rate of defects in the target intermediate process (mounting process).

[0120] Note that each of the above indexes does not necessarily need to be a value expressed as a percentage, and similar indexes can also be adopted for evaluating the inspection model not only for the post-mounting inspection but also for the post-printing inspection.

[0121] As in this modified example, by calculating the evaluation index and using it as an index for evaluating the accuracy of the inspection model, it becomes possible to objectively evaluate the accuracy of each inspection model using an absolute evaluation index.

[0122] <Embodiment 3> According to each of the above-described inspection management apparatuses (inspection management systems), by using the information on the intermediate inspection results and the information on the final inspection results, not only can the accuracy of the inspection model be evaluated, but also information useful for reviewing the inspection model (and the inspection criteria included therein) can be provided to the user. FIG. 14 is a block diagram showing an outline of the functional configuration of an inspection management apparatus 4 according to such an embodiment.

[0123] As shown in FIG. 14, the inspection management apparatus 4 according to the present embodiment includes, as functional units of the control unit 14, a report creation unit 141 and a UI creation unit 142. The report creation unit 141 uses the information on each inspection acquired by at least the post-printing inspection information acquisition unit 101, the post-mounting inspection information acquisition unit 102, and the post-reflow inspection information acquisition unit 103 to totalize the intermediate inspection results and the final inspection results for each predetermined period, and creates a report on the accuracy of the inspection model based on each of these totalizations.

[0124] Note that the predetermined period mentioned here does not necessarily refer to only one period. For example, reports may be created for each period of "daily", "weekly", "monthly", or "yearly".

[0125] In addition, the UI creation unit 142 creates a UI including the report created by the report creation unit 141. The created report (or the UI including this) is output from the display device or printer included in the output unit 20.

[0126] Hereinafter, examples of the report created by the report creation unit 141 and the UI created by the UI creation unit 142 will be described. FIG. 15 is a diagram showing an example of a daily report created by the report creation unit 141 at a fixed time every day, using the aggregation of in-process inspection results and the aggregation of final inspection results for each day. In the report shown in FIG. 15, as a report regarding the accuracy of the inspection model after mounting, for the substrates (and components) inspected during the report period (one day in the example of FIG. 15), the results of cross-tabulation of the number for each pass / fail determination of the automatic inspection after mounting, and the respective numbers of non-defective products and actual defective products during the post-reflow inspection are shown. Also, for each inspection program unit, the pass / fail determination results of the post-reflow inspection for non-defective judgment substrates during in-process inspection, and similarly the pass / fail determination results of the post-reflow inspection for defective reversal substrates are aggregated respectively, and the values of the indicators (four in the example of FIG. 15) calculated using these numerical values are shown. The values of the indicators (four in the example of FIG. 15) calculated using these numerical values are shown.

[0127] As shown in FIG. 15, using the data of the in-process inspection results and the final inspection results, the indicators related to the accuracy for each inspection program (intermediate NG judgment incorrect rate, intermediate post-reflow actual defective product overlooking rate, intermediate post-reflow actual defective product detection rate, intermediate NG judgment correct rate) are shown. Therefore, by referring to the values of the indicators, the accuracy of each inspection program (inspection model) shown in the list can be grasped. As a result, the user can easily recognize the program to be corrected and what kind of correction should be made.

[0128] Although the example shown in FIG. 15 was a daily report, the aggregation period for report generation is not particularly defined. As the number of data increases, the value of the index related to the accuracy of the inspection model becomes more reliable. FIG. 16 shows an example of the value of the index related to the inspection model accuracy when the number of data increases. Also, only one of the indexes related to the accuracy of the inspection model displayed in the report may be used. By showing only the values of the indexes that require particular attention, visibility and thus work efficiency can be improved.

[0129] In addition to simply displaying numerical values for each item as shown in FIGS. 15 and 16, it is also possible to create a report that includes a more graphical UI. FIGS. 17 and 18 show examples of such a graphical UI. In the example shown in FIG. 17, for each item of "defect type", "inspection program (ID)", "circuit number (targeted by the inspection program)", and "component part number", a bar graph showing the number of misjudgments for more specific items included in each item is displayed. Note that in the example of FIG. 17, the content of the misjudgment is what was judged as defective even though it should have been judged as good during the intermediate inspection. Such misjudgments are calculated for each item using data on whether it actually became defective in the final inspection rather than the visual inspection results during the intermediate inspection.

[0130] In this embodiment, "defect type", "inspection program ID", "circuit number", and "component part number" in FIG. 17 correspond to items of the first level of abstraction (hereinafter also referred to as middle items), and "side protrusion", "Prog0001", "Prog0001_C001", "CompNum0001", etc. correspond to items of the second level of abstraction (hereinafter also referred to as small items).

[0131] In the UI shown in FIG. 17, the small items in each middle item are shown in descending order from top to bottom in order of the number of misjudgments and are displayed as bar graphs. Further, the middle items (and the graphs of the small items related to their breakdown) are also sorted and displayed in descending order so that the middle item with the most misjudgments among the small items is located at the top. In this embodiment, the UI illustrated in FIG. 17 corresponds to the third user interface.

[0132] Also, in the UI illustrated in FIG. 17, for each sub-item where the number of misjudgments and / or the misjudgment rate exceeds a predetermined threshold, the bar of the bar graph is highlighted (displayed in a darker color in this embodiment). According to the UI as illustrated in FIG. 17, the user can visually and easily grasp the inspection programs, part numbers, etc. that should be corrected with particular priority.

[0133] Also, in the UI illustrated in FIG. 18, regarding the inspection model used for post-mount inspection, a graph is included that plots the numbers associated with each inspection program ID used in the intermediate inspection on a two-dimensional coordinate system with the number of cases where the intermediate inspection result is defective and the final inspection result is good on the X-axis, and the number of cases where the intermediate inspection result is good and the final inspection result is actually defective on the Y-axis.

[0134] By referring to such a graph, the user can determine whether the inspection criteria should be loosened or tightened while observing the relationship between "defective in intermediate inspection and good in final inspection" and "good in intermediate inspection and actually defective in final inspection". Specifically, for example, for the inspection program (1:Prog0001) plotted in the lower right of the graph illustrated in FIG. 18, it can be visually grasped that the number of "defective in intermediate inspection and good in final inspection" is large and there are almost no "good in intermediate inspection and actually defective in final inspection". Therefore, it can be easily grasped that misjudgments frequently occur because the current inspection criteria are too strict, and a review can be conducted in the direction of loosening the inspection criteria. For the inspection program (1:Prog0001) plotted in the lower right of the graph, it can be visually grasped that the number of "defective in intermediate inspection and good in final inspection" is large and there are almost no "good in intermediate inspection and actually defective in final inspection". Therefore, it can be easily grasped that misjudgments frequently occur because the current inspection criteria are too strict, and a review can be conducted in the direction of loosening the inspection criteria.

[0135] <Others> The description of the above embodiments is merely illustrative of the present invention, and the present invention is not limited to the above specific forms. The present invention can be variously modified and combined within the scope of its technical idea. For example, in each of the above examples, each functional unit is integrated in the inspection management device, but these functional units may be distributed to separate terminals and then communicatively connected. For example, the UI creation unit 107 and the teacher data update unit 108 may be provided in separate terminals, or the storage unit 40 may be separated as a data server. Further, the inspection management device according to Embodiment 2 and Embodiment 3 may be configured not to include the teacher data generation unit 105 and the data classification unit 106.

[0136] Also, regarding the processing flow of the inspection management device described in the embodiments, as long as the information necessary for generating teacher data can be obtained, there is no problem, so the processing from steps S101 to S104 may be in any order. Further, as long as the information on the target process of the teacher data and the information on the post-reflow inspection can be obtained, teacher data can be generated, so it is not necessary to obtain the inspection information of the processes that are not the target process. That is, it is not necessary to perform either of the processes in steps S101 and S102. Similarly, since repair information is not essential for generating teacher data, the process of obtaining repair information (i.e., step S104) does not have to be performed.

[0137] Also, in the above embodiment, the manufacturing line of the component mounting substrate is taken as an example for explanation (that is, the component mounting substrate is taken as an example of the product), but the present invention is naturally applicable to the manufacturing lines of products other than the component mounting substrate.

[0138] <Appendix 1> An inspection management system (100) used in a manufacturing line having a plurality of manufacturing processes related to the manufacture of a product, a plurality of manufacturing devices corresponding to the plurality of processes, and an inspection device that performs inspection based on an image of the product taken, Using the images used in each of the intermediate inspections related to one or more intermediate manufacturing processes in the manufacturing process and the final inspection result, which is the result of the final inspection related to the final manufacturing process of the product, teacher data generation means (105) for generating teacher data to be used for an inspection model corresponding to each of the intermediate inspections. An inspection management system.

[0139] <Appendix 2> The teacher data generation means When the final inspection result is a pass determination, assign a correct label of pass to the image used in the intermediate inspection that is the target of the teacher data generation. When the final inspection result is a fail determination, assign a correct label of fail to the image used in the intermediate inspection that is the target of the teacher data generation. The inspection management system according to Appendix 1, characterized in that.

[0140] <Appendix 3> The teacher data generation means Obtain repair presence / absence information on whether or not repair has been performed on the product when the result of the intermediate inspection is defective. When there is information indicating that repair has been performed on the product after the intermediate inspection that is the target of the teacher data generation, determine not to use the image of the product on which the repair has been performed for the teacher data. The inspection management system according to Appendix 1 or 2, characterized in that.

[0141] <Appendix 4> Display means (20) for at least displaying a first user interface for the user to check and correct the teacher data generated by the teacher data generation means. Input means (50) for receiving user operation inputs. The inspection management system according to any one of Appendices 1 to 3, characterized in that.

[0142] <Appendix 5> The first user interface includes a display that lists all the images used in the inspection for each manufacturing process of one of the products. The inspection management system according to supplementary note 4, characterized in that.

[0143] <Supplementary note 6> The inspection management system further includes classification means (106) for classifying to assist the user in confirming and correcting the teacher data with respect to the teacher data. The inspection management system according to supplementary notes 1 to 5, characterized in that.

[0144] <Supplementary note 7> Display means for at least displaying a second user interface for searching the generated teacher data, and Input means for receiving user operation inputs. The inspection management system according to supplementary note 6, characterized in that.

[0145] <Supplementary note 8> The second user interface includes a graph display that shows, for each inspection program to which at least the inspection model is applied, the quantity of the teacher data that can be used in the inspection model so that the quantity for each classification is known. The inspection management system according to supplementary note 7, characterized in that.

[0146] <Supplementary note 9> The graph display can display the teacher data filtered according to at least the type of defect specified in the inspection program, and The second user interface includes an area for displaying the images related to the teacher data displayed after filtering. The inspection management system according to supplementary note 8, characterized in that.

[0147] <Supplementary note 10> The intermediate inspection includes automatic inspection by a program and visual inspection by an inspector for the images. The classification means assigns a preset classification label to the teacher data to be classified according to a combination of the result of the automatic inspection and / or the result of the visual inspection performed using the image related to the teacher data to be classified, and the result of the corresponding final inspection. The inspection management system according to any one of Appendices 6 to 9, characterized in that.

[0148] <Appendix 11> The classification means is When both the result of the automatic inspection performed using the image related to the teacher data to be classified and the corresponding final inspection result are good product judgments, a classification label of "good product" is attached to the teacher data. When the result of the automatic inspection performed using the image related to the teacher data to be classified is a good product judgment and the corresponding final inspection result is a defective product judgment, a classification label of "to be confirmed for omission" is attached to the teacher data. When the result of the automatic inspection performed using the image related to the teacher data to be classified is a defective product judgment, the result of the visual inspection is a good product judgment, and the corresponding final inspection result is a good product judgment, a classification label of "overlooked" is attached to the teacher data. When the result of the automatic inspection performed using the image related to the teacher data to be classified is a defective product judgment, the result of the visual inspection is a good product judgment, and the corresponding final inspection result is a defective product judgment, a classification label of "visual result to be confirmed" is attached to the teacher data. When the result of the automatic inspection performed using the image related to the teacher data to be classified is a defective product judgment, the result of the visual inspection is a defective product judgment, and the corresponding final inspection result is a good product judgment, a classification label of "overlooked to be confirmed" is attached to the teacher data. When the result of the automatic inspection performed using the image related to the teacher data to be classified is a defective product judgment, the result of the visual inspection is a defective product judgment, and the corresponding final inspection result is a defective product judgment, a classification label of "actual defective" is attached to the teacher data. The inspection management system according to Appendix 10, characterized in that.

[0149] <Appendix 12> The intermediate inspection includes a formal inspection that treats the judgment result as a formal inspection result, and a provisional inspection that does not reflect the judgment result obtained by the inspection model. The inspection model evaluation means (110) further evaluates the accuracy of the inspection model by comparing the inspection result of the formal inspection with the judgment result in the provisional inspection. The inspection management system according to any one of Appendices 1 to 11, characterized in that.

[0150] <Appendix 13> Report creation means (141) that creates a report on the accuracy of the inspection model using the aggregation of the intermediate inspection results, which are the results of each of the intermediate inspections at predetermined intervals, and the aggregation of the final inspection results, which are the results of the final inspection. Output means (20) for outputting the report, further comprising. The inspection management system according to any one of Appendices 1 to 12, characterized in that.

[0151] <Appendix 14> The report includes, for each of a plurality of items of a first level of abstraction related to the intermediate inspection, an error judgment graph in which each of the items of a second level of abstraction included in the item of the first level of abstraction that was misjudged during the intermediate inspection is sorted by one of the number of misjudgments or the misjudgment rate, and the name of the item of the first level of abstraction for which one of the number of misjudgments or the misjudgment rate in the item of the second level of abstraction is larger and the error judgment graph for the item of the second level of abstraction related to the item are shown with higher priority, and includes a third user interface. The inspection management system according to Appendix 13, characterized in that.

[0152] <Appendix 15> The error judgment graph sorts one of the number of misjudgments or the misjudgment rate in descending order and highlights the items of the second level of abstraction for which the other of the number of misjudgments or the misjudgment rate exceeds a threshold value. The inspection management system according to Supplementary Note 14, characterized by the above.

[0153] <Supplementary Note 16> The report includes a diagram showing a plot on two-dimensional coordinates with the axis being the number of inspection models used in the intermediate inspection, the number of cases where the intermediate inspection result is defective and the final inspection result is non-defective, and the number of cases where the intermediate inspection result is non-defective and the final inspection result is defective. The inspection management system according to any one of Supplementary Notes 13 to 15, characterized by the above.

[0154] <Supplementary Note 17> Evaluation index calculation means (131) for calculating a predetermined index related to the accuracy evaluation of the inspection model, using a plurality of pieces of information on the intermediate inspection results, which are the results of each intermediate inspection, and information on the final inspection results, which are the results of the final inspection; And second inspection model evaluation means (132) for evaluating the accuracy of the inspection model based on the index. The inspection management system according to any one of Supplementary Notes 1 to 16, characterized by the above.

[0155] <Supplementary Note 18> The intermediate inspection and the final inspection include automatic inspection by a program and visual inspection by an inspector's visual judgment. The index includes a value based on a value obtained by dividing the number of inspection targets where the result of the automatic inspection in the intermediate inspection is defective and the result of the visual inspection in the final inspection is defective by the number of inspection targets where the result of the visual inspection in the final inspection is defective. The inspection management system according to Supplementary Note 17, characterized by the above.

[0156] <Supplementary Note 19> The intermediate inspection and the final inspection include automatic inspection by a program and visual inspection by an inspector's visual judgment. As the index, a value based on a value obtained by dividing the number of inspection targets for which the result of the automatic inspection in the intermediate inspection is defective and the result of the visual inspection in the final inspection is defective by the number of inspection targets for which the result of the automatic inspection in the intermediate inspection is defective is included. The inspection management system according to appended claim 17 or 18, characterized in that.

[0157] <Appended claim 20> The intermediate inspection and the final inspection include an automatic inspection by a program and a visual inspection by a visual determination of an inspector. As the index, a value based on a value obtained by dividing the number of inspection targets for which the result of the automatic inspection in the intermediate inspection is defective and the result of the final inspection is non-defective by the number of inspection targets for which the result of the automatic inspection in the intermediate inspection is defective is included. The inspection management system according to any one of appended claims 17 to 19, characterized in that.

[0158] <Appended claim 21> The intermediate inspection and the final inspection include an automatic inspection by a program and a visual inspection by a visual determination of an inspector. As the index, a value based on a value obtained by dividing the number of inspection targets for which the result of the automatic inspection in the intermediate inspection is non-defective and the result of the visual inspection in the final inspection is defective by the number of inspection targets for which the result of the visual inspection in the final inspection is defective is included. The inspection management system according to any one of appended claims 17 to 20, characterized in that.

[0159] <Appended claim 22> The product is a component mounting substrate. The final manufacturing process is a reflow process. The intermediate manufacturing process includes a printing process and a mounting process. The inspection management system according to any one of appended claims 1 to 21, characterized in that.

[0160] <Appended claim 23> An inspection management device (1, 2) used in a production line having a plurality of manufacturing processes related to the production of a product, a plurality of manufacturing apparatuses corresponding to the plurality of processes, and an inspection apparatus that performs inspection based on an image of the product, wherein: teacher data generation means (105) for generating teacher data used for an inspection model corresponding to each of the intermediate inspections, using the image used in each of the intermediate inspections related to one or more intermediate manufacturing processes in the manufacturing process and the final inspection result which is the result of the final inspection related to the final manufacturing process of the product; Inspection management device.

[0161] <Appendix 24> In a production line having a plurality of manufacturing processes related to the production of a product, a plurality of manufacturing apparatuses corresponding to the plurality of processes, and an inspection apparatus that performs inspection based on an image of the product, a method for generating teacher data for an inspection model used in each of the intermediate inspections related to one or more intermediate manufacturing processes in the manufacturing process, comprising: an image acquisition step (S101, S102) of acquiring the image used for inspection in each of the intermediate inspections; a final result acquisition step (S103) of acquiring a final inspection result which is the result of the final inspection related to the final manufacturing process; a generation step (S105) of generating the teacher data using the image and the final inspection result; Teacher data generation method.

[0162] <Appendix 25> In the generation step, when the final inspection result is a good product determination, a correct label of good product is assigned to the image used in the intermediate inspection which is the target of the teacher data generation; when the final inspection result is a defective product determination, a correct label of defective product is assigned to the image used in the intermediate inspection which is the target of the teacher data generation. The teacher data generation method according to Appendix 24, characterized in that.

[0163] <Appendix 26> A program for causing a computer to execute each step of the teacher data generation method described in Supplementary Note 24 or 25.

Explanation of Signs

[0164] 1, 2, 3, 4 ··· Inspection management device 10, 12, 13, 14 ··· Control unit 20 ··· Output unit 30 ··· Communication unit 40 ··· Storage unit 50 ··· Input unit 100 ··· Inspection management system X1 ··· Solder printing device X2 ··· Mounter X3 ··· Reflow oven Y1 ··· Post-solder printing inspection device Y2 ··· Post-mounting inspection device Y3 ··· Post-reflow inspection device S ··· Search item area P ··· Graph display area G ··· Inspection image display area

Claims

1. An inspection management system used in a production line having a plurality of manufacturing processes related to the production of a product, a plurality of manufacturing apparatuses corresponding to the plurality of processes, and an inspection apparatus for performing inspection based on an image obtained by photographing the product, comprising teacher data generation means for generating teacher data to be used for an inspection model corresponding to each of the intermediate inspections, using the images used in each of the intermediate inspections related to one or more intermediate manufacturing processes in the manufacturing process and the final inspection result which is the result of the final inspection related to the final manufacturing process of the product. Inspection management system.

2. The teacher data generation means, when the final inspection result is a pass determination, assigns a correct label of pass to the image used in the intermediate inspection which is the target of the teacher data generation, when the final inspection result is a fail determination, assigns a correct label of fail to the image used in the intermediate inspection which is the target of the teacher data generation. The inspection management system according to claim 1, characterized in that.

3. The teacher data generation means, acquires repair presence / absence information as to whether or not repair has been performed on the product when the result of the intermediate inspection is a fail, and when there is information indicating that repair has been performed on the product after the intermediate inspection which is the target of the teacher data generation, determines not to use the image of the product on which the repair has been performed for the teacher data. The inspection management system according to claim 1, characterized in that.

4. display means for at least displaying a first user interface for a user to confirm and correct the teacher data generated by the teacher data generation means, and input means for receiving an operation input of the user. The inspection management system according to claim 1, characterized in that.

5. The first user interface includes a display showing in a list all the images used in the inspections in each manufacturing process for one product. The inspection management system according to claim 4, characterized in that.

6. further comprising classification means for performing classification to assist the user in confirming and correcting the teacher data with respect to the teacher data. The inspection management system according to claim 1, characterized in that.

7. display means for at least displaying a second user interface for searching the generated teacher data. further comprising input means for receiving a user's operation input The inspection management system according to claim 6, characterized in that

8. The second user interface includes a graph display that shows, for at least each inspection program to which the inspection model is applied, the quantity of the teacher data that can be used in the inspection model so that the quantity for each classification is known. The inspection management system according to claim 7, characterized in that

9. The graph display can narrow down and display the teacher data according to at least the type of defect identified in the inspection program. The second user interface includes an area for displaying the image related to the teacher data that has been narrowed down and displayed. The inspection management system according to claim 8, characterized in that

10. The in-process inspection includes automatic inspection by a program and visual inspection by an inspector's visual judgment. The classification means assigns a preset classification label to the teacher data to be classified according to a combination of the result of the automatic inspection and / or the result of the visual inspection performed using the image related to the teacher data to be classified, and the result of the corresponding final inspection. The inspection management system according to claim 6, characterized in that

11. The classification means When both the result of the automatic inspection performed using the image related to the teacher data to be classified and the corresponding final inspection result are judged as good products, the classification label of "good product" is attached to the teacher data. When the result of the automatic inspection performed using the image related to the teacher data to be classified is judged as a good product and the corresponding final inspection result is judged as a defective product, the classification label of "missed defect to be confirmed" is attached to the teacher data. When the result of the automatic inspection performed using the image related to the teacher data to be classified is judged as a defective product, the result of the visual inspection is judged as a good product, and the corresponding final inspection result is judged as a good product, the classification label of "overlooked" is attached to the teacher data. When the result of the automatic inspection performed using the image related to the teacher data to be classified is judged as a defective product, the result of the visual inspection is judged as a good product, and the corresponding final inspection result is judged as a defective product, the classification label of "visual inspection result to be confirmed" is attached to the teacher data. ​ When the result of the automatic inspection performed using the image related to the teacher data to be classified is a failure determination, the result of the visual inspection is a failure determination, and the corresponding final inspection result is a pass determination, a classification label of "overlooked, need confirmation" is attached to the teacher data. When the result of the automatic inspection performed using the image related to the teacher data to be classified is a failure determination, the result of the visual inspection is a failure determination, and the corresponding final inspection result is a failure determination, a classification label of "actual failure" is attached to the teacher data. The inspection management system according to claim 10, characterized in that.

12. The intermediate inspection includes a main inspection that treats the determination result as an official inspection result, and a provisional inspection that makes a determination by the inspection model but does not reflect it as an inspection result. The inspection management system further includes inspection model evaluation means for evaluating the accuracy of the inspection model by comparing the inspection result of the main inspection and the determination result in the provisional inspection. The inspection management system according to claim 1, characterized in that.

13. Report creation means for creating a report on the accuracy of the inspection model using the aggregation of the intermediate inspection results, which are the results of each of the intermediate inspections, and the aggregation of the final inspection results, which are the results of the final inspection, for each predetermined period. The inspection management system further includes output means for outputting the report. The inspection management system according to claim 1, characterized in that.

14. The report shows, for each of a plurality of items of a first level of abstraction related to the intermediate inspection, a misjudgment graph in which each of the items of a second level of abstraction included in the item of the first level of abstraction that was misjudged during the intermediate inspection is sorted by one of the misjudgment count or the misjudgment rate, and the misjudgment count or the misjudgment rate in the items of the second level of abstraction. The inspection management system according to claim 13, characterized in that it includes a third user interface that preferentially shows the item name of the first level of abstraction for which one of the values of the misjudgment count or the misjudgment rate is larger and the misjudgment graph for the items of the second level of abstraction related to the item. The inspection management system according to claim 14, characterized in that.

15. The misjudgment graph sorts one of the misjudgment count or the misjudgment rate in descending order and highlights the items of the second level of abstraction for which the other of the misjudgment count or the misjudgment rate exceeds a threshold value. The inspection management system according to claim 14, characterized in that.

16. The report includes a diagram that plots, on two-dimensional coordinates with axes representing the number of inspection models used in the intermediate inspection that are identified by labels, the number of items with a non-conforming intermediate inspection result and a conforming final inspection result, and the number of items with a conforming intermediate inspection result and a non-conforming final inspection result. The inspection management system according to claim 13, characterized in that.

17. Evaluation index calculation means for calculating a predetermined index related to the accuracy evaluation of the inspection model, using a plurality of pieces of information on the intermediate inspection results, which are the results of each of the intermediate inspections, and information on the final inspection results, which are the results of the final inspection, respectively; Second inspection model evaluation means for evaluating the accuracy of the inspection model based on the index. The inspection management system according to claim 1, characterized in that.

18. The intermediate inspection and the final inspection include automatic inspection by a program and visual inspection by an inspector's visual judgment. The index includes a value based on the value obtained by dividing the number of inspection targets with a non-conforming result of the automatic inspection in the intermediate inspection and a non-conforming result of the visual inspection in the final inspection by the number of inspection targets with a non-conforming result of the visual inspection in the final inspection. The inspection management system according to claim 17, characterized in that.

19. The intermediate inspection and the final inspection include automatic inspection by a program and visual inspection by an inspector's visual judgment. The index includes a value based on the value obtained by dividing the number of inspection targets with a non-conforming result of the automatic inspection in the intermediate inspection and a non-conforming result of the visual inspection in the final inspection by the number of inspection targets with a non-conforming result of the automatic inspection in the intermediate inspection. The inspection management system according to claim 17, characterized in that.

20. The intermediate inspection and the final inspection include automatic inspection by a program and visual inspection by an inspector's visual judgment. The index includes a value based on the value obtained by dividing the number of inspection targets with a non-conforming result of the automatic inspection in the intermediate inspection and a conforming result of the final inspection by the number of inspection targets with a non-conforming result of the automatic inspection in the intermediate inspection. The inspection management system according to claim 17, characterized in that.

21. The intermediate inspection and the final inspection include automatic inspection by a program and visual inspection by an inspector's visual judgment. As the index, a value based on a value obtained by dividing the number of inspection targets for which the result of the automatic inspection in the intermediate inspection is a non-defective product and the result of the visual inspection in the final inspection is a defective product by the number of inspection targets for which the result of the visual inspection in the final inspection is a defective product is included. The inspection management system according to claim 17, characterized in that.

22. The product is a component mounting substrate, The final manufacturing process is a reflow process, The intermediate manufacturing process includes a printing process and a mounting process, The inspection management system according to any one of claims 1 to 21, characterized in that.

23. An inspection management device used in a manufacturing line having a plurality of manufacturing processes related to the manufacture of a product, a plurality of manufacturing devices corresponding to the plurality of processes, and an inspection device that performs inspection based on an image obtained by photographing the product, Teacher data generation means for generating teacher data to be used for an inspection model corresponding to each of the intermediate inspections, using the image used in each of the intermediate inspections related to one or more intermediate manufacturing processes in the manufacturing process and the final inspection result which is the result of the final inspection related to the final manufacturing process of the product. Inspection management device.

24. In a manufacturing line having a plurality of manufacturing processes related to the manufacture of a product, a plurality of manufacturing devices corresponding to the plurality of processes, and an inspection device that performs inspection based on an image obtained by photographing the product, a method for generating teacher data for an inspection model used in each of the intermediate inspections related to one or more intermediate manufacturing processes in the manufacturing process, An image acquisition step of acquiring the image used for inspection in each of the intermediate inspections, A final result acquisition step of acquiring a final inspection result which is the result of the final inspection related to the final manufacturing process, A generation step of generating the teacher data using the image and the final inspection result, including. Teacher data generation method.

25. In the generation step, When the final inspection result is a non-defective determination, a correct label of non-defective is assigned to the image used in the intermediate inspection that is the target of the teacher data generation, When the final inspection result is a defective determination, a correct label of defective is assigned to the image used in the intermediate inspection that is the target of the teacher data generation. The teacher data generation method according to claim 24, characterized in that.

26. A program for causing a computer to execute each step of the teacher data generation method according to claim 24 or 25.

Images