Learning system, defect prediction system, learned model, learning method, and defect prediction method

Abstract

To provide a learning system capable of predicting product defects which may occur in a product production line.SOLUTION: A learning system inputs, into an unlearned model 14, teaching data including change point information 22 indicating a change point regarding at least one of operators, production facilities, and work content in a production line which produces products, and defect information 24 indicating product defects having occurred in the production line with the change point. The learning system inputs the change point information 22 by performing machine learning of the unlearned model 14, and generates a learned model 15 which outputs defect prediction information indicating product defects which are predicted to occur at a change point indicated by the change point information 22.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present disclosure relates to a learning system, a defect prediction system, a learned model, a learning method, and a defect prediction method.

Background Art

[0002] Patent Document 1 discloses a technique for estimating the cause of a defect generated in a manufacturing process from knowledge data represented in a plurality of hierarchical structures.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, although the technique according to Patent Document 1 can estimate the cause of a defect such as a discovered defect, there is no device for preventing overlooking of product defects in the inspection process, and overlooking of abnormalities cannot be avoided.

[0005] In order not to overlook product defects in the inspection process, it is desirable to predict the occurrence and content of product defects before they are discovered in the inspection process or the like. However, since normal manufacturing conditions and the like are originally set so that product defects do not occur, it is difficult to predict the occurrence of product defects from normal manufacturing conditions and the like.

[0006] An object of the present disclosure is to provide a learning system, a defect prediction system, a learned model, a learning method, and a defect prediction method capable of predicting product defects that may occur in a product production line in view of the above-described problems.

Means for Solving the Problems

[0007] The learning system according to the present disclosure inputs teacher data including change point information indicating a change point related to at least one of an operator, production equipment, and work content in a production line for producing a product, and defect information indicating a defect of a product generated in the production line along with the change point, into a learning model, and machine-learns the learning model, so as to input the change point information and generate a learned model that outputs defect prediction information indicating a defect of a product that is predicted to occur with respect to the change point indicated by the change point information.

[0008] In this way, the learning system can machine-learn teacher data including the change point information and the defect information, and generate a learned model that outputs a defect of a product that is predicted to occur with respect to a change point. Therefore, according to the learning system, it is possible to obtain a learned model capable of predicting a defect of a product that may occur in a production line of a product when a change point occurs. And it is considered that product defects are likely to occur frequently when a change point occurs. Therefore, according to the learning system, it is possible to obtain a learned model capable of predicting a defect of a product that may occur in a production line.

[0009] Further, the defect information included in the teacher data may include information indicating a defect discovered after passing an inspection process included in the production line. Thereby, in the generated learned model, it becomes possible to predict also a defect that could not be discovered in the inspection process.

[0010] Further, the defect information included in the teacher data may include information indicating a defect of a product that occurred during a period determined according to the content of the change point indicated by the change point information. Thereby, information indicating a defect that occurred during a period according to the content of the change point can be machine-learned in association with the change point, and a learned model with higher prediction accuracy can be generated.

[0011] Alternatively, the production line may be a line including an assembly operation for assembling a product. A production line including an assembly operation is assumed to often have a change point, and it becomes easier to execute machine learning associating the change point with a defect. Therefore, with such a configuration, it is possible to obtain a learned model capable of accurately predicting possible defects in a production line including an assembly operation where change points often occur.

[0012] The defect prediction system according to the present disclosure inputs the change point information indicating a change point related to at least one of an operator, production equipment, and work content in a production line for producing a product into a learned model obtained by performing machine learning using teacher data including the change point information and defect information indicating a defect of a product generated in the production line along with the change point, and outputs defect prediction information indicating a defect of the product predicted to occur with respect to the change point indicated by the change point information.

[0013] As described above, the defect prediction system can output a defect of a product predicted to occur with respect to a change point by using a learned model obtained by performing machine learning on teacher data including the change point information and the defect information. Therefore, according to the defect prediction system, when a change point occurs, it is possible to predict a defect of a product that may occur in the production line of the product. And it is considered that defects of the product often occur when a change point occurs. Therefore, according to the defect prediction system, it is possible to predict a defect of a product that may occur in the production line.

[0014] Alternatively, the defect information included in the teacher data may include information indicating a defect discovered after passing through an inspection process included in the production line. Thereby, it becomes possible to predict also a defect that could not be discovered in the inspection process.

[0015] Further, the defective information included in the teacher data may include information indicating defects of products that occurred during a period determined according to the content of the change point indicated by the change point information. Thereby, it is possible to accurately predict defects using a learned model with higher prediction accuracy, in which information indicating defects that occurred during a period according to the content of the change point is machine-learned in association with the change point.

[0016] Further, the defective prediction information may be notified in a process after the change point indicated by the change point information input to the learned model. Thereby, the prediction result of the defect according to the change point can be notified at an appropriate time.

[0017] Further, when the defective prediction information includes information indicating a plurality of defects, information to be notified may be determined according to the importance of each of the plurality of defects, and the determined information may be notified. Thereby, even when a plurality of defects are predicted, necessary defective prediction information can be notified.

[0018] Further, the content of the inspection to be performed corresponding to the defective prediction information may be notified in a process after the change point indicated by the change point information input to the learned model. Thereby, the content of the inspection to be performed corresponding to the defect predicted according to the change point can be notified at an appropriate time.

[0019] Further, when there are a plurality of contents of the inspection to be performed corresponding to the defective prediction information, information to be notified may be determined according to the importance of the defect corresponding to each of the plurality of inspection contents, and the determined information may be notified. Thereby, even when there are a plurality of contents of the inspection to be performed for the predicted defect, necessary inspection contents can be notified.

[0020] Further, the content of the inspection to be performed corresponding to the defective prediction information may be notified, and the learned model may be updated based on the result of performing the inspection of the notified content. Thereby, the accuracy of the learned model can be improved.

[0021] Further, the production line may be a line including an assembly operation for assembling a product. It is assumed that a production line including an assembly operation often has a change point, and it becomes easier to execute machine learning associating the change point with a defect. Therefore, with such a configuration, it is possible to accurately predict a possible product defect in a production line including an assembly operation where change points often occur.

[0022] The learned model according to the present disclosure uses teacher data including change point information indicating a change point related to at least one of an operator, production equipment, and work content in a production line for producing a product, and defect information indicating a defect of a product generated in the production line along with the change point, inputs the change point information, and is machine-learned so as to output defect prediction information indicating a possible product defect with respect to the change point indicated by the change point information.

[0023] In this way, the learned model can be a model that machine-learns teacher data including the change point information and the defect information and outputs a possible product defect predicted for the change point. Therefore, according to the learned model, when a change point occurs, it is possible to predict a possible product defect in the production line of the product. And it is considered that product defects often occur when a change point occurs. Therefore, according to the learned model, it is possible to predict a possible product defect in the production line.

[0024] The learning method according to the present disclosure inputs teacher data including change point information indicating a change point related to at least one of an operator, production equipment, and work content in a production line for producing a product, and defect information indicating a defect of a product generated in the production line along with the change point, into a learning model, and machine-learns the learning model to generate a learned model that inputs the change point information and outputs defect prediction information indicating a possible product defect predicted for the change point indicated by the change point information.

[0025] According to the learning method, the same effect as that of the learning system can be expected.

[0026] The defective prediction method according to the present disclosure uses teacher data including change point information indicating a change point related to at least one of an operator, production equipment, and work content in a production line for producing a product, and defective information indicating a defect of a product generated in the production line along with the change point, and inputs the change point information into a learned model obtained by machine learning, and outputs defective prediction information indicating a defect of a product that is predicted to occur with respect to the change point indicated by the change point information.

[0027] According to the defective prediction method, the same effect as that of the defective prediction system can be expected.

Advantages of the Invention

[0028] According to the present disclosure, it is possible to provide a learning system, a defective prediction system, a learned model, a learning method, and a defective prediction method capable of predicting a defect of a product that may occur in a production line of a product.

Brief Description of the Drawings

[0029]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Mode for Carrying Out the Invention

[0030] Hereinafter, the embodiments will be described in detail with reference to the drawings, but the invention according to the claims is not limited to the following embodiments. Also, not all of the configurations described in the embodiments are essential as means for solving the problems. In each drawing, the same reference numerals are assigned to the same elements, and redundant descriptions are omitted as necessary for clarity of explanation.

[0031] (Embodiment) First, a configuration example of the defect prediction system according to the present embodiment will be described with reference to FIG. 1. FIG. 1 is a block diagram showing a configuration example of the defect prediction system according to the present embodiment. FIG. 2 is a schematic diagram showing an example of a production line, and FIG. 3 is a schematic diagram showing an example of the classification of product defects. Further, FIG. 4 is a diagram showing an example of the explanatory variables and target variables of the learned model used in the defect prediction system of FIG. 1.

[0032] The defect prediction system 100 shown in FIG. 1 is a system that predicts product defects in a production line for manufacturing products using a learned model 15. In the following, a configuration example in which the defect prediction system 100 includes a learning system for training an untrained model 14 will be described. However, if the learned model 15 is provided without including the learning system, it is possible to predict product defects. Note that an example in which the learning system and the defect prediction system are configured separately will be described later as another embodiment.

[0033] In the following, as the production line to be predicted, the production line 40 illustrated in FIG. 2 will be described as an example, but it is not limited to this. In the illustrated production line 40, there are change points 41 to 44 and 46 to 48 as locations that can be change points to be described later. Of course, there may be one or more locations that can be change points. Also, in the illustrated production line 40, there are inspection points 45 and 49 as locations that are inspection points. The inspection point 49 is the final inspection point, and it is sufficient that at least one inspection point is provided in the production line 40.

[0034] Also, the type of product and the production process (manufacturing process) are not limited. In the following, the description will be made on the premise that the product is a vehicle, but it may be other types of products such as electronic devices, foods, and furniture. Also, product defects can refer to product malfunctions, manufacturing scratches on the product, etc., which can be regarded as defects when the product is distributed in the market.

[0035] As illustrated in FIG. 3, defects that can occur in a vehicle can include tightening-related issues related to tightening and design-related issues related to design. The tightening-related issues can include diagonal insertion where parts are attached diagonally, hole misalignment where the holes between parts are misaligned, insufficient tightening torque, excessive baking during baking and painting, and unfastened where parts are not tightened. The design-related issues can include scratches, dirt, wrinkles, deformation, part lifting, etc.

[0036] As shown in FIG. 1, the defect prediction system 100 according to the present embodiment can include an information processing apparatus 10 that predicts defects in products, a vehicle production result database (DB) 20, and a plurality of terminals 31 to 36. The information processing apparatus 10, the vehicle production result DB 20, and the terminals 31 to 36 can be communicably connected via a network N.

[0037] Here, the network N is a communication line network such as the Internet, an intranet, a mobile phone network, or a LAN (Local Area Network). Although an example in which the defect prediction system 100 includes six terminals 31 to 36 will be described, terminals may be provided as long as they can input the change points described later, input defects, and view the prediction results. Therefore, the number of terminals may be one or more.

[0038] The information processing apparatus 10 is a main component of the defect prediction system 100. The information processing apparatus 10 can include a control unit 11 that controls the entire apparatus, a communication unit 12, and a storage unit 13 in which an unlearned model 14 or a learned model 15 is stored.

[0039] The information processing apparatus 10 can function as a learning apparatus in a learning stage for machine learning of the unlearned model 14, and can function as a defect prediction apparatus in an operation stage for executing prediction using the learned model 15. Note that the learning apparatus and the defect prediction apparatus can also be configured independently, and such a configuration example will be described later with reference to FIGS. 9 and 10.

[0040] The control unit 11 can be implemented by an integrated circuit, and can be implemented by, for example, a processor such as a CPU (Central Processing Unit), MPU (Micro Processor Unit), or GPU (Graphics Processing Unit), a working memory, and a non-volatile storage device. A control program executed by the processor is stored in this storage device, and the processor reads and executes the program in the working memory to perform the functions of the control unit 11. Note that this storage device can also be a part of the storage unit 13.

[0041] Also, as can be understood from the description of the control unit 11, the information processing device 10 can be configured by a general-purpose computer. When the information processing device 10 functions as a learning device, an unlearned model 14 can be stored in the general-purpose computer, and teacher data can be input to the unlearned model 14 for machine learning to generate a learned model 15. When the information processing device 10 functions as a failure prediction device, the learned model 15 can be stored in the general-purpose computer, information can be input to the learned model 15, and information indicating a prediction result can be output.

[0042] Also, the control unit 11 can include a first control unit with high processing power used during learning and a second control unit used during operation. The second control unit can keep the processing power lower than that of the first control unit.

[0043] The communication unit 12 can include a communication interface that communicates with the vehicle production record DB 20 and terminals 31 to 36 via the network N.

[0044] The memory unit 13 is a storage device in which the unlearned model 14 or the learned model 15 is stored. In the learning stage, the unlearned model 14 is stored in the memory unit 13. By inputting teacher data into the unlearned model 14 and performing machine learning, a learned model 15 can be generated, and the learned model 15 can be stored in the memory unit 13. The algorithm of the unlearned model 14, that is, the algorithm of the learned model 15, can be of any type as long as it performs supervised learning, and hyperparameters and the like are not limited either. The teacher data used for machine learning, the input data and output data during operation will be described later.

[0045] Although not shown in the figure, the vehicle production record DB 20 can be realized by combining one or more server computers and storage devices connected to the network N, or it may be incorporated in the information processing device 10, for example. The vehicle production record DB 20 reads and writes each data for access via the network N and returns the processing result to the access source.

[0046] The vehicle production record DB 20 is a database that manages by associating product information (that is, vehicle information) 21, which is information on vehicles produced on the vehicle production line 40, change point information 22 about the occurred change points, and defect information 23 indicating defects discovered by inspection of products on the production line 40. In the vehicle production record DB 20, change point information 22 and defect information 23 are associated with the vehicle indicated by the product information 21. However, for some vehicles, there may be no change points or defects during production, so there is also product information 21 with which at least one of the change point information 22 and the defect information 23 is not associated. In addition, the defect information 23 can include information indicating defects discovered not only after the product inspection on the production line 40 but also after the product has been distributed in the market.

[0047] The change point information 22 is information indicating a change point related to at least one of the operator, production equipment, and work content in the production line 40 for producing the product. Examples of change points of the operator include, for example, a new worker performing the work, or an operator temporarily replacing another during the normal operator's vacation. Therefore, the change point of the operator can refer to the point where the operator changes, or the point where the rank of the operator performing the work changes after ranking the operators by skill level or the like. The change point of the production equipment refers to the point where one of the equipment on the production line 40 is replaced. For example, even when a tool is replaced from an old one to a new one, it is treated as if a change point of the production equipment has occurred. The change point of the work content can refer to the point where the work procedure or work method is changed for the purpose of improving work efficiency or the like. Also, examples of the change point of the work content include a situation where a mistake occurs in the standard work and the correction is made.

[0048] Actually, examples of defects caused by these change points include, for example, a defect where parts are not properly assembled in the work performed by an inexperienced operator who replaced a vacationing operator. Also, examples of defects that occur include a defect where the scratches and dirt made by the operator were corrected, but another part got dirty during the correction. Also, examples of defects that occur include various defects such as a new worker who did not notice that the process had changed and performed the work as it was, resulting in unworked areas.

[0049] The terminals 31 to 36 can all be information processing devices such as a computer, a smartphone, or a dedicated processing terminal, and can access the vehicle production result DB 20 and the information processing device 10 via the network N.

[0050] The users U1, U2, U3, U4, and U5 are each persons who perform the work of inputting the change points of people, the change points of work, the change points of equipment and tools, the discovered defects, and the defects that have entered the market regarding the production line 40 for producing the vehicle.

[0051] In the learning stage, users U1, U2, U3, U4, and U5 respectively use terminals 31, 32, 33, 34, and 35 to input corresponding information, and each of terminals 31 to 35 can transmit the input information to the vehicle production performance DB20 via the network N as information for generating learning teacher data.

[0052] Also, in the operation stage, users U1, U2, and U3 respectively use terminals 31, 32, and 33 to input corresponding information, and each of terminals 31 to 33 can transmit the input information to the information processing device 10 as input information for performing defect prediction in the operation stage. Also, in the operation stage, simultaneously with or as needed for such transmission, each of terminals 31 to 33 can transmit the input information to the vehicle production performance DB20 via the network N as information for updating the learning teacher data.

[0053] Also, user U6 is a person who views the prediction results in the operation stage. In the operation stage, terminal 36 can receive a notification of the prediction results from the information processing device 10 or access the information processing device 10 to view the prediction results.

[0054] First, the case where the information processing device 10 functions as a learning device will be described. In this case, the defect prediction system 100 will function as a learning system.

[0055] The control unit 11 inputs teacher data including the change point information 22 and the defect information indicating the defects of the products generated on the production line 40 accompanied by the change point indicated by the change point information 22 among the change point information 22 and the defect information 23 into the unlearned model 14.

[0056] The input of teacher data can be implemented by acquiring information from the vehicle performance DB20 via the communication unit 12. Among the defect information 23, the defect information indicating the defects of products that occurred in the production line 40 along with the change points indicated by the change point information 22 can be the defect information 23 associated with the change point information 22 in the vehicle performance DB20, and hereinafter, it is referred to as defect information 24 for the sake of distinction. Also, the teacher data can include product information 21.

[0057] That is, the teacher data can include product information 21 and change point information 22 as explanatory variables 201a, and defect information 24 as a target variable 201b, like the teacher data 201 exemplified in FIG. 4.

[0058] However, as described above, the defects of products that occurred in the production line 40 can include the defects of vehicles discovered after being distributed in the market. That is, the defect information 24 included in the teacher data can include information indicating defects discovered after passing through the inspection processes included in the production line 40. This inspection process includes the finished product inspection process at the inspection point 49 for inspecting the product as a finished product, but can also include the in-process inspection process at the inspection point 45 for inspecting the product during production. Thereby, in the generated learned model 15, it becomes possible to predict defects that could not be discovered in the inspection processes employed in the production line 40, such as defects discovered after being distributed in the market.

[0059] In addition, the defective information 24 included in the teacher data can include information indicating defects in products that occurred during a period determined according to the content of the change point indicated by the change point information 22. In this way, according to the content of the change point, the occurrence period of product defects to be adopted as the defective information 24 in the teacher data can be determined, that is, the occurrence period of defects to be adopted as teacher data can be changed according to the content of the change point. Determining the period to be adopted as the defective information 24 in the teacher data means determining the period of the defective information 24 used for learning, and as a result, means determining the period of the defective information (defect prediction information) to be predicted. In fact, depending on the content of the change point, there are cases where it is sufficient to learn the defects of the products at the timing of the change point, and cases where it is necessary to learn the defects of the products during a predetermined time after the occurrence of the change point. Therefore, such a configuration can be said to be beneficial.

[0060] As a result, information indicating defects that occurred during a period corresponding to the content of the change point can be machine-learned in association with the change point, and a learned model 15 with higher prediction accuracy can be generated. Then, the prediction accuracy can be improved by predicting product defects using such a learned model 15.

[0061] Moreover, the production line 40 may be a line including an assembly operation for assembling products. The assembly operation can refer to an operation of assembling parts. It is assumed that production lines including assembly operations are likely to have change points, and it becomes easier to execute machine learning associating change points with defects. Therefore, with such a configuration, it is possible to obtain a learned model 15 that can accurately predict possible product defects for the production line 40 including assembly operations where change points often occur. Then, the prediction accuracy can be improved by predicting product defects using such a learned model 15.

[0062] Further, the teacher data can also include at least one of the change point information 22 not associated with the defective information 23 and the defective information 23 not associated with the change point information 22. For example, this is because it can be learned that the change point indicated by the change point information 22 does not affect the occurrence of defects or that the defective information 23 indicates a defect that occurs independently of the change point.

[0063] Then, the control unit 11 generates a learned model 15 that, during operation, inputs the change point information and outputs defective prediction information indicating a product defect that is predicted to occur for the change point indicated by the change point information, by performing machine learning on the unlearned model 14 into which the above-described teacher data is input. The defective prediction information output from the learned model 15 is information corresponding to the defective information 24, and can be information equivalent to the content included in the defective information 24.

[0064] In this way, the information processing apparatus 10 functioning as a learning apparatus can perform machine learning on teacher data including the change point information 22 and the defective information 24, and generate a learned model 15 that outputs a product defect that is predicted to occur for a change point. Therefore, according to the information processing apparatus 10, it is possible to obtain a learned model 15 that can predict a product defect that may occur in the product production line when a change point occurs. And it is considered that product defects are likely to occur when a change point occurs. Thus, it can be said that according to the information processing apparatus 10, it is possible to obtain a learned model 15 that can predict a product defect that may occur in the production line 40.

[0065] That is, the trained model 15 obtained in this way is a model that performs machine learning on the teacher data including the change point information 22 and the defect information 24, and outputs the predicted product defects that may occur for the change point. Therefore, the trained model 15 can predict the product defects that may occur in the product production line 40 when a change point occurs. And it is considered that product defects are likely to occur when a change point occurs. Thus, according to the trained model 15, the product defects that may occur in the production line 40 can be predicted.

[0066] Next, the case where the information processing apparatus 10 functions as a defect prediction apparatus will be described. The control unit 11 inputs the change point information into the trained model 15 generated as described above, and outputs defect prediction information indicating the product defects that are predicted to occur for the change point indicated by the change point information. As described above, the defect prediction information output from the trained model 15 is information corresponding to the defect information 24, and can be information equivalent to the content included in the defect information 24. In this way, the control unit 11 can function as a prediction unit that performs defect prediction.

[0067] In this way, the trained model 15 can be a model for causing the computer to function so as to input the change point information 22 and output defect prediction information indicating the product defects that are predicted to occur for the change point indicated by the change point information 22.

[0068] In addition, the control unit 11 can notify the defect prediction information via the communication unit 12 at a process after the change point indicated by the change point information 22 input into the trained model 15. The notification destination can be determined as a predetermined notification destination such as the terminal 36. Thereby, the prediction result of the defect corresponding to the change point can be notified at an appropriate time.

[0069] Further, when the defect prediction information includes information indicating a plurality of defects, the control unit 11 may determine information to be notified according to the importance of each of the plurality of defects via the communication unit 12, and notify the determined information. The notification destination here can also be determined as a predetermined notification destination such as the terminal 36. Thereby, even when a plurality of defects are predicted, only necessary defect prediction information can be notified. Also in this case, the timing of notification can be a process after the change point.

[0070] Further, when there are a plurality of input change points, the control unit 11 may notify the defect prediction information in a process after the last change point, or may notify the defect prediction information about that change point in a process after at least the corresponding change point. The notification destination here can also be determined as a predetermined notification destination such as the terminal 36. Also, the control unit 11 may change the notification destination according to the importance.

[0071] Further, the control unit 11 can also notify the content of the inspection to be performed corresponding to the defect prediction information in a process after the change point indicated by the change point information 22 input to the learned model 15. The notification destination here can also be determined as a predetermined notification destination such as the terminal 36. Thereby, the content of the inspection to be performed corresponding to the defect predicted according to the change point can be notified at an appropriate time, and thereby the inspector can receive the notification and perform the inspection.

[0072] Further, when there are a plurality of inspection contents to be performed corresponding to the defect prediction information, the control unit 11 may determine information (inspection content) to be notified according to the importance of the defect corresponding to each of the plurality of inspection contents, and notify the determined information. The notification destination here can also be determined as a predetermined notification destination such as the terminal 36. Thereby, even when there are a plurality of inspection contents to be performed for the predicted defect, necessary inspection contents can be notified. Also in this case, notification in a process after the change point as described above may be adopted.

[0073] Further, in a configuration where the control unit 11 notifies the content of the inspection to be performed in response to the failure prediction information, the control unit 11 may update (re-learn) the learned model 15 based on the result of performing the inspection of the notified content. The result of performing the inspection can be input from the terminal 36 or the like and registered in the vehicle performance DB 20. Thereby, the accuracy of the learned model 15 can be improved.

[0074] As described above, the failure prediction system 100 can output a product failure that is predicted to occur for a change point by using the learned model 15 obtained by machine learning the teacher data including the change point information 22 and the failure information 24. Therefore, according to the failure prediction system 100, it is possible to exclude from the data used for prediction the failures not caused by the change point, and when a change point occurs, it is possible to predict the product failures (failures with a high possibility of occurring) that may occur in the product production line. And it is considered that product failures occur frequently when a change point occurs. Therefore, according to the failure prediction system 100, it is possible to predict the product failures that may occur in the production line 40. Further, thereby, it is possible to prevent overlooking product failures in the inspection process, for example, before they are discovered in the inspection process or the like.

[0075] Next, an example of the learning process in the failure prediction system 100 will be described with reference to FIG. 5. FIG. 5 is a flowchart for explaining an example of the processing (learning process) in the learning stage executed by the information processing device 10 in the failure prediction system 100. Hereinafter, an example will be described in which the production line 40 is an assembly line configured by an assembly process, but the example is not limited thereto.

[0076] First, users U1 to U3 each output and record change point information 22 indicating change points in the assembly process from terminals 31 to 33 in the vehicle production results DB 20, and user U4 outputs and records defect information 24 indicating defects generated from terminal 34 in the vehicle production results DB 20 (step S1). The recording of change points can also be executed, for example, when the operator of the assembly work acts as any one of users U1 to U3 when a vehicle comes by. Also, the recording of defects can be executed, for example, when the operator of the assembly work acts as user U4 when a vehicle comes by.

[0077] Also, in step S1, after distributing the product to the market, the process of outputting and recording defect information 24 indicating defects occurring in the market from terminal 35 by user U5 in the vehicle production results DB 20 can also be included.

[0078] Next, the information processing device 10 refers to the vehicle production results DB 20 and determines whether the vehicle information 21 has defect information 24 (step S2). When the information processing device 10 determines that the vehicle information 21 has defect information 24 (YES in step S2), it extracts the defect information 24 and presents it to terminal 36, and user U6 classifies the type of the defect, the stage at which the defect was discovered, and the severity (importance) of the defect from the presented information (step S3).

[0079] In step S3, by classifying the defects, they can be labeled for each classification. For this classification, the classification method illustrated in FIG. 3 or the like can be adopted. By classifying the defects as in step S3, first, effects such as improving the accuracy of machine learning to be performed later can be achieved.

[0080] However, the classification method is not limited to using the classification system illustrated in FIG. 3, but the generated defects can be classified according to the operations in which they occurred and the types of defects, and further classified for each detailed phenomenon of the defects. By classifying the defects by type, it becomes possible to reflect the tendency of the defects in the learned model 15 and to instruct an additional inspection method for predicted defects later.

[0081] Also, by performing classification as in step S3, secondly, an effect can be achieved such that it can be reflected in the ranking of the priority of performing the prediction inspection. That is, at the stage where a defect is found, the priority of predicted defects to be found especially in additional inspections can be determined according to the severity of the defect. Here, as described above, there are roughly two types of defects: defects found during inspection on the production line 40 (i.e., within the factory), and defects that have leaked into the market. In particular, the latter type of defect that has leaked into the market will cause trouble to vehicle users, so it can be said that it is a defect of particularly high importance for vehicle manufacturers. Furthermore, among the defects that have leaked into the market, those with a shorter time from the completion of the vehicle to the discovery can be positioned as being of extremely high importance in order to maintain the credibility of the vehicle. On the other hand, defects found within the factory are often those discovered by basic pre-determined inspections, and it can be said that they have a lower priority among the defects.

[0082] Next, the severity of the defect will be explained. The importance of the defect can be roughly classified into four levels, for example, R1, R2, R3, and R4. Here, R1 can be regarded as something necessary for safety with the risk of accidents, fires, etc. Also, R2 can be those that do not fall under R1, (1) important failures of the functions of the vehicle and those with the risk thereof, or (2) failures that conflict with safety standards and relevant laws and regulations related to the safety of the country or region. Also, R3 can be failures of functions and those with the risk thereof, and R4 can be problems that do not fall under any of R1 to R3.

[0083] Following step S3, the user U6 registers (records) the information indicating the classified defects as defect information 23 in the vehicle production result DB20 using the terminal 36 (step S4). Note that the operations in steps S3 and S4 can be appropriately performed using a user interface that supports such operations on the terminal 36.

[0084] After the process of step S4 and when the result in step S2 is NO, the information processing apparatus 10 refers to the vehicle production result DB20 and associates the product (in this example, the vehicle) in which the change point is involved with the defect in that vehicle (step S5). That is, in step S5, in the vehicle production result DB20, the recorded change point information 22 and the defect information 24 are associated with each other. This association can be automatically performed based on the product information 21. Note that the association for the information determined as NO in step S2 will not be performed.

[0085] In step S5, by associating the vehicle in which the change point is involved with the defect in that vehicle, their causal relationship can be learned, and it becomes possible to predict defects. The association can be executed based on the product information 21. Regarding the setting of the change point, as described above, it is preferable to perform the association in this step S5 so as to be beneficial, that is, so that accurate prediction is possible.

[0086] For example, regarding a defect where the assembly method of a certain part is incorrect, when removing the dirt on the part discovered during assembly, the part may be held in an unusual way, and the cause may be that the attachment method was incorrect. Therefore, in order to address this, as the change point, it is possible to adopt the point where an unusual cleaning operation of the part was performed. Also, regarding a defect where a certain operation has not been carried out, the cause may be that an unskilled worker whose work has changed due to equipment changes has rushed through the work and the procedure has been disrupted and the operation has been skipped. Therefore, in order to address this, as the change point, it is possible to adopt the point where there has been a change in the operator (especially a change to an operator with low proficiency) or a change in the equipment.

[0087] After step S5, the information processing apparatus 10 inputs the associated change point information 22 and defect information 24 as teacher data into the unlearned model 14, executes machine learning (step S6), saves it as the generated learned model 15 (step S7), and ends the process.

[0088] Next, an example of defect prediction processing in the defect prediction system 100 will be described with reference to FIG. 6. FIG. 6 is a flowchart for explaining an example of defect prediction processing, which is processing in the operation stage in the defect prediction system 100.

[0089] First, users U1, U2, and U3 use terminals 31, 32, and 33 respectively to input known change points before or during production on production line 40 and transmit them to information processing device 10. Information processing device 10 inputs information indicating the change points (information corresponding to change point information 22) into learned model 15 (step S11). The change points to be input can be change points that have occurred but not been input and change points that are known to occur.

[0090] Next, information processing device 10 lists up defect prediction information indicating defects that may occur from the change points indicated by the output result from learned model 15 (step S12), ranks the list of possible defects based on the importance of the defects, and assigns priorities (step S13).

[0091] Next, information processing device 10 selects, from among a plurality of previously stored methods of additional inspection (contents indicating additional inspection) capable of detecting the defect for each type of predicted defect, the necessary amount (step S14).

[0092] In step S14, by selecting a method of additional inspection capable of detecting the defect for each type of predicted defect, it is possible to output an additional inspection corresponding to the type of defect. For example, in the case of a defect related to tightening, it is possible to select a method of additional inspection such as retightening the target bolt or visually checking whether the bolt is seated in a state where it is firmly tightened to the bottom. Also, in the case of a scratch related to the design, it is possible to select a method of additional inspection such as actually touching and checking by hand, or in the case of dirt, visually checking while changing the angle of light and shining it. Further, at the time of additional inspection, not only the inspection procedure but also the parts and locations of the vehicle where the change point has occurred are clearly shown on a monitor or the like, so that the inspector can respond to the additional inspection for each vehicle.

[0093] Next, following step S14, the information processing apparatus 10 calculates the time required for the additional inspection when the selected additional inspection method is adopted (step S15). Thereafter, the information processing apparatus 10 selects predicted defects and additional inspections within the range that falls within the predetermined inspection time in descending order of priority, and transmits information indicating the content of the selected additional inspections and the predicted defects to the terminal 36 and presents it to the user U6 (step S16), and ends the process. Thereafter, the user U6 instructs the operator to perform the inspection, obtains the inspection result, and when a defect occurs, registers it in the vehicle production result DB 20 as information for updating the learned model 15 from the terminal 36 so that it can be used at the time of the next model update.

[0094] In the present embodiment, as described in step S16, predicted defects and additional inspections are selected in descending order of priority within the range that falls within the predetermined inspection time and presented to the operator, that is, the time required for the additional inspection is calculated in advance and only the additional inspections that can be performed within the predetermined inspection time can be carried out. By doing so, it is possible to perform additional inspections on predicted defects during the flow operation and prevent an increase in the burden on the operator on the production line 40.

[0095] Next, with reference to FIGS. 7 and 8, an example of the overall processing procedure in the operation stage will be described. FIG. 7 is a flowchart for explaining an example of the overall processing procedure in the operation stage in the defect prediction system 100, and FIG. 8 is a flowchart for explaining an example of the post-distribution defect countermeasure processing procedure for the products distributed according to the processing procedure of FIG. 7.

[0096] When assembling a vehicle in the assembly process, first, the information processing apparatus 10 obtains information related to the assembly work (including such information if there are change points) from a vehicle assembly management system (not shown), and determines whether there are change points based on the information (step S21).

[0097] If the answer in step S21 is NO, the assembly process is carried out, and normal predetermined inspections are carried out in the inspection process included therein (step S22). As will be described below, when at least a defect occurs, the user U4 inputs the inspection result from the terminal 34.

[0098] Next to step S22, the terminal 34 refers to the input information and determines whether a defect has occurred (step S23). If no defect has occurred (if the answer in step S23 is NO), the information processing device 10 performs a process of taking the product (vehicle) off the line (step S24), and ends the process. By step S24, the target vehicle will be distributed in the market. If a defect has occurred (if the answer in step S23 is YES), the user U4 uses the terminal 34 to register the discovered defect in the vehicle production record DB20 (step S28), and notifies the operator to repair the discovered defect. Upon receiving this notification, the operator performs the repair (step S29). After the process of step S29, the process also proceeds to step S24.

[0099] On the other hand, if the answer in step S21 is YES, any one of the users U1 to U3 uses the terminals 31 to 33 respectively to input the change points that occur and transmit them to the information processing device 10, and the information processing device 10 temporarily records the change point information indicating the change points (step S25). Also, at this time, the information processing device 10 or the terminals 31 to 33 can also register the change point information 22 in the vehicle production record DB20.

[0100] Next to step S25, the information processing device 10 inputs the recorded change point information into the learned model 15, and outputs defect prediction information indicating possible defects from the change points that occurred in real time, thereby predicting possible defects caused by the occurred change points, and presenting corresponding additional inspections to the terminal 36 (step S26). For the process of step S26, for example, the process described with reference to FIG. 6 can be used.

[0101] Next, the normal predetermined inspection described in step S22 and the additional inspection for the predicted defect presented in step S26 are carried out, and at least when a defect occurs, the user U4 inputs the inspection result from the terminal 34 (step S27). Next to step S27, the processes of step S23 and subsequent steps are executed.

[0102] In step S24, after the target vehicle has been distributed in the market, the user U5 collects information on defects (outflow defects that have flowed into the market) obtained from the vehicle user or the selling dealer of the vehicle, etc., using the terminal 356. The user U5 checks whether there are any outflow defects for each vehicle (step S31), and if there are no outflow defects for a vehicle, the process ends. In this case, the vehicle user can continue to use the vehicle.

[0103] In the case of YES in step S31, that is, for vehicles with outflow defects, the user U5 registers (records) the defect information 23 in the vehicle production record DB20 (step S32), notifies the operator to repair the defect found by the operator, and the operator performs the repair (step S33). The process of step S32 may be automatically registered by obtaining the defect information from the seller. Also, the operator in step S33 may be a repairer. Further, the defect information 23 registered in step S32 can be associated with the change point information 24 if there is a change point by referring to the vehicle information 21 for the target vehicle.

[0104] As described above, according to the defect prediction system 100, it is possible to predict defects in products that may occur in the production line 40. Also, thereby, it is possible to prevent overlooking product defects in the inspection process, for example, before they are discovered in the inspection process or the like.

[0105] For example, in existing methods for investigating the causes of defects, when a defect occurs, the cause is identified by referring to the record of changes at the time of the occurrence, and countermeasures are taken accordingly. In contrast, in the present embodiment, it is possible to prevent the outflow of defects by predicting possible defects from the learning results based on the changes input before or during the operation.

[0106] Moreover, many of the defects that occur in the production process of products are caused by changes in people, operations, tools, and equipment in the process. Even if it is empirically recognized that such defects are likely to occur when a certain change point occurs, since there is no known inspection method corresponding to the change point, the defects will flow out into the market. This problem becomes particularly prominent when there are many change points occurring in the process, and also becomes prominent for products such as vehicles where many people share the work and assemble one product, because defects that are unexpected in machine operations may occur.

[0107] In contrast, in the present embodiment, by subjecting the change points in past processes and the defects caused by those change points to machine learning, and inputting the change points occurring in the current process, it is possible to predict what kind of defects will occur. Then, for each defect predicted in this way, by instructing the operator to conduct inspections in addition to general inspections, it becomes possible to discover defects that would otherwise flow out into the market.

[0108] (Other embodiments) Next, with reference to FIGS. 9 and 10, a configuration example in which the learning device and the defect prediction device are each independently configured will be described. However, in the present embodiment as well, various application examples of the above-described embodiment can be applied. FIG. 9 is a block diagram showing a configuration example of a learning system that generates a learned model used in the defect prediction system according to the present embodiment, and FIG. 10 is a block diagram showing a configuration example of the defect prediction system according to the present embodiment.

[0109] As shown in FIG. 9, the learning system 50 can include a control unit 51, an input unit 52, and a storage unit 53. The learning system 50 can be constructed using, for example, a computer such as a PC for AI (Artificial Intelligence) learning. However, the learning system 50 may be configured as a single device or may have its functions distributed among a plurality of devices.

[0110] The control unit 51 controls the entire learning system 50. The control unit 51 can be realized, for example, by an integrated circuit, and can be realized by, for example, a processor such as a CPU or MPU or GPU, a working memory, and a non-volatile storage device. A control program to be executed by the processor is stored in this storage device, and the processor reads and executes the program in the working memory to perform the functions of the control unit 51. This control program can include a learning program for executing learning. Note that this storage device can also use the storage unit 53.

[0111] The input unit 52 acquires and inputs a dataset of teacher data 54 necessary for learning from, for example, the vehicle production results DB 20 in FIG. 1, and stores it in the storage unit 53 so that it can be referred to during learning. The input unit 52 can be configured to include a communication unit. The storage unit 53 can store this teacher data 54 and can store a learning model 55 as an unlearned model.

[0112] The processing by the learning system 50 is as described with reference to FIG. 1 and the like. The control unit 51 may perform machine learning on the learning model 55 as an unlearned model based on the teacher data 54 to make the learning model 55 a learned model. Also, the learning model 55 can be relearned based on a newly prepared dataset when relearning is required.

[0113] As shown in FIG. 10, the defect prediction device 60 can include a control unit 61, a communication unit 62, and a storage unit 63 that stores a learned model 65 learned by the learning system 50. The defect prediction device 60 can be constructed as an information processing device such as a computer as described with reference to FIG. 1. However, the defect prediction device 60 may be configured as a single device or may have its functions distributed among a plurality of devices.

[0114] The control unit 61 controls the entire defect prediction device 60. The control unit 61 can be realized by, for example, an integrated circuit, and can be realized by, for example, a processor such as a CPU, MPU, or GPU, a working memory, and a non-volatile storage device. A control program executed by the processor is stored in this storage device, and the processor reads and executes the program in the working memory to perform the functions of the control unit 61. This control program can include a prediction program that inputs information including change point information serving as input parameters to the learned model 65 and obtains defect prediction information as an output result. Note that this storage device can also utilize the storage unit 63.

[0115] The communication unit 62 is an example of an input unit that receives and inputs change point information from any one or more of the terminals 31 to 35, and is also an example of an output unit that outputs by transmitting defect prediction information (and information indicating additional inspection contents) to the terminal 36.

[0116] The processing by the defect prediction device 60 is as described with reference to FIG. 1 and the like. The control unit 61 receives information including change point information via the communication unit 62, inputs it to the learned model 65, obtains defect prediction information from the learned model 65, and presents it to the terminal 36. Further, the control unit 61 can also present information indicating additional inspection contents corresponding to the defects indicated by the defect prediction information to the terminal 36 together with the defect prediction information.

[0117] (Alternative examples, etc.) In each of the above-described embodiments, an example has been given in which various change points in the process that have a large influence as factors causing defects are used as explanatory variables. However, the explanatory variables are not limited to this. As explanatory variables, for example, other types of information such as the workability of work in each production process (difficulty of work, working posture, etc.), characteristics of workers (age, experience, personality, etc.), time zone for work (before a long holiday when people tend to slack off, time zone when fatigue sets in, etc.) can also be applied. The workability of work, or workability and time zone, can also be expressed, for example, by indices related to ergonomics. Thus, by increasing the information adopted as explanatory variables, it becomes possible to perform defect prediction with higher accuracy.

[0118] Also, in each of the above-described embodiments, the description was made on the premise that the product is a vehicle. However, it may be other products, and the described systems and devices are not limited to the illustrated configurations, as long as their functions can be fulfilled.

[0119] Also, the systems and devices described in each of the above-described embodiments can all be constructed either as a system in which functions are arbitrarily distributed to a plurality of devices or as a single device. Any of these plurality of devices or a single device can be configured by general-purpose or dedicated circuitry, or as a device including a processor, a memory, and an interface.

[0120] As this processor, a CPU, MPU, GPU, FPGA (Field-Programmable Gate Array), quantum processor (quantum computer control chip), etc. can be used. This interface can include an interface with an operating device for receiving user operations, an interface with a display device for displaying information, a communication interface for sending and receiving information, etc. The functions in the above-described plurality of devices or a single device are realized by the processor reading a control program stored in the memory and performing operations while exchanging necessary information via the interface.

[0121] This program includes a set of instructions (or software code) that cause a computer to perform one or more of the functions described in the embodiments when loaded into the computer. The program may be stored on a non-transitory computer-readable medium or a tangible storage medium. By way of example and not limitation, a computer-readable medium or a tangible storage medium includes random-access memory (RAM), read-only memory (ROM), flash memory, solid-state drive (SSD), or other memory technologies, CD-ROM, digital versatile disc (DVD), Blu-ray (registered trademark) disc, or other optical disc storage, magnetic cassette, magnetic tape, magnetic disk storage, or other magnetic storage devices. The program may also be transmitted on a transient computer-readable medium or a communication medium. By way of example and not limitation, a transient computer-readable medium or a communication medium includes electrical, optical, acoustic, or other forms of propagated signals.

[0122] Note that the present invention is not limited to the above embodiments and can be appropriately modified without departing from the spirit thereof.

Explanation of Reference Numerals

[0123] U1 to U6 Users N Network 10 Information Processing Device 11 Control Unit 12 Communication Unit 13 Storage Unit 14 Unlearned Model 15 Learned Model 20 Vehicle Production Record DB 21 Product Information 22 Change Point Information 23, 24 Defect Information 31 to 36 Terminals 40 Production Line 41 to 44, 46 to 48 Change Points 45, 49 Inspection Points 50 Learning System 51 Control Unit 52 Input Unit 53 Memory Unit 54 Teacher Data 55 Learning Model 60 Defect Prediction Device 61 Control Unit 62 Communication Unit 63 Memory Unit 65 Trained Model 100 Defect Prediction System 201 Teacher Data 201a Explanatory Variable 201b Objective Variable

Claims

1. Input teacher data including change point information indicating changes related to workers, or related to workers and production equipment, or related to workers and work content, or related to workers, production equipment, and work content in a production line for producing products, and defect information indicating defects of products generated in the production line along with the change points, into a learning model, Generate a trained model that outputs defect prediction information indicating defects of products that are predicted to occur for the change points indicated by the change point information by performing machine learning on the learning model, The defect information included in the teacher data includes information indicating defects of products that occurred during a period determined according to the content of the change points indicated by the change point information, The teacher data includes the defect information not associated with the change point information, A learning system.

2. The defect information included in the teacher data includes information indicating defects discovered after passing the inspection process included in the production line, The learning system according to claim 1.

3. The production line is a line including an assembly operation for assembling products, The learning system according to claim 1 or 2.

4. Input the change point information into a trained model obtained by performing machine learning using teacher data including change point information indicating changes related to workers, or related to workers and production equipment, or related to workers and work content, or related to workers, production equipment, and work content in a production line for producing products, and defect information indicating defects of products generated in the production line along with the change points, Output defect prediction information indicating defects of products that are predicted to occur for the change points indicated by the change point information, The defect information included in the teacher data includes information indicating defects of products that occurred during a period determined according to the content of the change points indicated by the change point information, The teacher data includes the defect information not associated with the change point information, A defect prediction system.

5. The defect information included in the teacher data includes information indicating defects discovered after passing the inspection process included in the production line, The defect prediction system according to claim 4.

6. Notify the defect prediction information in a process after the change point indicated by the change point information input into the trained model, The defect prediction system according to claim 4 or 5.

7. When the defect prediction information includes information indicating a plurality of defects, information to be notified is determined according to the importance of each of the plurality of defects, and the determined information is notified. The defect prediction system according to claim 4 or 5.

8. Notify the content of the inspection to be performed corresponding to the defect prediction information in a process after the change point indicated by the change point information input to the learned model. The defect prediction system according to claim 4 or 5.

9. When there are a plurality of inspection contents to be performed corresponding to the defect prediction information, information to be notified is determined according to the importance of the defects corresponding to each of the plurality of inspection contents, and the determined information is notified. The defect prediction system according to claim 4 or 5.

10. Notify the content of the inspection to be performed corresponding to the defect prediction information, Update the learned model based on the result of performing the inspection of the notified content. The defect prediction system according to claim 4 or 5.

11. The production line is a line including an assembly operation for assembling a product. The defect prediction system according to claim 4 or 5.

12. A learned model obtained by machine learning so as to input the change point information and output defect prediction information indicating a product defect that is predicted to occur with respect to the change point indicated by the change point information, using teacher data including change point information indicating a change point regarding an operator, or regarding an operator and production equipment, or regarding an operator and work content, or regarding an operator, production equipment, and work content in a production line for producing a product, and defect information indicating a product defect that occurred in the production line along with the change point, A learned model for causing a computer to function so as to input the change point information and output the defect prediction information, The defect information included in the teacher data includes information indicating a product defect that occurred during a period determined according to the content of the change point indicated by the change point information. The teacher data includes the defect information not associated with the change point information. Learned model.

13. The computer is Input teacher data including change point information indicating changes related to an operator, or related to an operator and production equipment, or related to an operator and work content, or related to an operator, production equipment, and work content in a production line for producing a product, and defect information indicating defects of products occurring in the production line along with the change points, into a learning model. By machine learning the learning model, generate a learned model that inputs the change point information and outputs defect prediction information indicating defects of products that are predicted to occur for the change points indicated by the change point information. The defect information included in the teacher data includes information indicating defects of products that occurred during a period determined according to the content of the change points indicated by the change point information. The teacher data includes the defect information not associated with the change point information. Learning method.

14. A computer Inputs the change point information into a learned model obtained by machine learning using teacher data including change point information indicating changes related to an operator, or related to an operator and production equipment, or related to an operator and work content, or related to an operator, production equipment, and work content in a production line for producing a product, and defect information indicating defects of products occurring in the production line along with the change points. Outputs defect prediction information indicating defects of products that are predicted to occur for the change points indicated by the change point information. The defect information included in the teacher data includes information indicating defects of products that occurred during a period determined according to the content of the change points indicated by the change point information. The teacher data includes the defect information not associated with the change point information. Defect prediction method.

Images