Abnormality determination device and abnormality determination method

Abstract

An abnormality determination device includes: a learning unit that creates an abnormality detection model for detecting an abnormality of a device by machine learning using log data indicating an actual operation state of the device; a determination unit that inputs input data indicating the actual operation state of the target device into the abnormality detection model to determine an abnormality of the target device; and an output unit for outputting a determination result by the determination unit, in which the learning unit updates the abnormality detection model on the basis of an evaluation result when the determination result is different from the evaluation result for the target device obtained by the field confirmation.

Description

Technical Field

[0001] The present disclosure relates to an abnormality determination device and an abnormality determination method. Background Art

[0002] In recent years, the application of machine learning in various fields has been explored, and research and development of technologies for improving the accuracy of machine learning have been carried out (for example, refer to Patent Documents 1 and 2).

[0003] Prior Art Documents

[0004] Patent Documents

[0005] Patent Document 1: Japanese Unexamined Patent Application Publication No. 2021-32115

[0006] Patent Document 2: Japanese Unexamined Patent Application Publication No. 2016-143352 Summary of the Invention

[0007] It is desired to improve the accuracy in the case of determining an abnormality of a device by using machine learning.

[0008] An abnormality determination device according to one aspect of the present disclosure includes: a learning unit that creates an abnormality detection model for detecting an abnormality of the device by machine learning using log data indicating the actual operation state of the device; a determination unit that inputs input data indicating the actual operation state of the target device into the abnormality detection model to determine the abnormality of the target device; and an output unit that outputs the determination result of the determination unit. When the determination result is different from the evaluation result for the target device obtained through on-site confirmation, the learning unit updates the abnormality detection model based on the evaluation result.

[0009] An abnormality determination method according to one aspect of the present disclosure includes the following steps: creating an abnormality detection model for detecting an abnormality of the device by machine learning using log data indicating the actual operation state of the device; inputting input data indicating the actual operation state of the target device into the abnormality detection model to determine the abnormality of the target device; and outputting the determination result obtained in the determination step. In the creation step, when the determination result is different from the evaluation result for the target device obtained through on-site confirmation, the abnormality detection model is updated based on the evaluation result.

[0010] In addition, one aspect of the present disclosure can be implemented as a program that causes a computer to execute the above-described abnormality determination method. Alternatively, one aspect of the present disclosure can also be implemented as a non-transitory computer-readable recording medium storing the program.

[0011] According to the present disclosure, anomalies can be determined with high precision. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 is a diagram showing the configuration of the anomaly determination system according to the embodiment.

[0013] Figure 2 is a block diagram showing the configuration of the anomaly determination device according to the embodiment.

[0014] Figure 3 is a block diagram showing the configuration of the input / output device according to the embodiment.

[0015] Figure 4 is a flowchart showing the processing in the learning stage of the anomaly determination system according to the embodiment.

[0016] Figure 5 is a diagram for explaining the processing of the anomaly determination device according to the embodiment.

[0017] Figure 6 is a diagram showing an example of learning data for machine learning.

[0018] Figure 7 is a flowchart showing the processing in the usage stage of the anomaly determination system according to the embodiment.

[0019] Figure 8 is a diagram showing an example of re-learning data for machine learning.

[0020] Figure 9 is a flowchart showing the processing related to on-site confirmation of the anomaly determination system according to the embodiment.

[0021] Figure 10 is a diagram showing an example of a notification screen displaying the determination result of an anomaly.

[0022] Figure 11 is a diagram showing an example of evaluation data representing the evaluation result obtained through on-site confirmation. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0023] (SUMMARY OF THE PRESENT DISCLOSURE)

[0024] The abnormality determination device according to the first aspect of the present disclosure includes: a learning unit that creates an abnormality detection model for detecting an abnormality of the device by machine learning using log data indicating the actual operating condition of the device; a determination unit that inputs input data indicating the actual operating condition of the target device into the abnormality detection model to determine the abnormality of the target device; and an output unit that outputs the determination result of the determination unit. When the determination result is different from the evaluation result obtained through on-site confirmation for the target device, the learning unit updates the abnormality detection model based on the evaluation result.

[0025] Accordingly, the abnormality detection model is updated based on the evaluation result obtained through on-site confirmation. Therefore, the true device state that sometimes cannot be obtained from the data of the target device can be reflected in the abnormality detection model. As a result, the accuracy of the abnormality detection model is improved. Therefore, according to the abnormality determination device according to this aspect, an abnormality can be determined with high accuracy.

[0026] In addition, regarding the abnormality determination device according to the second aspect of the present disclosure, in the abnormality determination device according to the first aspect, a comparison unit is provided. The comparison unit is used to compare the determination result with the evaluation result. When the determination result is different from the evaluation result, the learning unit updates the abnormality detection model based on the evaluation result.

[0027] Accordingly, the operator only needs to perform an evaluation through on-site confirmation and input the evaluation result, and there is no need to compare the evaluation result with the determination result. Therefore, according to the abnormality determination device according to this aspect, the burden on the operator can be reduced, and thus both convenience and improvement in abnormality determination accuracy can be achieved.

[0028] In addition, regarding the abnormality determination device according to the third aspect of the present disclosure, in the abnormality determination device according to the second aspect, when the determination result indicates that the target device is abnormal, the comparison unit performs the comparison. When the determination result indicates that the target device is normal, the comparison unit does not perform the comparison.

[0029] Accordingly, it is possible to prompt the operator to perform on-site confirmation only when it is determined to be abnormal. The burden on the operator can be reduced by reducing the number of on-site confirmations. Therefore, both convenience and improvement in abnormality determination accuracy can be achieved.

[0030] In addition, regarding the abnormality determination device according to the fourth aspect of the present disclosure, in the abnormality determination device according to any one of the first to third aspects, when the determination result indicates that the target device is abnormal, the output unit outputs the determination result, and when the determination result indicates that the target device is normal, the output unit does not output the determination result.

[0031] Accordingly, it is possible to prompt the operator to perform on-site confirmation only when it is determined to be abnormal. By reducing the number of on-site confirmations, the burden on the operator can be reduced, so that both convenience and improvement in abnormality determination accuracy can be achieved simultaneously.

[0032] In addition, regarding the abnormality determination device according to the fifth aspect of the present disclosure, in the abnormality determination device according to any one of the first to fourth aspects, an acquisition unit is provided, and the acquisition unit acquires the evaluation result.

[0033] Accordingly, the evaluation result obtained through on-site confirmation is acquired, so that it is possible to digitalize the actual device state data that sometimes cannot be obtained from the data of the target device. Therefore, by digitalizing the evaluation result and accumulating the evaluation results, the accuracy of the abnormality detection model can be improved.

[0034] In addition, regarding the abnormality determination device according to the sixth aspect of the present disclosure, in the abnormality determination device according to any one of the first to fifth aspects, a storage unit is provided, and the storage unit is used to store the log data and the abnormality detection model.

[0035] Accordingly, access to the data required for the production and use of the abnormality detection model becomes easy, so that it is possible to contribute to high-speed processing and reduction of power consumption.

[0036] In addition, the abnormality determination method according to the seventh aspect of the present disclosure includes the following steps: producing an abnormality detection model for detecting an abnormality of the device through machine learning using log data representing the actual operating condition of the device; inputting input data representing the actual operating condition of the target device into the abnormality detection model to determine the abnormality of the target device; and outputting the determination result obtained in the determination step, wherein in the production step, when the determination result is different from the evaluation result for the target device obtained through on-site confirmation, the abnormality detection model is updated based on the evaluation result.

[0037] Accordingly, the anomaly detection model is updated based on the evaluation results obtained through on-site confirmation. Therefore, the true state that sometimes cannot be obtained from the data of the target device can be reflected in the anomaly detection model. As a result, the accuracy of the anomaly detection model is improved. Therefore, according to the anomaly determination method according to this embodiment, anomalies can be determined with high accuracy in the same manner as the anomaly determination devices according to the above-described respective embodiments.

[0038] In addition, the program according to the eighth embodiment of the present disclosure is a program that causes a computer to execute the anomaly determination method according to any one of the first to seventh embodiments.

[0039] Accordingly, anomalies can be determined with high accuracy in the same manner as the anomaly determination devices according to the above-described respective embodiments.

[0040] Hereinafter, embodiments will be specifically described with reference to the drawings.

[0041] In addition, the embodiments described below all show general or specific examples. The numerical values, shapes, materials, constituent elements, arrangement positions and connection methods of the constituent elements, steps, order of steps, etc. shown in the following embodiments are examples and are not intended to limit the present disclosure. In addition, regarding the constituent elements in the following embodiments that are not described in the independent claims, they are described as arbitrary constituent elements.

[0042] In addition, the respective drawings are schematic diagrams and are not necessarily drawn strictly. Therefore, for example, the scales and the like are not necessarily the same in each drawing. In addition, in each drawing, the same reference numerals are assigned to substantially the same structures, and repeated descriptions are omitted or simplified.

[0043] (Embodiment)

[0044] [Outline]

[0045] First, use Figure 1 to describe the outline of the anomaly determination system according to the embodiment. Figure 1 is a diagram showing the outline of the anomaly determination system 1 according to the present embodiment.

[0046] Figure 1 The shown anomaly determination system 1 is a system used in a production system such as a factory for determining anomalies of equipment. Specifically, the anomaly determination system 1 uses machine learning to create an anomaly detection model, and uses the created anomaly detection model to determine anomalies of the target device.

[0047] As Figure 1As shown, the abnormality determination system 1 includes an abnormality determination device 100, an input / output device 200, and a plurality of devices 10. The abnormality determination device 100, the input / output device 200, and the plurality of devices 10 are connected in such a manner that they can communicate via a network 20. The communication is performed by wireless communication, wired communication, or a combination thereof.

[0048] The plurality of devices 10 are respectively manufacturing devices that execute one process among a plurality of processes for manufacturing a product. The device 10 is, for example, a component mounter, a processing device, an assembly device, etc., but is not particularly limited. The device 10 produces components by executing a process and outputs the produced components.

[0049] Here, the component is, for example, a part included in the final product (i.e., the product) or a semi-finished product during the manufacturing of the final product, but is not limited thereto. The component is an article used for producing a part or a semi-finished product and may not be included in the final product. In addition, the device 10 only needs to be a device related to the manufacturing of the product and may also be an inspection device for inspecting components, semi-finished products, or products.

[0050] In this specification, "manufacturing" not only refers to producing a final product but also includes processing, assembling, inspecting, etc. of components (parts or semi-finished products). For example, the component manufactured by the device 10 refers to the component output by the device 10 after executing the process (processing, assembling, inspecting, etc.) assigned to the device 10. In addition, "manufacturing" is an example of "producing", and when the final product is an industrial product, "manufacturing" is used in the same meaning as "producing". Furthermore, the final product is not limited to being an industrial product and may also be an article manufactured or produced in a food factory, a plant factory, etc., for example.

[0051] In this embodiment, the plurality of devices 10 respectively manufacture a product by executing a prescribed process. In addition, the number of the devices 10 is not particularly limited. The number of the devices 10 may also be one. That is, a product may also be manufactured by one device 10. A device 10 being normal means that the device 10 executes a prescribed process. A device 10 being abnormal means that the device 10 executes a process that is not a prescribed process or does not execute a prescribed process.

[0052] The abnormality determination device 100 is a device that determines the abnormality of the device 10. Specifically, the abnormality determination device 100 uses machine learning to determine whether an abnormality has occurred in each of the plurality of devices 10.

[0053] The abnormality determination device 100 is one or more computer devices including a processor and a memory. The processor executes processing related to the determination of abnormality by reading out a program stored in the memory and executing the program. In addition, at least a part of the processing executed by the abnormality determination device 100 may also be executed by a dedicated circuit.

[0054] The input / output device 200 is a device that presents (outputs) information to the operator 30 (refer to Figure 2 ) and acquires (inputs) information from the operator 30. The operator 30 is, for example, a person who performs maintenance, management, etc. of the equipment 10. The input / output device 200 is an operation terminal possessed by the operator 30, such as a mobile terminal like a tablet personal computer or a smart phone.

[0055] In addition, the input / output device 200 is not limited to being a mobile terminal, and may also be a fixed computer device. For example, the input / output device 200 may also be configured integrally with the abnormality determination device 100. Additionally, the input / output device 200 may also be such that the input device and the output device are separate devices.

[0056] In order to improve the productivity of the product, it is sought to improve the operation rate of the equipment 10 by detecting abnormalities of the equipment 10 and quickly responding to the detected abnormalities. An abnormality detection model created through machine learning is used in the detection of abnormalities. By improving the accuracy of the abnormality detection model, it is possible to suppress the occurrence of false detections and missed detections of abnormalities, and contribute to the improvement of the operation rate of the equipment 10.

[0057] In order to improve the accuracy of the abnormality detection model, it is sought to obtain appropriate feedback for the determination result and utilize it for updating the abnormality detection model. In the equipment 10, generally one or more sensors for detecting the operating condition of the equipment are provided. Log data representing the actual operating condition of the equipment is generated based on the sensor values output from the sensors, and this log data is utilized in machine learning. Therefore, by using the data based on the sensor values determined to be abnormal for feedback, it is possible to update the abnormality detection model.

[0058] However, the abnormalities of the equipment 10 may occur due to various factors, so the data based on the sensor values does not necessarily appropriately present the abnormal conditions of the equipment 10. In this case, the abnormality detection model cannot be appropriately updated by using the feedback of the sensor values.

[0059] Therefore, the abnormality determination device 100 according to the present embodiment updates the abnormality detection model based on the evaluation result for the equipment 10 obtained through on-site confirmation. On-site confirmation is the confirmation of the condition of the equipment 10 performed by the operator 30 on-site. The "site" is the place where the equipment 10 is actually installed. On-site confirmation includes confirmation of the equipment 10 by visual inspection of the equipment 10 by the operator 30 and confirmation of the operating condition of the equipment 10 using an inspection device, etc. The operator 30 inputs the evaluation result obtained through on-site confirmation by means of the input / output device 200.

[0060] Thus, in the present embodiment, the anomaly detection model is updated based on the evaluation results obtained through on-site confirmation. Therefore, the true device state that cannot sometimes be obtained from the data of the device 10 can be reflected in the anomaly detection model. As a result, the accuracy of the anomaly detection model is improved, and thus, according to the anomaly determination device 100, anomalies can be determined with high accuracy.

[0061] [Anomaly determination device]

[0062] Next, Figure 2 is used to describe the specific structure of the anomaly determination device 100. Figure 2 is a block diagram showing the structure of the anomaly determination device 100 according to the present embodiment.

[0063] As Figure 2 shown, the anomaly determination device 100 includes a device information acquisition unit 110, a storage unit 120, a learning unit 130, a determination unit 140, an output unit 150, an evaluation result acquisition unit 160, and a comparison unit 170.

[0064] The device information acquisition unit 110 acquires device data of each device 10 from a plurality of devices 10. The device data includes the above-mentioned sensor values. The device information acquisition unit 110 generates log data 122 representing the actual operating condition of the device 10 based on the acquired device data, and stores the log data 122 in the storage unit 120. A specific example of the log data 122 will be described later using Figure 6 is used.

[0065] In addition, the device information acquisition unit 110 acquires device data from the target device, where the device data includes sensor values that are the source of the input data used in the determination of anomalies. The target device is one or more devices that are the object of the determination of anomalies, and is one, a plurality, or all of the plurality of devices 10.

[0066] The storage unit 120 is a storage unit for storing data used by the anomaly determination device 100. As Figure 2 shown, the storage unit 120 stores log data 122, comparison results 124, and an anomaly detection model 126.

[0067] The learning unit 130 creates an anomaly detection model 126 for detecting anomalies in the device 10 through machine learning using the log data 122. In addition, when the determination result of the determination unit 140 is different from the evaluation result for the device 10 obtained through on-site confirmation, the learning unit 130 updates the anomaly detection model 126 based on the evaluation result.

[0068] As Figure 2 shown, the learning unit 130 includes a request unit 132 and a model creation unit 134.

[0069] The request unit 132 requests the stored data from the storage unit 120. For example, the request unit 132 requests the log data 122 required for creating the anomaly detection model 126. Additionally, the request unit 132 requests the log data 122 and the comparison result 124 required for updating the anomaly detection model 126.

[0070] The model creation unit 134 creates the anomaly detection model 126 through machine learning using the log data 122. Additionally, the model creation unit 134 updates the anomaly detection model 126 based on the log data 122 and the comparison result 124.

[0071] The determination unit 140 inputs the input data representing the actual operating condition of the target device into the anomaly detection model 126 to determine the anomaly of the target device. The determination unit 140 outputs the determination result of each device 10 to the output unit 150.

[0072] The determination result includes information indicating whether the target device is abnormal (whether it is normal). Additionally, when it is determined that the target device is abnormal, the determination result may also include information for determining the factors causing the anomaly.

[0073] The output unit 150 outputs the determination result of the determination unit 140. The output unit 150 outputs the determination result to the input / output device 200 and the comparison unit 170. Alternatively, the output unit 150 may output the determination result to an output device such as a display device different from the input / output device 200 instead of outputting it to the input / output device 200. For example, it may be that the output unit 150 outputs the determination result to the monitor screen provided in the factory to enable the factory worker 30 to know the determination result.

[0074] When the determination result indicates that the target device is abnormal, the output unit 150 outputs the determination result. When the determination result indicates that the target device is normal, the output unit 150 does not output the determination result. Thereby, it is possible to prompt the worker 30 to perform on-site confirmation only when the target device is abnormal. Generally, the number of times determined to be abnormal is less than the number of times determined to be normal, so it is possible to reduce the frequency of on-site confirmation, thereby reducing the burden on the worker 30.

[0075] The evaluation result acquisition unit 160 acquires the evaluation result for the target device obtained through on-site confirmation. In the present embodiment, the evaluation result acquisition unit 160 acquires the evaluation result from the input / output device 200. The evaluation result includes, for example, the result obtained through on-site confirmation as to whether the target device determined to be abnormal is actually abnormal or normal, and the result obtained by evaluating the occurrence status of multiple factors that may cause the anomaly. A specific example of the evaluation result will be described later. Figure 11 to illustrate a specific example of the evaluation result.

[0076] The comparison unit 170 compares the determination result of the determination unit 140 with the evaluation result obtained by the evaluation result acquisition unit 160. The comparison result 124 of the comparison unit 170 is stored in the storage unit 120. The comparison result 124 is information indicating the agreement or disagreement between the determination result and the evaluation result. The comparison result 124 may also include the determination result and the evaluation result.

[0077] In addition, the device information acquisition unit 110 is implemented by a communication interface or the like capable of communicating with sensors provided in a plurality of devices 10. The storage unit 120 is a non-volatile storage device such as a magnetic disk such as an HDD (Hard Disk Drive) or a semiconductor memory such as an SDD (Solid State Drive). The output unit 150 and the evaluation result acquisition unit 160 are respectively implemented by a communication interface or the like capable of communicating with the input / output device 200.

[0078] The learning unit 130, the determination unit 140, and the comparison unit 170 are respectively implemented by, for example, an LSI (Large Scale Integration) as an integrated circuit (IC: Integrated Circuit). In addition, the integrated circuit is not limited to an LSI, and may be a dedicated circuit or a general-purpose processor. Further, for example, the learning unit 130, the determination unit 140, and the comparison unit 170 may also be a programmable FPGA (Field Programmable Gate Array), or a reconfigurable processor in which the connection and setting of circuit units in the LSI can be reconfigured. At least a part of the functions executed by the learning unit 130, the determination unit 140, and the comparison unit 170 can be implemented by software or by hardware. The learning unit 130, the determination unit 140, and the comparison unit 170 may also be implemented by common hardware resources.

[0079] [Input / Output Device]

[0080] Next, Figure 3 is used to describe the specific structure of the input / output device 200. Figure 3 is a block diagram showing the structure of the input / output device 200 according to the present embodiment.

[0081] As Figure 3 shown, the input / output device 200 includes a communication unit 210, a display control unit 220, a display unit 230, a reception unit 240, and a signal processing unit 250.

[0082] The communication unit 210 transmits and receives information by communicating with the abnormality determination device 100. Specifically, the communication unit 210 obtains the determination result from the output unit 150 of the abnormality determination device 100. In addition, the communication unit 210 sends the evaluation result to the evaluation result acquisition unit 160 of the abnormality determination device 100. The communication unit 210 is implemented by a communication interface that communicates in a wired or wireless manner.

[0083] The display control unit 220 controls the display unit 230. Specifically, the display control unit 220 generates a notification screen for notifying the operator 30 of the determination result obtained by the communication unit 210, and causes the notification screen to be displayed on the display unit 230. In addition, the display control unit 220 generates an input reception screen for receiving the evaluation result obtained by the on-site confirmation by the operator 30, and causes the input reception screen to be displayed on the display unit 230.

[0084] The display control unit 220 is implemented, for example, by an LSI (Large Scale Integration) as an integrated circuit. The display control unit 220 is implemented, for example, by a dedicated integrated circuit, a microcontroller, a processor, or the like. Alternatively, the display control unit 220 may be a programmable FPGA (Field Programmable Gate Array), or a reconfigurable processor in which the connection and setting of circuit units in the LSI can be reconfigured. At least a part of the functions executed by the display control unit 220 can be implemented by software or hardware.

[0085] The display unit 230 displays the image generated by the display control unit 220. Specifically, the display unit 230 displays the notification screen and the input reception screen. The display unit 230 is, for example, a liquid crystal display device or an organic EL (Electro luminescenct) display device.

[0086] The reception unit 240 receives the operation input from the operator 30. Specifically, the reception unit 240 receives the input related to the evaluation of the target device by on-site confirmation performed by the operator 30. The reception unit 240 is, for example, a touch sensor or a physical button. The reception unit 240 may also be implemented in the form of a touch panel display together with the display unit 230.

[0087] The signal processing unit 250 processes the input received by the reception unit 240. Specifically, the signal processing unit 250 generates an evaluation result based on the input received by the reception unit 240, and sends the evaluation result to the abnormality determination device 100 via the communication unit 210.

[0088] The signal processing unit 250 is implemented by, for example, an LSI (Large Scale Integration) which is an integrated circuit. The signal processing unit 250 is implemented by, for example, an application-specific integrated circuit, a microcontroller, or a processor, etc. Alternatively, the signal processing unit 250 can also be a programmable FPGA (Field Programmable Gate Array), or a reconfigurable processor in which the connection and setting of circuit units within an LSI can be reconfigured. At least a part of the functions performed by the signal processing unit 250 can be implemented by software or by hardware. The signal processing unit 250 and the display control unit 220 can also be implemented by common hardware resources.

[0089] [Operation]

[0090] Next, the operation of the abnormality determination system 1 according to the present embodiment will be described.

[0091] The operation of the abnormality determination system 1 includes processing that is generally divided into two stages: a learning stage of creating an abnormality detection model through machine learning and a usage stage of using the created abnormality detection model.

[0092] [Learning Stage]

[0093] First, use Figure 4 to illustrate the processing in the learning stage. Figure 4 is a flowchart showing the processing in the learning stage of the abnormality determination system 1 according to the present embodiment.

[0094] As Figure 4 shown, first, the device information acquisition unit 110 acquires device data from a plurality of devices 10 and stores the acquired device data as log data 122 in the storage unit 120 (S10). The log data 122 is data representing the actual operation status of each process (for each device). For example, as Figure 5 shown, the actual operation status is the production quantity, operation time, stop time for each factor, and production information, etc. of each process (for each device). In addition, Figure 5 is a diagram for explaining the processing of the abnormality determination device 100 according to the present embodiment. Figure 5 Corresponding to the structure of the abnormality determination device 100 shown in Figure 2 but the illustration of the storage unit 120 and the request unit 132 is omitted.

[0095] In addition, Figure 6 is a diagram showing an example of the log data 122 for machine learning. In Figure 6In the example shown, a record (one line of data) is generated for each line, process, and batch number (batch No). The equipment information acquisition unit 110 generates, for each batch, the manufacturing line, process (equipment), batch start time, batch end time, operation time, input quantity, output quantity, product type information, stop occurrence time, stop end time, and stop factor, etc., as one line of data based on the sensor values detected by the sensors of each equipment 10, and stores it in the storage unit 120. In addition, the one line of data is not limited to the Figure 6 example shown, and may also include other elements such as the number of stops and management time.

[0096] Next, as Figure 4 shown, the learning unit 130 creates an anomaly detection model 126 through machine learning using the log data 122, and stores the anomaly detection model 126 in the storage unit 120 (S12). Specifically, the request unit 132 requests the storage unit 120 to read the log data 122. Accordingly, the model creation unit 134 reads the log data 122 from the storage unit 120. The model creation unit 134 creates an anomaly detection model 126 by performing machine learning using the read log data 122.

[0097] The anomaly detection model 126 is a learning model used in the determination of anomalies in the target equipment. As Figure 5 shown, the anomaly detection model 126 corresponds to the probability distribution of the production rhythm. The probability distribution is defined by its type and the value of the parameter. The production rhythm is the so-called takt time, which is the time required to manufacture one product. For example, an anomaly detection model 126 is created for each product (each manufacturing line). In addition, an anomaly detection model 126 can also be created for each equipment (each process).

[0098] The types of probability distributions are normal distribution, lognormal distribution, zero-inflated exponential distribution, gamma distribution, etc. The parameter type of the probability distribution is determined according to the type of probability distribution. For example, in the case of a normal distribution, the parameters of the probability distribution are the mean μ and the standard deviation σ. The values of the parameters are generated based on the past production actual situation, that is, the log data 122.

[0099] The parameters of the learning model can be obtained based on Bayesian estimation. For example, they can be obtained by sampling methods such as the Markov chain Monte Carlo method (MCMC) or variational estimation algorithms such as the VB-EM (Variational Bayesian-Expectation Maximization) algorithm.

[0100] As Figure 5As shown, the anomaly detection model 126 is a model obtained by integrating multiple learning models. Specifically, the multiple learning models include a production quantity model corresponding to the probability distribution of the production quantity (manufacturing quantity) during the operation time and a stop time model corresponding to the probability distribution of the stop time during the operation time. In addition, the stop time model is created based on the probability distribution of the stop time for each stop factor. Furthermore, the method for creating the anomaly detection model 126 is not limited to the above example.

[0101] [Usage Phase]

[0102] Next, use Figure 7 to illustrate the processing in the usage phase. Figure 7 is a flowchart showing the processing in the usage phase of the anomaly determination system 1 according to the present embodiment.

[0103] As Figure 7 shown, first, the device information acquisition unit 110 acquires device data of the target device (S20). The device information acquisition unit 110 generates input data representing the actual operation status of the target device based on the acquired device data, and outputs the input data to the determination unit 140. The input data is, for example, equivalent to one line of the log data 122. The input data may also be stored in the storage unit 120 as part of the log data 122.

[0104] Next, the determination unit 140 inputs the input data into the anomaly detection model 126 to determine the anomaly of the target device (S22). Specifically, the determination unit 140 calculates the anomaly degree of the target device. As Figure 5 shown, the anomaly degree is equivalent to the area of the region on the right side of the measured value in the probability distribution corresponding to the anomaly detection model 126 (also referred to as the upper probability). The measured value is the production cycle calculated based on the input data. When the upper probability is smaller than the threshold value, the determination unit 140 determines that the target device is abnormal (anomaly is detected). When the upper probability is larger than the threshold value, the determination unit 140 determines that the target device is normal (no anomaly is detected). Furthermore, the method for determining the anomaly is not limited to this.

[0105] As Figure 7 shown, when no anomaly of the target device is detected (S24 is "No"), the anomaly determination process ends. Alternatively, it may also return to step S20 to continue the anomaly determination process based on the device data of different target devices.

[0106] When an anomaly of the target device is detected (S24 is "Yes"), the output unit 150 outputs the determination result (S26). The output unit 150 outputs the determination result to the input / output device 200. The processing performed by the input / output device 200 will be described later using Figure 9 as an example.

[0107] After outputting the determination result, the abnormality determination device 100 stands by until the evaluation result obtained through on-site confirmation is obtained. Alternatively, during the waiting period, steps S20 to S26 may be repeatedly performed using other device data to output a plurality of determination results.

[0108] Next, the evaluation result acquisition unit 160 acquires the evaluation result for the target device obtained through on-site confirmation (S28). Specifically, the evaluation result acquisition unit 160 acquires the evaluation result sent from the input / output device 200.

[0109] Next, the comparison unit 170 compares the determination result with the evaluation result (S30). The comparison unit 170 determines whether the determination result and the evaluation result are consistent by performing the comparison. Specifically, it is determined whether the on-site confirmation result of the target device detected as abnormal according to the determination result actually has an abnormality (consistent) or actually does not have an abnormality (inconsistent). The comparison result of the comparison unit 170 is stored in the storage unit 120 as the comparison result 124.

[0110] When the determination result and the evaluation result are consistent (S32 is "yes"), the learning unit 130 ends the abnormality determination process without updating the abnormality detection model 126 stored in the storage unit 120. Alternatively, it may also be possible to return to step S20 and continue the abnormality determination process based on the device data of different target devices.

[0111] When the determination result and the evaluation result are inconsistent (S32 is "no"), the learning unit 130 updates the abnormality detection model 126 based on the evaluation result and stores the updated abnormality detection model 126 in the storage unit 120 (S34). Specifically, as Figure 5 shown, the learning unit 130 changes the threshold for determining as abnormal based on the probability distribution corresponding to the abnormality detection model 126. Alternatively, the learning unit 130 may update the parameters of the probability distribution by performing re-learning.

[0112] Figure 8 FIG. is an example of re-learning data for machine learning. As Figure 8 shown, the re-learning data includes, in addition to the learning data shown in Figure 6 shown, the determination result and the comparison result. By using the information on the consistency or inconsistency of the determination result and the comparison result in re-learning, the accuracy of the abnormality detection model 126 can be improved. Thus, for example, if the determination result of the abnormality detection model 126 is incorrect in a certain case, a correct determination result can be obtained if the same case occurs next.

[0113] In addition, in Figure 8Among them, for the case where the determination result is normal, the comparison result is also shown. Thus, it can also be that the abnormality determination device 100 makes the operator 30 conduct on-site confirmation to obtain an evaluation result even when the determination result is normal. For example, it can also be that the output unit 150 outputs the determination result even when no abnormality is detected. Or, it can also be that the operator 30 conducts on-site confirmation regularly regardless of the determination result.

[0114] In addition, the update of the abnormality detection model 126 (S34) can be performed either each time the comparison (S30) is executed or after obtaining a plurality of comparison results. Similarly, the comparison (S30) can be performed either each time an evaluation result is obtained or after obtaining a plurality of evaluation results. When the abnormality detection model 126 is updated each time a comparison is made, the abnormality detection model 126 can always be kept in the latest state, so the determination accuracy of abnormalities can be improved. In addition, when the abnormality detection model 126 is updated after obtaining a certain amount of comparison results, more data can be used in the update, so the accuracy of the updated abnormality detection model 126 can be further improved. Therefore, the determination accuracy of abnormalities can be improved.

[0115] [Evaluation Process by On-Site Confirmation]

[0116] Next, use Figure 9 to illustrate the evaluation process by on-site confirmation performed by the operator 30.

[0117] Figure 9 is a flowchart showing the process related to on-site confirmation of the abnormality determination system 1 according to the present embodiment. Figure 9 The process shown is mainly executed by the input / output device 200.

[0118] First, in the input / output device 200, when the communication unit 210 obtains the determination result from the abnormality determination device 100, the display unit 230 displays the determination result (S40).

[0119] Figure 10 is a diagram showing an example of a notification screen displaying the determination result of an abnormality. For example, as Figure 10 shown, the display screen includes the determination results of normal and abnormal for each production line and the determination results of normal and abnormal for each process (equipment) in each production line. In the Figure 10 example, an abnormality is detected in process B (equipment B) in production line 2 and production line 3 respectively. The operator 30 can observe the display screen to determine the equipment (process) where the abnormality has occurred.

[0120] Therefore, as Figure 9As shown, the operator 30 performs on-site confirmation (S42). Specifically, the operator 30 evaluates the abnormal situation by actually going to the installation site of the equipment and confirming the state of the equipment.

[0121] Next, the display unit 230 displays an input reception screen for inputting the content of on-site confirmation (S44). The input reception screen includes, for example, a GUI (Graphical User Interface) object for allowing the operator 30 to input whether or not the occurrence of each abnormal factor item exists. As the GUI object, there are text boxes, selection buttons, etc., which are not particularly limited.

[0122] The reception unit 240 receives the input from the operator 30 via the input reception screen displayed on the display unit 230 (S46). Figure 11 It is a diagram showing an example of evaluation data representing the evaluation result obtained through on-site confirmation. As Figure 11 shown, "Date and Time" and "Lot Number" are identification information for identifying the target equipment. Similarly, Figure 6 it may also include the production line and process, etc.

[0123] Items such as "dust adhesion", "foreign object mixing", "wiring breakage", etc. are items representing the occurrence factors of abnormalities. The types of items are preset by the abnormality determination system 1. Therefore, the operator 30 only needs to input the result obtained through on-site confirmation for the content of the items. "True" for each item means that the content of the corresponding item has occurred, and "False" means that the content of the corresponding item has not occurred. For example, for the lot number "L001", it is known that the results of on-site confirmation are that "dust adhesion" and "wiring breakage" have occurred, while "foreign object mixing", "tool abnormality", and "configuration abnormality" have not occurred. Even if one of the multiple items is "True", that is, when an abnormality is confirmed for at least one factor, the evaluation result of the target equipment is "abnormal". When all items are "False", that is, when no abnormality is confirmed for any factor, the evaluation result of the target equipment is "normal". In addition, Figure 11 the content of the item "Evaluation" can be input by the operator 30 or can be the result generated by the signal processing unit 250 based on the input results for each item representing the abnormal factors.

[0124] As Figure 9 shown, after receiving the input, the signal processing unit 250 generates an evaluation result based on the input content and sends the evaluation result to the abnormality determination device 100 via the communication unit 210 (S48).

[0125] As described above, the on-site confirmation performed by the operator 30 is carried out according to pre-specified items, and the input of the confirmation results for each item is accepted. The items are specified to improve the accuracy of the anomaly detection model 126. Therefore, the evaluation results obtained through the on-site confirmation by the operator 30 can be obtained as quantitative data.

[0126] Generally speaking, the operator 30 in charge of the equipment 10 is mostly not an expert in machine learning. Therefore, there is a concern that the feedback from the operator 30 is likely to become inappropriate, and even if the feedback is given, the accuracy of the anomaly detection model 126 cannot be improved. In this case, for the anomaly determination system whose accuracy has not been improved despite the feedback, the operator 30 is likely to be dissatisfied, the motivation to give feedback decreases, and it gradually becomes an environment where it is difficult to improve the accuracy of the anomaly detection model.

[0127] In contrast, according to the present embodiment, it is possible to quantify the evaluation content according to the knowledge of the operator 30 by dividing the input content input by the operator 30 in advance. It is possible to effectively improve the accuracy by obtaining information useful for improving the accuracy of the anomaly detection model 126.

[0128] (Other Embodiments)

[0129] As described above, the anomaly determination device and the anomaly determination method related to one or more embodiments have been described based on the embodiments. However, the present disclosure is not limited to these embodiments. As long as it does not deviate from the gist of the present disclosure, the embodiments obtained by applying various modifications that occur to those skilled in the art to the present embodiment and the embodiments constructed by combining the constituent elements in different embodiments are also included in the scope of the present disclosure.

[0130] For example, in the above embodiment, an example in which the anomaly determination device includes a comparison unit is shown, but it is not limited thereto. The comparison can also be performed by other devices or by the operator. The anomaly determination device can also obtain the comparison result 124 by communicating with other devices or by accepting the input of the comparison result 124 from the operator.

[0131] In addition, for example, the model production unit 134 can also perform weighting of the data to be used when producing and updating the anomaly detection model 126. For example, the model production unit 134 can also change the weight of the comparison result based on the date and time of the comparison result. For example, it can be that when the date and time of obtaining the comparison result is information earlier than a specified threshold, that is, old information, the weight of this comparison result is reduced, and when the date and time of obtaining the comparison result is information later than the specified threshold, that is, new information, the weight of the comparison result is increased. That is, multiple determination results and multiple evaluation results are obtained at multiple date and times, and multiple comparison results obtained by comparing the multiple determination results with the multiple evaluation results respectively are obtained. When the determination result and the evaluation result in each comparison result are different, the learning unit 130 updates the anomaly detection model 126 based on the multiple evaluation results. In this case, when updating the anomaly detection model 126, the weight of the evaluation result obtained at a date and time later than the specified threshold among the multiple evaluation results is set to be larger than the weight of the evaluation result obtained before this threshold.

[0132] Alternatively, the model production unit 134 can also change the weight of the comparison result according to the proficiency of the operator 30 who performs the on-site confirmation as the source of the comparison result. The proficiency is a specified parameter based on the number of working days of the operator 30, work experience, evaluation from the manager, etc. In this case, for example, information indicating the proficiency of each operator is stored in the storage unit 120. The evaluation result acquisition unit 160 acquires information for determining proficiency, such as the identification number of the operator 30 who has performed the on-site confirmation, together with the evaluation result. It can also be that the model production unit 134 increases the weight of the comparison result based on the on-site confirmation performed by the operator 30 with a proficiency higher than the threshold, and reduces the weight of the comparison result based on the on-site confirmation performed by the operator 30 with a proficiency lower than the threshold. That is, multiple evaluation results of multiple operators are obtained. When the determination result and the evaluation result in the comparison result are different, the learning unit 130 updates the anomaly detection model 126 based on the multiple evaluation results. In this case, when updating the anomaly detection model 126, the weight of the evaluation result based on the on-site confirmation performed by the operator 30 with a proficiency higher than the threshold among the multiple evaluation results is increased, and the weight of the evaluation result based on the on-site confirmation performed by the operator 30 with a proficiency lower than the threshold is set to be smaller than the weight of the above evaluation result.

[0133] In addition, there is no particular limitation on the communication method between the devices described in the above embodiments. When wireless communication is performed between the devices, the wireless communication method (communication standard) is, for example, short-range wireless communication such as ZigBee (registered trademark), Bluetooth (registered trademark), or wireless LAN (Local Area Network), or the wireless communication method (communication standard) may also be communication via a wide-area communication network such as the Internet. In addition, wired communication may be performed between the devices instead of wireless communication. Specifically, the wired communication is power line transmission communication (PLC) or communication using a wired LAN, etc.

[0134] In addition, in the above embodiments, the processing executed by a specific processing unit may also be executed by other processing units. In addition, the order of multiple processes may be changed, or multiple processes may be executed in parallel. In addition, the allocation of the components included in the job notification system among multiple devices is an example. For example, the components included in one device may also be included in other devices.

[0135] For example, the processing described in the above embodiments may be implemented by centralized processing using a single device (system), or may also be implemented by distributed processing using multiple devices. In addition, the processor that executes the above program may be single or multiple. That is, either centralized processing or distributed processing may be performed.

[0136] Specifically, the abnormality determination device 100 may also execute at least a part of the functions of the input / output device 200. For example, it may be that the input / output device 200 is an input-only device, and the abnormality determination device 100 has the output function of the input / output device 200. In this case, the output unit 150 of the abnormality determination device 100 executes the functions of the display control unit 220 and the display unit 230.

[0137] In addition, for example, it may be that the input / output device 200 is an output-only device, and the abnormality determination device 100 has the input function of the input / output device 200. In this case, the evaluation result acquisition unit 160 of the abnormality determination device 100 executes the functions of the reception unit 240 and the signal processing unit 250. In addition, it may be that in this case, the abnormality determination device 100 has the display control unit 220 and the display unit 230, and the display unit 230 assists the input of the evaluation result by the operator.

[0138] Alternatively, the abnormality determination device 100 may also execute all the functions of the input / output device 200. That is, the abnormality determination device 100 and the input / output device 200 may be an integrated single device.

[0139] In addition, in the above-described embodiments, all or part of the constituent elements such as the control unit may be configured by dedicated hardware, or may be implemented by executing a software program suitable for each constituent element. Each constituent element may also be implemented by a program execution unit such as a CPU (Central Processing Unit) or a processor reading a software program recorded in a recording medium such as an HDD or a semiconductor memory and executing the software program.

[0140] In addition, constituent elements such as the control unit may be constituted by one or more electronic circuits. Each of the one or more electronic circuits may be a general-purpose circuit or a dedicated circuit.

[0141] The one or more electronic circuits may include, for example, semiconductor devices, ICs, or LSIs. The IC or LSI may be integrated on one chip or may be integrated on multiple chips. Here, it is called an IC or an LSI, but the naming method changes according to the degree of integration, and it may be called a system LSI, a VLSI (Very Large Scale Integration), or a ULSI (Ultra Large Scale Integration). In addition, an FPGA programmed after manufacturing the LSI can also be used for the same purpose.

[0142] In addition, the whole or a specific aspect of the present disclosure may be implemented by a system, a device, a method, an integrated circuit, or a computer program. Alternatively, it may be implemented by a computer-readable non-transitory recording medium such as an optical disc, an HDD, or a semiconductor memory storing the computer program. In addition, it may be implemented by any combination of a system, a device, a method, an integrated circuit, a computer program, and a recording medium.

[0143] In addition, various changes, substitutions, additions, omissions, etc. can be made within the scope of the claims or their equivalents in the above-described embodiments.

[0144] Industrial Applicability

[0145] The present disclosure can be used as an abnormality determination device and an abnormality determination method capable of accurately determining an abnormality, and can be used, for example, in a management system and a production system of a factory.

[0146] Description of Reference Numerals

[0147] 1: Abnormal determination system; 10: Device; 20: Network; 30: Operator; 100: Abnormal determination device; 110: Device information acquisition unit; 120: Storage unit; 122: Log data; 124: Comparison result; 126: Abnormality detection model; 130: Learning unit; 132: Request unit; 134: Model production unit; 140: Determination unit; 150: Output unit; 160: Evaluation result acquisition unit; 170: Comparison unit; 200: Input / output device; 210: Communication unit; 220: Display control unit; 230: Display unit; 240: Reception unit; 250: Signal processing unit.

Claims

1. An abnormality determination device, comprising: a learning unit that creates an abnormality detection model for detecting an abnormality of the device by machine learning using log data indicating the actual operating condition of the device; a determination unit that inputs input data indicating the actual operating condition of the target device into the abnormality detection model to determine the abnormality of the target device; and an output unit that outputs the determination result of the determination unit, wherein, when the determination result is different from the evaluation result for the target device obtained through on-site confirmation, the learning unit updates the abnormality detection model based on the evaluation result.

2. The abnormality determination device according to claim 1, wherein it further comprises a comparison unit that compares the determination result with the evaluation result, and when the determination result is different from the evaluation result, the learning unit updates the abnormality detection model based on the evaluation result.

3. The abnormality determination device according to claim 2, wherein when the determination result indicates that the target device is abnormal, the comparison unit performs the comparison, and when the determination result indicates that the target device is normal, the comparison unit does not perform the comparison.

4. The abnormality determination device according to any one of claims 1 to 3, wherein when the determination result indicates that the target device is abnormal, the output unit outputs the determination result, and when the determination result indicates that the target device is normal, the output unit does not output the determination result.

5. The abnormality determination device according to any one of claims 1 to 4, wherein it further comprises an acquisition unit that acquires the evaluation result.

6. The abnormality determination device according to any one of claims 1 to 5, wherein it further comprises a storage unit that stores the log data and the abnormality detection model.

7. An abnormality determination method, comprising the following steps: creating an abnormality detection model for detecting an abnormality of the device by machine learning using log data indicating the actual operating condition of the device; inputting input data indicating the actual operating condition of the target device into the abnormality detection model to determine the abnormality of the target device; and outputting the determination result obtained in the determination step, wherein, in the creating step, when the determination result is different from the evaluation result for the target device obtained through on-site confirmation, the abnormality detection model is updated based on the evaluation result.

8. A program that causes a computer to execute the abnormality determination method according to claim 7.

9. A recording medium that records a program that causes a computer to execute the abnormality determination method according to claim 7.

Images