Method of detecting abnormality and program

Abstract

To efficiently detect abnormality in various steps of an operation.SOLUTION: A method of detecting abnormality allows an arithmetic device connected for data communication with a management system for operation management to input a gap between a first number of processing and a second number of processing of execution-based operations of a plurality of steps calculated by discrete simulation to an AI model to output a value indicative of degrees of abnormality in respective ones of the plurality of steps, in response to an acquisition from the management system of the first number of processing of execution-based operations of the plurality of steps of operations, determine whether or not abnormalities are occurring in respective ones of the plurality of steps based on values corresponding to respective ones of the plurality of steps outputted from the AI model, and notify, if there is a step detected an abnormality occurring out of the plurality of steps, that an abnormality is occurring in that step.SELECTED DRAWING: Figure 5

Description

[Technical field]

[0001] The present disclosure relates to an anomaly detection method and a program. [Background technology]

[0002] Conventionally, computer-based simulations have been used to analyze abnormalities in business processes. For example, Patent Document 1 discloses a method for analyzing the root cause of retention of slabs in a production line using a logistics simulation. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2017-4149 A Summary of the Invention [Problem to be solved by the invention]

[0004] For example, in order to avoid unexpected stoppage of equipment installed in a factory or warehouse, or to perform proper maintenance of the equipment, it is necessary to detect the occurrence of an abnormality in the equipment and estimate the cause of the abnormality. Detecting abnormalities in various processes related to the work, as well as abnormalities in the equipment, enables efficient management of the work. In the analysis method disclosed in Patent Document 1, the flow of slabs in a production line is simulated to analyze the root cause of slab retention, and the analysis results can be reflected in the design of the production plant. However, since a simulation is required to analyze the root cause of slab retention, there is room for improvement in efficiently detecting in which process, among the various processes performed at the slab production site, an abnormality may occur.

[0005] The present disclosure has been devised in consideration of the above-mentioned conventional situation, and aims to efficiently detect abnormalities in various processes related to business. [Means for solving the problem]

[0006] The present disclosure provides an anomaly detection method for detecting an anomaly that may occur in at least one of a plurality of processes related to a business operation, executed by a computing device connected to a management system that manages the business operation and capable of data communication, the anomaly detection method comprising: inputting a gap between a first processing number of the business operation based on the execution of the plurality of processes, the first processing number being obtained from the management system, and a second processing number of the business operation based on the execution of the plurality of processes, the second processing number being calculated by a discrete simulation, into an AI model that outputs a value indicating the degree of anomaly in each of the plurality of processes; determining whether or not an anomaly has occurred in each of the plurality of processes based on the value corresponding to each of the plurality of processes output from the AI ​​model; and, if any of the plurality of processes is determined to have an anomaly, notifying the user that an anomaly has occurred in that process.

[0007] The present disclosure also provides a program for causing a computing device, which is a computer, connected to a management system that manages a business operation so as to be capable of data communication, to input a gap between a first processing number of the business operation based on the execution of the multiple processes related to the business operation, calculated by discrete simulation, into an AI model that outputs a value indicating the degree of abnormality at each of the multiple processes in response to obtaining from the management system a first processing number of the business operation based on the execution of the multiple processes related to the business operation, and determining whether or not an abnormality has occurred at each of the multiple processes based on the value corresponding to each of the multiple processes output from the AI ​​model, and if any of the multiple processes is determined to have an abnormality, notifying that an abnormality has occurred at that process.

[0008] Any combination of the above components, and conversion of the present disclosure into a method, device, system, storage medium, computer program, etc. are also valid aspects of the present disclosure. Effect of the Invention

[0009] According to the present disclosure, abnormalities in various processes related to business can be efficiently detected. [Brief description of the drawings]

[0010] [Figure 1] FIG. 1 is a block diagram showing a configuration example of a calculation device according to a first embodiment. [Diagram 2] FIG. 1 is a schematic diagram for explaining a process of a production operation according to a first embodiment; [Diagram 3] FIG. 1 is a schematic diagram for explaining a processing time of a process of a production operation according to the first embodiment; [Figure 4] FIG. 1 is a graph illustrating a change in the number of processes over time according to the first embodiment. [Diagram 5] FIG. 1 is a schematic diagram for explaining input and output of an AI model according to a first embodiment. [Figure 6] 1 is a flowchart of a process for generating learning data according to the first embodiment; [Figure 7] Flowchart of abnormality detection processing according to the first embodiment DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] Hereinafter, with reference to the drawings as appropriate, an embodiment specifically disclosing an anomaly detection method and a program according to the present disclosure will be described in detail. However, more detailed explanation than necessary may be omitted. For example, detailed explanation of already well-known matters and duplicate explanation of substantially the same configuration may be omitted. This is to avoid the following explanation becoming unnecessarily redundant and to facilitate understanding by those skilled in the art. Note that the attached drawings and the following explanation are provided to enable those skilled in the art to fully understand the present disclosure, and are not intended to limit the subject matter described in the claims.

[0012] (Embodiment 1) [Example of the configuration of the computing device] First, a configuration example of a calculation device 1 according to the first embodiment will be described with reference to Fig. 1. Fig. 1 is a block diagram showing a configuration example of a calculation device 1 according to the first embodiment.

[0013] The arithmetic device 1 is a device that executes various processes for detecting abnormalities that may occur in at least one of a plurality of steps related to a business. In the description of this embodiment, the business means a production business. The production business will be described later. Furthermore, the abnormalities that may occur include abnormalities that have actually occurred. The arithmetic device 1 is configured using a general-purpose computer device such as a personal computer or a server computer. The arithmetic device 1 includes a processor 2, a memory 3, an input device 4, a communication device 5, and a display device 6. Each part of the arithmetic device 1 is connected to be able to communicate with each other via an internal bus 7.

[0014] The processor 2 is configured using, for example, a Central Processing Unit (hereinafter referred to as "CPU"), a Graphics Processing Unit (hereinafter referred to as "GPU"), a Micro Processing Unit (hereinafter referred to as "MPU"), a Digital Signal Processor (hereinafter referred to as "DSP"), or a Field Programmable Gate Array (hereinafter referred to as "FPGA") etc. The processor 2 realizes various functions by reading and executing various data and programs etc. stored in the memory 3.

[0015] The memory 3 is a storage area for storing and holding various data, information, programs, etc. The memory 3 is composed of, for example, a Read Only Memory (hereinafter referred to as "ROM"), which is a non-volatile storage area, a Hard Disk Drive (hereinafter referred to as "HDD"), and a Random Access Memory (hereinafter referred to as "RAM"), which is a volatile storage area. The RAM is, for example, a work memory used during the operation of the arithmetic device 1. The ROM stores and holds, for example, programs for controlling the arithmetic device 1 in advance.

[0016] The input device 4 includes a keyboard, a mouse, a touch panel, or other input devices. The input device 4 accepts input of various data and information. The input device 4 is operated by a user such as a production operation manager.

[0017] The communication device 5 communicates with the production management system 8 via a network (not shown) and transmits and receives various data or information. The communication device 5 may be compatible with either wired or wireless communication. The communication method used by the communication device 5 may be, for example, a Wide Area Network (hereinafter referred to as "WAN"), a Local Area Network (hereinafter referred to as "LAN"), Long Term Evolution (hereinafter referred to as "LTE"), mobile communication such as 5G, power line communication, short-range wireless communication such as Wi-Fi (registered trademark) and Bluetooth (registered trademark), or a combination of these.

[0018] The display device 6 displays data generated by the processor 2. For example, the display device 6 displays a process in which it is determined that an abnormality has occurred as a result of various processes performed by the processor 2. The display device 6 is, for example, a display. The display device 6 may be provided separately from the arithmetic device 1.

[0019] A production management system 8, which is an example of a business management system, is a system for managing business carried out in a factory 9. The production management system 8 may be connectable to an in-factory computing device 10 for managing the factory 9. The production management system 8 may have the same configuration as the computing device 1, for example.

[0020] The production management system 8 manages the inventory status of goods stored in the factory 9, the transportation of goods in and out, etc. The goods may include products, materials, equipment, and device parts, etc. Information related to the production work carried out in the factory 9 is registered in the production management system 8. Information registered in the production management system 8 includes, for example, the number of production work processes based on the execution of each process related to the production work, and the processing time of each process related to the production work. Specific examples of production work and each process related to the production work will be described later. The number of processes and the processing time will also be described later.

[0021] The in-factory computing device 10 is installed, for example, in the factory 9, and records the inventory status of items managed in the factory 9, the number of items processed and the processing time of production work carried out in the factory 9, etc. The in-factory computing device 10 may receive various data from a terminal such as a handheld terminal, and record the inventory status of items, the number of items processed and the processing time of production work carried out in the factory 9, etc. in real time. The in-factory computing device 10 is also connectable to the production management system 8, and can transmit the recorded information, etc. to the production management system 8. The in-factory computing device 10 may have the same configuration as the computing device 1, for example.

[0022] 1 shows an example in which the production management system 8 manages one factory 9, but the production management system 8 may be capable of managing multiple factories. Also, in FIG. 1, an example in which the calculation device 1 communicates with one production management system 8, but the calculation device 1 may be capable of communicating with multiple production management systems.

[0023] [Production-related processes] Next, the steps of the production business, the number of processes, etc. will be described with reference to Figures 2, 3, and 4. Figure 2 is a schematic diagram for explaining the steps of the production business according to the first embodiment.

[0024] In this embodiment, "production work" refers to a series of work that is composed of one or more processes for some purpose and is carried out in a factory. An example of production work is the manufacture of products. In order to manufacture products, multiple processes such as inspection, processing, and assembly are carried out. In the following explanation, the processes related to production work and the processes of production work may be referred to as the processes that make up the production work.

[0025] For the sake of explanation, the production work composed of the various processes shown in FIG. 2 is the manufacture of a product. For example, an inspection process such as an inspection process PI1 is a process for performing various inspections of parts, etc. For example, a molding process such as a molding process PM1 is a process for forming a material into a predetermined shape. For example, a processing process such as a processing process PP1 is a process for processing materials or parts, etc. For example, an assembly process such as an assembly process PA1 is a process for assembling parts, etc. For example, a completion process such as a completion process PC1 is a process for performing various inspections of an assembled product. Note that these various processes are examples for explanation, and are not intended to limit the processes related to the manufacture of a product. Also, the various processes shown in FIG. 2 may be a part of the processes constituting the production work of the manufacture of a product. In other words, more inspection processes, molding processes, processing processes, etc. may constitute the production work. However, in the following explanation, for the sake of simplicity, attention is focused only on the various processes shown in FIG. 2.

[0026] In this embodiment, the number of products obtained by a certain process is called the "number of processes." However, it is assumed that a certain process is executed repeatedly. For example, in a processing process, after processing one part, another part is processed. In this way, the number of processes increases over time.

[0027] The product obtained by the completion process is the target product of the production work. Therefore, the number of processes in the completion process is sometimes referred to as the number of processes in the production work. In the example of Figure 2, the number of processes in the production work means the number of finished products manufactured in each completion process.

[0028] In this embodiment, at least the processing number of production tasks is sensed and registered in the production management system 8. That is, the processing number of the completing process PC1, the processing number of the completing process PC2, the processing number of the completing process PC3, and the processing number of the completing process PC4 are each sensed and registered in the production management system 8. Note that, although details will be described later using FIG. 4, changes in the processing number over time are sensed.

[0029] Next, the processing time of each process constituting a production job will be described with reference to Fig. 3. The processing time of a process is, in other words, the time required to execute the process and obtain one processing number. Fig. 3 is a schematic diagram for explaining the processing time of a process of a production job according to the first embodiment.

[0030] In this embodiment, the arithmetic device 1 treats the processing time of each process as a probability distribution. Although details will be described later, the arithmetic device 1 needs to set the processing time of each process in order to detect anomalies. Processes related to production work may vary in processing time whether they are automated by machines or performed by humans. Thus, by expressing the processing time of a process as a probability distribution, the arithmetic device 1 can handle the processing time of a process while taking such variations into account. Note that in this embodiment, the probability distribution will be explained using a Gaussian distribution as an example of its expression. The parameters of the probability distribution include μ and σ. Note that μ is the mean of the probability distribution, and σ is the mean of the probability distribution. 2 is called the variance of the probability distribution.

[0031] For example, the processing time of inspection process PI1 is represented by a probability distribution 20 characterized by parameter 20A. Also, for example, the processing time of molding process PM1 is represented by a probability distribution 21 characterized by parameter 21A. Also, for example, the processing time of machining process PP1 is represented by a probability distribution 22 characterized by parameter 22A. Also, for example, the processing time of inspection process PI2 is represented by a probability distribution 23 characterized by parameter 23A. Also, for example, the processing time of machining process PP2 is represented by a probability distribution 24 characterized by parameter 24A. Also, for example, the processing time of inspection process PI3 is represented by a probability distribution 25 characterized by parameter 25A.

[0032] Next, with reference to FIG. 4, the change in the number of processes over time will be described. FIG. 4 is a graph for explaining the change in the number of processes over time according to Embodiment 1.

[0033] In FIG. 4, the vertical axis represents the number of processes, and the horizontal axis represents time. Also, in FIG. 4, the magnitude relationship among B1, B2, B3, and B4 is B4 < B1 < B3 < B2. Also, in FIG. 4, the magnitude relationship between T1 and T2 is T1 < T2.

[0034] Characteristic 26 and characteristic 27 respectively show the change in the number of processes over time between time T1 and T3. For the sake of explanation, it is assumed that the number of processes shown in characteristic 26 and characteristic 27 is the number of processes in completion process PC1. Characteristic 28 is the difference between the number of processes shown in characteristic 26 and the number of processes shown in characteristic 27.

[0035] T1 is, for example, the sensing start time of the number of processes in completion process PC1. Or, T1 is, for example, the start time of completion process PC1. T3 is, for example, the sensing end time of the number of processes in completion process PC1. Or, T3 is, for example, the end time of completion process PC1.

[0036] Characteristic 26 represents the change in the number of processes over time when the processing times of the completion process PC1 and the various processes preceding the completion process PC1 are all average values. With reference to Fig. 3, it has been explained that the processing times of processes related to production operations are represented by probability distributions. When the processing times of a certain process and the various processes preceding that process are each the average of the probability distributions representing the processing times of each process, the characteristic representing the change in the number of processes over time in that process has a constant slope, as shown by characteristic 26.

[0037] In characteristic 26, the number of processes increases at a constant rate over time in the range from T1 to T3. In characteristic 26, the number of processes at time T1 is B1. B1 is, for example, 0. In characteristic 26, the number of processes at time T2 is B2, and thereafter, the number of processes increases until the time reaches T3.

[0038] Characteristic 27 is the change in the number of processes over time when the processing time in the completion process PC1 or any of the various processes before the completion process PC1 is an abnormal value. Whether or not the processing time is an abnormal value is determined by the calculation device 1, for example, based on the degree to which the processing time is different from the average of the probability distribution representing the processing time. Hereinafter, a processing time that is not an abnormal value may be referred to as a normal value. When the processing time is an abnormal value, the processing time may be longer or shorter than the average of the probability distribution representing the processing time. In other words, when the processing time of a process is an abnormal value, it may take longer than the average to execute the process and obtain one processing number, or it may take shorter than the average.

[0039] The time allocated to a production job may be predetermined. Therefore, if the processing time of any of the processes constituting the production job is an abnormal value longer than the average, the time for executing the processes after the process whose processing time is abnormal may be shorter than, for example, a predetermined time. In this case, the number of processes based on the execution of the processes after the process whose processing time is abnormal may also be less than, for example, a predetermined number of processes.

[0040] Furthermore, when the processing time in a certain process is an abnormal value longer than the average, the number of processes based on the execution of that process will be smaller than when the processing time in that process is normal. As a result, the number of processes based on the execution of processes after that process may also be smaller. For example, when the processing time of a processing process PP1 before a completion process PC1 is an abnormal value longer than the average, the processing process PP1 can process fewer parts than a predetermined number of parts. As a result, in an assembly process PA1 after the processing process PP1, for example, only a number of parts can be assembled that is smaller than a predetermined number. And the number of processes in a completion process such as the completion process PC1, i.e., the number of processes in a production operation, may also be smaller than the predetermined number of processes.

[0041] In characteristic 27, the number of processes at time T1 is B1, which is the same as in characteristic 26. However, because the processing time in completion process PC1 and in any of the various processes prior to completion process PC1 is an abnormal value, the manner in which the number of processes changes over time is different from that of characteristic 26. For example, in characteristic 27, the number of processes at time T2 is B3.

[0042] Characteristic 28 shows the difference between the number of processes over time shown in characteristic 26 and the number of processes over time shown in characteristic 27. The difference in the number of processes is calculated by subtracting the number of processes shown in characteristic 26 from the number of processes shown in characteristic 27. The difference in the number of processes is calculated by the calculation device 1. In characteristic 28, the number of processes at time T1 is B1. This is because there is no difference between characteristic 26 and characteristic 27 at time T1. In characteristic 28, the number of processes at time T2 is B4. This is the difference between the number of processes B2 shown in characteristic 26 at time T2 and the number of processes B3 shown in characteristic 27. When B1 is 0, B4 is a negative value. This indicates that the number of processes obtained at time T2 is smaller by B4 when the processing time of any process is abnormal compared to when the processing times of the completion process PC1 and the various processes before the completion process PC1 are all average values.

[0043] In the following description, the difference between the number of processes based on the execution of a plurality of processes when it is assumed that no abnormality occurs in any of the processes (for example, the second number of processes described later) and the number of processes other than the second number of processes over time may be referred to as the gap in the number of processes. The number of processes other than the second number of processes is, for example, the number of processes based on the execution of a plurality of processes when it is assumed that an abnormality occurs in one or more of the processes. In addition, the number of processes other than the second number of processes is, for example, the number of processes registered in the production management system 8 (for example, the first number of processes described later). The number of processes registered in the production management system 8 may include both the number of processes based on the execution of a plurality of processes when no abnormality occurs in any of the processes and the number of processes based on the execution of a plurality of processes when an abnormality occurs in one or more of the processes. One example of the gap is the difference in the number of processes over time indicated by the characteristic 28.

[0044] By dealing with the difference in the number of processes over time, rather than the difference in the number of processes at a specific timing, the calculation device 1 can avoid overlooking an abnormality in cases where there is almost no difference in the number of processes at a specific timing, but a large gap occurs before and after that timing.

[0045] For simplicity of explanation, the transaction numbers shown in characteristics 26 and 27 have been described as the transaction numbers of the completing process PC1. This is to ensure consistency in the explanation, because at least the transaction numbers of a production job, i.e., the transaction numbers of each completing process, are sensed and registered in the production management system 8. Specifically, when obtaining a transaction number gap for a certain process using the transaction number registered in the production management system 8 (for example, a first transaction number described below), the transaction number of the process needs to be sensed and registered in the production management system 8. Therefore, the explanation has been given using the transaction number of the completing process PC1, which is the sensing target and registered in the production management system 8, as an example.

[0046] However, the number of processes for obtaining the gap is not limited to the number of processes in a completing process such as the completing process PC1, in other words, the number of finished products, but may be the number of processes in each process before the completing process, in other words, the number of semi-finished products. For example, the numbers of processes in the molding process PM2, the processing process PP5, the inspection process PI14, etc. may be sensed and registered in the production management system 8, and a gap for the number of processes in each process may be calculated based on the number of processes in each process.

[0047] [AI model generation] In this embodiment, a trained AI model is used for efficient anomaly detection. The AI ​​model may be generated by an arbitrary learning algorithm so as to output a value indicating the degree of anomaly for each process related to the production work when a gap in the number of processes is input. The learning algorithm for generating the trained AI model according to this embodiment is not particularly limited, but for example, a Multi Layer Perceptron (hereinafter referred to as "MLP") may be used.

[0048] Generation of the AI ​​model 32 will be described with reference to Fig. 5 and Fig. 6. First, input data and output data of the AI ​​model 32 will be described with reference to Fig. 5. Fig. 5 is a schematic diagram for explaining input and output of the AI ​​model 32 according to the first embodiment.

[0049] The input to the AI ​​model 32 is the gap between the first processing number registered in the production management system 8 and the second processing number calculated by the discrete simulation. The first processing number registered in the production management system 8 is the change in the processing number over time in a certain process based on the execution of the process and various processes before the process. The second processing number calculated by the discrete simulation is the change in the processing number over time in a certain process, assuming that the processing times of the certain process and various processes before the process are all normal values. In other words, the second processing number is the processing number based on the execution of the multiple processes related to the production work, assuming that no abnormality occurs in any of the processes. An example of the second processing number is the change in the processing number over time shown in the characteristic 26 of FIG. 4. The computing device 1 can obtain a large amount of the second processing number at high speed by the discrete simulation. In the description of this specification, the expressions "first", "second" and "third" described later are used merely to distinguish from other elements, and are not intended to be interpreted in a limited manner.

[0050] The first process number needs to be sensed and registered in the production management system 8. In other words, the process number being sensed is treated as the first process number. For example, when the process number for a completion process PC1 is sensed, the process number for the completion process PC1 is treated as the first process number. When the process numbers for multiple processes are sensed, the process numbers for the multiple processes are treated as the first process number.

[0051] If the processing numbers of all the processes constituting the production work can be sensed, anomalies can be detected, for example, by monitoring the sensing results in all the processes. However, as described with reference to FIG. 2, the processes constituting the production work are diverse, and it may not be practical to sense the processing numbers of all the processes. Therefore, in this embodiment, in order to detect anomalies, at least the processing numbers of the production work, that is, the processing numbers of each completion process, are sensed and registered in the production management system 8. In addition to the processing numbers of the production work, the processing numbers of various processes constituting the production work other than the completion process, such as the inspection process PI4, the assembly process PA5, or the assembly process PA6, are sensed and treated as the first processing numbers, so that the accuracy of the output of the AI ​​model 32 described later can be improved.

[0052] In order to calculate the gap between the first and second process numbers, the first and second process numbers need to be the process numbers of the same process. For example, gap 29 is the first process number of a completion process PC1 registered in the production management system 8 and the second process number of a completion process PC1 calculated by the discrete simulation. For example, gap 30 is the first process number of a completion process PC2 registered in the production management system 8 and the second process number of a completion process PC2 calculated by the discrete simulation. For example, gap 31 is the first process number of an assembly process PA1 registered in the production management system 8 and the second process number of a assembly process PA1 calculated by the discrete simulation.

[0053] The AI ​​model 32 outputs a value corresponding to each process constituting a production operation in response to an input of the gap between the first processing count and the second processing count. The value output by the AI ​​model 32 indicates the degree of abnormality in the corresponding process. The AI ​​model 32 outputs the value by linking it to the process. Each process is made distinguishable from the other processes. For example, as shown in FIG. 5, each process may be assigned a process number. In this case, for example, process number 1 may be an inspection process PI1. Also, for example, process number 2 may be a molding process PM1.

[0054] The value output by the AI ​​model 32 may be a numerical value as shown in FIG. 5, or may be a character such as an English letter. Whether the value output by the AI ​​model 32 is a numerical value or a character may be determined by a user who generates the AI ​​model 32. When the value output by the AI ​​model 32 is a numerical value, the range of the numerical value is, for example, from 0 to 1. The numerical value indicates the degree of abnormality, the possibility that an abnormality has occurred, or both, depending on the magnitude of the numerical value. Whether the numerical value indicates the degree of abnormality, the possibility that an abnormality has occurred, or both, may be determined in advance, for example, by a user or the like when generating the AI ​​model 32. When the numerical value indicates the degree of abnormality, the larger the value of the numerical value, i.e., the closer to 1, the larger the degree of abnormality in the process corresponding to the numerical value. When the numerical value indicates the possibility of an abnormality, the larger the value of the numerical value, i.e., the closer to 1, the higher the possibility of an abnormality in the process corresponding to the numerical value. When the value output by the AI ​​model 32 is a character such as an English letter, the character may indicate the degree of abnormality, the possibility that an abnormality has occurred, or both. When an English letter is output from the AI ​​model 32, the English letter may be defined as increasing in ascending alphabetical order, for example. The English letter may play a role similar to that of a numerical value when the value output from the AI ​​model 32 is a numerical value. For example, the closer an alphabet is to Z, the greater the degree of abnormality in the process corresponding to the alphabet. Also, for example, the closer an alphabet is to Z, the higher the possibility that an abnormality has occurred in the process corresponding to the alphabet. In the following description, the value output by the AI ​​model 32 is a numerical value. Also, in the following description, the calculation device 1 determines whether or not an abnormality has occurred in each process based on the numerical value.

[0055] A threshold value for the arithmetic device 1 to determine whether an abnormality occurs in each process based on the numerical value output from the AI ​​model may be specified in advance by a user or the like. For example, when the threshold value is set to 0.7, the arithmetic device 1 determines that an abnormality occurs in a process whose numerical value is 0.7 or more. In this case, in the example of FIG. 5, the arithmetic device 1 determines that an abnormality occurs in the processes whose process numbers are 3, 8, and 14. Note that, in the example of FIG. 5, the numerical values ​​are shown to two decimal places, but this is not limited to this. In addition, the threshold value may be set uniformly for all the processes, or may be set for each process so as to correspond to each of the processes.

[0056] The threshold value for the arithmetic device 1 to determine whether or not an abnormality has occurred may be determined based on data such as the processing time of each process constituting a production job, which is registered in the production management system 8. For example, the threshold value for a certain process may be determined based on the average value of the processing time of the process registered in the production management system 8.

[0057] Next, generation of learning data for generating the AI ​​model 32 will be described with reference to Fig. 6. Fig. 6 is a flowchart of a process for generating learning data. Each process in the flowchart of Fig. 6 is executed by the arithmetic device 1. Note that the arithmetic device 1 will be described as using a discrete simulation for each process in the flowchart of Fig. 6. By using the discrete simulation, the arithmetic device 1 can quickly generate learning data even when a large amount of learning data is to be generated.

[0058] The arithmetic device 1 sets parameters of a probability distribution representing the processing time of each process constituting the production work (step St100). The parameters of the probability distribution are set based on the processing time of each process registered in the production management system 8. For example, the arithmetic device 1 may calculate the parameters of the probability distribution by a discrete simulation or the like based on the processing time of each process registered in the production management system 8. For example, the arithmetic device 1 may create a simulation model of each process constituting the production work and exploratory find the parameters by simulation. For example, the arithmetic device 1 may use a trained AI model for finding the parameters of the probability distribution. For example, the arithmetic device 1 may find the parameters of the probability distribution by various maximum likelihood estimation methods.

[0059] The arithmetic device 1 sets the processing time of each process constituting the production work as a normal value based on the parameters of the probability distribution representing the processing time of each process set in step St100, and calculates the second processing number of each process based on the set processing time of each process (step St101). The arithmetic device 1 randomly sets the processing time of each process within the range from the mean μ of the probability distribution to 1σ as the normal value, for example. The arithmetic device 1 also calculates the second processing number of each process based on the set processing time of each process.

[0060] The arithmetic device 1 sets the processing time of one or more of the processes constituting the production operation as an abnormal value based on the parameters of the probability distribution representing the processing time of each process set in step St100 (step St102). For example, the arithmetic device 1 sets the processing time for the assembly process PA6 as an abnormal value. For example, the arithmetic device 1 sets a value that is farther than 1σ from the average μ of the probability distribution representing the processing time of the assembly process PA6 as the abnormal value of the processing time of the assembly process PA6. In addition, the arithmetic device 1 sets the processing time for each of the machining process PP1 and the assembly process PA2 as an abnormal value.

[0061] The computing device 1 sets the processing time of each process constituting the production business to a normal value, except for the process whose processing time was set to an abnormal value in step St102, and calculates a third processing number for each process constituting the production business based on the processing time of each process set to a normal value or an abnormal value (step St103). The third processing number is the processing number based on the execution of multiple processes related to the production business when it is assumed that an abnormality has occurred in one or more of the processes.

[0062] The arithmetic device 1 calculates the gap between the second processing number calculated in step St101 and the third processing number calculated in step St103 for each process constituting the production work (step St104). An example will be described in which the arithmetic device 1 sets the processing times of the processing process PP1, the assembly process PA2, and the assembly process PA6 as abnormal values ​​in step St102. In this case, the third processing number of processes after the processes whose processing times are set to abnormal values, such as the inspection process PI2, the assembly process PA3, and the completion process PC1, is different from the second processing number when the processing times of each process constituting the production work are normal. Therefore, a gap such as that shown in the characteristic 28 of FIG. 4 is obtained between the process whose processing time is set to an abnormal value and the process after the process whose processing time is set to an abnormal value.

[0063] On the other hand, for processes before the processes whose processing times are set to abnormal values, such as the molding process PM1, the processing process PP3, and the inspection process PI12, the third processing number is equal to or almost equal to the second processing number when the processing time of each process constituting the production work is normal. Therefore, for processes before the processes whose processing times are set to abnormal values, a gap is obtained in which there is almost no difference in the processing number over time.

[0064] The calculation device 1 calculates a numerical value corresponding to each process based on the processing time of each process set as an abnormal value or a normal value in steps St102 and St103 (step St105). For example, the calculation device 1 calculates a numerical value based on how many σ the processing time of a certain process set is away from the average μ of the probability distribution representing the processing time of the process. The numerical value is calculated within the range from 0 to 1, and the greater the numerical value is, the farther the processing time is from the average of the probability distribution.

[0065] The computing device 1 prepares learning data consisting of pairs of input data representing the gap between the second processing number and the third processing number for each process constituting the production operation and output data representing a numerical value corresponding to each process.

[0066] The arithmetic device 1 determines whether the number of pieces of learning data is sufficient (step St106). Note that a specified value of the number of pieces of learning data required may be set in advance by a user. For example, the specified value of the number of pieces of learning data required may be set by verifying in advance how much learning data is required to generate the AI ​​model 32 in order to obtain a desired accuracy.

[0067] When the calculation device 1 determines that the number of learning data is insufficient (step St106; NO), the calculation device 1 returns to step St102 and repeats the process. In order to realize the AI ​​model 32 with high accuracy, it is preferable that learning data in which the processing time is set to an abnormal value is generated for each of all processes constituting the production work. The calculation device 1 can generate a large amount of learning data at high speed by using a discrete simulation for each process of this processing flow.

[0068] When the calculation device 1 determines that the number of pieces of learning data is sufficient (step St106; YES), it ends this processing flow.

[0069] [Processing flow] Next, a process of the arithmetic device 1 performing anomaly detection will be described with reference to Fig. 7. Fig. 7 is a flowchart of the anomaly detection process according to the embodiment 1. Each process in the flowchart of Fig. 7 is executed by the arithmetic device 1.

[0070] As described with reference to FIG. 6, the arithmetic device 1 generates learning data for generating the AI ​​model 32 (step St200).

[0071] The arithmetic device 1 generates the AI ​​model 32 by using the learning data generated in the process of step St200 (step St201).

[0072] The arithmetic device 1 acquires the first process number registered in the production management system 8 (step St202). The first process number acquired by the arithmetic device 1 in step St202 includes at least the process number of the production job, that is, the process number of the completion process.

[0073] The arithmetic device 1 inputs the gap between the first processing number acquired in step St202 and the second processing number generated in step St200, more precisely, the second processing number calculated in step St101 of the processing flow shown in FIG. 6, to the AI ​​model 32 (step St203). As the second processing number, the processing numbers of all processes constituting the production work are calculated by discrete simulation. However, the first processing number acquired by the arithmetic device 1 in step St202 is the processing number of the process to be sensed, and is not necessarily the first processing number of all processes. Therefore, the gap calculated in step St203 is also the gap for the process to be sensed. In this embodiment, at least each completion process is a process to be sensed.

[0074] The arithmetic device 1 acquires numerical values ​​corresponding to each of the processes constituting the production work, which are output from the AI ​​model 32 in response to the input of the gap in the process of step St203 (step St204).

[0075] The arithmetic device 1 compares the numerical values ​​corresponding to each process, acquired in step St204, with a specified threshold value (step St205).

[0076] If the comparison result in the process of step St205 indicates that a numerical value is equal to or greater than a prescribed threshold value, the arithmetic device 1 determines that an abnormality has occurred in the process corresponding to that numerical value (step St206).

[0077] The arithmetic device 1 outputs the numerical values ​​corresponding to each process to the display device 6 together with the determination result in the process of step St206 so that the link between the numerical values ​​and the processes can be understood (step St207). The arithmetic device 1 outputs the numerical values ​​and the processes to the display device 6 in a table format, for example, as in the output data shown in Fig. 5. In this case, the arithmetic device 1 enables the user or the like to confirm the process in which an abnormality has occurred, for example, by changing the color of the square in the table in which the process in which it has been determined that an abnormality has occurred and the numerical value corresponding to the process are written.

[0078] The arithmetic device 1 determines whether to end the abnormality detection (step St208). The determination of whether to end the abnormality detection by the arithmetic device 1 may be determined according to an input by a user such as a production manager via the input device 4, for example.

[0079] When the calculation device 1 determines not to end the abnormality detection (step St208; NO), it refers to data such as the processing time of each process constituting the production work and the number of processes to be sensed, which are registered in the production management system 8, and determines whether or not a sufficient amount of data has been accumulated in the production management system 8 (step St209). The criterion by which the calculation device 1 determines whether or not a sufficient amount of data has been accumulated and registered in the production management system 8 may be set in advance by a user, etc.

[0080] The parameters of the probability distribution set in step St100 of the process flow shown in Fig. 6 are set by the calculation device 1 based on data registered in the production management system 8. Therefore, the more data registered in the production management system 8 that the calculation device 1 refers to for setting the parameters, the more accurately the parameters are set.

[0081] In step St208, when the number of data registered in the production management system 8 for setting the parameters of the probability distribution is sufficiently accumulated compared to the number of data at the time when the parameters were set to generate the AI ​​model 32, the calculation device 1 refers to the data registered in the production management system 8 in order to perform the processing of steps St200 and St201 to generate a more accurate AI model.

[0082] When the calculation device 1 determines that a sufficient amount of data is not accumulated in the production management system 8 (step St209; NO), the calculation device 1 returns to step St202 and repeats the process. In this case, the calculation device 1 continues to use the AI ​​model 32 for anomaly detection.

[0083] When the arithmetic device 1 determines that a sufficient amount of data has been accumulated in the production management system 8 (step St209; YES), the arithmetic device 1 returns to step St200 and repeats the process. In this case, the arithmetic device 1 executes step St200, i.e., the process flow shown in FIG. 6, to generate learning data. At this time, the parameters of the probability distribution set in step St100 of the process flow shown in FIG. 6 become more accurate because the amount of data registered in the production management system 8 for parameter setting has increased. As a result, the output of the newly generated AI model is also realized with high accuracy. The arithmetic device 1 can perform anomaly detection with higher accuracy by using the newly generated AI model instead of the AI ​​model 32.

[0084] When the arithmetic device 1 determines to end the abnormality detection (step St208; YES), it ends this processing flow.

[0085] (Other variations) In the above-described first embodiment, an example has been shown in which the arithmetic device 1 outputs the process in which an abnormality has occurred to the display device 6 in the process of step St207 in the processing flow shown in Fig. 7. However, this is not limited to this, and for example, the arithmetic device 1 may cause a device that emits a sound, such as an alarm device, to notify a user, etc., of the process in which an abnormality has occurred. In this way, a user, such as a manager of the factory 9, can know the process in which an abnormality has occurred by the sound without checking the display device 6.

[0086] In the above-mentioned first embodiment, the arithmetic device 1 has output the numerical values ​​and the processes in a table format to the display device 6 as in the output data shown in FIG. 5 in the process of step St207 of the process flow shown in FIG. 7. In the table format output data shown in FIG. 5, a process number is assigned to each process, and a numerical value corresponding to each process is shown in the order of the process number. However, this is not limited to this, and the arithmetic device 1 may display each process and a numerical value corresponding to each process on the display device 6 based on the magnitude of the numerical value. The arithmetic device 1 may display each process on the display device 6 in the order of the largest numerical value, for example. This allows a user such as a manager of the factory 9 to check each process in the order of the largest numerical value. This allows a user to efficiently check a process in which a larger abnormality occurs, for example, when the numerical value indicates the degree of abnormality of the process.

[0087] In the above-described first embodiment, when the calculation device 1 determines that a sufficient amount of data has been accumulated in the production management system 8 (step St209: YES in the process flow shown in FIG. 7), the calculation device 1 returns to the process of setting parameters of the probability distribution representing the processing time of each process (step St100 in the process flow shown in FIG. 6), generates learning data, and further generates an AI model. However, this is not limited to this, and the calculation device 1 may update the AI ​​model as needed. In other words, the AI ​​model is not limited to a state at any point in time. Updating includes generating a new AI model different from the AI ​​model used by the calculation device 1. The calculation device 1 may update the AI ​​model as needed in response to each process constituting the production work being executed in the factory 9 and data such as the processing time of each process being registered in the production management system 8. For example, a method such as transfer learning or fine tuning may be used to update the AI ​​model.

[0088] In the above embodiment 1, the production management system 8 has been described as an example of a management system, but the present invention is not limited thereto, and the technical idea of ​​the above embodiment 1 can also be applied to, for example, a warehouse management system or a logistics management system. A warehouse management system may be called a Warehouse Management System (hereinafter, referred to as "WMS"). For example, the production management system 8 in the description of embodiment 1 may be read as a warehouse management system. In this case, the factory may be read as a warehouse, the in-factory computing device may be read as a warehouse computing device, and the production work may be read as a warehouse work. Examples of warehouse work include simple processing work and item picking work.

[0089] In the above embodiment 1, the number of deliverables obtained by a certain process has been described as the number of processes. In embodiment 1, since the manufacturing of products is taken as an example of a production operation, the number of processes is the number of finished products or semi-finished products. In other words, the deliverables are tangible. However, this is not limited to this, and the deliverables may be intangible. For example, the deliverables may be data.

[0090] (Summary of the first embodiment) At least the following techniques are disclosed by the above description of the first embodiment. Note that, in parentheses, examples of components corresponding to the first embodiment are shown, but the present invention is not limited to these.

[0091] (Technology 1) A calculation device (e.g., the calculation device 1) connected to a management system (e.g., a production management system 8) for managing a business and capable of data communication detects an abnormality that may occur in at least one of a plurality of processes (e.g., an inspection process PI1, a processing process PP1, an assembly process PA1, a completion process PC1, etc.) related to the business, a gap (e.g., gap 29) between the first processing number and a second processing number of the task based on the execution of the multiple steps calculated by a discrete simulation is input to an AI model (e.g., AI model 32) that outputs a value indicating the degree of abnormality in each of the multiple steps in response to acquisition of a first processing number of the task based on the execution of the multiple steps from the management system; Based on the values ​​corresponding to each of the multiple processes output from the AI ​​model, it is determined whether or not an abnormality has occurred in each of the multiple processes; If it is determined that an abnormality has occurred in any of the multiple processes, a notification is given that an abnormality has occurred in that process. Execute the anomaly detection method.

[0092] As a result, the calculation device can efficiently detect an abnormality that may occur in at least one of the multiple processes related to the business by using an AI model that outputs a value indicating the degree of abnormality in each of the multiple processes related to the business when a gap between the first processing count and the second processing count is input. Furthermore, when the calculation device determines that an abnormality has occurred in a process, it notifies the user that an abnormality has occurred in the process. As a result, a user such as a manager of a factory or warehouse can efficiently check in which process of the business an abnormality has occurred.

[0093] (Technology 2) In the anomaly detection method described in Technology 1, the computing device determines, for each of a plurality of processes, whether or not a value corresponding to each of the plurality of processes output from the AI ​​model is equal to or greater than a specified threshold value corresponding to each of the plurality of processes. If the computing device determines that the value of any of the plurality of processes is equal to or greater than the threshold value, it notifies the user that an anomaly has occurred in that process.

[0094] This allows the computing device to determine that an abnormality has occurred in a process when a value output from the AI ​​model corresponding to that process is equal to or greater than a specified threshold value. This allows a user, such as a manager of a factory or warehouse, to set a threshold value to determine whether a process is abnormal or normal.

[0095] (Technology 3) In the anomaly detection method described in Technology 1 or 2, the second processing number is the processing number of a task based on the execution of multiple processes when it is assumed that an abnormality has occurred in none of the multiple processes, and the calculation device calculates, by discrete simulation, a third processing number of a task based on the execution of the multiple processes when it is assumed that an abnormality has occurred in one or more of the multiple processes, and values ​​corresponding to each of the multiple processes in that case, and generates an AI model using the gap between the second processing number and the third processing number calculated by the discrete simulation and the values ​​corresponding to each of the multiple processes as learning data.

[0096] This enables the computing device to generate large amounts of learning data for creating an AI model quickly through discrete simulation.

[0097] (Technology 4) In the anomaly detection method described in any one of Techniques 1 to 3, when the computing device determines that an abnormality has occurred in any of the multiple processes based on values ​​corresponding to each of the multiple processes output from the AI ​​model, the computing device notifies the degree of the abnormality occurring in that process.

[0098] This allows the computing device to notify the degree of anomaly occurring in a certain process based on the value output from the AI ​​model, so that, for example, when anomalies occur in multiple processes, a user such as a manager of a factory or warehouse can prioritize how to deal with each anomaly based on the degree of the anomaly in each process.

[0099] (Technology 5) In the anomaly detection method described in Technique 1 or 2, the gap is a difference between a second processing number based on the execution of the multiple processes when it is assumed that no abnormality has occurred in any of the multiple processes, and a first processing number based on the execution of the multiple processes registered in the management system.

[0100] As a result, the calculation device can obtain the difference between the second processing number and the first processing number, using the second processing number as a reference for the processing number when no abnormality occurs in any of the processes.

[0101] (Technology 6) In the anomaly detection method described in Technique 3 or 4, the processing times of each of the multiple processes are represented by a probability distribution (e.g., probability distribution 20) based on data registered in the management system, and the calculation device calculates the second processing number and the third processing number based on the probability distribution by discrete simulation.

[0102] As a result, even if there may be variation in the processing time for each of the multiple processes related to the business, by representing the processing time as a probability distribution, the processing time for each of the multiple processes related to the business can be represented including such variation.

[0103] (Technology 7) In the anomaly detection method described in Technique 6, the probability distribution is updated in response to the accumulation of data in the management system, and the computing device generates learning data in response to the update of the probability distribution through discrete simulation, and updates the AI ​​model in response to the generation of the learning data.

[0104] This enables the computing device to update the AI ​​model in response to data accumulation in the management system, enabling efficient anomaly detection with high accuracy.

[0105] (Technology 8) In the anomaly detection method according to any one of Techniques 2 to 7, the threshold is determined based on the processing time of each of a plurality of processes registered in the management system.

[0106] This allows the calculation device to determine the threshold value used to determine whether or not an abnormality has occurred based on the data registered in the management system, making it possible to make judgments tailored to each process.

[0107] (Technology 9) In the anomaly detection method described in any one of Techniques 3, 6, and 7, when the number of pieces of learning data is less than a specified value of the number of pieces of learning data required to generate an AI model, the computing device calculates, by discrete simulation, a third processing number of a task based on the execution of multiple steps when it is assumed that an abnormality has occurred in one or more steps among the multiple steps, and a value indicating the degree of abnormality in each of the multiple steps in that case, and adds the gap between the second processing number and the third processing number and the value as learning data.

[0108] This allows the computing device to repeat the process until the required amount of training data for generating an AI model is obtained.

[0109] (Technology 10) The program causes an arithmetic device, which is a computer connected to a management system that manages business operations so as to be capable of data communication, to input the gap between the first processing number and a second processing number of a business operation based on the execution of multiple processes related to the business, calculated by discrete simulation, into an AI model that outputs a value indicating the degree of abnormality in each of the multiple processes in response to obtaining from the management system a first processing number of the business operation based on the execution of multiple processes related to the business operation, and determines whether or not an abnormality has occurred in each of the multiple processes based on the values ​​corresponding to each of the multiple processes output from the AI ​​model, and if it is determined that an abnormality has occurred in any of the multiple processes, notifies the user that an abnormality has occurred in that process.

[0110] This allows the program to achieve the same effect as Technique 1.

[0111] The functions of the above-described embodiments can also be realized by supplying programs and applications for realizing the functions of the above-described embodiments to a system or device via a network or storage medium, and having one or more processors in a computer of the system or device read and execute the programs.

[0112] Furthermore, the functions of the above-described embodiments may be realized by a circuit that realizes one or more functions (for example, an Application Specific Integrated Circuit (hereinafter, referred to as "ASIC") or an FPGA).

[0113] Although the embodiments of the present disclosure have been described above with reference to the drawings, it goes without saying that the present disclosure is not limited to such examples. It is clear that a person skilled in the art can think of various modifications, corrections, substitutions, additions, deletions, and equivalents within the scope of the claims, and it is understood that these also naturally belong to the technical scope of the present disclosure. In addition, the components in the above-mentioned embodiments may be arbitrarily combined within the scope of the invention. [Industrial Applicability]

[0114] The present disclosure is useful as an anomaly detection method and program. [Explanation of symbols]

[0115] 1 Computing device 2 processors 3. Memory 4 Input Devices 5. Communications equipment 6 Display device 8 Factory Management System 9. Factory 10 Factory computing device

Claims

1. An anomaly detection method for detecting an anomaly that may occur in at least one of a plurality of processes related to a business, the anomaly detection method being executed by a computing device connected to a management system that manages the business so as to be capable of data communication, the method comprising: an AI model that outputs a value indicating the degree of anomaly in each of the plurality of processes in response to acquisition of a first processing number of the task based on the execution of the plurality of processes from the management system, the AI ​​model inputting a gap between the first processing number and a second processing number of the task based on the execution of the plurality of processes calculated by a discrete simulation; determining whether or not an abnormality has occurred in each of the plurality of processes based on the values ​​corresponding to each of the plurality of processes output from the AI ​​model; If it is determined that an abnormality has occurred in any of the plurality of processes, a notification is given that an abnormality has occurred in the process. Anomaly detection methods.

2. Determining whether or not the value corresponding to each of the plurality of processes output from the AI ​​model is equal to or greater than a specified threshold value corresponding to each of the plurality of processes, for each of the plurality of processes; If it is determined that the value is equal to or greater than the threshold value among the plurality of processes, a notification is given that an abnormality has occurred in the process. The anomaly detection method according to claim 1 .

3. the second processing number is a processing number of the job based on execution of the plurality of steps on the assumption that no abnormality occurs in any of the plurality of steps, calculating, by the discrete simulation, a third processing number of the task based on the execution of the plurality of steps when it is assumed that an abnormality has occurred in one or more steps among the plurality of steps, and the value corresponding to each of the plurality of steps in that case; generating the AI ​​model using the gap between the second number of processes and the third number of processes calculated by the discrete simulation and the values ​​corresponding to each of the plurality of processes as learning data; The anomaly detection method according to claim 1 .

4. If it is determined that an abnormality has occurred in any of the plurality of processes based on the values ​​corresponding to each of the plurality of processes output from the AI ​​model, notifying the degree of the abnormality occurring in that process. The anomaly detection method according to claim 1 .

5. The gap is a difference between a second processing number based on the execution of the plurality of processes when it is assumed that no abnormality occurs in any of the plurality of processes, and a first processing number based on the execution of the plurality of processes registered in the management system. The anomaly detection method according to claim 1 .

6. a processing time for each of the plurality of processes is represented by a probability distribution based on data registered in the management system; calculating the second number of processes and the third number of processes based on the probability distribution by the discrete simulation; The anomaly detection method according to claim 3 .

7. The probability distribution is updated in response to accumulation of the data in the management system; generating the learning data according to the update of the probability distribution by the discrete simulation; Updating the AI ​​model in response to generation of the learning data. The anomaly detection method according to claim 6.

8. The threshold value is determined based on the processing times of the plurality of processes registered in the management system. The anomaly detection method according to claim 2 .

9. If the number of the learning data is less than a specified value of the number of the learning data required to generate the AI ​​model, a third processing number of the warehouse operation based on the execution of the plurality of processes when it is assumed that an abnormality has occurred in one or more of the plurality of processes, and a value indicating the degree of abnormality in each of the plurality of processes in that case are calculated by the discrete simulation, and the gap between the second processing number and the third processing number and the value are added as the learning data. The anomaly detection method according to claim 3 .

10. A computing device, which is a computer, is connected to a management system that manages business operations so as to be capable of data communication. inputting a gap between a first processing number of the task based on the execution of the plurality of steps, which is calculated by a discrete simulation, into an AI model that outputs a value indicating a degree of abnormality in each of the plurality of steps in response to acquisition of a first processing number of the task based on the execution of the plurality of steps related to the task from the management system; determining whether or not an abnormality has occurred in each of the plurality of processes based on the values ​​corresponding to each of the plurality of processes output from the AI ​​model; If it is determined that an abnormality has occurred in any of the plurality of processes, a notification is made that an abnormality has occurred in the process. Program for.

Images