Process analysis device, method, and computer-readable storage medium

Abstract

Provided is a process analysis device, method, and computer-readable storage medium that can easily determine a factor of an abnormality occurring in a manufacturing line. The device includes: a normal data acquisition section that acquires first state data related to normal states of a plurality of mechanisms; an abnormal data acquisition section that acquires second state data related to abnormal states of the mechanisms; a normal-time analysis section that analyzes the first state data and determines an inter-mechanism connection state as a first connection state; an abnormal-time analysis section that analyzes the second state data and determines an inter-mechanism connection state as a second connection state; a normal-time relationship determination section that determines an inter-mechanism causal relationship based on the first connection state as a normal-time causal relationship; an abnormal-time relationship determination section that determines an inter-mechanism causal relationship based on the second connection state as an abnormal-time causal relationship; and a special causal relationship determination section that compares the normal-time and abnormal-time causal relationships and determines an inter-mechanism causal relationship associated with the occurrence of an abnormality as a special causal relationship.

Description

Technical Field

[0001] This invention relates to a process analysis apparatus, a process analysis method, and a computer-readable storage medium for process analysis. Background Technology

[0002] Manufacturing lines in factories and similar facilities include multiple mechanisms such as conveyors and robotic arms. If any mechanism on the manufacturing line malfunctions, production will cease, potentially causing significant damage. Therefore, in factories and similar facilities, maintenance personnel regularly inspect the manufacturing lines to check for any abnormalities or signs of impending problems.

[0003] When an anomaly or its precursor is detected within a manufacturing line, the true cause of the anomaly may lie in a mechanism preceding the one that detected it. Therefore, to determine the true cause of the anomaly, it is crucial to understand the causal relationships between the various mechanisms within the manufacturing line. However, due to the large number of mechanisms constituting a manufacturing line and the fact that the operating conditions of each mechanism may change daily, it is difficult to accurately determine the causal relationships of all mechanisms.

[0004] Therefore, in the past, skilled maintenance personnel would rely on their experience and investigation to understand the causal relationships between the various mechanisms that make up the manufacturing line, and detect anomalies or their precursors within the line. To enable less skilled maintenance personnel to perform this maintenance work, there is a need to develop a technology that can visualize the causal relationships between the various mechanisms that make up the manufacturing line.

[0005] For example, in the system disclosed in Patent Document 1, domain knowledge is acquired when constructing a causal relationship model. Furthermore, the constructed causal relationship model is compared with the domain knowledge, and if there is a contradiction between the two, the causal relationship model is reconstructed.

[0006] Existing technical documents

[0007] Patent documents

[0008] Patent Document 1: Japanese Patent Application Publication No. 2018-181158 Summary of the Invention

[0009] The problem that the invention aims to solve

[0010] However, the system described in Patent Document 1 relies on the flexible use of domain knowledge, making it difficult for unskilled users to utilize, and also considered difficult to use when dealing with newly occurring phenomena. The present invention was made to solve these problems, and its object is to provide, for example, a process analysis apparatus, a process analysis method, and a process analysis program that enable even unskilled users to easily determine the factors causing anomalies occurring on a manufacturing line.

[0011] Technical means to solve the problem

[0012] The process analysis apparatus of the present invention includes: a normal data acquisition unit that acquires multiple pieces of first state data related to the normal state of multiple mechanisms constituting a manufacturing line; an abnormal data acquisition unit that acquires multiple pieces of second state data related to the state of the multiple mechanisms when an abnormality occurs; a normal state analysis unit that analyzes the acquired multiple pieces of first state data to determine a connection state between the multiple mechanisms as a first connection state; an abnormal state analysis unit that analyzes the acquired multiple pieces of second state data to determine a connection state between the multiple mechanisms as a second connection state; a normal state relationship determination unit that, based on the first connection state, determines a causal relationship between the multiple mechanisms in a process implemented in the manufacturing line as a normal state causal relationship; an abnormal state relationship determination unit that, based on the second connection state, determines a causal relationship between the multiple mechanisms in a process implemented in the manufacturing line as an abnormal state causal relationship; and a special causal relationship determination unit that compares the normal state causal relationship with the abnormal state causal relationship to determine a causal relationship between the multiple mechanisms associated with the occurrence of an abnormality as a special causal relationship.

[0013] It can be configured as follows: In the process analysis device, the normal time analysis unit and the abnormal time analysis unit determine the connection state between the plurality of mechanisms through the following steps: (1) calculate the characteristic quantity based on the state data of each component; (2) calculate the correlation coefficient or partial correlation coefficient between the characteristic quantities, the special causal relationship determination unit calculates the deviation degree between the characteristic quantity in each mechanism of the normal time causal relationship and the characteristic quantity in each mechanism of the abnormal time causal relationship, selects the mechanism with the deviation degree greater than a specified value as a node, when comparing the normal time causal relationship and the abnormal time causal relationship, when the connection state between the mechanisms meets the specified conditions, selects the connection between the mechanisms as an edge, and determines the special causal relationship through the selected node and the edge.

[0014] The process analysis device can be configured as follows: it may further include: a control program acquisition unit for acquiring a control program used to control the operation of the manufacturing line; a control program analysis unit for determining the sequential relationship of the plurality of mechanisms by analyzing the acquired control program; and a constraint model generation unit for generating a constraint model of the plurality of mechanisms based on the sequential relationship of the plurality of mechanisms, wherein the special causal relationship determination unit determines the special causal relationship based on the constraint model.

[0015] The process analysis device may further include: a mechanism data acquisition unit, which acquires mechanism data related to at least one of the relative positional relationship of the devices contained in the plurality of mechanisms and the order in which the devices participate in the process; and a mechanism data analysis unit, which analyzes the acquired mechanism data, models the process implemented in the manufacturing line, thereby determining a process model representing the sequential relationship of the plurality of mechanisms, and the constraint model generation unit can generate the constraint model based on the sequential relationship of the plurality of mechanisms and the process model.

[0016] It can be configured as follows: In the process analysis device, the control program analysis unit determines the sequential relationship of the plurality of mechanisms based on log data obtained by operating the manufacturing line using the control program.

[0017] It can be configured as follows: In the process analysis device, the control program analysis unit determines the sequential relationship of the plurality of mechanisms through the following steps: (1) constructing an abstract syntax tree according to the control program; (2) extracting variables and conditional branches related to each mechanism from the constructed abstract syntax tree; (3) obtaining log data when the manufacturing line is operating normally using the control program; (4) sorting the variables based on the execution results of the conditional branches with reference to the obtained log data.

[0018] The process analysis device may further include: a first experimental plan data acquisition unit, which acquires first experimental plan data during the operation of the manufacturing line to determine the adjustment amount of the adjustment items of the manufacturing line used to achieve the specified quality; and a control causal relationship model generation unit, which generates a control causal relationship model based on the first experimental plan data and the constraint model, the causal relationship between the adjustment items and the mechanism, and the causal relationship between the plurality of mechanisms.

[0019] The process analysis device may further include a quality adjustment causality determination unit, which determines the quality adjustment causality by comparing the control causality with the special causality, wherein the quality adjustment causality includes causal relationships between multiple mechanisms associated with the occurrence of the abnormality, and causal relationships between the mechanism and the adjustment item.

[0020] The process analysis device may further include: a second experimental plan data acquisition unit, which acquires second experimental plan data for determining the relationship between the adjustment amount of the adjustment item and the quality as determined by the causal relationship of the quality adjustment; and a quality prediction model generation unit, which generates a quality prediction model based on the second experimental plan data, wherein the quality prediction model determines the relationship between the adjustment amount of the adjustment item and the quality in the manufacturing line.

[0021] The process analysis device may further include an adjustment amount calculation unit, which calculates the adjustment amount of the adjustment item to achieve the desired quality based on the quality prediction model.

[0022] In the process analysis device, the status data of each component can represent at least one of torque, speed, acceleration, temperature, current, voltage, air pressure, pressure, flow rate, position, size, area, light intensity, and on / off state.

[0023] The process analysis method of the present invention comprises the following steps performed by a computer: acquiring multiple pieces of first state data related to the normal state of multiple mechanisms constituting a manufacturing line; acquiring multiple pieces of second state data related to the state of the multiple mechanisms when an anomaly occurs; determining a connection state between the multiple mechanisms as a first connection state by analyzing the acquired multiple pieces of first state data; determining a connection state between the multiple mechanisms as a second connection state by analyzing the acquired multiple pieces of second state data; determining a causal relationship between the multiple mechanisms in a process implemented in the manufacturing line as a normal state causal relationship based on the first connection state; determining a causal relationship between the multiple mechanisms in a process implemented in the manufacturing line as an anomaly state causal relationship based on the second connection state; and comparing the normal state causal relationship with the anomaly state causal relationship to determine a causal relationship between the multiple mechanisms associated with the occurrence of an anomaly as a specific causal relationship.

[0024] The process analysis computer-readable storage medium of the present invention enables a computer to perform the following steps: acquiring multiple pieces of first state data related to the normal state of multiple mechanisms constituting a manufacturing line; acquiring multiple pieces of second state data related to the state of the multiple mechanisms when an anomaly occurs; determining a connection state between the multiple mechanisms as a first connection state by analyzing the acquired multiple pieces of first state data; determining a connection state between the multiple mechanisms as a second connection state by analyzing the acquired multiple pieces of second state data; determining a causal relationship between the multiple mechanisms in a process implemented in the manufacturing line as a normal state causal relationship based on the first connection state; determining a causal relationship between the multiple mechanisms in a process implemented in the manufacturing line as an anomaly state causal relationship based on the second connection state; and comparing the normal state causal relationship with the anomaly state causal relationship to determine a causal relationship between the multiple mechanisms associated with the occurrence of an anomaly as a specific causal relationship.

[0025] The effects of the invention

[0026] According to the present invention, even an unskilled person can easily identify the factors causing abnormalities in the manufacturing line. Attached Figure Description

[0027] Figure 1 This is a schematic diagram illustrating an example of an application scenario for the process analysis apparatus according to an embodiment of the present invention.

[0028] Figure 2 This is an example of a block diagram representing the hardware structure of a process analysis device.

[0029] Figure 3 This is an example of a block diagram representing the hardware structure of a PLC.

[0030] Figure 4 This is an example of a block diagram representing the software structure of a process analysis device.

[0031] Figure 5 This is an example of a block diagram representing the functional structure of the constraint model generation module.

[0032] Figure 6 This is an example of a flowchart illustrating the processing flow of the constraint model generation module.

[0033] Figure 7 This is an example of a flowchart representing the analysis and processing of a control program.

[0034] Figure 8 It is an example of an abstract syntax tree constructed by a control program.

[0035] Figure 9A This is an example of a scenario where the initialization of sorting of variables is performed.

[0036] Figure 9B This is an example of the sequential relationships between multiple mechanisms determined by the analysis of the control program.

[0037] Figure 10 This is an example of the process of analyzing institutional data.

[0038] Figure 11 This is an example of a block diagram representing the functional structure of the event causality generation module.

[0039] Figure 12 This is an example of a flowchart representing the processing flow of the event causality generation module.

[0040] Figure 13 This is an example of a flowchart illustrating the processing flow when analyzing state data.

[0041] Figure 14 This is an example of the process of analyzing state data under normal conditions.

[0042] Figure 15 This is an example of a directed graph representing causal relationships between multiple organizations under normal circumstances.

[0043] Figure 16 This is an example of the process of analyzing state data during anomalies.

[0044] Figure 17 This is an example of a directed graph representing causal relationships between multiple organizations under normal circumstances.

[0045] Figure 18 This is an example of a flowchart illustrating the process of generating causal relationships between events.

[0046] Figure 19 This is a table representing examples of the deviation degree of characteristic quantities.

[0047] Figure 20 This is an example of a causal relationship between events.

[0048] Figure 21 This is an example of a block diagram representing the functional structure of the module that controls the generation of causal relationships.

[0049] Figure 22 This is an example of a flowchart representing the processing flow of the module that controls the generation of causal relationships.

[0050] Figure 23 This is an example of a controlled causal model.

[0051] Figure 24 This is an example of a block diagram representing the functional structure of the quality adjustment causal model generation module.

[0052] Figure 25 This is a flowchart illustrating the processing flow of the quality adjustment causal model generation module.

[0053] Figure 26 It is a diagram showing the process of creating a quality adjustment causal model based on the corresponding event causal model and the control causal relationship model.

[0054] Figure 27 It is a diagram showing the process of creating a quality adjustment causal model based on the corresponding event causal model and the control causal relationship model.

[0055] Figure 28 It is a diagram showing the process of creating a quality adjustment causal model based on the corresponding event causal model and the control causal relationship model.

[0056] Figure 29 This is a diagram illustrating other examples of the process of creating a quality-adjusted causal model based on the corresponding event causal model and the control causal relationship model.

[0057] Figure 30 This is a diagram illustrating other examples of the process of creating a quality-adjusted causal model based on the corresponding event causal model and the control causal relationship model.

[0058] Figure 31 This is an example of a block diagram representing the functional structure of the quality prediction model generation module.

[0059] Figure 32 This is an example of a flowchart illustrating the processing flow of the quality prediction model generation module.

[0060] Figure 33 This is a diagram representing other examples of how causal relationships between events are generated.

[0061] [Explanation of Symbols]

[0062] 1: Process Analysis Device

[0063] 11: Control Department

[0064] 221: Control Procedure

[0065] 222: Status Data

[0066] 223: Institutional Data

[0067] 3: Manufacturing line

[0068] 31: Institution Detailed Implementation

[0069] Hereinafter, an embodiment of the process analysis apparatus of the present invention will be described with reference to the accompanying drawings.

[0070] <1. Application Examples>

[0071] First, use Figure 1 An example of a scenario in which the present invention is applicable will be described. Figure 1 An example of a usage scenario for the process analysis device 1 of this embodiment is illustrated schematically.

[0072] like Figure 1As shown, the process analysis apparatus 1 of this embodiment acquires multiple item status data 222 and mechanism data 223 related to the status of the multiple mechanisms 31 constituting the manufacturing line 3. The manufacturing line 3 can be capable of manufacturing any article and may include multiple devices or a single device such as a packaging machine. Furthermore, each mechanism 31 can be capable of performing any processing step of the manufacturing process and may include one or more devices or a part of a device. When one mechanism 31 includes a part of a device, multiple mechanisms 31 may include one device. Moreover, when multiple processes are performed on the same device, they can be considered as different mechanisms 31. For example, when a first process and a second process are performed on the same device, the device performing the first process can be considered as the first mechanism 31, and the device performing the second process can be considered as the second mechanism 31. Furthermore, the status data 222 may include all kinds of data related to the status of each mechanism 31 constituting the manufacturing line 3.

[0073] Furthermore, the process analysis device 1 of this embodiment acquires a control program 221 for controlling the operation of the manufacturing line 3. The control program 221 may include all types of programs that control the operation of each mechanism 31 constituting the manufacturing line 3. The control program 221 may include one program or multiple programs. In addition, in this embodiment, the operation of the manufacturing line 3 is controlled by a programmable logic controller (PLC) 2. Therefore, the process analysis device 1 acquires multiple status data 222, mechanism data 223, and control program 221 from the PLC 2.

[0074] Next, the process analysis device 1 of this embodiment performs statistical analysis on the acquired multi-item state data 222 to determine the strength of the relationship between the multiple mechanisms 31 within the manufacturing line 3. The strength of the relationship is an example of the "connection state" of this invention. Furthermore, the process analysis device 1 of this embodiment determines the sequential relationship of the multiple mechanisms 31 within the manufacturing line 3 by analyzing the acquired control program 221 and mechanism data 223. Furthermore, the process analysis device 1 of this embodiment determines the causal relationship between the multiple mechanisms 31 in the processes executed in the manufacturing line 3 based on the determined strength of the relationship and the sequential relationship.

[0075] As described above, in this embodiment, during the analysis of the causal relationships among multiple mechanisms 31, a control program 221 is used to determine the sequential relationships among the multiple mechanisms 31. The control program 221 specifies the actions of each mechanism 31, and therefore, by utilizing the control program 221, the sequential relationships among the mechanisms 31 can be determined more accurately. Therefore, according to this embodiment, the causal relationships among the multiple mechanisms 31 constituting the manufacturing line 3 can be accurately modeled.

[0076] Furthermore, as explained below, when determining the strength of the relationship between the mechanisms 31 using state data 222, two causal relationships are obtained: one when using state data 222 when the manufacturing line 3 is operating normally and the other when using state data 222 when the manufacturing line 3 is operating abnormally. By comparing these causal relationships, an event causal relationship model is generated that extracts only the mechanisms 31 associated with the occurrence of the abnormality.

[0077] <2. Hardware Structure of Process Analysis Device>

[0078] Next, use Figure 2 An example of the hardware structure of the process analysis device 1 in this embodiment will be described. Figure 2 An example of the hardware structure of the process analysis device 1 of this embodiment is illustrated schematically.

[0079] like Figure 2 As shown, the process analysis device 1 in this embodiment is a computer electrically connected to a control unit 11, a storage unit 12, a communication interface 13, an input device 14, an output device 15, and a driver 16. Furthermore, in Figure 2 In this context, the communication interface is recorded as "Communication I / F".

[0080] The control unit 11 includes a central processing unit (CPU), random access memory (RAM), read-only memory (ROM), etc., which are hardware processors, and controls each component according to information processing. The storage unit 12 is, for example, an auxiliary storage device such as a hard disk drive or a solid-state drive, which stores the process analysis program 121 executed by the control unit 11.

[0081] Process analysis program 121 is used to enable process analysis device 1 to perform the following process (using multi-part status data 222, mechanism data 223, and control program 221) to analyze the causal relationships between multiple mechanisms 31 in the manufacturing process implemented on manufacturing line 3. Figure 5 The procedure will be described in detail later.

[0082] The communication interface 13 is, for example, a wired local area network (LAN) module or a wireless LAN module, used for wired or wireless communication via a network. The process analysis device 1 can perform data communication between PLCs 2 via the network through the communication interface 13. Furthermore, the type of network can be appropriately selected from, for example, the Internet, wireless communication networks, mobile communication networks, telephone networks, private networks, etc.

[0083] Input device 14 is, for example, a mouse, keyboard, or other device used for input. Output device 15 is, for example, a display, speaker, or other device used for output. The operator can operate the process analysis device 1 via input device 14 and output device 15.

[0084] The drive 16, such as a compact disc (CD) drive or a digital versatile disc (DVD) drive, is a device used to read programs stored in the storage medium 91. The type of drive 16 can be appropriately selected according to the type of storage medium 91. The process analysis program 121 can be stored in the storage medium 91.

[0085] Storage medium 91 is a medium that stores recorded program information through electrical, magnetic, optical, mechanical, or chemical means, in a manner that allows devices such as computers or machines to read the information. Process analysis device 1 can obtain the process analysis program 121 from storage medium 91.

[0086] Here, in Figure 2 In this example, storage medium 91 is exemplified by disc-type storage media such as CDs and DVDs. However, the type of storage medium 91 is not limited to disc type, and it may not be a disc type. Examples of storage media other than disc type include semiconductor memories such as flash memory.

[0087] Furthermore, the specific hardware structure of the process analysis device 1 can be appropriately omitted, replaced, or have additional components added depending on the implementation method. For example, the control unit 11 may include multiple processors. The process analysis device 1 may include multiple information processing devices. Moreover, in addition to being designed as an information processing device specifically for the service provided, the process analysis device 1 may also be a general-purpose server device, a personal computer (PC), etc.

[0088] <3. PLC>

[0089] Next, use Figure 3 An example of the hardware structure of PLC2, which controls the operation of manufacturing line 3, will be described. Figure 3 An example of the hardware structure of the PLC2 in this embodiment is illustrated schematically.

[0090] like Figure 3 As shown, the PLC2 is a computer electrically connected to a control unit 21, a storage unit 22, an input / output interface 23, and a communication interface 24. Thus, the PLC2 is configured to control the operation of each mechanism 31 of the manufacturing line 3. Furthermore, in Figure 3In this document, the input / output interface and the communication interface are respectively referred to as "Input / Output I / F" and "Communication I / F".

[0091] The control unit 21 includes a CPU, RAM, ROM, etc., and is configured to perform various information processing based on programs and data. The storage unit 22 includes, for example, RAM, ROM, etc., and stores control program 221, status data 222, mechanism data 223, and log data 224. The control program 221 is a program used to control the operation of the manufacturing line 3. The status data 222 is data related to the status of each mechanism 31. The mechanism data 223 is data related to at least one of the relative positional relationships of the devices included in each of the multiple mechanisms 31 and the order in which the devices participate in the process. In addition, the mechanism data 223 may include, for example, a list of devices, a list of sensors that monitor the operation of the devices, information indicating the order of the process, and information indicating the installation positions of the sensors. Furthermore, the log data 224 is log data indicating the operation of the manufacturing line 3.

[0092] The input / output interface 23 is an interface for connecting to external devices, and its configuration varies depending on the external device being connected. In this embodiment, the PLC2 is connected to the manufacturing line 3 via the input / output interface 23. Furthermore, if different status data can be obtained for a single device, the single device can be considered as multiple mechanisms 31, or it can be considered as a single mechanism 31. Therefore, the number of input / output interfaces 23 can be the same as or different from the number of mechanisms 31 constituting the manufacturing line 3.

[0093] The communication interface 24 is, for example, a wired LAN module or a wireless LAN module, and is used for wired or wireless communication. The PLC2 can communicate data with the process analysis device 1 through the communication interface 24.

[0094] Furthermore, the specific hardware structure of the PLC2 can be appropriately omitted, replaced, or have additional components added depending on the implementation method. For example, the control unit 21 may include multiple processors. The storage unit 22 may include the RAM and ROM contained in the control unit 21. The storage unit 22 may include auxiliary storage devices such as hard disk drives and solid-state drives. Moreover, in addition to being designed as an information processing device dedicated to the services provided, the PLC2 can also be replaced with a general-purpose desktop PC, tablet PC, etc., depending on the object being controlled.

[0095] <4. Software Structure of Process Analysis Device>

[0096] Next, use Figure 4 An example of the software structure of the process analysis device 1 in this embodiment will be described. Figure 4An example of the software structure of the process analysis device 1 of this embodiment is illustrated schematically.

[0097] The control unit 11 of the process analysis device 1 expands the process analysis program 121 stored in the storage unit 12 into RAM. Then, the control unit 11 interprets and executes the process analysis program 121 expanded into RAM via the CPU, thereby controlling each component. Thus, as... Figure 4 As shown, the process analysis apparatus 1 of this embodiment is configured as a computer including a constraint model generation module 101, an event causal relationship model generation module 102, a control causal relationship model generation module 103, a quality adjustment causal relationship model generation module 104, and a quality prediction model generation module 105 as software modules. These will be described in sequence below.

[0098] <4-1. Constraint Model Generation Module>

[0099] <4-1-1. Structure of the Restricted Model Generation Module>

[0100] First, the constraint model generation module 101 will be explained. For example... Figure 5 As shown, the constraint model generation module 101 further includes a first acquisition unit (control program acquisition unit) 111, a second acquisition unit (mechanism data acquisition unit) 112, a first analysis unit (control program analysis unit) 113, a second analysis unit (mechanism data analysis unit) 114, and a first relationship determination unit (constraint model generation unit) 115 as functional blocks.

[0101] The first acquisition unit 111 acquires a control program 221 used to control the operation of the manufacturing line 3. The first analysis unit 113 analyzes the acquired control program 221 to determine the sequential relationship of the multiple mechanisms 31. In this embodiment, the first analysis unit 113 uses log data 224 obtained by executing the control program 221 to determine the sequential relationship of the multiple mechanisms 31. The second acquisition unit 112, as described above, acquires data related to the relative positional relationships of the devices included in each of the multiple mechanisms 31. The second analysis unit 114 analyzes the acquired mechanism data 223 to determine a process model. Furthermore, the first relationship determination unit 115 determines the causal relationship between the multiple mechanisms 31 in the process implemented in the manufacturing line 3 based on the determined sequential relationship and the process model.

[0102] The functional structures 111 to 115 of the process analysis device 1 will be described in detail through the operational examples described below. Furthermore, in this embodiment, an example will be described where each software module of the process analysis device 1 is implemented by a general-purpose CPU. However, some or all of the above software modules may also be implemented by one or more dedicated hardware processors. Moreover, regarding the software structure of the process analysis device 1, software modules may be appropriately omitted, replaced, or added depending on the implementation method. This is also true for each of modules 102 to 105 described below.

[0103] <4-1-2. Restricting the Actions of the Model Generation Module>

[0104] Next, use Figure 6 An example of the action of the constraint model generation module 101 is explained. Figure 6 This example illustrates the processing flow of the constraint model generation module. However, the processing flow described below is only one example, and each process can be modified as much as possible. Moreover, regarding the processing flow described below, steps can be appropriately omitted, replaced, or added depending on the implementation method. The same applies to the operation of modules 102 to 105 described below.

[0105] [Step S101]

[0106] In step S101, the control unit 11 operates as the first acquisition unit 111, acquiring the control program 221 from the PLC2. The control program 221 can be executed by the PLC2, for example, using at least any one of ladder diagram language, function block diagram language, structured text language, instruction list language, sequential function chart language, and C language. After the control program 221 is acquired, the processing of the control unit 11 proceeds to the next step S102.

[0107] [Step S102]

[0108] In step S102, the control unit 11 operates as the second acquisition unit 112, acquiring the mechanism data 223 from the PLC2. After the acquisition of the mechanism data 223 is completed, the processing of the control unit 11 proceeds to the next step S103. Furthermore, this step S102 can be executed simultaneously with step S101, or it can be executed before step S101.

[0109] [Step S103]

[0110] In step S103, the control unit 11 operates as the first analysis unit 113, analyzing the control program 221 acquired in step S101 to determine the sequential relationship of the multiple mechanisms 31 within the manufacturing line 3. In this embodiment, the control unit 11 determines the sequential relationship of the multiple mechanisms 31 based on log data 224 obtained by operating the manufacturing line 3 using the control program 221. After the analysis and processing of the control program 221 is completed, the processing of the control unit 11 proceeds to the next step S104. Furthermore, this step S103 can be executed at any time as long as it follows step S101. For example, if step S102 is executed before step S101, this step S103 can be executed before step S102.

[0111] <Analysis and Processing of Control Programs>

[0112] Here, further use Figure 7 The processing of the analysis control program 221 is described in detail. Figure 7 An example of the processing flow of the analysis control program 221 is given. In the following description, for ease of explanation, the manufacturing line 3 includes four mechanisms F1 to F4 as multiple mechanisms 31, and in step S101, it is assumed that the control unit 11 acquires the control program 221 using variables v1 to v4 corresponding to the four mechanisms F1 to F4.

[0113] (Step S1401)

[0114] First, in step S1401, the control unit 11 performs syntax analysis on the acquired control program 221, and constructs an abstract syntax tree (AST). The AST can be constructed using known syntax analysis methods that utilize top-down or bottom-up parsing. For example, the AST can be constructed using a parser that processes strings according to a specific formal grammar. After the AST is constructed, the control unit 11 proceeds to the next step, S1402.

[0115] Figure 8 Example: Abstract syntax tree 2211 obtained according to the syntax "if(a>0)[v1=a;]else[v2=-a;]" within control program 221. Figure 8 As illustrated, an abstract syntax tree is a data structure that expresses the structure of a program in a tree structure to explain the meaning of the program.

[0116] Specifically, the control unit 11 omits parentheses and other words in the terms (clauses) used by the program that are not necessary for interpreting the meaning of the program, and extracts words that are relevant to interpreting the meaning of the program. Then, the control unit 11 maps operators such as conditional branches to nodes and operands such as variables to leaves. By performing syntax analysis of the control program 221 in this manner, the control unit 11 is able to construct... Figure 8 The illustrated abstract syntax tree. Furthermore, the abstract syntax tree constructed in this manner represents variables, operators, and relationships between nodes (such as the relationship between operators and operands). The structure of the abstract syntax tree can be appropriately modified, altered, or omitted within the scope of its represented content remaining unchanged.

[0117] (Step S1402)

[0118] In step S1402, the control unit 11 extracts variables (v1 to v4) related to each mechanism 31 and operators including conditional branches and substitution operations from the abstract syntax tree constructed in step S1401. For example, in obtaining Figure 8 In the case of the abstract syntax tree 2211 shown, the control unit 11 extracts ("if", "v1", "=", "a", "v2", "=", and "-a") from the abstract syntax tree 2211. After the extraction is completed, the processing of the control unit 11 proceeds to the next step S1403.

[0119] (Steps S1403 and S1404)

[0120] In step S1403, the control unit 11 sequentially lists the variables and operators extracted in step S1402, limiting the object of monitoring the execution order to the parts associated with the variables (v1 to v4) of each mechanism 31 that become the object of determining the causal relationship. In the example above, the control unit 11 further extracts ("if", "v1", "=", "a", "v2", "=", and "-a") from ("if", "v1", "=", and "v2").

[0121] Then, in step S1404, a conditional branch is randomly selected, and the execution of control program 221 is tested, thereby initializing the order of the defined variables. Specifically, during the test execution of control program 221, control unit 11 monitors the order in which the defined variables are used, thereby enabling the initialization of the order of the variables.

[0122] Figure 9A An example of the initialization described is given. Figure 9A The following (1) to (5) sequence relationships are shown in Figure 2212 as the result of initialization.

[0123] (1) Among the variables “v1” to “v4”, we first use variable “v1”.

[0124] (2) After variable “v1”, use variable “v2” with a probability of 0.5 and variable “v3” with a probability of 0.5.

[0125] (3) After variable “v2”, use variable “v3” with a probability of 1.

[0126] (4) After variable “v3”, use variable “v4” with a probability of 1.

[0127] (5) Among the variables “v1” to “v4”, the variable “v4” is used last.

[0128] After this initialization is completed, the processing of the control unit 11 proceeds to the next step S1405.

[0129] (Step S1405)

[0130] In step S1405, the control unit 11 acquires log data 224 from the PLC2, representing the execution results of the manufacturing line 3 operating normally using the control program 221. For example, the PLC2 creates log data 224 representing the execution results of the control program 221 simultaneously when collecting the status data 222. At this time, the log data 224 records timestamps, values ​​of the variables used, etc. Furthermore, for example, a debug mode can be set in the control program 221 to collect information such as the frequency of execution of each line of code, the lines of code executed, and the computation time consumed by each segment of the code. At this time, the PLC2 can create log data 224 recording this information by executing the control program 221 in debug mode. The control unit 11 can acquire the log data 224 created in this manner from the PLC2. After acquiring the log data 224, the processing of the control unit 11 proceeds to the next step S1406.

[0131] (Step S1406)

[0132] In step S1406, the control unit 11 refers to the log data 224 obtained in step S1405 and sorts the variables (v1 to v4) based on the execution result of the conditional branch, thereby determining the sequential relationship of each mechanism 31 in the manufacturing line 3.

[0133] Figure 9B An example of the results of the sorting is shown below. Figure 9B Chart 2213 shows the following order relationships (A) to (E) as the result of sorting using log data 224.

[0134] (A) Among the variables “v1” to “v4”, we first use variable “v1”.

[0135] (B) After variable “v1”, use variable “v2” with a probability of 0.01 and variable “v3” with a probability of 0.99.

[0136] (C) After variable “v2”, use variable “v3” with a probability of 1.

[0137] (D) After variable “v3”, use variable “v4” with a probability of 1.

[0138] (E) Among the variables “v1” to “v4”, the variable “v4” is used last.

[0139] Based on the above, after the sequential relationship of the multiple mechanisms 31 is determined, the analysis and processing of the control program 221 of this embodiment is completed, and the processing of the control unit 11 proceeds to the next step S104.

[0140] [Step S104]

[0141] In step S104, the control unit 11 operates as the second analysis unit 114, for example, as Figure 10 As shown, the mechanism data 223 obtained in step S102 is analyzed to determine the process model 252. For example, the control unit 11 refers to the mechanism data 223 to determine the sequential relationship between two mechanisms 31, and inputs the determined result into the corresponding cell of table 251 representing the relationship between the two mechanisms 31. By completing table 251, the control unit 11 can determine the process model 252 as a directed graph. After the analysis and processing of the control program 221 is completed, the processing of the control unit 11 proceeds to the next step S105. In addition, this step S104 can be executed at any time as long as it is after step S102. For example, if step S102 is executed before step S101, this step S104 can be executed before step S101.

[0142] [Step S105]

[0143] return Figure 6 In the next step S105, the control unit 11 operates as the first relationship determination unit 115. That is, the control unit 11 determines the causal relationship between the multiple mechanisms 31 in the process implemented in the manufacturing line 3 based on the sequential relationship between the multiple mechanisms 31 in the manufacturing line 3 determined by steps S103 and S104 respectively and the process model 252.

[0144] In this embodiment, the control unit 11 determines the causal relationships between multiple mechanisms 31 in the processes implemented in the manufacturing line 3 based on the sequential relationships of each mechanism 31 determined in step S1406 and the process model 252 determined in step S104. For example, the presence or absence of a relationship between each mechanism 31 can be determined by the product of the sequential relationships of each mechanism 31 and the process model 252. Then, based on the presence or absence of the relationship, the nodes are hierarchically sorted and arranged to generate a constraint model.

[0145] [Step S106]

[0146] In step S106, the control unit 11 outputs the constraint model created in step S105. For example, the control unit 11 outputs the created constraint model in image form to an output device 15 such as a display. After this, the control unit 11 concludes the processing of this operation example.

[0147] <4-2. Event Causal Relationship Model Generation Module>

[0148] <4-2-1. Structure of the Event Causal Relationship Model Generation Module>

[0149] Next, the event causal relationship model generation module 102 will be described. For example... Figure 11 As shown, the event causality generation module further includes a third acquisition unit 116, a fourth acquisition unit 117, a third analysis unit 118, a fourth analysis unit 119, a second relationship determination unit 1110, a third relationship determination unit 1111, and a fourth relationship determination unit 1112 as functional blocks.

[0150] The third acquisition unit 116 acquires multiple pieces of status data 222 related to the states of the multiple mechanisms 31 constituting the manufacturing line 3. At this time, the acquired status data 222 is the data when the manufacturing line 3 is operating normally. The fourth acquisition unit 117, like the third acquisition unit 116, also acquires multiple pieces of status data 222 related to the states of the multiple mechanisms 31, but in the fourth acquisition unit 117, it acquires status data when the manufacturing line 3 experiences an anomaly. Both the third analysis unit 118 and the fourth analysis unit 119 perform statistical analysis on the acquired multiple pieces of status data 222 to determine the strength of the relationships between the multiple mechanisms 31.

[0151] The second relationship determination unit 1110 determines the normal causal relationships between multiple mechanisms 31 in the processes implemented in the manufacturing line 3 based on the analysis results in the third analysis unit 118 and the constraint model. The third relationship determination unit 1111 determines the abnormal causal relationships between multiple mechanisms 31 in the processes implemented in the manufacturing line 3 based on the analysis results in the fourth analysis unit 119 and the constraint model. The fourth relationship determination unit 1112 generates a specific specialized causal relationship, i.e., an event causal model, for the factors of the generated abnormality (event), based on the normal and abnormal causal relationships generated in the second and third relationship determination units 1110 and 1111.

[0152] <4-2-2. Action Examples of the Event Causality Model Generation Module>

[0153] Next, use Figure 12 The action examples of the event causal relationship model generation module 102 are explained. Figure 12 This is an example of the processing flow of the event causality generation module.

[0154] [Step S201]

[0155] First, in step S201, the control unit 11 operates as the third acquisition unit 116, acquiring multiple status data 222 related to the status of each mechanism 31 from the PLC 2. Each mechanism 31 may include, for example, a conveyor, a robotic arm, a servo motor, a barrel (forming machine, etc.), an adsorption pad, a cutting device, a sealing device, or a part thereof. Moreover, each mechanism 31 may be, for example, a composite device such as a printing machine, an mounting machine, a reflow oven, or a substrate inspection device. Furthermore, in addition to any of the devices accompanied by physical movement as described above, each mechanism 31 may also include, for example, a device for detecting certain information through various sensors, a device for acquiring data from various sensors, a device for detecting certain information from the acquired data, and a device for information processing of the acquired data, etc., which perform internal processing. As a specific example, in a manufacturing line that includes an optical sensor that detects a mark assigned to an object flowing on a conveyor, the optical sensor and the device utilizing the information detected by the optical sensor can be regarded as each mechanism 31. Furthermore, the status data 222 can be, for example, data representing at least one of torque, speed, acceleration, temperature, current, voltage, air pressure, pressure, flow rate, position, size (height, length, width), area, light intensity, and on / off state. This status data 222 can be obtained using known measuring devices such as sensors and cameras. For example, flow rate can be obtained using a float sensor. Moreover, position, size, and area can be obtained using an image sensor.

[0156] Additionally, the status data 222 may include data obtained from one or more measuring devices. Furthermore, the status data 222 may be data directly obtained from the measuring devices, or it may be data that can be obtained by applying a certain processing to the data obtained from the measuring devices, such as position data obtained from image data. Each piece of status data 222 is obtained corresponding to each mechanism 31.

[0157] Each measuring device is appropriately configured on each mechanism 31 of the manufacturing line 3 in a manner capable of monitoring. The PLC 2 operates the manufacturing line 3 and collects status data 222 of each part from each measuring device. The control unit 11 acquires status data 222 related to the status of each mechanism 31 during normal operation of the manufacturing line 3 from the PLC 2. Thus, the control unit 11 can acquire status data 222 of multiple parts. After acquiring the status data 222 of multiple parts under normal conditions, the processing of the control unit 11 proceeds to the next step S202.

[0158] [Step S202]

[0159] In step S202, the control unit 11 operates as the fourth acquisition unit 117, acquiring multiple status data 222 related to the status of each mechanism 31 from the PLC2. The difference from the operation of the third acquisition unit 116 in step S201 is that it acquires status data 222 related to the status of each mechanism 31 when the manufacturing line 3 is operating abnormally. After acquiring the status data 222 of multiple abnormal states, the control unit 11 proceeds to the next step, S203.

[0160] [Step S203]

[0161] In step S203, the control unit 11 operates as the third analysis unit 118, performing statistical analysis on the multiple status data 222 acquired in step S201 to determine the strength of the relationships between the multiple mechanisms 31 within the manufacturing line 3. After the analysis and processing of the multiple status data 222 is completed, the control unit 11 proceeds to the next step, S204. Furthermore, this step S203 can be executed at any time as long as it follows step S201. For example, this step S203 can be executed before step S202.

[0162] <Analysis and Processing of Status Data>

[0163] Here, further use Figure 13 as well as Figure 14 The processing of multiple state data 222 in statistical analysis is explained in detail. Figure 13 This is an example of a process for analyzing and processing state data 222 in a normal state. Figure 14 Schematic example of passing Figure 13The process shown is an example of analyzing the status data 222. In the following description, for ease of explanation, the manufacturing line 3 includes four mechanisms (e.g., four servo motors) F1 to F4 as multiple mechanisms 31. In step S101, it is assumed that the control unit 11 acquires the status data 222 of each mechanism F1 to F4.

[0164] (Step S2301)

[0165] First, in step S2301, the control unit 11 calculates the feature quantity 2221 based on the status data 222 of each component acquired in step S201. The type of feature quantity 2221 is not particularly limited and can be appropriately selected according to the implementation method. Furthermore, the method for calculating the feature quantity 2221 can be appropriately determined according to the implementation method.

[0166] As a specific example, in this embodiment, the control unit 11 calculates the feature quantity 2221 based on the state data 222 using the following method. First, in order to define the processing range for calculating the feature quantity 2221, the control unit 11 divides the acquired state data 222 frame by frame. The length of each frame can be appropriately set according to the embodiment.

[0167] The control unit 11 can, for example, divide the status data 222 of each component into frames of fixed duration. However, the manufacturing line 3 may not operate at fixed time intervals. Therefore, if the status data 222 of each component is divided into frames of fixed duration, there is a possibility that the actions of each mechanism 31 reflected in each frame may be deviated.

[0168] Therefore, the control unit 11 can divide the status data 222 into frames every cycle time. The cycle time is the time taken for the manufacturing line 3 to produce a specified number of products. The cycle time can be determined based on signals that control the manufacturing line 3, such as control signals from the PLC 2 that control the actions of each mechanism 31 of the manufacturing line 3.

[0169] use Figure 15 The relationship between control signals and cycle time is explained. Figure 15 This illustrates the relationship between control signals and cycle time. For example... Figure 15 As shown, the control signal for manufacturing line 3, which produces repetitive products, becomes a pulse signal that periodically indicates "on" and "off" based on the production of a specified number of products.

[0170] Therefore, the control unit 11 can obtain the control signal from the PLC2 and set the time from the rise ("on") of the obtained control signal to the next rise ("on") as the cycle time. Furthermore, as... Figure 15 As illustrated, the control unit 11 divides the status data 222 into frames every clock cycle.

[0171] Next, the control unit 11 calculates the value of the feature quantity 2221 based on each frame of the state data 222. When the state data 222 is continuous data such as measurement data, the control unit 11 can calculate, for example, the amplitude, maximum value, minimum value, average value, dispersion value, standard deviation, and instantaneous value (one-point sample) within the frame as the feature quantity 2221. Furthermore, when the state data 222 is discrete data such as detection data, the control unit 11 can calculate, for example, the "on" time, "off" time, load ratio, number of "on" times, and number of "off" times within each frame as the feature quantity 2221. After the calculation of each feature quantity 2221 is completed, the processing of the control unit 11 proceeds to the next step, S2302.

[0172] (Step S2302)

[0173] return Figure 13 as well as Figure 14 Continuing with the explanation, in step S2302, the control unit 11 calculates the correlation coefficient or partial correlation coefficient between each characteristic quantity 2221. The correlation coefficient can be calculated using the formula in formula 1 below. Furthermore, the partial correlation coefficient can be calculated using the formula in formula 2 below.

[0174] [Number 1]

[0175]

[0176] In addition, r ij Let x represent the element in the i-th row and j-th column of matrix 2222. i and x j This corresponds to the data representing the characteristic quantity 2221 calculated based on the state data 222 of each component. X i and X j They represent x respectively i and x j The average value of the specimen. n represents the number of each feature used in the relevant calculations.

[0177] [Number 2]

[0178]

[0179] Additionally, matrix R(r) ij The inverse matrix representation of ) is R -1 (r ij ), r ij This represents the element in the i-th row and j-th column of the inverse matrix of matrix 2222.

[0180] Therefore, the control unit 11 can obtain a matrix 2222 with correlation coefficients or partial correlation coefficients as each element. The correlation coefficients and partial correlation coefficients between each feature quantity 2221 represent the strength of the relationship between the corresponding mechanisms 31. That is, the strength of the relationship between the corresponding mechanisms 31 is determined based on each element of the matrix 2222. After the correlation coefficients or partial correlation coefficients between each feature quantity 2221 are calculated, the processing of the control unit 11 proceeds to the next step S2303.

[0181] (Step S2303)

[0182] In step S2303, the control unit 11 constructs undirected graph information 2223 representing the strength of the relationship between corresponding mechanisms 31 based on the correlation coefficient or partial correlation coefficient between each feature quantity 2221.

[0183] For example, the control unit 11 creates nodes corresponding to each mechanism 31. Furthermore, if the correlation coefficient or partial correlation coefficient calculated between two mechanisms 31 is above a threshold, the control unit 11 connects the corresponding two nodes with an edge. Conversely, if the correlation coefficient or partial correlation coefficient calculated between two mechanisms 31 is below the threshold, the control unit 11 does not connect the corresponding two nodes with an edge. The threshold can be a fixed value specified in the process analysis program 121, or it can be a set value that can be changed by the operator. Moreover, the thickness of the edge can be determined according to the magnitude of the corresponding correlation coefficient or partial correlation coefficient.

[0184] Therefore, it is possible to produce such Figure 14 The example undirected graph information is 2223. In Figure 14 In the example, four nodes corresponding to the four mechanisms F1 to F4 were created. Furthermore, edges were formed between the nodes of mechanisms F1 and F2, between the nodes of mechanisms F1 and F3, between the nodes of mechanisms F2 and F3, and between the nodes of mechanisms F3 and F4. Moreover, the correlation between mechanisms F1 and F3 and between mechanisms F3 and F4 is greater than the correlation between other mechanisms, so the edges between the nodes of mechanisms F1 and F3 and between the nodes of mechanisms F3 and F4 are formed in a manner that is thicker than other edges. After the construction of the undirected graph information 2223 representing the strength of the relationships between the multiple mechanisms 31 is completed, the analysis and processing of the state data 222 of this embodiment is completed, and the processing of the control unit 11 proceeds to the next step S204.

[0185] In addition, Figure 14In the example, the undirected graph information 2223 is expressed as an image representing the undirected graph. However, the output form of the undirected graph information 2223 is not limited to an image; it can also be expressed through text or the like. Furthermore, in the example described, edges are not formed between nodes (mechanisms 31) with weak relationships by comparing the correlation coefficient or partial correlation coefficient with a threshold. However, the method for removing edges between nodes with weak relationships is not limited to this example. For instance, after forming a graph that connects all nodes with edges, the control unit 11 can sequentially delete the edges of the formed graph, starting from the edges with small correlation coefficients or partial correlation coefficients, in a manner where the fitting index (Goodness of Fit Index (GFI), standardized root mean square residual (SRMR), etc.) representing the degree of deviation does not exceed a threshold.

[0186] [Step S204]

[0187] return Figure 12 In step S204, the control unit 11 operates as the second relationship determination unit 1110. That is, the control unit 11 determines the normal causal relationship between the multiple mechanisms 31 in the process implemented in the manufacturing line 3 based on the constraint model and the strength of the relationship between the multiple mechanisms 31 in the manufacturing line 3 determined by step S2303.

[0188] In this embodiment, the control unit 11 creates directed graph information representing the causal relationships between the mechanisms 31 by applying the sequence relationships of each mechanism 31 determined by the constraint model to the undirected graph information 2223 constructed through step S2303. At this time, the control unit 11 can determine sequences (transitions) in the sequence relationships of each mechanism 31 whose probability of occurrence is below a threshold, and delete edges corresponding to the determined sequences from the edges constituting the undirected graph information 2223. Furthermore, the threshold can be a fixed value specified in the process analysis program 121, or a set value that can be changed by an operator or the like.

[0189] Figure 15 Here is an example of a directed graph of information that has been created. Figure 15 The directed graph information 122 represents the causal relationships (a) to (d) below.

[0190] (a) Among mechanisms “F1” to “F4”, first use mechanism “F1” or mechanism “F2”.

[0191] (b) After mechanism “F1” and mechanism “F2”, use mechanism “F3”.

[0192] (c) After mechanism “F3”, use mechanism “F4”.

[0193] (d) Among mechanisms “F1” to “F4”, mechanism “F4” is used last.

[0194] In addition, Figure 15 In the example, the probability of the sequence from variable "v1" to variable "v2" occurring is low, so the edges between mechanism "F1" and mechanism "F2" in the undirected graph information 2223 are deleted. Furthermore, the thickness of each edge in the directed graph information 122 is set to correspond to the thickness of each edge in the undirected graph information 2223. After the directed graph information is created, the control unit 11 proceeds to the next step, S206.

[0195] In addition, Figure 15 In the example, the directed graph information is expressed as an image (directed graph information 122). However, the output format of the directed graph information is not limited to an image; it can also be expressed through text, etc. Furthermore, in the example described, the strength of the relationship between nodes (mechanism 31) is expressed by the thickness of the edges. However, the method for expressing the strength of the relationship between nodes is not limited to this example. The strength of the relationship between nodes can also be expressed by labeling the vicinity of each edge with numbers.

[0196] [Step S205]

[0197] In step S205, the control unit 11 operates as the fourth analysis unit 119, and performs statistical analysis on the state data 222 obtained in step S202 during multiple abnormal events to determine the strength of the relationships between the multiple mechanisms 31 within the manufacturing line 3. Except for the state data 222 used, which differs from that of the third analysis unit 118, the processing in step S205 is the same as in step S203, therefore a detailed description is omitted. In the fourth analysis unit 119, since the state data during abnormal events is used, for example, a... Figure 16 The undirected graph information shown is 2226.

[0198] [Step S206]

[0199] In step S206, the control unit 11 operates as the third relationship determination unit 1111. That is, the control unit 11 determines the causal relationship between the multiple mechanisms 31 in the process implemented in the manufacturing line 3 when anomalies occur, based on the constraint model and the strength of the relationship between the multiple mechanisms 31 in the manufacturing line 3 as determined in step S205. The processing flow is the same as the method shown in step S204, so detailed description is omitted. For example, generating as... Figure 17 The directed graph information shown is 123.

[0200] [Step S207]

[0201] In step S207, the control unit 11 operates as the fourth relationship determination unit 1112. That is, the control unit 11 determines the relationship based on... Figure 15 The directed graph information 122 shown in normal condition is... Figure 17 The directed graph information 123 shown in the figure generates an event causal model when an anomaly occurs.

[0202] (Generation and processing of event causal models)

[0203] Here, further use Figures 18-20 The method for generating event causal models is explained in detail. Figure 18 This is a diagram illustrating an example of the generation process of an event causal model. Figure 19 as well as Figure 20 This is an example of an event causal model.

[0204] (Step S2701)

[0205] First, in step S2701, the control unit 11 calculates the degree of deviation between the characteristic quantity acquired in step S203 under normal conditions and the characteristic quantity acquired in step S205 under abnormal conditions. The degree of deviation represents the extent of difference between the corresponding characteristic quantities under normal and abnormal conditions. The degree of deviation can be calculated, for example, using the difference in averages, the difference in medians, or the KL distance. For example, calculating... Figure 19 The deviation shown.

[0206] (Step S2702)

[0207] In step S2702, the control unit 11 selects an edge by comparing the directed graph information 122 for normal conditions obtained in step S204 with the directed graph information 123 for abnormal conditions obtained in step S206. For example, it can select an edge that has a large change in intensity between two graph information 122 or graph information 123, or an edge that exists in one of them but not in the other.

[0208] (Step S2703)

[0209] In step S2703, the control unit 11 selects nodes. Here, nodes whose deviation calculated in step S2701 is greater than a predetermined threshold are selected. Thus, a node is generated from the selected edges and nodes, for example... Figure 19 The event causal model shown illustrates the occurrence of an anomaly. According to this model, the mechanisms that cause the anomaly are F1, F3, and F4. Alternatively, step S2703 can be performed before step S2702.

[0210] (Step S2704)

[0211] In step S2704, the control unit 11 sorts the candidate anomalies based on the event causal model generated in S2703. An example of the factor order is shown below. Figure 19 Therefore, in Figure 19 In the event causal model shown, F1, F3, and F4 are maintained sequentially according to the orientation of the edges, thereby eliminating anomalies. Thus, the order of anomaly factors can be calculated based on the orientation of the nodes and edges appearing in the event causal model. However, in the event causal model, when there are nodes not connected by edges, the factor order can be calculated based on the order of the constraint model, or based on the deviation degree and the order of the constraint model. For example, it is possible to generate... Figure 20 The event causal model shown.

[0212] <4-3. Control Causal Relationship Model Generation Module>

[0213] <4-3-1. Structure of the Control Causal Relationship Model Generation Module>

[0214] Next, the controlled causal relationship model generation module 103 will be described. The controlled causal relationship model generation module 103 creates a new causal relationship by applying the experimental plan data from the conditions used to determine the start-up of manufacturing line 3 to the constraint model. Here, this new causal relationship is referred to as the controlled causal relationship model. Figure 21 As shown, the control causal relationship model generation module 103 also includes a fifth acquisition unit 1113, a fifth analysis unit 124, and a fifth relationship determination unit 125 as functional blocks.

[0215] The fifth acquisition unit 1113 acquires the first experimental plan data 225 for starting the manufacturing line 3. The first experimental plan data 225 is data related to the experimental plan used to determine the conditions for operating the manufacturing line 3, such as data used to start the manufacturing line 3 while changing the adjustment amounts of various adjustment items to achieve the target quality. This first experimental plan data is stored, for example, in a database within the process analysis device 1 or external to it. The first experimental plan data 225 includes, for example, status data, quality data, and adjustment item data for each mechanism 31 when operating the manufacturing line 3, including multiple results obtained from changing the adjustment item data. Here, quality data includes, for example, image inspection results and characteristic inspection results such as resistance values. Adjustment item data includes, for example, set speed, set temperature, and arm angle. Furthermore, the first experimental plan data 225 mentioned here may not necessarily be acquired from the experimental plan used to operate the manufacturing line 3; data from manufacturing lines operating under multiple conditions in the past may also be used.

[0216] The fifth analysis unit 124 determines the strength of the relationships between the multiple mechanisms 31 by performing statistical analysis on the acquired first experimental plan data 225. Based on the analysis results in the fifth analysis unit 124 and the aforementioned constraint model, the fifth relationship determination unit 125 determines the causal relationships between the multiple mechanisms 31 in the processes implemented in the manufacturing line 3.

[0217] <4-3-2. Example of actions for controlling the causal relationship generation module>

[0218] Next, use Figure 22 The action examples for the module that controls the generation of causal relationships are explained. Figure 22 This example illustrates the processing flow of the module that controls the causal model generation.

[0219] [Step S301]

[0220] First, in step S301, the control unit 11 operates as the fifth acquisition unit 1113, acquiring feature quantities of each state from the first experimental plan data 225. After the feature quantities are acquired, the processing of the control unit 11 proceeds to the next step S302.

[0221] [Step S302]

[0222] In steps S302 and S303, the control unit 11 operates as the fifth analysis unit 124, performing statistical analysis on the feature quantities acquired in step S301 to determine the strength of the relationships between the multiple mechanisms 31 within the manufacturing line 3. Specifically, it determines the presence or absence of directed edges and generates an undirected graph. When determining the presence or absence of directed edges, it derives partial correlation coefficients and performs thresholding to determine the presence or absence of conditional independence. The analysis performed here can be conducted using the same method as the analysis and processing of the state data 222.

[0223] [Step S303]

[0224] In step S304, the control unit 11 operates as the fifth relationship determination unit 125. That is, the control unit 11 determines the causal relationships between the multiple mechanisms 31 in the process implemented in the manufacturing line 3 based on the constraint model and the strength of the relationships between the multiple mechanisms 31 in the manufacturing line 3 as determined in step S302.

[0225] In this embodiment, the control unit 11 applies the sequential relationship of each mechanism 31 determined in the constraint model to the undirected graph information constructed in step S302, thereby creating, for example, a... Figure 23 The diagram shown is a directed graph 2227 representing the causal relationships between the various organizations 31. The directed graph 2227 includes adjustment item data A and quality data Q. The method for creating the directed graph 2227 is as described above.

[0226] [Step S304]

[0227] Finally, as needed, directed edges are added based on orientation rules. Regarding orientation rules, for example, when a local directed graph is provided, edges are added according to the following three rules.

[0228] (1) In a local directed graph, among three nodes X, Y and Z, if there is an edge from X to Y, Y and Z are adjacent by an undirected edge, and there is no edge between X and Z, add an edge from Y to Z.

[0229] (2) In a local directed graph, among three nodes X, Y and Z, when there are edges from X to Y and from Y to Z, and X and Z are adjacent by undirected edges, add an edge from X to Z.

[0230] (3) In a local directed graph, among the four nodes X, Y, Z and W, when there are edges from Y and W toward Z respectively, and X is adjacent to Y, Z and W with undirected edges, add an edge from X toward Z.

[0231] In this way, a control causal relationship model is generated.

[0232] <4-4. Quality Adjustment Causal Relationship Model Generation Module>

[0233] <4-4-1. Structure of the Quality Adjustment Causal Relationship Model Generation Module>

[0234] Next, the quality adjustment causal relationship model generation module 104 will be described. The quality adjustment causal relationship model is created based on the control causal relationship model and the event causal relationship model, and has the function of narrowing down the scope of adjustment items when an anomaly occurs. For example... Figure 24 As shown, the quality adjustment causal relationship model generation module 104 also includes a sixth acquisition unit 126 and a sixth relationship determination unit 127 as functional blocks.

[0235] <4-4-2. Action Examples of the Quality Adjustment Causal Relationship Generation Module>

[0236] Next, use Figure 25 This section explains the action examples of the quality adjustment causal relationship generation module. Figure 25 This is an example of the processing flow of the quality causal model generation module.

[0237] [Step S401]

[0238] In step S401, the control unit 11 operates as the sixth acquisition unit 126, acquiring the control causal relationship model and the event causal relationship model from the storage unit 12. After acquiring these models, the processing of the control unit 11 proceeds to the next step S402.

[0239] [Step S402]

[0240] In step S402, the control unit 11 operates as the sixth relationship determination unit 127, generating a quality adjustment causal relationship model based on the control causal relationship model and the event causal relationship model. (Refer to...) Figures 26-28 The aforementioned aspects will be explained. Figure 26 This illustrates the control causality model and the event causality model. Furthermore, among the nodes in the control causality model, nodes common to the event causality model are selected. For example, in... Figure 26 In the example, mechanisms F1, F3, and F4 are common nodes. That is, mechanisms F1, F3, and F4 are the nodes where anomalies occur. Here, as an example, let's assume mechanism F4 is a sensor used to monitor quality, and its characteristic quantity is related to quality Q1.

[0241] Next, as Figure 27 As shown, tracing back the causal relationship and selecting adjustment item data that affects the state data of the event causal relationship model. For example, in Figure 27 In the control causal relationship model, the adjustment item connected to the upstream side of mechanism F1 in the edge connected to mechanism F1 is A1. Similarly, the adjustment item connected to the upstream side of mechanism F2 in the edge connected to mechanism F2 is A2. Next, as... Figure 28 As shown, extract Figure 26 as well as Figure 27 The selected nodes and edges are shown. This explicitly demonstrates the adjustment items for quality anomalies. Specifically, it explicitly shows how adjusting adjustment items A1 and A2 affects the characteristic quantity of the quality-related mechanism F4, thereby improving quality.

[0242] As other examples, such as in generating... Figure 29 When generating event causality models and control causality models, the following is generated: Figure 30 The quality adjustment causal relationship model.

[0243] <4-5. Quality Prediction Model Generation Module>

[0244] <4-5-1. Structure of the Quality Prediction Model Generation Module>

[0245] Next, the quality prediction model generation module will be described. The quality prediction model generation module 105 is a module that generates a quality prediction model based on the adjustment items determined through the quality adjustment causal relationship model. For example... Figure 31 As shown, the quality prediction model generation module 105 also includes a second experimental plan generation unit 128, a second experimental data acquisition unit 129, and a quality prediction model generation unit 130 as functional blocks.

[0246] <4-5-2. Action Examples of the Quality Prediction Model Module>

[0247] Next, use Figure 32 The action examples of the quality prediction model module are explained. Figure 32 This is an example of the processing flow of the quality prediction model generation module.

[0248] [Step S501]

[0249] In step S501, the control unit 11 operates as the experiment plan generation unit 128. Here, firstly, adjustment items are extracted from the quality adjustment causal relationship model generated in the quality adjustment causal relationship model generation module 104. Figure 28 In the example, it is determined that adjustment items A1 and A2 affect the occurrence of anomalies. Therefore, while changing the adjustment amounts of these adjustment items A1 and A2, an experimental plan is generated to determine the conditions for the normal operation of manufacturing line 3, i.e., the manufacture of products of specified quality.

[0250] [Step S502]

[0251] In step S502, the control unit 11 operates as the experimental data acquisition unit 129. Here, based on the experimental plan generated in step S501, the manufacturing line 3 is operated to acquire data on the relationship between the adjustment amount of the adjustment items and the quality of the product. At this time, firstly, the adjustment items are extracted from the quality adjustment causal relationship model generated in the quality adjustment causal relationship model generation module 104. Figure 28 In the example, it was determined that adjustments A1 and A2 affected the occurrence of quality-related anomalies. Therefore, while adjusting the amounts of these adjustments A1 and A2, an experimental plan was generated to determine the conditions for normal operation in manufacturing line 3. That is, the changes in the characteristic quantities of mechanism F4 when adjustments A3 and A4 were made were acquired and stored as second experimental data.

[0252] [Step S503]

[0253] In step S503, the control unit 11 operates as the quality prediction model generation unit 130. Here, a quality prediction model is generated based on the second experimental data generated in step S502. The quality prediction model is modeled by the relationship between the adjustment amount of the adjustment item and quality, for example, using a multiple regression model or support vector regression.

[0254] [Step S504]

[0255] After generating the quality prediction model, the optimal adjustment amount that yields the best quality in the selected adjustment items can be derived using the model. This derivation can be achieved using, for example, Newton's method or the Nelder-Mead method.

[0256] <5. Characteristics>

[0257] As described above, the following effects can be obtained according to this embodiment.

[0258] (1) In this embodiment, the generation of the event causal relationship model uses both a causal relationship model generated based on data from normal times and a causal relationship model generated based on data from abnormal times. The reason for this is as follows. For example, a causal relationship model generated using data from abnormal times accurately includes not only the causal relationships during abnormal times but also a portion of the causal relationships during normal times. Therefore, it is difficult to accurately determine the organization responsible for the abnormality using only the causal relationship model generated using data from abnormal times. Therefore, in this embodiment, by comparing the causal relationship model during abnormal times with the causal relationship model during normal times, the organization related to the occurrence of the abnormality can be accurately determined. Therefore, even those who are not skilled can easily and quickly identify the factors contributing to the abnormality.

[0259] (2) The data used to generate the event causal relationship model during normal and abnormal times are basically obtained from the manufacturing line 3, which operates under a set condition (one condition) after the determination of the conditions. Therefore, when the adjustment items are changed, it is difficult to use the generated event causal relationship model to identify the abnormal factors or to adjust the quality. Therefore, in this embodiment, a control causal relationship model is generated using the first experimental plan data conducted before the manufacturing line 3 is started. That is, the first experimental plan data includes the following conditions: while the manufacturing line 3 is running, the adjustment amount of the adjustment items is changed, and the manufacturing line 3 is run through various settings to achieve the target quality. Therefore, by using the first experimental plan data, a restrictive causal relationship model corresponding to the adjustment amount of multiple adjustment items can be generated.

[0260] (3) Furthermore, by using the restricted causal relationship model and the event causal model corresponding to multiple adjustment items, a quality adjustment causal relationship model that can determine the adjustment items that affect quality can be generated.

[0261] (4) Furthermore, by changing the adjustment amount of the determined adjustment item and operating manufacturing line 3, second experimental plan data is generated to determine how the adjustment amount of the adjustment item affects quality, and a quality prediction model can be generated based on this to predict the relationship between the adjustment amount of the adjustment item and quality. Thus, the adjustment amount of the adjustment item can be optimized in a way that quality becomes the desired benchmark.

[0262] <6. Variations>

[0263] The present invention has been described in detail above with respect to one embodiment, but the present invention is not limited to the described embodiment. Various modifications or variations can be made as long as the spirit of the invention is not departed from. In addition, the following variations and the described embodiment can be appropriately combined.

[0264] <6-1>

[0265] In the described embodiment Figures 14-20 The example provided is a simple model, but in slightly more complex models with an increased number of nodes, for example, it might be possible to... Figure 33 In this way, an event causal relationship model is generated. The method for generating the event causal relationship model is the same as that described in the implementation method.

[0266] <6-2>

[0267] In the described embodiment, the control program 221 and the mechanism data 223 are analyzed in the constraint model generation module 101. However, it is also possible to generate causal relationships using only the analysis of the control program 221 and apply them to the event causal relationship model generation module 102. That is, in Figure 5 In this case, the second acquisition unit 112 and the second analysis unit 114 can also be omitted. Alternatively, only the institutional data 223 can be used to generate causal relationships, and these relationships can be applied to the event causal relationship model generation module 102. That is, in Figure 5 In this context, the first acquisition unit 111 and the first analysis unit 113 can also be omitted.

[0268] <6-3>

[0269] In the event causality model generation module 102 of the above embodiment, the event causality is generated using the constraint model generated in the constraint model generation module 101, but the constraint model may not necessarily be used.

[0270] <6-4>

[0271] In the described embodiment, by Figure 11The event causality model generation module 102 shown generates an event causality model, but it is not limited to this; other methods can be used to generate event causality models. For example, regarding node selection, nodes connected to the selected edge can be selected without using divergence.

[0272] In the event causality model generation module 102 of the described embodiment, causal relationships are generated using state data under normal conditions and state data under abnormal conditions, respectively. However, for example, causal relationships using only state data under abnormal conditions may also be generated. In this case, a causal relationship using state data under normal conditions is not constructed, but a deviation degree is calculated based on the feature quantity calculated in the third analysis unit 118 and the feature quantity calculated in the third analysis unit 118, and nodes are selected based on the deviation degree. Furthermore, a sensing partial graph is generated based on the nodes. This becomes the event causality model. In addition, the sensing partial graph is a type of partial graph, which refers to a graph from which a portion of vertices are taken and the presence or absence of edges between the vertex pairs is consistent with the original graph. For example, nodes with a deviation degree higher than a predetermined threshold are selected according to the abnormal causality model, and a graph is created in which the nodes and edges included between these nodes are the same as those in the original abnormal causality model, and this graph is set as the event causality model.

[0273] <6-5>

[0274] In generating the quality prediction model, to improve the model's accuracy, data acquired outside of the time of anomaly occurrence can be used flexibly. That is, in the above embodiment, the model is generated by changing the adjustment amount of adjustment items that affect the occurrence of anomalies, but it is also possible to add adjustment items unrelated to the occurrence of anomalies and generate the model by changing their adjustment amounts.

Claims

1. A process analysis device, comprising: The normal data acquisition unit acquires multiple first state data related to the normal state of multiple mechanisms constituting the manufacturing line; An abnormal data acquisition unit acquires multiple pieces of second state data related to the state at the time of the abnormality of the multiple mechanisms; Under normal circumstances, the analysis unit analyzes the acquired first state data of the multiple components, thereby determining the connection state between the multiple mechanisms as the first connection state; In case of anomaly, the analysis unit analyzes the acquired multiple second state data to determine the connection state between the multiple mechanisms as the second connection state. The normal-time relationship determination unit, based on the first connection state, determines the causal relationship between the multiple mechanisms in the process implemented in the manufacturing line as first graph information expressing the causal relationship in normal time; The abnormal situation relationship determination unit determines the causal relationship between the multiple mechanisms in the process implemented in the manufacturing line based on the second connection state as a second chart information expressing the causal relationship in the abnormal situation; and The special causal relationship determination unit compares the first chart information with the second chart information, selects the edge with a large change in intensity, or the edge that exists only in one of the first chart information and the second chart information, thereby determining the causal relationship between the multiple institutions associated with the occurrence of the anomaly as a special causal relationship.

2. The process analysis device according to claim 1, wherein... The normal state analysis unit and the abnormal state analysis unit determine the connection status between the multiple mechanisms through the following steps: (1) calculating the characteristic quantity based on the status data of each component; (2) calculating the correlation coefficient or partial correlation coefficient between the characteristic quantities. The special causal relationship determination part Calculate the degree of deviation between the characteristic quantity in each mechanism of the causal relationship under normal conditions and the characteristic quantity in each mechanism of the causal relationship under abnormal conditions. The mechanism whose deviation is greater than the specified value is selected as the node. When comparing the causal relationship under normal conditions with the causal relationship under abnormal conditions, if the connection state between the mechanisms meets the specified conditions, the connection between the mechanisms is selected as the edge. The specific causal relationship is determined by the selected nodes and edges.

3. The process analysis apparatus according to claim 1 or 2 further includes: The control program acquisition unit acquires a control program used to control the operation of the manufacturing line; The control program analysis unit determines the sequential relationship of the multiple mechanisms by analyzing the acquired control program; and The constraint model generation unit generates constraint models for the multiple mechanisms based on their sequential relationships. The special causal relationship determination unit determines the special causal relationship based on the constraint model.

4. The process analysis device according to claim 3 further includes: The mechanism data acquisition unit acquires mechanism data related to at least one of the relative positional relationships of the devices included in the plurality of mechanisms and the order in which the devices participate in the process; and The mechanism data analysis department analyzes the acquired mechanism data, models the processes implemented in the manufacturing line, and thereby determines a process model representing the sequential relationship of the multiple mechanisms. The constraint model generation unit generates the constraint model based on the sequential relationship of the multiple mechanisms and the process model.

5. The process analysis apparatus according to claim 3, wherein... The control program analysis unit determines the sequential relationship of the multiple mechanisms based on log data obtained by operating the manufacturing line using the control program.

6. The process analysis apparatus according to claim 5, wherein... The control program analysis unit determines the sequential relationship of the multiple mechanisms through the following steps: (1) constructing an abstract syntax tree based on the control program; (2) extracting variables and conditional branches related to each mechanism from the constructed abstract syntax tree; (3) obtaining log data when the manufacturing line is operating normally using the control program; and (4) sorting the variables based on the execution results of the conditional branches, referring to the obtained log data.

7. The process analysis apparatus according to claim 3 further includes: The first experimental plan data acquisition unit acquires first experimental plan data during the operation of the manufacturing line to determine the adjustment amount of the adjustment items of the manufacturing line used to achieve the specified quality. and The control causal relationship model generation unit generates, based on the first experimental plan data and the constraint model, the causal relationship between the adjustment project and the institution, as well as the causal relationship among the multiple institutions, as a control causal relationship model.

8. The process analysis apparatus according to claim 7, further comprising: The quality adjustment causality determination unit determines the quality adjustment causality by comparing the control causality with the specific causality. The quality adjustment causality includes causal relationships among the multiple agencies associated with the occurrence of the anomaly, as well as causal relationships between the agencies and the adjustment items.

9. The process analysis apparatus according to claim 8, further comprising: The second experimental plan data acquisition unit acquires second experimental plan data used to determine the relationship between the adjustment amount of the adjustment item and the quality, as determined by the causal relationship of the quality adjustment; and The quality prediction model generation unit generates a quality prediction model based on the second experimental plan data. The quality prediction model determines the relationship between the adjustment amount of the adjustment item in the manufacturing line and the quality.

10. The process analysis apparatus according to claim 9, further comprising: The adjustment calculation unit calculates the adjustment amount of the adjustment item to achieve the desired quality based on the quality prediction model.

11. The process analysis apparatus according to any one of claims 1 or 2, wherein The status data of each component represents at least one of torque, speed, acceleration, temperature, current, voltage, air pressure, pressure, flow rate, position, size, area, light intensity, and on / off status.

12. A process analysis method, wherein a computer performs the following steps: The steps for acquiring multiple first-state data related to the normal state of multiple mechanisms constituting a manufacturing line; The step of acquiring multiple pieces of second state data related to the state at the time of the anomaly of the multiple mechanisms; The step of determining the connection state between the multiple mechanisms as the first connection state by analyzing the acquired multiple first state data; The step of determining the connection state between the multiple mechanisms as the second connection state by analyzing the acquired multiple second state data; Based on the first connection state, the step of determining the causal relationship between the multiple mechanisms in the process implemented in the manufacturing line as a first graphical information expressing the causal relationship under normal conditions; Based on the second connection state, the step of determining the causal relationships between the multiple mechanisms in the processes implemented in the manufacturing line as second graphical information expressing the causal relationships in case of anomalies; and The first chart information is compared with the second chart information, and the edge with a large change in intensity, or the edge that exists only in one of the first chart information and the second chart information, is selected as the step of determining the causal relationship between the multiple institutions associated with the occurrence of the anomaly as a special causal relationship.

13. A process analysis computer-readable storage medium for causing a computer to perform the following steps: The steps for acquiring multiple first-state data related to the normal state of multiple mechanisms constituting a manufacturing line; The step of acquiring multiple pieces of second state data related to the state at the time of the anomaly of the multiple mechanisms; The step of determining the connection state between the multiple mechanisms as the first connection state by analyzing the acquired multiple first state data; The step of determining the connection state between the multiple mechanisms as the second connection state by analyzing the acquired multiple second state data; Based on the first connection state, the step of determining the causal relationship between the multiple mechanisms in the process implemented in the manufacturing line as a first graphical information expressing the causal relationship under normal conditions; Based on the second connection state, the step of determining the causal relationships between the multiple mechanisms in the processes implemented in the manufacturing line and expressing them as second graphical information of causal relationships when an anomaly occurs; and The first chart information is compared with the second chart information, and the edge with a large change in intensity, or the edge that exists only in one of the first chart information and the second chart information, is selected as the step of determining the causal relationship between the multiple institutions associated with the occurrence of the anomaly as a special causal relationship.

Images