Data analysis apparatus and data analysis method

Abstract

To accurately and efficiently specify data related to an analysis object in a manufacturing process.SOLUTION: The analysis device includes a variable group management part for managing a plurality of variables related to a manufacturing process by dividing them into a plurality of groups, an inter-group causal relation management part for managing inter-group causal relation information which is information on a causal relation between a plurality of groups, an inter-variable causal relation management part for managing variable inter-causal relation information which is information on the presence / absence of a causal relation between a predetermined variable and one or more other variables, an analysis part for specifying one or more specific groups related to the group to which the predetermined variable belongs on the basis of the inter-group causal relation information and specifying candidates for a variable to be analyzed on the basis of information on the specified groups and the inter-variable causal relation information, and an analysis part for generating a model using at least the variable candidates to be analyzed as explanatory variables and the predetermined variable as an objective variable.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention is suitable for application to a data analysis device and a data analysis method that perform data analysis on input data used as a learning sample when a model is machine-learned. [Background technology]

[0002] In data analysis of a manufacturing process, there are techniques such as those described below that identify candidates for explanatory variables that have a causal relationship with a certain target variable. First, Patent Document 1 describes a technique for narrowing down explanatory variables using, for example, sparse modeling. Furthermore, Patent Document 2 describes a technique for narrowing down explanatory variables using, for example, a linear regression model. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2023-151754 [Patent Document 2] Japanese Patent Application Publication No. 2019-191634 Summary of the Invention [Problem to be solved by the invention]

[0004] For example, data analysis of a manufacturing process may be performed for on-site improvement. In such cases, analysis may be performed using the data to be analyzed and data related to the data to be analyzed. However, when analyzing the data to be analyzed, it can be difficult to accurately identify the data related to the data to be analyzed. This is because a wide variety of data is associated with the manufacturing process, and managing the relationships between all of the data results in a huge amount of information, which is then difficult to manage. Furthermore, even when existing methods are used to narrow down the data related to the data to be analyzed as explanatory variables, sufficient sample data may not be obtained in the manufacturing process, resulting in a small number of input data to serve as learning samples. This can lead to variations in the candidates for explanatory variables narrowed down depending on the content of the input data, making it difficult to efficiently and accurately narrow down the candidates for explanatory variables. Furthermore, the accuracy of identifying data related to the data to be analyzed may affect the accuracy of data analysis, such as the prediction accuracy of a manufacturing process model. For these reasons, it is desirable to be able to accurately and efficiently identify data related to the data to be analyzed in a manufacturing process. Ultimately, it is desirable to improve the efficiency and accuracy of data analysis, such as model predictions in manufacturing processes.

[0005] The present invention has been made in consideration of the above points, and aims to propose a data analysis device and a data analysis method that enable accurate and efficient identification of data related to an analysis target in a manufacturing process. [Means for solving the problem]

[0006] In order to solve this problem, the present invention includes a variable group management unit that divides a plurality of variables related to a manufacturing process into a plurality of groups and manages them; an inter-group causal relationship management unit that manages inter-group causal relationship information, which is information on the causal relationships between a plurality of groups; an inter-variable causal relationship management unit that manages variable causal relationship information, which is information on the presence or absence of a causal relationship between a predetermined variable and one or more other variables; a candidate selection unit that identifies one or more specific groups related to the group to which the predetermined variable belongs based on the inter-group causal relationship information, and identifies candidate variables to be analyzed based on information on the identified specific groups and the variable causal relationship information; and an analysis unit that generates a model in which at least the candidate variables to be analyzed are used as explanatory variables and the predetermined variable is used as a target variable.

[0007] Furthermore, in the present invention, the method includes a variable group management step in which a variable group management unit divides a plurality of variables related to the manufacturing process into a plurality of groups and manages them; an inter-group causal relationship management step in which an inter-group causal relationship management unit manages inter-group causal relationship information, which is information regarding the causal relationship between the plurality of groups; an inter-variable causal relationship management step in which an inter-variable causal relationship management unit manages variable causal relationship information, which is information regarding the presence or absence of a causal relationship between a predetermined variable and one or more other variables; a candidate selection step in which a candidate selection unit identifies one or more specific groups associated with a group to which the predetermined variable belongs based on the inter-group causal relationship information, and identifies candidate variables to be analyzed based on information on the identified specific groups and the variable causal relationship information; and an analysis step in which an analysis unit generates a model using at least the candidate variables to be analyzed as explanatory variables and the predetermined variable as a target variable. [Effects of the Invention]

[0008] According to the present invention, it is possible to accurately and efficiently identify data related to an analysis target in a manufacturing process. [Brief explanation of the drawings]

[0009] [Figure 1] 1 is a system configuration diagram showing an example of the configuration of a manufacturing execution system according to a first embodiment. [Figure 2] FIG. 2 is a block diagram illustrating an example of a functional configuration of the data analysis apparatus. [Figure 3] FIG. 10 is a diagram illustrating an example of a variable group management table according to the first embodiment. [Figure 4] FIG. 3 is a diagram illustrating an example of an inter-group causal relationship management table in the first embodiment. [Figure 5] FIG. 3 is a diagram illustrating an example of an inter-variable causal relationship management table according to the first embodiment. [Figure 6] 10 is a flowchart illustrating an example of a procedure for data analysis processing according to the first embodiment. [Figure 7] FIG. 10 is a diagram showing an example of a KPI display screen displayed on a client's display device as an analysis result. [Figure 8] FIG. 10 is a system configuration diagram illustrating an example of the configuration of an MES system including a data analysis device according to a second embodiment. [Figure 9] FIG. 11 is a diagram illustrating an example of a variable group management table according to the second embodiment. [Figure 10] FIG. 11 is a diagram illustrating an example of an inter-group causal relationship management table according to the second embodiment. [Figure 11] FIG. 11 is a diagram illustrating an example of an inter-variable causal relationship management table according to the second embodiment. [Figure 12] 10 is a flowchart illustrating an example of a procedure for data analysis processing according to the second embodiment. [Figure 13] FIG. 10 is a diagram showing an example of a plant operation status screen showing the operation status of the plant displayed on the client's display device as an analysis result. [Figure 14] FIG. 11 is a system configuration diagram showing an example of the configuration of an analysis target variable generation unit of a data analysis apparatus according to a third embodiment. [Figure 15] 10 is a flowchart illustrating an example of a procedure for an explanatory variable extraction process. [Figure 16] FIG. 10 is a diagram illustrating an example of iteration. [Figure 17]FIG. 10 is a diagram illustrating an example of iteration. [Figure 18] FIG. 10 is a diagram illustrating an example of iteration. [Figure 19] FIG. 10 is a diagram illustrating an example of iteration. [Figure 20] FIG. 10 is a diagram illustrating an example of iteration. [Figure 21] FIG. 10 is a diagram illustrating an example of iteration. DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings. (1) First embodiment FIG. 1 is a system configuration diagram showing an example of the configuration of a manufacturing execution system (hereinafter referred to as "MES: Manufacturing Execution System") 1 according to a first embodiment. The MES system 1 includes an MES server 100, a plant information system 200, a client 300, a data analysis device 400, and a network 500. The MES server 100, the plant information system 200, the client 300, and the data analysis device 400 are connected to one another via the network 500, and can exchange information such as data. The data analysis device 400 includes a display device (not shown).

[0011] The MES server 100 is, for example, a computer, and includes a display screen generation unit 101, a KPI (Key Performance Indicators) generation unit 102, a control unit 103, a KPI 104, process information 105, manufacturing performance 106, and a quality control unit 107. The control unit 103 controls the operation of each of these elements.

[0012] The display screen generation unit 101 generates data for a display screen to be displayed on a display device (not shown), and causes the display device to display a display screen based on the data for the display screen. The display screens are a KPI display screen (see FIG. 7) and a plant operating status display screen (see FIG. 13), which will be described later.

[0013] The KPI generation unit 102 generates data for displaying KPIs related to the manufacturing process based on the KPI 104 related to the KPIs of the manufacturing process, and causes the display device to display a KPI display screen based on the data for displaying KPIs.

[0014] Process information 105 represents information about each step in the manufacturing process. Manufacturing performance 106 is various time-series data including information about the manufacturing performance at each step in the manufacturing process. Quality control unit 107 performs quality control in the manufacturing process and manages the quality control data that is the result of the quality control.

[0015] The plant information system 200 is, for example, a computer, and includes a plant information collection unit 201 and time-series data 202. The plant information collection unit 201 collects various time-series data 202 generated in conjunction with manufacturing, for example, targeting the manufacturing process of a plant in a factory.

[0016] The client 300 is a computer for operating the MES server 100. The client 300 includes a KPI display unit 301 and a plant operation status display unit 302.

[0017] The KPI display unit 301 controls the display of KPIs as described below based on the time-series data 202 acquired from the MES server 100. The plant operation status display unit 302 controls the display of the operation status of the plant as described below based on the time-series data 202 acquired from the MES server 100. The contents of these displays will be described later.

[0018] The data analysis device 400 has a function of analyzing time-series data accumulated in the manufacturing process. The data analysis device 400 will be described in detail later.

[0019] 2 is a block diagram showing an example of the functional configuration of a data analysis apparatus 400. The data analysis apparatus 400 includes a factor analysis result 411, an analysis target variable generation unit 401, a factor analysis result generation unit 402, a variable group management table 421, an inter-group causal relationship management table 422, and an inter-variable causal relationship management table 423.

[0020] In this embodiment, variables (hereinafter sometimes referred to as "parameters") are managed in groups, which facilitates information management, and narrowing down the search results by group when creating a model makes processing more efficient.

[0021] First, the difference between a "process" and a "variable" in this embodiment will be explained. A "process" refers to a group that includes multiple related "variables (parameters)." Examples of "variables" include time-series data, predetermined parameter items, etc.

[0022] The process classification may be set arbitrarily by an administrator, or may be set automatically based on information on the flow of a manufacturing plan or information from a manual, etc. The process classification may be set based on the classification of a plurality of work steps that make up the manufacturing process, for example.

[0023] Next, the "objective variable" and the "explanatory variable" will be described. In this embodiment, among the multiple parameters included in the above-mentioned process, the parameter to be analyzed is the "objective variable." On the other hand, in this embodiment, among the multiple parameters included in the above-mentioned process, other parameters included in the process to be analyzed and other parameters included in other processes are the "explanatory variables."

[0024] The variable group management table 421 is an example of a variable group management unit, and manages multiple variables related to the manufacturing process, i.e., the variables included in multiple processes that generate time-series data accumulated in the manufacturing process, by dividing them into multiple groups. In the variable group management table 421, the variables included in multiple processes are grouped into groups for each tag name and managed as processes. Details of the variable group management table 421 will be described later.

[0025] The inter-group causal relationship management table 422 is an example of an inter-group causal relationship management unit, and manages inter-group causal relationship information, which is information relating to the causal relationships between a plurality of grouped groups. The inter-group causal relationship management table 422 will be described in detail later. The variable group management table 421 and the inter-group causal relationship management table 422 can be defined in advance, but in this embodiment, for example, they are defined mainly based on information managed on the MES server 100 side.

[0026] The inter-variable causal relationship management table 423 is an example of a variable causal relationship management unit, and manages, for example, variable causal relationship information, which is information on the presence or absence of a causal relationship between a predetermined variable and one or more other variables. In the inter-variable causal relationship management table 423, the presence or absence of a causal relationship between a plurality of processes is set in advance.

[0027] The analysis target variable generation unit 401 is an example of a candidate selection unit, which identifies one or more specific groups related to the group to which a predetermined variable belongs based on the inter-group causal relationship information, and identifies candidate variables to be analyzed based on the information on the identified specific groups and the variable causal relationship information.

[0028] Here, an upstream group of a certain group means a group related to a process located upstream of the process indicated by that group in the manufacturing process. Note that "same level" refers to a positional relationship in which a certain process and another process are parallel, and neither can be said to be upstream or downstream.

[0029] In factor analysis, variables that belong to a group upstream of or at the same level as the dependent variable are used as explanatory variables. In the illustrated inter-group causal relationship management table 422, for example, in the case of analysis in which a variable included in a process hatched from the upper left to the lower right is used as the dependent variable, other variables included in the process hatched from the upper left to the lower right and variables included in the process hatched from the upper right to the lower left are used as explanatory variables.

[0030] The factor analysis result generating unit 402 is an example of an analysis unit, and generates a model using at least candidate variables to be analyzed as explanatory variables and a predetermined variable as a response variable, and performs an analysis process using the generated model 450. For example, the factor analysis result generating unit 402 analyzes the causes of defects that may occur in the predetermined variable using the generated (e.g., machine learning) model 450 using one or more other variables included in a group to which the predetermined variable belongs as explanatory variables, and obtains a factor analysis result 411 as the analysis result. In the embodiment, performing machine learning in this manner is also referred to as a "training process." Note that the training process may be performed by either the analysis target variable generating unit 401 or the factor analysis result generating unit 402. In the present embodiment, for example, a predetermined variable included in an intermediate inspection process is the response variable, and a narrowed-down set of variables included in processes upstream of the predetermined variable are the explanatory variables. In addition, for example, other variables included in the intermediate inspection process other than the predetermined variable can also be candidates for the explanatory variables. The factor analysis result generating unit 402 is an example of an analysis unit, and generates a model in which a candidate variable to be analyzed and one or more other variables included in the group to which a predetermined variable belongs are used as explanatory variables, and the predetermined variable is used as a target variable.

[0031] Fig. 3 is a diagram showing an example of a variable group management table 421 in the first embodiment. The variable group management table 421 manages the group names of process groups, each of which groups multiple processes, for each tag name. In Fig. 3, each variable is managed by group.

[0032] The variable group management table 421 manages tag IDs for each group. A tag ID is an identification ID for a variable. The tag ID may be an identification ID for a variable used in an MES or DCS. In this way, information acquired by an MES can be managed by directly associating it with a process group, making information management easier. Groups may be classified by process steps, such as "Process 1" and "Process 2." Groups may also be classified by categories other than processes, such as "Temperature" and "Flow Rate." Each group includes one or more variables. In this embodiment, for example, one tag ID may be classified into multiple groups.

[0033] In variable group management table 421, for example, a group named "Process 1" manages variables with tag IDs named "T123," "T232," and "P232," a group named "Process 2" manages a variable with tag ID named "T332," a group named "Process 3" manages a variable with tag ID named "T345," a group named "Intermediate Inspection Process" manages a variable with tag ID named "T223," a group named "Process 4" manages variables with tag IDs named "P522," "T645," and "T874," and a group named "Final Inspection Process" manages a variable with tag ID named "L224."

[0034] In the illustrated example, variable group management table 421 shows that processes are arranged from upstream to downstream, from the process at the top to the process at the bottom. Therefore, it can be seen that the process with the group name "Intermediate Inspection Process" (including a variable with a tag ID named "L223") is a downstream process of the process in the group named "Process 1," which includes variables with tag IDs named "T123," "T232," and "P232," a variable with a tag ID named "T332" managed by the group name "Process 2," and a variable with a tag ID named "T345" managed by the group name "Process 3."

[0035] 4 is a diagram showing an example of an inter-group causal relationship management table 422 in the first embodiment. The inter-group causal relationship management table 422 shows that a process managed by a group named "Process 1" and a process managed by a group named "Process 2" are upstream of a process managed by a group named "Process 3," and that the process managed by a group named "Process 3" is upstream of a process managed by a group named "Intermediate Inspection Process." In the case of an analysis in which variables included in processes hatched from upper left to lower right are used as target variables, variables (tag IDs) included in processes hatched from upper left to lower right and processes hatched from upper right to lower left are used as explanatory variables.

[0036] Furthermore, the inter-group causal relationship management table 422 indicates that the process managed by the group named "Intermediate Inspection Process" is upstream of the process managed by the group named "Process 4," which is in turn upstream of the process managed by the group named "Final Inspection Process."

[0037] Figure 4 above shows the first stage of narrowing down candidate variables based on the causal relationships between groups, while Figure 5 below shows further narrowing down or adjustment after the first stage of narrowing down. Figure 4 makes it possible to manage the relationships between variables on a group-by-group basis, making management easier.

[0038] Fig. 5 is a diagram showing an example of the inter-variable causal relationship management table 423 in the first embodiment. The illustrated example shows exceptions to the variables managed in Fig. 4. In Fig. 5, for example, it is possible to exclude variables that have no causal relationship (for example, "T123" described later) from the multiple variables narrowed down at the stage shown in Fig. 4. Furthermore, in Fig. 5, it is also possible to set variables that should always be added to the explanatory variables (for example, "T232" described later) when searching for explanatory variables, etc.

[0039] In the inter-variable causal relationship management table 423, a predetermined process including a variable to be analyzed, for example, a process with tag name "L223", is defined as having a causal relationship with another process, for example, a variable with tag name "T232" (included in process 1), and is defined as not having a causal relationship with, for example, a variable with tag name "T123" (included in process 1). Also, a predetermined variable to be analyzed, for example, a variable with tag name "P522", is defined as having a causal relationship with another variable, for example, a variable with tag name "T645" (included in process 4).

[0040] 4. For example, in the inter-variable causal relationship management table 423, it is also possible to define that a predetermined process to be analyzed, for example, a process with tag name "L223", has a causal relationship with another process, for example, a variable with tag name "P522" (included in process 4), to be added to the items that are determined to have a causal relationship in the inter-variable causal relationship management table 423 (corresponding to "causal relationship present" in the figure).

[0041] That is, in the inter-variable causal relationship management table 423, by defining whether or not there is a causal relationship between a predetermined objective variable to be analyzed and an explanatory variable, it is possible to prevent explanatory variables that are unlikely to contribute to improving the analysis accuracy of the model 450 from being input into the model 450, among the other variables input as explanatory variables for machine learning. Also, it is possible to input explanatory variables that are likely to contribute to improving the analysis accuracy of the model 450, among the other variables to be excluded as explanatory variables for machine learning.

[0042] The MES system 1 including the data analysis device 400 is configured as described above. Next, an example of the operation of the MES system 1, particularly the data analysis method using the data analysis device 400, will be described. The data analysis method includes a variable group management step in which a variable group management table 421 manages multiple variables related to the manufacturing process by dividing them into multiple groups (e.g., "Process 1," "Process 2," and "Process 4"), an inter-group causal relationship management step in which an inter-group causal relationship management table 422 manages inter-group causal relationship information, which is information about causal relationships between multiple groups, an inter-variable causal relationship management step in which an inter-variable causal relationship management table 423 manages variable causal relationship information, which is information about the presence or absence of a causal relationship between a predetermined variable and one or more other variables, a candidate selection step in which an analysis target variable generation unit 401 identifies one or more specific groups related to the group to which the predetermined variable belongs based on the inter-group causal relationship information and identifies candidate variables to be analyzed based on information about the identified specific groups and the variable causal relationship information, and an analysis step in which an analysis unit (e.g., a factor analysis result generation unit 402) generates a model using at least candidate variables to be analyzed as explanatory variables and a predetermined variable as a response variable. The following describes these steps in detail.

[0043] 6 is a flowchart showing an example of the procedure of the data analysis process according to the first embodiment. The illustrated data analysis process is, for example, data analysis process relating to factor analysis.

[0044] In step S1, the user selects the analysis content, for example, "factor analysis." In step S2, the user inputs the variable "L223" (for example, included in "intermediate inspection process") as the target variable, i.e., the variable of the specified process to be analyzed.

[0045] In step S3, the analysis target variable generation unit 401 narrows down the explanatory variables based on the causal relationships. Specifically, the analysis target variable generation unit 401 selects, as an explanatory variable, at least one variable included in process 1, process 2, or process 3 that is upstream (in other words, "higher level") than the variable with tag name "L223" (intermediate inspection process) as a predetermined variable to be analyzed, in accordance with the inter-variable causal relationship management table 423.

[0046] In step S4, the analysis target variable generation unit 401 narrows down candidates for explanatory variables as variables of other processes that have a causal relationship with the variable of the specified process for each input target variable as a variable of the specified process. The analysis target variable generation unit 401 narrows down the additional explanatory variables based on the inter-variable causal relationship management table 423.

[0047] In step S5, the analysis target variable generation unit 401 acquires the target data to be analyzed. The analysis target variable generation unit 401 acquires the time-series data 202 from the plant information system 200.

[0048] Here, a training process is executed for the model 450. In this embodiment, performing machine learning in this manner is also referred to as a "training process." The training process may be performed by either the analysis target variable generation unit 401 or the factor analysis result generation unit 402. Details of the training process have already been explained, so further explanation will be omitted.

[0049] In step S6, the factor analysis result generating unit 402 generates a model 450 in which at least candidates for the variables to be analyzed are used as explanatory variables and the predetermined variable is used as a target variable, and executes an analysis process using the generated model 450. The factor analysis result generating unit 402 analyzes factors of defects that may occur in the predetermined variable using the generated model 450 in which, for example, one or more other variables included in the group to which the predetermined variable belongs are used as explanatory variables, and obtains a factor analysis result 411 as the analysis result.

[0050] In step S7, the client 300 acquires the factor analysis result 411 as the analysis result by the factor analysis result generating unit 402, and the user of the client 300 evaluates the analysis result. When the user evaluates the analysis result, the following display is displayed.

[0051] 7 is a diagram showing an example of a KPI display screen 301A displayed as an analysis result on the display device of the client 300. The KPI display screen 301A is displayed by the KPI display unit 301 of the client 300 based on the factor analysis result 411.

[0052] The KPI display screen 301A displays past performance data, for example, for each lot indicated by a lot number, including the lot, quality index, and the first and second factors resulting from the above-mentioned analysis process. In other words, a user viewing the KPI display screen 301A can understand, for each lot, how variables of other processes affect the variables of a specific process that are input as explanatory variables as possible factors.

[0053] (2) Second embodiment The MES system including the data analysis device according to the second embodiment has almost the same configuration and operation as the MES system 1 including the data analysis device 400 according to the first embodiment, and therefore the following description will focus on the differences from the first embodiment. Unlike the first embodiment, the second embodiment performs anomaly detection instead of cause analysis.

[0054] 8 is a system configuration diagram showing an example of the configuration of an MES system 1A including a data analysis apparatus 400A according to the second embodiment. The data analysis apparatus 400A according to the second embodiment is provided with an anomaly detection result generation unit 403 instead of the factor analysis result generation unit 402.

[0055] The analysis target variable generation unit 401 is an example of a candidate selection unit, which identifies variables of multiple specific processes that belong to other groups within a specified range upstream and downstream (for example, two stages upstream, at the same level, or two stages downstream) of a specified group to which a specified variable belongs in the manufacturing process, and narrows down the candidates for variables of the specific processes to be analyzed as candidates for explanatory variables depending on whether or not there is a causal relationship between the process including the specified variable and the multiple specific processes. The candidates for variables of the specific processes to be analyzed are candidates for explanatory variables that have a causal relationship with the target variable.

[0056] The anomaly detection result generation unit 403 is an example of an analysis unit, and by inputting, for example, a predetermined variable as an explanatory variable and a candidate variable for the specific process as a target variable, the generated model 450 is used to detect anomalies that may be caused by other variables with respect to the predetermined variable.

[0057] 9 is a diagram showing an example of a variable group management table 421 in the second embodiment. The variable group management table 421 manages the group names of process groups, each of which groups together variables for multiple processes, for each tag name used to identify the variables of each process. Note that a description of the same content in the variable group management table 421 in the second embodiment as in the variable group management table 421 in the first embodiment will be omitted.

[0058] In the illustrated example, variable group management table 421 shows that processes are arranged from upstream to downstream, moving from the top process to the bottom process. Therefore, it can be seen that a process in a group named "Process 4" (including, for example, a variable with a tag name "T645") is a downstream process of a process including a variable with a tag name "L233" managed in a group named "Intermediate Inspection Process" and a process including a variable with a tag name "P522" managed in a group named "Process 4", and is an upstream process of a process including a variable with a tag name "T872" managed in a group named "Process 4" and a process including a variable with a tag name "L244" managed in a group named "Final Inspection Process".

[0059] 10 is a diagram showing an example of an inter-group causal relationship management table 422 in the second embodiment. The inter-group causal relationship management table 422 indicates that a process managed by a group named "Process 4" is downstream of a process managed by a group named "Intermediate Inspection Process" and upstream of a process managed by a group named "Final Inspection Process." The other processes are the same as those in the inter-group causal relationship management table 422 shown in FIG. 4 described above, and therefore will not be described again.

[0060] 11 shows an example of the inter-variable causal relationship management table 423 in the second embodiment. The inter-variable causal relationship management table 423 in the second embodiment is similar to the inter-variable causal relationship management table 423 in the first embodiment, and therefore a description thereof will be omitted.

[0061] 12 is a flowchart showing an example of a procedure of a data analysis process according to the second embodiment. The illustrated data analysis process is, for example, a data analysis process related to anomaly detection. Note that in the data analysis process according to the second embodiment, a description of procedures that are substantially the same as those of the first embodiment shown in FIG. 6 will be omitted.

[0062] In step S1A, the user selects, for example, "anomaly detection" as the content of analysis. In step S2A, the user inputs the variable "T645" as the target variable, that is, the predetermined variable to be analyzed.

[0063] In step S3, the analysis target variable generation unit 401 narrows down the explanatory variables based on the causal relationships. Specifically, the analysis target variable generation unit 401 selects, as an explanatory variable, at least one variable that belongs to the intermediate inspection process and process 4 on the upstream side, and process 4 and final inspection process on the downstream side, within a predetermined range centered on the process (process 4) that includes the variable with the tag name "T645" as a variable of the predetermined process to be analyzed, in accordance with the inter-variable causal relationship management table 423.

[0064] In step S4A, the analysis target variable generation unit 401 narrows down the explanatory variables for each objective variable. The analysis target variable generation unit 401 narrows down the additional explanatory variables based on the inter-variable causal relationship management table 423.

[0065] In step S5A, the analysis target variable generation unit 401 acquires the target data to be analyzed. The analysis target variable generation unit 401 acquires the time-series data 202 from the plant information system 200.

[0066] Here, a training process is executed for the model 450. The training process may be performed by either the analysis target variable generation unit 401 or the anomaly detection result generation unit 403. Details of the training process have already been explained, and therefore will not be explained again.

[0067] In step S6A, the anomaly detection result generating unit 403 performs an analysis process using the generated model 450 to detect an anomaly that may be caused by other variables with respect to the predetermined variable.

[0068] In step S7A, client 300 acquires anomaly detection result 412 as the analysis result from anomaly detection result generation unit 403, and the user of client 300 evaluates the analysis result. When the user evaluates the analysis result, the following display is displayed.

[0069] 13 is a diagram showing an example of a plant operation status screen 302A showing the operation status of the plant as an analysis result, which is displayed on the display device of the client 300. The plant operation status screen 302A displays, for example, a normal message 302A1 as the real-time operation status of the manufacturing process of the plant based on the time-series data 202, and also displays an abnormality message 302A2 based on the abnormality detection result 412 by the plant operation status display unit 302 of the client 300.

[0070] On the plant operation status screen 302A, for example, a normal message 302A1 is displayed in real time, and an abnormality message 302A2 is displayed in a conspicuous display manner based on an abnormality detection result 412 created by analysis processing. The conspicuous display manner referred to here is, for example, a display manner in which the message is displayed in white on a black background and includes, for example, the word "(Caution)".

[0071] (3) Third embodiment The MES system including data analysis apparatus 400B according to the third embodiment has substantially the same configuration and operation as MES system 1 including data analysis apparatus 400 according to the first embodiment and MES system including data analysis apparatus 400A according to the second embodiment, and therefore the following description will focus on the differences from the first and second embodiments. The third embodiment is configured to be able to select explanatory variable candidates even with an even smaller amount of input data than the first and second embodiments. This will be described in detail below.

[0072] 14 is a system configuration diagram showing an example of the configuration of the analysis target variable generation unit 401 of the data analysis apparatus 400B according to the third embodiment. In the data analysis apparatus 400B according to the third embodiment, the analysis target variable generation unit 401 includes a candidate verification unit 404.

[0073] The candidate verification unit 404 is an example of an explanatory variable candidate verification unit, and when narrowing down the explanatory variable candidates in the above-mentioned analysis process (step S6), compared to the first and second embodiments, verifies a combination of multiple processes upstream and / or downstream of a process including a specific variable to be analyzed, for variables of a specific process that should be used as an explanatory variable having a causal relationship with a specific variable to be analyzed. Hereinafter, these combinations are referred to as "explanatory variable process candidates," and the process of sequentially verifying them is referred to as "iteration." Specific details of this verification will be described.

[0074] When verifying variables of at least one process upstream of a process including a predetermined variable to be analyzed as in the first embodiment, the candidate verification unit 404 creates combinations of variables of multiple types of specific processes that belong to other groups upstream of or at the same level as the predetermined group to which the predetermined variable belongs in the manufacturing process, and narrows down candidates for variables of the specific process to be analyzed as explanatory variable candidates depending on whether or not there is a causal relationship between the predetermined variable as the objective variable and the combination of variables of multiple types of specific processes (hereinafter also referred to as "explainer variable process candidates"). Note that the candidates for variables of the specific process to be analyzed are candidates for explanatory variables that have a causal relationship with the objective variable.

[0075] In each iteration, the candidate verification unit 404 selects explanatory variables for each group based on the inter-group causal relationship management table 422 shown in FIG. 4, generates a model, and evaluates it. In the inter-group causal relationship management table 422 shown in FIG. 4, processes in groups hatched from the upper right to the lower left are "explainer variable process candidates," and variables in groups hatched from the upper left to the lower right are target variables. The candidate verification unit 404 adopts the "explainer variable process candidate" in the iteration with the highest evaluation result as the analysis result. The evaluation method may be, for example, based on the accuracy of the model. Alternatively, the continuity of contributions based on the analysis results using explainable artificial intelligence (XAI) may also be used. The analysis process may be terminated, for example, when a specified number of iterations have been performed, when a target evaluation value has been reached, or when a specified time has elapsed.

[0076] Fig. 15 is a flowchart showing an example of the procedure for the explanatory variable extraction process. The explanatory variable extraction process is performed, for example, as part of the analysis process of step S6 in Fig. 6 and step S6A in Fig. 12. In step S21, the candidate verification unit 404 narrows down the explanatory variables based on the causal relationships. Specifically, the candidate verification unit 404 narrows down the explanatory variable candidates based on the above-mentioned causal relationships by performing verification for each iteration as shown in Figs. 16 to 21.

[0077] In step S22, the analysis target variable generation unit 401 generates a model 450. In step S23, the candidate verification unit 404 evaluates the model 450. In step S24, the candidate verification unit 404 determines whether or not a termination condition is met, and if the termination condition is met, the process ends, whereas if the termination condition is not met, the process does not end and the process returns to step S21 and is executed. The details of the analysis process will be further explained below.

[0078] In this embodiment, for example, using the inter-group causal relationship management table 422 shown in Figure 4, an example will be explained of a process managed by a group named "Process 1", a process managed by a group named "Process 2", a process managed by a group named "Process 3", and a process managed by a group named "Intermediate Inspection Process".

[0079] The process managed by the group named "Process 1" and the process managed by the group named "Process 2" are upstream of the process managed by the group named "Process 3", which in turn is upstream of the process managed by the group named "Intermediate Inspection Process".

[0080] The process (objective variable) managed by the group named "Intermediate Inspection Process" can be influenced by any of the processes in the upstream groups because the process (explanatory variable) managed by the group named "Process 1" and the process managed by the group named "Process 2" are upstream of the process (explanatory variable) managed by the group named "Process 3." In this embodiment, various iterations are used to analyze which of the processes in the three groups is more likely to have a greater influence. In this embodiment, the analysis process is performed using, for example, the first to sixth iterations.

[0081] In the first iteration shown in Figure 16, the processes managed by the group named "Process 3" are used as explanatory variables, i.e., "candidate explanatory variable processes," and the analysis process is performed using the variables included in the group named "Intermediate Inspection Process" as the objective variables.

[0082] In the second iteration shown in Figure 17, the variables included in the process managed by the group named "Process 1" and the process managed by the group named "Process 3" are used as explanatory variables, i.e., "candidate explanatory variable processes," and the analysis process is performed using the variables included in the group named "Intermediate Inspection Process" as the target variable.

[0083] In the third iteration shown in Figure 18, the variables included in the process managed by the group named "Process 2" and the process managed by the group named "Process 3" are used as explanatory variables, i.e., "candidate explanatory variable processes," and the analysis process is performed using the variables included in the group named "Intermediate Inspection Process" as the objective variables.

[0084] In the fourth iteration shown in Figure 19, the variables included in the process managed by the group named "Process 1," the process managed by the group named "Process 2," and the process managed by the group named "Process 3" are used as explanatory variables, i.e., "candidate explanatory variable processes," and analysis processing is performed using the variables included in the group named "Intermediate Inspection Process" as the objective variable.

[0085] In the fifth iteration shown in Figure 20, the variables contained in the processes managed by the group named "Process 1" are used as explanatory variables, i.e., "candidate explanatory variable processes," and the analysis process is performed using the variables contained in the group named "Intermediate Inspection Process" as the objective variables.

[0086] In the sixth iteration shown in Figure 21, the variables included in the process managed by the group named "Process 2" are used as explanatory variables, i.e., "candidate explanatory variable processes," and the analysis process is performed using the variables included in the group named "Intermediate Inspection Process" as the objective variables.

[0087] That is, the candidate verification unit 404 analyzes each iteration to identify other groups that are likely to affect the specified group to which the specified process being analyzed belongs, and narrows down the explanatory variables to the variables of the processes belonging to those other groups.

[0088] On the other hand, when verifying multiple processes within a predetermined range upstream and downstream of a specific process to be analyzed as in the second embodiment, the candidate verification unit 404 creates combinations of variables for multiple types of specific processes that belong to other groups within a predetermined range upstream and downstream (for example, two levels higher, the same level, or two levels lower) centered on a predetermined group to which the specific variable belongs in the manufacturing process, and verifies and narrows down candidates for variables for the specific process to be analyzed as candidates for explanatory variables depending on whether there is a causal relationship between the predetermined variable as the objective variable and the combination of variables for multiple types of specific processes. Candidates for variables for the specific process to be analyzed are candidates for explanatory variables that have a causal relationship with the objective variable.

[0089] The candidate verification unit 404 verifies and narrows down candidate variables for specific processes to be analyzed as candidates for explanatory variables by changing the combination of variables for multiple types of specific processes that belong to other groups in a predetermined range upstream and downstream (for example, "two levels higher, the same level, or two levels lower") centered on the predetermined group to which the specific process belongs in the manufacturing process, for example, as in the iterations shown in FIGS. 16 to 21, and verifies the combination in the same way as in the first embodiment described above.

[0090] The data analysis apparatus 400 according to this embodiment includes: a variable group management table 421 that divides a plurality of variables related to a manufacturing process into a plurality of groups and manages them; an inter-group causal relationship management table 422 that manages inter-group causal relationship information, which is information about the causal relationships between a plurality of groups; an inter-variable causal relationship management table 423 that manages variable causal relationship information, which is information about the presence or absence of a causal relationship between a predetermined variable and one or more other variables; an analysis target variable generation unit 401 that identifies one or more specific groups related to the group to which the predetermined variable belongs based on the inter-group causal relationship information, and identifies candidate variables to be analyzed based on information about the identified specific groups and the variable causal relationship information; and an analysis unit that generates a model 450 in which at least candidate variables to be analyzed are used as explanatory variables and the predetermined variable is used as a response variable.

[0091] According to this embodiment, a wide variety of variables are managed in groups, each group consisting of variables with related meanings, and narrowing down the search results by group during model creation makes it possible to accurately and efficiently identify variables related to the target variable for analysis while facilitating information management compared to managing individual variables. In other words, managing variables by group facilitates information management, and narrowing down the search results by group during model creation improves processing efficiency. In this way, data related to the target variable for analysis in the manufacturing process can be accurately and efficiently identified. This ultimately improves the efficiency and accuracy of data analysis, such as model prediction, in the manufacturing process.

[0092] In this embodiment, the factor analysis result generating unit 402 uses one or more other variables included in the group to which the predetermined variable belongs as explanatory variables to analyze the factors of defects that may occur in the predetermined variable using the generated model 450. Therefore, even if the number of input data serving as learning samples is small, it is possible to efficiently narrow down candidates for explanatory variables, and it is possible to generate a model 450 with high prediction accuracy even from a small amount of input data for, for example, a wide variety of products.

[0093] In this embodiment, the analysis target variable generation unit 401 identifies variables of at least one specific process that is upstream of or belongs to another group at the same level as the specific group to which the specific variable belongs in the manufacturing process, and narrows down candidate variables of the specific process to be analyzed as candidate explanatory variables depending on whether or not there is a causal relationship between the process including the specific variable and the at least one specific process. In this way, it is possible to efficiently narrow down candidate explanatory variables even if the number of input data serving as learning samples is small, and it is possible to generate a model 450 with high prediction accuracy even from a small amount of input data for, for example, a wide variety of products.

[0094] In this embodiment, the anomaly detection result generating unit 403 uses the generated model 450 to detect an anomaly that may occur with respect to a predetermined variable due to other variables.

[0095] In this embodiment, the analysis target variable generation unit 401 identifies multiple specific variables that belong to other groups within a specified range upstream and downstream of a specific group to which a process including a specific variable belongs in the manufacturing process, and narrows down the candidate variables for the specific process to be analyzed as candidate explanatory variables depending on whether or not there is a causal relationship between the process including the specific variable and the multiple specific processes. In this way, it is possible to efficiently narrow down the candidate explanatory variables even if the number of input data used as learning samples is small, and it is possible to generate a model 450 with high prediction accuracy even from a small amount of input data.

[0096] In this embodiment, the analysis target variable generation unit 401 includes a candidate verification unit 404 that verifies combinations of multiple variables upstream and / or downstream of a predetermined variable to be analyzed, for variables of a specific process to be analyzed as candidates for explanatory variables. In this way, even when the number of input data serving as learning samples is small and the product is diverse, it is possible to efficiently narrow down the candidates for explanatory variables, and it is possible to generate a model 450 with high prediction accuracy even from a small amount of input data.

[0097] In this embodiment, the candidate verification unit 404 creates combinations of variables for multiple specific processes that belong to other groups upstream of or at the same level as the specific group to which the specific variable belongs in the manufacturing process, and verifies and narrows down candidates for the specific process variables to be analyzed as explanatory variable candidates depending on whether or not there is a causal relationship between the specific variable as the objective variable and the combination of variables for multiple specific processes. In this way, even when the number of input data used as learning samples is small and the product is diverse, it is possible to efficiently narrow down the candidates for explanatory variables, and it is possible to generate a model 450 with high prediction accuracy even from a small amount of input data.

[0098] In this embodiment, the candidate verification unit 404 creates combinations of variables for multiple specific processes that belong to other groups within a predetermined range upstream and downstream of a predetermined group to which a specific variable belongs in the manufacturing process, and verifies and narrows down candidates for the specific process variables to be analyzed as explanatory variable candidates depending on whether or not there is a causal relationship between the predetermined variable as the objective variable and the combination of variables for multiple specific processes. In this way, even when the number of input data used as learning samples is small and the product is diverse, it is possible to efficiently narrow down the candidates for explanatory variables, and it is possible to generate a model 450 with high prediction accuracy even from a small amount of input data.

[0099] In this embodiment, the factor analysis result generating unit 402 is an example of an analysis unit, and generates a model in which a candidate variable to be analyzed and one or more other variables included in the group to which a predetermined variable belongs are used as explanatory variables, and the predetermined variable is used as a response variable.

[0100] The present invention is not limited to the above-described embodiments, and includes various modifications and equivalent configurations within the spirit and scope of the appended claims. For example, the above-described embodiments have been described in detail to clearly explain the present invention, and the present invention is not necessarily limited to those including all of the described configurations. Furthermore, the elements described in parallel in the present embodiment may be configured such that at least one of the elements is connected in series to the other elements. [Industrial Applicability]

[0101] The present invention can be applied to, for example, a data analysis device related to a technology for performing data analysis on input data used as a learning sample when machine learning a model. [Explanation of symbols]

[0102] 100...MES server, 200...factory server, 300...MES client, 400...data analysis device, 401...analysis target variable generation unit, 402...factorial analysis result generation unit, 403...anomaly detection result generation unit, 421...variable group management table, 422...inter-group causal relationship management table, 423...inter-variable causal relationship management table

Claims

1. a variable group management unit that manages a plurality of variables related to the manufacturing process by dividing them into a plurality of groups; an inter-group causal relationship management unit that manages inter-group causal relationship information, which is information about the causal relationships between the plurality of groups; an inter-variable causal relationship management unit that manages variable causal relationship information, which is information regarding the presence or absence of a causal relationship between a predetermined variable and one or more other variables; a candidate selection unit that identifies one or more specific groups related to the group to which the predetermined variable belongs based on the inter-group causal relationship information, and identifies candidate variables to be analyzed based on information on the identified specific groups and the variable causal relationship information; an analysis unit that generates a model using at least the candidate variables to be analyzed as explanatory variables and the predetermined variable as a response variable; A data analysis device comprising:

2. The analysis unit The generated model is used to analyze factors causing defects that may occur in the predetermined variables. The data analysis device according to claim 1 .

3. The candidate selection unit At least one specific variable belonging to another group upstream of or at the same level as the specific group to which the specific variable belongs in the manufacturing process is identified, and candidates for variables of the specific process to be analyzed as candidates for explanatory variables are narrowed down depending on whether or not there is a causal relationship between a process including the specific variable and the at least one specific process. The data analysis device according to claim 2 .

4. The analysis unit an anomaly detection result generation unit that uses the generated model to detect anomalies that other variables may cause with respect to the predetermined variable; The data analysis device according to claim 1 .

5. The candidate selection unit Identifying variables of a plurality of specific processes that belong to a predetermined group to which the predetermined variable belongs in the manufacturing process and that are located within a predetermined range on the upstream and downstream sides of the predetermined group; and narrowing down candidates for variables of the specific processes to be analyzed as candidates for explanatory variables depending on whether or not there is a causal relationship between the process including the predetermined variable and the plurality of specific processes.

5. The data analysis device according to claim 4.

6. The candidate selection unit The system includes an explanatory variable candidate verification unit that verifies a combination of multiple processes at least on either the upstream side or downstream side of a process including a specific variable to be analyzed, for the variable of the specific process to be analyzed as a candidate explanatory variable. The data analysis device according to claim 1 .

7. The explanatory variable candidate verification unit A combination of variables of a plurality of types of specific processes that belong to another group upstream of or at the same level as the predetermined group to which the predetermined variable belongs in the manufacturing process is created, and candidates for the specific process variables to be analyzed as candidates for explanatory variables are verified and narrowed down depending on whether or not there is a causal relationship between the predetermined variable as a target variable and candidate explanatory variable processes as combinations of the variables of the plurality of types of specific processes.

7. The data analysis device according to claim 3 or 6.

8. The explanatory variable candidate verification unit Combinations of variables of multiple types of specific processes that belong to other groups in a predetermined range upstream and downstream of a predetermined group to which the predetermined variable belongs in the manufacturing process are created, and candidates for variables of the specific processes to be analyzed as candidates for explanatory variables are verified and narrowed down depending on whether or not there is a causal relationship between the predetermined variable as a target variable and the combination of variables of the multiple types of specific processes.

7. The data analysis device according to claim 5 or 6.

9. The analysis unit A model is generated using the candidate variable to be analyzed and one or more other variables included in the group to which the predetermined variable belongs as explanatory variables, and the predetermined variable as a response variable. The data analysis device according to claim 1 .

10. a variable group management step in which a variable group management unit divides a plurality of variables related to the manufacturing process into a plurality of groups and manages them; an inter-group causal relationship management step in which an inter-group causal relationship management unit manages inter-group causal relationship information, which is information about the causal relationships between the plurality of groups; an inter-variable causal relationship management step in which an inter-variable causal relationship management unit manages variable causal relationship information, which is information regarding the presence or absence of a causal relationship between a predetermined variable and one or more other variables; a candidate selection step in which a candidate selection unit identifies one or more specific groups related to the group to which the predetermined variable belongs based on the inter-group causal relationship information, and identifies candidates for variables to be analyzed based on information on the identified specific groups and the variable causal relationship information; an analysis step in which an analysis unit generates a model using at least the candidates for the variables to be analyzed as explanatory variables and the predetermined variable as a response variable; A data analysis method comprising:

Images