Determination device, determination method, and determination program

Abstract

The present invention realizes high-precision factor analysis. The determination device of the present application is characterized by including: an estimation unit that estimates the correlation between the outliers measured in each variable category affecting product quality in each industrial process and the quality of the product; a determination unit that determines the variable category associated with quality based on the comparison result of the variable category list that selects the variable categories having a correlation and the variable category list provided by the user; a generation unit that generates a quality prediction model for the product based on multivariate analysis; and a decision unit that uses the quality prediction model generated by the generation unit to determine the variable category associated with quality.

Description

Technical Field

[0001] The present invention relates to a determination device, a determination method, and a determination program. Background Art

[0002] Conventionally, in various industrial processes, in order to determine factor categories (hereinafter sometimes referred to as variable categories) that affect the quality of a product, a method using multivariate analysis techniques has been known.

[0003] Prior Art Documents

[0004] Patent Documents

[0005] Patent Document 1: Japanese Patent Laid-Open No. 2006-323523 Summary of the Invention

[0006] Problems to be Solved by the Invention

[0007] However, in the above prior art, in various industrial processes, it is sometimes impossible to correctly determine the variable categories that affect the quality of the product. For example, it is generally known that outliers have a great influence on the analysis results. Here, an outlier refers to a value that deviates from a set composed of a main data group. For example, in the case where a small number of outliers are generated in both the data of one variable category and the data of a quality index representing the quality of the product, it may be easily determined that the one variable category affects the quality.

[0008] In addition, in the case of generating a quality prediction model for predicting the quality of a product from data including outliers, since the quality prediction model is generally affected by outliers, it is possible to generate a quality prediction model that outputs results lacking in correctness. That is, the quality prediction model may output a variable category different from the variable category that originally affects the quality of the product.

[0009] On the other hand, a method is known in which, before generating a quality prediction model, a process of removing outliers included in the data of each variable category (hereinafter sometimes referred to as an outlier process) is performed, thereby reducing the influence of outliers on the quality prediction model.

[0010] However, in the case of performing an outlier process on the data of all variable categories, it is possible to remove variable categories that originally affect the quality due to the outlier process. Thus, in various industrial processes, it is sometimes difficult to correctly determine variable categories having a correlation with quality.

[0011] The present application has been completed to solve such problems, and an object thereof is to achieve highly accurate factor analysis.

[0012] Technical Means for Solving the Problems

[0013] The determining device of the present application is characterized by comprising: an estimation unit that estimates the correlation between the outliers measured in each variable category affecting the quality of a product in each industrial process and the quality of the product; a determination unit that determines the variable categories associated with quality based on the comparison result between the variable category list selecting the variable categories having the correlation and the variable category list provided by the user; a generation unit that generates a quality prediction model of the product based on multivariate analysis; and a decision unit that uses the quality prediction model generated by the generation unit to decide the variable categories associated with quality.

[0014] In the above determining device, the variable categories common to the variable category list of the variable categories having the correlation and the variable category list provided by the user are determined as the variable categories associated with quality.

[0015] In the above determining device, the variable category list provided by the user is a list of variable categories having the correlation with the quality of the product preselected by the user.

[0016] In the above determining device, for each variable category affecting the quality of the product in each industrial process, an outlier process for removing outliers is performed, and based on the execution result, a quality prediction model of the product is generated.

[0017] In the above determining device, an outlier is a value measured in a variable category and is a value deviated from the set composed of values included within a specified range.

[0018] Effect of the Invention

[0019] According to the above determining device, the correlation between the outliers measured in each variable category affecting the quality of the product in each industrial process and the quality of the product is estimated, and the variable categories associated with quality are determined based on the comparison result between the variable category list selecting the variable categories having the correlation and the variable category list provided by the user. Then, according to the above determining device, a quality prediction model of the product is generated based on multivariate analysis, and the variable categories associated with quality are decided using the generated quality prediction model. Thus, the determining device can achieve high-precision factor analysis. Description of the Drawings

[0020] Figure 1 It is a diagram showing a configuration example of the determining system of the embodiment.

[0021] Figure 2 It is a diagram showing an example of the data storage unit of the embodiment.

[0022] Figure 3 It is a diagram showing an example of the variable category list storage unit of the embodiment.

[0023] Figure 4 It is a conceptual diagram of the estimation process executed by the determination device of the embodiment.

[0024] Figure 5 It is a conceptual diagram of the generation process executed by the determination device of the embodiment.

[0025] Figure 6 It is a flowchart showing an example of the process of the determination process executed by the determination device of the embodiment.

[0026] Figure 7 It is a flowchart showing an example of the process of the decision process executed by the determination device of the embodiment. Detailed Embodiment

[0027] Next, the embodiments will be described with reference to the drawings. In the following description, the same reference numerals are given to the constituent elements common to the respective embodiments, and redundant descriptions are omitted.

[0028] [Principle]

[0029] Among the causes of outliers included in the data of various industrial processes, there are causes not associated with quality (for example, anomalies in the values represented by the information detected by sensor 10, etc.) and causes associated with quality (for example, anomalies in industrial processes that rarely occur, etc.). Among the causes not associated with quality, if the variable category of the data containing outliers is determined to be a variable category associated with quality, a variable category different from the true variable category may sometimes be determined.

[0030] On the other hand, it is required to be able to detect early data containing outliers caused by reasons associated with quality. However, outliers (quality anomaly data) known to have a correlation with quality are only extremely rarely generated. Therefore, it is difficult to distinguish between outliers caused by reasons associated with quality and outliers caused by reasons not associated with quality, and it is difficult to determine the cause of outlier generation based on only a small number of data. In response to such a problem, a solution using the knowledge of various variable categories related to industrial processes possessed by the user has been conceived. Thus, in the present disclosure, on the basis of estimating the correlation between outliers and quality, a method of determining the variable category of data containing outliers having a correlation with quality and a method of using a quality prediction model to determine the variable category having a correlation with quality are combined to solve the above problem.

[0031] [Embodiment]

[0032] [1. Outline of the Embodiment]

[0033] First, the outline of the process executed by the determination device of the embodiment will be described. The determination device executes a determination process and a decision process.

[0034] First, the determination process performed by the determination device will be described. For example, the determination device determines whether the data of each variable category contains an outlier. Thus, the determination device can determine the variable category of the data containing the outlier. Here, the so-called variable category refers to information detected by sensors, information measured by various devices, information related to the category of physical quantities, information indicating the trend of data containing outliers, and the like. Below, an example where the variable category is information detected by various sensors will be described.

[0035] Furthermore, the determination device compares the variable category list after listing the determined variable categories with the variable category list after listing the variable categories that have a relevance to quality and are variable categories based on the knowledge of various variable categories related to the industrial process that the user has and are selected based on the user's knowledge. In addition, below, as an example of the industrial process, the chemical industrial process will be described.

[0036] Then, the determination device determines the variable category having a relevance to quality based on the comparison result. Thus, the determination device can accurately determine the factor associated with quality abnormality.

[0037] Next, the decision process performed by the determination device will be described. For example, the determination device determines whether the data of each variable category contains an outlier. Then, when the determination device determines that the data of each variable category contains an outlier, it performs outlier processing on the data of each variable category.

[0038] Furthermore, the determination device generates a quality prediction model based on the prior art related to multivariate analysis. In this case, the determination device generates a quality prediction model based on the data of each variable category on which outlier processing has been performed. Then, the determination device uses the generated quality prediction model to decide the variable category having a relevance to quality. Thus, the determination device can appropriately decide the variable category based on the data on which outlier processing has been performed and that has a relevance to quality.

[0039] [2. Configuration of the Determination System]

[0040] Use Figure 1 An example of the configuration of the determination system 1 will be described. Figure 1 is a diagram showing an example of the configuration of the determination system of the embodiment. In Figure 1 In the example, the determination system 1 includes a sensor 10, an administrator terminal 20, and a determination device 100. The sensor 10, the administrator terminal 20, and the determination device 100 can be communicably connected via the network N by wire or wirelessly. In addition, in Figure 1In the determination system 1 shown, multiple sensors 10, multiple manager terminals 20, and multiple determination devices 100 may also be included.

[0041] The sensor 10 of the embodiment is a sensor that detects various information. For example, the sensor 10 is a sensor that detects temperature, humidity, pressure, volume, flow rate, flow, liquid level, concentration of a reaction solution, concentration of a specified gas, pH, redox potential, resistance, or conductivity, etc.

[0042] The manager terminal 20 of the embodiment is an information processing device used by a manager who manages an industrial process. For example, the manager terminal 20 is a desktop PC (Personal Computer), a notebook PC, a tablet terminal, a mobile phone, a PDA (Personal Digital Assistant), etc.

[0043] The determination device 100 of the embodiment is an information processing device capable of communicating with various devices via the network N, and is implemented, for example, by a server device or a cloud system, etc. For example, the determination device 100 is communicably connected to other various devices via the network N.

[0044] [3. Configuration of the determination device]

[0045] Hereinafter, an example of the functional configuration of the above-described determination device 100 will be described. As Figure 1 shown, the determination device 100 of the embodiment includes a communication unit 110, a storage unit 120, and a control unit 130.

[0046] (Regarding the communication unit 110)

[0047] The communication unit 110 is implemented, for example, by a NIC (Network Interface Card), etc. And the communication unit 110 is connected to the network N by wire or wirelessly, and information is transmitted and received between it and other various devices.

[0048] (Regarding the storage unit 120)

[0049] The storage unit 120 is implemented, for example, by semiconductor storage elements such as RAM (Random Access Memory), flash memory, or storage devices such as hard disks and optical discs. The storage unit 120 includes a data storage unit 121, a variable category list storage unit 122, and a quality prediction model 123.

[0050] (Regarding the data storage unit 121)

[0051] The data storage unit 121 stores various data of each variable category. Here, Figure 2An example of the data storage unit 121 of the embodiment is shown. In Figure 2 In the example shown, the data storage unit 121 has items such as "Dataset ID (Identifier)", "Variable Category ID", "Data", "Date and Time", and "Quality Index".

[0052] The "Dataset ID" is an identifier that identifies a dataset, which is a set of data. The "Variable Category ID" is an identifier that identifies a variable category. The "Data" is data corresponding to the "Variable Category ID". The "Date and Time" is information about the date and time when the data associated with the "Data ID" was acquired. The "Quality Index" is information about a quality index that represents the quality of the product corresponding to the "Data ID".

[0053] For example, in Figure 2 "D1" identified by the dataset ID has a variable category ID of "VT1", data of "DA1", a date and time of "DT1", and a quality index of "QI1".

[0054] In addition, in Figure 2 In the example shown, data, etc. are represented by abstract symbols such as "DA1", but data, etc. can also be specific numerical values, specific character strings, file forms of files containing various information representing data, etc.

[0055] (Regarding the variable category list storage unit 122)

[0056] The variable category list storage unit 122 stores various information related to the variable category list. Here, Figure 3 An example of the variable category list storage unit 122 of the embodiment is shown. In Figure 3 In the example shown, the variable category list storage unit 122 has items such as "Variable Category ID", "Relevance to Quality", and "Relevance to Quality Represented by the Variable Category List Provided by the User".

[0057] The "Variable Category ID" is an identifier that identifies a variable category. The "Relevance to Quality" is information about the relevance between outliers included in the data of the variable category corresponding to the "Variable Category ID" and the quality of the product.

[0058] For example, in the item of "Relevance to Quality", "1" is stored when there is a relevance to quality. In addition, in the item of "Relevance to Quality", "0" is stored when there is no relevance to quality.

[0059] "Relevance to quality represented by a list of variable categories provided by the user" is information on the relevance between the variable categories selected based on the user's knowledge corresponding to "variable category ID" and the quality of the product.

[0060] For example, in the item of "Relevance to quality represented by a list of variable categories provided by the user", if there is relevance to quality, "1" is stored. Additionally, in the item of "Relevance to quality represented by a list of variable categories provided by the user", if there is no relevance to quality, "0" is stored.

[0061] For example, in Figure 3 "VT1" identified by the variable category ID has a relevance to quality of "1", and the relevance to quality represented by the list of variable categories provided by the user is "1".

[0062] (Regarding the control unit 130)

[0063] The control unit 130 is a controller, which is implemented, for example, by a CPU (Central Processing Unit), an MPU (Micro Processing Unit), etc., using the RAM as a working area to execute various programs (an example of a determination program) stored in the storage device inside the determination device 100. Additionally, the control unit 130 is a controller, which is implemented, for example, by an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).

[0064] As Figure 1 shown, the control unit 130 has an acquisition unit 131, a determination unit 132, an estimation unit 133, a determination unit 134, a generation unit 135, a decision unit 136, and a transmission unit 137, and realizes or executes the functions and actions of the information processing described below. Additionally, the internal structure of the control unit 130 is not limited to Figure 1 the structure shown, and as long as it is a structure for performing the information processing described later, it can also be other structures. Additionally, the connection relationship between the respective processing units of the control unit 130 is not limited to Figure 1 the connection relationship shown, and it can also be other connection relationships.

[0065] (Regarding the acquisition unit 131)

[0066] The acquisition unit 131 acquires various types of information. Specifically, the acquisition unit 131 acquires data of each variable category detected by the sensor 10. For example, the acquisition unit 131 acquires the characteristic quantities of the data of each variable category detected by the sensor 10. Here, the characteristic quantity of the data refers to statistical values such as the average value and variance, etc.

[0067] For example, the acquisition unit 131 acquires the data of each variable category, the information about the date and time when the data is acquired, and the information about the quality index of the product corresponding to the data. Then, the acquisition unit 131 stores the various types of information acquired in the data storage unit 121.

[0068] In addition, the acquisition unit 131 may also acquire the raw data detected by the sensor 10 as data. In this case, the acquisition unit 131 may also calculate the characteristic quantities of the data based on the acquired raw data. For example, the acquisition unit 131 may calculate the average value, variance, etc. of the data of each variable category based on the acquired raw data. Then, the acquisition unit 131 may store the calculated average value, variance, etc. in the item of "data" in the data storage unit 121.

[0069] Furthermore, the acquisition unit 131 acquires information about the variable category list from the user. For example, assume that the user is a technician related to an industrial process and has proficient skills, knowledge, etc. In this case, the acquisition unit 131 acquires information related to the variable category list obtained by listing the variable categories selected based on the user's knowledge from the user terminal used by the user, etc. Then, the acquisition unit 131 stores the variable category list in the variable category list storage unit 122.

[0070] (Regarding the determination unit 132)

[0071] The determination unit 132 determines whether the data of various variable categories contains outliers. Here, use Figure 4 To explain the determination process performed by the determination unit 132. Figure 4 It is a conceptual diagram of the estimation process performed by the determination device of the embodiment. In addition, assume that Figure 4 Outliers are included in the variable categories VT11 to VT14 shown.

[0072] In Figure 4 In the example of, the horizontal axis of the graph FI11 is the data of the variable category VT11, and the vertical axis is the quality index corresponding to the data of the variable category VT11. In Figure 4 In the example of, the graphs FI11 to FI14 respectively show the correlation diagrams between each variable category VT11 to VT14 and the quality index.

[0073] In Figure 4In the example, the graph FI11 represents a data set composed of main data (hereinafter sometimes referred to as data set AS11) and a data set composed of two data located at positions far from the main data (hereinafter sometimes referred to as data set AS12). In this case, the determination unit 132 uses an existing technique such as DBSCAN (Density-based spatial clustering of applications with noise) to detect the data set AS12 as an outlier. Then, the determination unit 132 determines that the data of the variable category VT11 contains an outlier and the quality index contains an outlier (quality abnormal data).

[0074] In addition, the horizontal axis of the graph FI12 is the data of the variable category VT12, and the vertical axis is the quality index corresponding to the data of the variable category VT12. In this case, for the data set represented by the graph FI12, the determination unit 132 uses DBSCAN to detect outliers. Then, the determination unit 132 determines that the data of the variable category VT12 contains an outlier and the quality index contains an outlier.

[0075] In addition, the horizontal axis of the graph FI13 is the data of the variable category VT13, and the vertical axis is the quality index corresponding to the data of the variable category VT13. In this case, for the data set represented by the graph FI13, the determination unit 132 uses DBSCAN to detect outliers. Then, the determination unit 132 determines that the data of the variable category VT13 contains an outlier and the quality index contains an outlier.

[0076] In addition, the horizontal axis of the graph FI14 is the data of the variable category VT14, and the vertical axis is the quality index corresponding to the data of the variable category VT14. In this case, for the data set represented by the graph FI14, the determination unit 132 uses DBSCAN to detect outliers. Then, the determination unit 132 determines that the data of the variable category VT14 contains an outlier and the quality index contains an outlier.

[0077] (Regarding the estimation unit 133)

[0078] The estimation unit 133 estimates the correlation between the outliers measured in each variable category that affects the quality of the product in each industrial process and the quality of the product. Here, Figure 4 An explanation of the estimation process performed by the estimation unit 133 is given.

[0079] For example, the estimation unit 133 estimates the correlation between the outliers of the variable category and the quality based on a specified criterion. For example, the specified criterion is that when all the data determined to be quality abnormal by the determination unit 132 are outliers, it is estimated that the variable category is related to the quality. In this case,Figure 4 In the example, the presumption unit 133 presumes that the variable category VT11 is related to quality. In addition, the presumption unit 133 presumes that the variable category VT12 has no relation to quality. In addition, the presumption unit 133 presumes that the variable category VT13 is related to quality. In addition, the presumption unit 133 presumes that the variable category VT14 has no relation to quality.

[0080] In this way, as shown in Table TA11, the presumption unit 133 presumes the relation to quality for each variable category. For example, in Table TA11, since the variable category VT11 has a relation to quality, "1" is indicated in the item of the relation to quality. In addition, since the variable category VT12 has no relation to quality, "0" is indicated in the item of the relation to quality. In addition, since the variable category VT13 has a relation to quality, "1" is indicated in the item of the relation to quality. In addition, since the variable category VT14 has no relation to quality, "0" is indicated in the item of the relation to quality. Then, the presumption unit 133 stores the information on the relation between the presumed variable category and quality in the variable category list storage unit 122.

[0081] In addition, in the above example, an example is described in which when all the data determined to be quality anomalies are outliers, it is presumed that this variable category is related to quality, but it is not limited to this. For example, it may be presumed that this variable category is related to quality when a specified number of the multiple data with quality anomalies are outliers.

[0082] (Regarding the determination unit 134)

[0083] The determination unit 134 determines the variable category related to quality based on the comparison result between the variable category list selecting the variable category with a relation and the variable category list provided by the user.

[0084] For example, the determination unit 134 determines the variable category that is common to the variable category list of the variable category having a relation to quality and the variable category list selected based on the user's knowledge as the variable category related to quality. In Figure 3In the example, among the correlations between the variable categories and quality estimated by the estimation unit 133, for "VT1" identified by the variable category ID, it is "1", for "VT2" it is "0", and for "VT3" it is "1". Additionally, among the correlations between the variable categories and quality represented by the variable category list provided by the user, for "VT1" it is "1", for "VT2" it is "0", and for "VT3" it is "0". In this case, the determination unit 134 determines the variable category ID "VT1" that is common between the correlation between the variable categories and quality estimated by the estimation unit 133 and the correlation between the variable categories and quality represented by the variable category list provided by the user as the variable category associated with quality.

[0085] (Regarding the generation unit 135)

[0086] The generation unit 135 performs an outlier process for removing outliers for each variable category that affects the quality of the product in each industrial process. Then, the generation unit 135 generates a quality prediction model 123 of the product based on the execution result. In this case, the generation unit 135 generates the quality prediction model 123 based on multivariate analysis.

[0087] Here, use Figure 5 to explain the generation process performed by the generation unit 135. Figure 5 is a conceptual diagram of the generation process executed by the determination device of the embodiment. Additionally, assume that Figure 5 in the data of the variable category VT21 shown, there are outliers.

[0088] In Figure 5 the example, the horizontal axis of the graph FI21 is the data of the variable category VT21, and the vertical axis is the quality index corresponding to the data of the variable category VT21. In Figure 5 the example, assume that the determination unit 132 detects the data set AS21 as an outlier. In this case, the generation unit 135 performs an outlier process to remove the data set AS21.

[0089] To give a more specific example, the generation unit 135 performs an outlier process such as deleting the data set AS21 or replacing the data included in the data set AS21 with the average value calculated based on the data of the variable category VT21. Then, the generation unit 135 performs the outlier process for other variable categories in the same way as for the variable category VT21.

[0090] Then, the generation unit 135 generates the quality prediction model 123 based on the data of each variable category from which outliers have been removed, using existing techniques related to multivariate analysis. For example, the generation unit 135 uses regression analysis to generate the quality prediction model 123 based on the data of each variable category from which outliers have been removed. Then, the generation unit 135 stores this quality prediction model 123 in the storage unit 120.

[0091] (Regarding determination unit 136)

[0092] The determination unit 136 uses the quality prediction model 123 generated by the generation unit 135 to determine the variable categories associated with quality. For example, the determination unit 136 uses the quality prediction model 123 stored in the storage unit 120 to determine the variable categories having an association with quality in a data set composed of data of each variable category.

[0093] (Regarding transmission unit 137)

[0094] The transmission unit 137 transmits various information. Specifically, the transmission unit 137 transmits information about the variable categories associated with quality to the manager terminal 20. For example, it is assumed that a set of data sets composed of data of each variable category is specified by the manager who manages the manager terminal 20. In this case, the transmission unit 137, according to the acquisition request of the manager terminal 20, transmits information related to the result obtained from the determination process and the decision process performed on the data set to the manager terminal 20.

[0095] [4. Processing steps (1) Determination process]

[0096] Next, use Figure 6 To describe the steps of the determination process executed by the determination device 100 of the embodiment. Figure 6 It is a flowchart showing an example of the flow of the determination process executed by the determination device 100 of the embodiment.

[0097] As Figure 6 shown, the acquisition unit 131 acquires data (step S101). For example, the acquisition unit 131 acquires the feature amounts of the data of each variable category detected by the sensor 10. In addition, the acquisition unit 131 also acquires information about the date and time when the data of each variable category is acquired, and information about the quality index of the product corresponding to the data of each variable category.

[0098] Then, the determination unit 132 determines whether the data of one variable category contains outliers (step S102). For example, it is assumed that the data of the variable category VT31 and the data set DS1 of the quality index contain outliers related to quality anomalies, and the data of the variable category VT32 and the data set DS2 of the quality index do not contain outliers related to quality anomalies. In this case, the determination unit 132 uses DBSCAN to detect outliers in the variable category VT31 and the data set DS1 of the quality index. At this time, the determination unit 132 determines that the data set DS1 contains outliers related to quality anomalies.

[0099] On the other hand, the determination unit 132 uses DBSCAN to detect outliers in the data set DS2. In this case, no outliers are detected in the data set DS2, so the determination unit 132 determines that the data set DS2 does not contain outliers.

[0100] In Figure 6 the example of, when the determination unit 132 determines that the data of a variable category does not contain outliers (step S102: No), the generation unit 135 skips the process of performing outlier processing. On the other hand, when the determination unit 132 determines that the data of a variable category contains outliers (step S102: Yes), the generation unit 135 performs outlier processing (step S103). For example, assume that the data set AS21 is detected as an outlier by the determination unit 132. In this case, the generation unit 135 performs outlier processing to remove the data set AS21.

[0101] To give a more specific example, the generation unit 135 deletes the data set AS21, or replaces the data included in the data set AS21 with an average value, etc., thereby performing outlier processing.

[0102] Then, regardless of whether outlier processing is performed or not, the generation unit 135 temporarily holds the data of the variable category in the storage unit 120 (step S104). Then, the determination device 100 repeats the above process for all the data of the variable categories (step S105). Specifically, when the determination device 100 determines that there is other acquired data (step S105: Yes), it repeats the above process.

[0103] Then, when the estimation unit 133 determines that there is no other acquired data (step S105: No), it estimates the relevance between the outliers of each variable category and the quality (step S106). For example, the estimation unit 133 estimates the relevance between the outliers of the variable category based on a specified criterion. For example, the specified criterion is that when all the data determined to be of abnormal quality are outliers, it is estimated that this variable category is relevant to the quality. In this case, the estimation unit 133 estimates whether there is a relevance to the quality for each variable category.

[0104] Then, the determination unit 134 determines the variable categories related to the quality based on the comparison result between the list of variable categories with relevance and the list of variable categories provided by the user (step S107).

[0105] For example, assume that in the relevance between the variable category and quality deduced by the deduction unit 133, for "VT1" identified by the variable category ID, it is "1", for "VT2" it is "0", and for "VT3" it is "1". Additionally, assume that in the relevance between the variable category and quality represented by the variable category list provided by the user, for "VT1" it is "1", for "VT2" it is "0", and for "VT3" it is "0". In this case, the determination unit 134 determines the variable category ID "VT1" that is common in the relevance between the variable category and quality deduced by the deduction unit 133 and the relevance between the variable category and quality represented by the variable category list provided by the user as the variable category associated with quality.

[0106] [5. Processing Step (2) Decision Processing]

[0107] Next, use Figure 7 to illustrate the steps of the decision processing executed by the determination device 100 of the embodiment. Figure 7 is a flowchart showing an example of the process of the decision processing executed by the determination device 100 of the embodiment.

[0108] As Figure 7 shown, the acquisition unit 131 acquires data (step S201). Then, the determination unit 132 determines whether the data of a variable category contains outliers (step S202). Specifically, when the determination unit 132 determines that the data of a variable category does not contain outliers (step S202: No), the generation unit 135 skips the process of performing outlier processing.

[0109] On the other hand, when the determination unit 132 determines that the data of a variable category contains outliers (step S202: Yes), the generation unit 135 performs outlier processing (step S203). For example, assume that the data set AS21 is detected as an outlier by the determination unit 132. In this case, the generation unit 135 performs outlier processing to remove the data set AS21.

[0110] To give a more specific example, the generation unit 135 deletes the data set AS21, or replaces the data contained in the data set AS21 with an average value, etc., thereby performing outlier processing.

[0111] Then, regardless of whether there is outlier processing, the generation unit 135 temporarily stores the data of the variable category in the storage unit 120 (step S204). Then, the determination device 100 repeats the above process for the data of all variable categories (step S205). Specifically, when the determination device 100 determines that there is other acquired data (step S205: Yes), it repeats the above process.

[0112] Then, when it is determined that there is no other acquired data (step S205: No), the generation unit 135 generates the quality prediction model 123 based on multivariate analysis (step S206). For example, the generation unit 135 uses regression analysis to generate the quality prediction model 123 according to the data of each variable category with outliers removed.

[0113] Then, the determination unit 136 uses the quality prediction model 123 to determine the variable category associated with quality (step S207). For example, the determination unit 136 uses the quality prediction model 123 stored in the storage unit 120 to determine the variable category having an association with quality in a set of data composed of the data of each variable category.

[0114] [6. Effect]

[0115] As described above, the determination device 100 of the embodiment uses DBSCAN to detect outliers in the data of variable categories. For example, when the determination device 100 detects an outlier in the data of a variable category, it determines that the data of that variable category contains an outlier.

[0116] Then, the determination device 100 estimates the association between the outliers measured in each variable category that affects the quality of the product in each industrial process and the quality of the product. For example, when it is determined that all the data with abnormal quality are outliers, the determination device 100 estimates that the outliers of the variable category are associated with quality. Thus, the determination device 100 can appropriately estimate the association between the variable category and quality.

[0117] Furthermore, the determination device 100 determines the variable category associated with quality based on the comparison result between the variable category list selecting the variable categories having an association and the variable category list provided by the user.

[0118] For example, the determination device 100 compares the variable category list after the determined variable categories are listed, and the variable category list after the variable categories having an association with quality and being selected based on the user's knowledge, which are various variable categories related to the industrial process that the user has, are listed. Then, the determination device 100 determines the variable category having an association with quality based on the comparison result. Thus, the determination device 100 can accurately determine the factor associated with quality abnormality.

[0119] In addition, when the determination device 100 determines that the data of each variable category contains an outlier, it performs outlier processing on the data of each variable category. Then, the determination device 100 generates a quality prediction model based on multivariate analysis. In this case, the determination device 100 generates a quality prediction model based on the data of each variable category on which outlier processing has been performed.

[0120] Furthermore, the determination device 100 uses the generated quality prediction model to determine the variable categories having a correlation with quality. Thus, the determination device 100 can appropriately determine the variable categories based on the data subjected to the outlier processing and that are variable categories having a correlation with quality. Therefore, the determination device 100 of the embodiment can achieve high-precision factor analysis.

[0121] [7. Modification Example]

[0122] In addition to the above-described embodiment, the determination device 100 can be implemented in various different ways. Therefore, other embodiments of the determination device 100 will be described below.

[0123] [7-1. Application Example]

[0124] Examples of performing the processing shown in the above-described embodiment in a chemical industrial process have been illustrated, but are not limited thereto. For example, the processing shown in the above-described embodiment can be performed in any industrial process. For example, the processing shown in the above-described embodiment can also be performed in a metal manufacturing process, a metal processing process, a semiconductor manufacturing process, a pharmaceutical manufacturing process, a medical device manufacturing process, and the like.

[0125] [7-2. Feature Quantities of Data]

[0126] In the above-described embodiment, examples in which the determination unit 132, the estimation unit 133, the determination unit 134, and the generation unit 135 of the determination device 100 perform various information processes on the feature quantities of data have been illustrated, but are not limited thereto. For example, the determination unit 132, the estimation unit 133, the determination unit 134, and the generation unit 135 can also perform various information processes on the raw data.

[0127] [7-3. Correlation with Quality]

[0128] In the above-described embodiment, as Figure 3 shown, the correlation with quality and the correlation with quality represented by the variable category list provided by the user are expressed as "1" or "0", but are not limited thereto. For example, the correlation with quality and the correlation with quality represented by the variable category list provided by the user can also be the degree of correlation with quality. In this case, in the variable category list storage unit 122, information related to the score representing the correlation with quality can also be stored as the degree of correlation with quality.

[0129] In addition, when estimating the correlation between the outlier generated in each variable category and quality, the estimation unit 133 can also output a score representing the correlation between the outlier generated in each variable category and quality.

[0130] [7-4. Combination of Output Results]

[0131] Furthermore, the decision unit 136 of the determination device 100 can also combine the variable categories determined by the determination unit 134 and the variable categories determined by using the quality prediction model 123, or determine whether the variable categories are the same variable category, thereby determining the variable category. Thus, the decision unit 136 can achieve a higher-precision factor analysis.

[0132] [8. Others]

[0133] In addition, all or part of the processes described as automatically performed in each of the processes described in the above embodiments and modification examples can also be performed manually, or all or part of the processes described as manually performed can be automatically performed by a known method. In addition, information including the processing steps, specific names, various data, or parameters shown in the above documents and drawings can be arbitrarily changed except in the case of special records. For example, the various information shown in each figure is not limited to the information shown in the figure.

[0134] In addition, each component of each device shown in the figure is a functional conceptual component, and it is not necessary to be physically configured as shown in the figure. That is, the specific form of dispersion and integration of each device is not limited to the way shown in the figure, and all or part of it can be functionally or physically dispersed and integrated in any unit according to various loads, usage conditions, etc.

[0135] In addition, the above embodiments and modification examples can be appropriately combined within the scope where the processing contents do not conflict.

[0136] In addition, the above "section, module, unit" can be replaced with "method" or "circuit", etc. For example, the determination unit can be replaced with a determination method or a determination circuit.

[0137] The above several embodiments of the present application have been described in detail based on the drawings, but these are only examples, and the present invention can be implemented in other ways with various deformations and improvements based on the knowledge of those skilled in the art, starting from the manner described in the disclosure column of the invention.

[0138] Symbol Explanation

[0139] N Network

[0140] 1 Determination System

[0141] 10 Sensor

[0142] 20 Manager Terminal

[0143] 100 Determination Device

[0144] 110 Communication Unit

[0145] 120 Storage Unit

[0146] 121 Data Storage Unit

[0147] 122 Variable Category List Storage Unit

[0148] 123 Quality Prediction Model

[0149] 130 Control Unit

[0150] 131 Acquisition Unit

[0151] 132 Judgment Unit

[0152] 133 Estimation Unit

[0153] 134 Determination Unit

[0154] 135 Generation Unit

[0155] 136 Decision Unit

[0156] 137 Transmission Unit.

Claims

1. A determination device, characterized in that, it comprises: a determination unit that determines whether data of various variable categories contains outliers and determines whether the outliers are quality abnormal data; an estimation unit that estimates the correlation between the outliers measured for each variable category affecting the quality of the product in each industrial process and the quality of the product; a determination unit that determines the variable categories common to the variable category list of the variable categories having the correlation and the variable category list provided by the user as the variable categories associated with quality; a generation unit that generates a quality prediction model for the product based on multivariate analysis; and a decision unit that uses the quality prediction model generated by the generation unit to determine the variable categories associated with the quality, when the determination unit determines that all the quality abnormal data of a variable category are outliers, the estimation unit estimates that the variable category is associated with quality.

2. The determination device according to claim 1, characterized in that, the variable category list provided by the user is a list of variable categories preselected by the user having a correlation with the quality of the product.

3. The determination device according to claim 1, characterized in that, the generation unit performs outlier processing to remove outliers for each variable category affecting the quality of the product in each industrial process, and generates the quality prediction model for the product based on the execution result.

4. The determination device according to claim 1, characterized in that, the outlier is a value measured in a variable category and is a value that deviates from a set composed of values included within a specified range.

5. A determination method, which is a determination method executed by a computer, characterized in that it includes: a determination process that determines whether data of various variable categories contains outliers and determines whether the outliers are quality abnormal data; an estimation process that estimates the correlation between the outliers measured in each variable type affecting the quality of the product in each industrial process and the quality of the product; a determination process that determines the variable categories common to the variable category list of the variable categories having the correlation and the variable category list provided by the user as the variable categories associated with quality; a generation process that generates a quality prediction model for the product based on multivariate analysis; and a decision process that uses the quality prediction model generated by the generation process to determine the variable categories associated with the quality, when the determination process determines that all the quality abnormal data of a variable category are outliers, the estimation process estimates that the variable category is associated with quality.

6. A computer program product, which implements when running on a computer: a determination step that determines whether data of various variable categories contains outliers and determines whether the outliers are quality abnormal data; an estimation step that estimates the correlation between the outliers measured in each variable type affecting the quality of the product in each industrial process and the quality of the product; A determination step that determines, as variable categories associated with quality, the variable categories common to the list of variable categories having the said relevance and the list of variable categories provided by the user; A generation step that generates a quality prediction model for the said product based on multivariate analysis; And A decision step that uses the quality prediction model generated by the said generation step to decide the variable categories associated with the said quality, When the determination step determines that all of the data of quality anomalies of a variable category are outliers, the presumption step presumes that this variable category is associated with quality.

Images