WO2020027207A1

WO2020027207A1 - Abnormality detecting method, information processing device, and abnormality detecting system

Info

Publication number: WO2020027207A1
Application number: PCT/JP2019/030043
Authority: WO
Inventors: 小掠　哲義; 瞳嶺岸; 佐々木　幸紀; 多鹿　陽介
Original assignee: パナソニックＩｐマネジメント株式会社
Priority date: 2018-08-03
Filing date: 2019-07-31
Publication date: 2020-02-06

Abstract

An abnormality detecting method according to one aspect of the present disclosure includes: a step (S10) of acquiring a plurality of items of time-series data obtained from a plurality of elements relating to the manufacture of a manufactured object, and attribute information for each of the plurality of items of time-series data; a step (S11) of analyzing correlations between the plurality of items of time-series data; a step (S12) of correcting the attribute information on the basis of the results of the analysis; a step (S13) of detecting an abnormality occurring in the plurality of elements, on the basis of the corrected attribute information and the plurality of items of time-series data; and a step (S14) of outputting information indicating the detected abnormality and the time at which the abnormality occurred.

Description

Abnormality detection method, information processing apparatus and abnormality detection system

The present disclosure relates to an abnormality detection method, an information processing device, and an abnormality detection system.

Patent Document 1 discloses a method of diagnosing a failure. In the method disclosed in Patent Document 1, when a relative contribution of one or more process variables contributing to a fault is within a relative contribution range of a matching failure sign, the failure sign is associated with the failure. Judge that you are.

Japanese Patent Publication No. 2010-501091

However, according to the above-described conventional method, if data collected for diagnosing a fault includes an error, the fault is diagnosed based on the erroneous data, and thus the fault cannot be detected with high accuracy.

Therefore, the present disclosure provides an abnormality detection method, an information processing device, and an abnormality detection system that can accurately detect an abnormality.

In order to solve the above problem, an abnormality detection method according to an aspect of the present disclosure provides a plurality of time-series data obtained from a plurality of elements related to manufacturing of a product, and attribute information of each of the plurality of time-series data. Acquiring, and analyzing the interrelationship of the plurality of time-series data, correcting the attribute information based on the result of the analysis, corrected attribute information and the plurality of time-series data And a step of outputting information indicating the detected abnormality and the time of occurrence of the abnormality based on the above.

In addition, the information processing apparatus according to an aspect of the present disclosure includes an acquisition unit configured to acquire a plurality of time-series data obtained from a plurality of elements related to manufacturing of a product and attribute information of each of the plurality of time-series data. An analysis unit that analyzes the mutual relationship between the plurality of time-series data, a correction unit that corrects the attribute information based on the result of the analysis, and the corrected attribute information and the plurality of time-series data. A detection unit that detects an abnormality occurring in the plurality of elements based on the detected abnormality, and an output unit that outputs information indicating the detected abnormality and a time when the abnormality occurs.

Further, the abnormality detection system according to an aspect of the present disclosure includes the information processing device, a data storage device for storing a plurality of time-series data and a plurality of attribute information acquired by the acquisition unit, and the output unit. And a presentation device for presenting the information output from.

According to the present disclosure, an abnormality can be detected with high accuracy.

FIG. 1 is a diagram illustrating a configuration of a factory that is a target of the abnormality detection device according to the first embodiment. FIG. 2 is a block diagram illustrating a configuration of the abnormality detection device according to the first embodiment. FIG. 3 is a diagram illustrating an example of time-series data acquired by the abnormality detection device according to the first embodiment. FIG. 4 is a diagram illustrating an example of attribute information acquired by the abnormality detection device according to the first embodiment. FIG. 5 is a flowchart illustrating an operation of the abnormality detection device according to the first embodiment. FIG. 6 is a diagram illustrating an example of a result of the cross-correlation analysis performed by the abnormality detection device according to the first embodiment. FIG. 7 is a flowchart illustrating a process of correcting attribute information in the operation of the abnormality detection device according to the first embodiment. FIG. 8 is a diagram illustrating an example of correction of attribute information by the abnormality detection device according to the first embodiment. FIG. 9 is a diagram illustrating an example of attribute information before correction for each group by the abnormality detection device according to the first embodiment. FIG. 10 is a diagram illustrating an example of an abnormality detection process performed by the abnormality detection device according to the first embodiment. FIG. 11 is a diagram illustrating an example of information output by the abnormality detection device according to the first embodiment. FIG. 12 is a diagram illustrating a configuration example of a manufacturing apparatus determined by cross-correlation analysis by the abnormality detection device according to the modification of the first embodiment. FIG. 13 is a diagram illustrating an example of an abnormality detection process performed by the abnormality detection device according to the modification of the first embodiment. FIG. 14 is a diagram illustrating an example of information output by the abnormality detection device according to the modification of the first embodiment. FIG. 15 is a diagram illustrating a configuration of a plurality of compressors that are targets of the abnormality detection device according to the second embodiment. FIG. 16 is a block diagram illustrating a configuration of the abnormality detection device according to the second embodiment. FIG. 17 is a diagram illustrating an example of attribute information acquired by the abnormality detection device according to the second embodiment. FIG. 18 is a flowchart illustrating the operation of the abnormality detection device according to the second embodiment. FIG. 19 is a diagram illustrating an example of a combination created by the abnormality detection device according to the second embodiment. FIG. 20 is a diagram illustrating an example of an abnormality detection process performed by the abnormality detection device according to the second embodiment. FIG. 21 is a diagram illustrating an example of information output by the abnormality detection device according to the second embodiment. FIG. 22 is a diagram illustrating a configuration of a factory that is a target of the abnormality detection device according to the third embodiment. FIG. 23 is a block diagram illustrating a configuration of the abnormality detection device according to the third embodiment. FIG. 24 is a flowchart illustrating the operation of the abnormality detection device according to the third embodiment. FIG. 25 is a diagram illustrating an example of a result of a multiple regression analysis performed by the abnormality detection device according to the third embodiment. FIG. 26 is a diagram illustrating an example of correction of the attribute information by the abnormality detection device according to the third embodiment. FIG. 27 is a diagram illustrating an example of an abnormality detection process performed by the abnormality detection device according to the third embodiment. FIG. 28 is a diagram illustrating an example of information output by the abnormality detection device according to the third embodiment. FIG. 29 is a diagram illustrating an example of an abnormality detection process performed by the abnormality detection device according to the comparative example. FIG. 30 is a diagram illustrating an example of information output by the abnormality detection device according to the comparative example. FIG. 31 is a diagram illustrating a configuration of a thermal model that is a target of the abnormality detection device according to the modification of the third embodiment. FIG. 32 is a diagram illustrating an example of an abnormality detection process performed by the abnormality detection device according to the modification of the third embodiment. FIG. 33 is a diagram illustrating an example of information output by the abnormality detection device according to the modification of the third embodiment. FIG. 34 is a block diagram illustrating a configuration of an abnormality detection system that executes the abnormality detection method according to the fourth embodiment. FIG. 35 is a block diagram illustrating another example of the configuration of the abnormality detection system that executes the abnormality detection method according to the fourth embodiment. FIG. 36 is a block diagram illustrating a configuration of the abnormality detection system according to the fourth embodiment.

(Summary of the present disclosure)
In order to solve the above-described problem, an abnormality detection method according to an embodiment of the present disclosure provides a plurality of time-series data obtained from a plurality of elements related to manufacturing of a product, and an attribute of each of the plurality of time-series data. Obtaining information, analyzing the interrelationship of the plurality of time-series data, correcting the attribute information based on a result of the analysis, and modifying the attribute information and the plurality of time-series data. Detecting an abnormality occurring in the plurality of elements based on the data; and outputting information indicating the detected abnormality and the time of occurrence of the abnormality.

Thus, even if an error exists in the attribute information, the error can be detected by analyzing the correlation between a plurality of time-series data. For example, when different attribute information is associated with the two time-series data despite the strong relationship between the two time-series data, it is determined that at least one of the attribute information is incorrect. can do. As described above, since an error in the attribute information can be detected, the attribute information can be corrected. Further, since the abnormality is detected based on the corrected attribute information, the abnormality can be detected with high accuracy.

Also, for example, the analysis may be a statistical analysis.

(4) Since an error in the attribute information can be detected with high accuracy, an abnormality can be detected with high accuracy.

Further, for example, the statistical analysis may be a cross-correlation analysis.

Thereby, the mutual relationship (for example, the strength of the relationship) between a plurality of time-series data can be represented by a numerical value such as a cross-correlation coefficient, so that the mutual relationship between the time-series data is easily and accurately determined. be able to. For this reason, an error in the attribute information corresponding to the time-series data can be easily and accurately detected, so that an abnormality can be accurately detected.

Further, for example, in the step of performing the analysis, a cross-correlation coefficient of the plurality of time-series data is calculated, and the plurality of time-series data is classified into a plurality of groups based on the calculated cross-correlation coefficient. In the correcting step, for each classified group, a plurality of time-series data included in the group is classified into a minority and a majority based on corresponding attribute information, and the time-series data belonging to the minority The attribute information may be modified to the attribute information of the time-series data belonging to the majority.

(4) By this, by classifying the time-series data into a minority group and a majority group in the group, it is possible to easily detect an error in the attribute information and easily correct the error.

In particular, the smaller the number of time-series data belonging to the minority is smaller than the number of time-series data belonging to the majority, the higher the possibility that the attribute information of the time-series data belonging to the minority is wrong, and Is more likely to be correct for the attribute information of the time-series data belonging to.

Therefore, for example, in the correcting step, when the number of the time-series data belonging to the minority is less than half the number of the time-series data belonging to the majority, the attribute of the time-series data belonging to the minority The information may be modified to attribute information of the time-series data belonging to the majority.

(4) Thereby, an error in the attribute information can be more easily and accurately detected, so that an abnormality can be detected more accurately.

Further, for example, the statistical analysis may be a multiple regression analysis.

This makes it possible to determine an appropriate model by multiple regression analysis. Therefore, even if there is an error in the correspondence between the time-series data and the attribute information, the error can be accurately detected by comparing with the appropriate model. it can.

In addition, for example, the plurality of elements include a plurality of compressors whose outputs from each are integrated, and in the obtaining step, a plurality of pressure data, which are time-series data of pressures from each of the plurality of compressors, A plurality of flow rate data which is time series data of flow rate, and integrated flow rate data which is time series data of flow rate and integrated pressure data which is time series data of integrated pressure, are obtained as the plurality of time series data, In the step of performing the analysis, a plurality of combinations of the plurality of pressure data and the plurality of flow rate data, and a plurality of attribute information candidates corresponding to each of the plurality of pressure data and the plurality of flow rate data were created and created. For each combination, a plurality of pressure data and a plurality of flow rate data constituting the combination are made independent variables, and the integrated pressure data and the integrated The amount data as the dependent variable, may be performed multiple regression analysis.

This makes it possible to correct an error in the correspondence between the time-series data from the compressor and the attribute information. Therefore, it is possible to accurately detect an abnormality occurring in the compressor.

In addition, for example, the plurality of elements include an air conditioner disposed in each of a plurality of spaces, and a plurality of manufacturing devices each disposed in any one of the plurality of spaces, and the acquisition is performed. In the step, acquiring the time series data of the power consumption of each of the plurality of air conditioners and the time series data of the parameters of each of the plurality of manufacturing apparatuses as the plurality of time series data, and performing the analysis. Then, multiple regression analysis may be performed using the time series data of the parameter as an independent variable and the time series data of the power consumption as a dependent variable.

電力 There is a strong correlation between the power consumption of the air conditioner and the parameters of the device installed in the same space as the space where the air conditioner is installed. According to this aspect, by analyzing the correlation, an error in the correspondence between each device and the installation position can be corrected. By correcting each device and the installation position so as to have a correct correspondence, for example, it is possible to propose a device that can operate in place of the device in which the abnormality has occurred. For example, in a case where a plurality of devices of the same type are present in the same room, even if an abnormality is detected for one device, a reduction in production efficiency is suppressed by performing another operation of another device of the same type. Can be.

For example, in the outputting step, information indicating the cause of the detected abnormality may be output.

(4) As a result, information indicating the cause of the abnormality is output, so that a user (eg, a factory manager or the like) involved in the manufacture of the product can study a countermeasure based on the cause of the abnormality that has occurred. For this reason, for example, maintenance work can be performed promptly, or production of a product can be continued using an alternative device or the like, so that a decrease in production efficiency can be suppressed. As described above, according to the abnormality detection method according to this aspect, it is possible to assist in suppressing a decrease in production efficiency.

In addition, for example, in the outputting step, information indicating a countermeasure for the abnormality, which is determined according to the detected abnormality, may be further output.

(4) As a result, the countermeasure plan is output, so that the user involved in the manufacture of the product can easily determine what action should be taken for the abnormality that has occurred. For this reason, for example, maintenance work can be performed promptly, or production of a product can be continued using an alternative device or the like, so that a decrease in production efficiency can be suppressed. As described above, according to the abnormality detection method according to this aspect, it is possible to assist in suppressing a decrease in production efficiency.

In addition, the information processing apparatus according to an aspect of the present disclosure includes an acquisition unit configured to acquire a plurality of time-series data obtained from a plurality of elements related to manufacturing of a product and attribute information of each of the plurality of time-series data. An analysis unit that analyzes the mutual relationship between the plurality of time-series data, a correction unit that corrects the attribute information based on the result of the analysis, and the corrected attribute information and the plurality of time-series data. A detection unit that detects an abnormality occurring in the plurality of elements, and an output unit that outputs information indicating the detected abnormality and an occurrence time of the abnormality.

This makes it possible to detect an abnormality with high accuracy, similarly to the abnormality detection method described above.

Further, the abnormality detection system according to an aspect of the present disclosure includes the information processing device, a data storage device for storing a plurality of time-series data and a plurality of attribute information acquired by the acquisition unit, and the output unit. And a presentation device for presenting the information output from the device.

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

Note that each of the embodiments described below shows a comprehensive or specific example. Numerical values, shapes, materials, constituent elements, arrangement positions and connection forms of constituent elements, steps, order of steps, and the like shown in the following embodiments are merely examples, and do not limit the present disclosure. Further, among the components in the following embodiments, components not described in the independent claims are described as arbitrary components.

図 Moreover, each drawing is a schematic diagram and is not necessarily strictly illustrated. Further, in each of the drawings, substantially the same configuration is denoted by the same reference numeral, and redundant description will be omitted or simplified.

(Embodiment 1)
Hereinafter, an abnormality detection device that executes the abnormality detection method according to the first embodiment will be described. The abnormality detection device according to the present embodiment detects, for example, an abnormality that occurs in a device included in a manufacturing system in a factory or the like in which a manufacturing system that manufactures a product is installed.

FIG. 1 is a diagram showing a configuration of a factory 1 which is a target of the abnormality detection device 100 according to the present embodiment. In the factory 1, a plurality of devices related to the manufacture of a product are arranged in each of the three rooms A to C. Specifically, as shown in FIG. 1, in a room A, a mounting machine A, a molding machine A, and an air conditioner A are arranged. In the room B, a molding machine B, a coating machine B, a welding machine B, a press machine B, and an air conditioner B are arranged. In the room C, a press C and an air conditioner C are arranged.

The mounting machine A, the molding machines A and B, the coating machine B, the welding machine B, the press machines B and C, and the air conditioners A to C are each an example of a plurality of elements related to the manufacture of a product. The plurality of elements include elements other than the manufacturing apparatus that directly manufactures products, such as the air conditioners A to C. The plurality of elements may include, for example, a repeater that relays communication between the devices. Further, the plurality of elements may include a compressor.

The abnormality detection device 100 acquires a plurality of time-series data obtained from the plurality of elements, and detects an abnormality occurring in the plurality of elements based on the acquired plurality of time-series data. The time-series data represents a temporal change of a numerical value of one or more parameters of each element. The one or more parameters include, for example, power consumption, the number of operations, the operation time, the number of inserted components, the number of output components, and the like. Each of the plurality of elements is provided with a sensor or a measuring device for detecting a numerical value of the parameter. The abnormality detection device 100 is communicably connected by wire or wirelessly to a sensor or a measuring device provided for each of the plurality of elements.

The abnormality detection device 100 is realized by, for example, a computer device. Specifically, the abnormality detection device 100 is realized by a nonvolatile memory in which a program is stored, a volatile memory that is a temporary storage area for executing the program, an input / output port, a processor that executes the program, and the like. You.

FIG. 2 is a block diagram showing a configuration of the abnormality detection device 100 according to the present embodiment. As illustrated in FIG. 2, the abnormality detection device 100 includes an acquisition unit 110, an analysis unit 120, a correction unit 130, a detection unit 140, and an output unit 150. Further, the abnormality detection device 100 includes a storage unit 160 for storing various data and information.

The acquisition unit 110 acquires a plurality of time-series data obtained from a plurality of elements related to manufacturing of a product and attribute information of each of the plurality of time-series data. The acquiring unit 110 is realized by, for example, an input interface or a communication interface to which output data from a sensor or a measuring instrument attached to each device such as the manufacturing apparatus and the air conditioner illustrated in FIG. 1 is input. The acquiring unit 110 acquires a plurality of time-series data by periodically acquiring output data from each sensor or measuring instrument.

The acquisition unit 110 is further realized by a user interface or the like that receives an operation input by an administrator of the factory 1 or the like. The acquisition unit 110 acquires attribute information by receiving an operation input. That is, in the present embodiment, the attribute information is information input by a person such as a manager of the factory 1. For this reason, an error may occur in the correspondence relationship with the time-series data due to a human error or omission in input.

The plurality of time-series data and attribute information acquired by the acquisition unit 110 are stored in the storage unit 160. In the storage unit 160, a plurality of time-series data is managed in a time-series data DB (database) 162, and attribute information is managed in an attribute information DB 164.

The analysis unit 120 analyzes the cross-correlation of the plurality of time-series data acquired by the acquisition unit 110. The analysis unit 120 specifically refers to the time-series data DB 162 stored in the storage unit 160, and analyzes the mutual relationship using the plurality of time-series data read from the storage unit 160. The analysis unit 120 outputs the result of the analysis to the correction unit 130.

In the present embodiment, the analysis unit 120 performs a statistical analysis. Specifically, the analysis unit 120 performs a cross-correlation analysis. More specifically, analysis section 120 calculates a cross-correlation coefficient of a plurality of time-series data, and classifies the plurality of time-series data into a plurality of groups based on the calculated cross-correlation coefficient.

{For example, the analysis unit 120 selects two time series data from the plurality of time series data, and calculates a cross-correlation coefficient of the selected time series data. When N is a natural number of 2 or more, there are N (N-1) / 2 combinations for selecting two time series data from the N pieces of time series data. Therefore, N (N-1) / 2 cross-correlation coefficients are calculated.

The cross-correlation coefficient is represented by a value between −1 and +1. The closer the cross-correlation coefficient is to +1, the stronger the positive correlation between the two time-series data, and the closer the cross-correlation coefficient is to -1, the stronger the negative correlation between the two time-series data. ing. The closer the cross-correlation coefficient is to 0, the closer the two time-series data are to no correlation.

In the present embodiment, the analysis unit 120 classifies the plurality of time-series data into a plurality of groups according to the strength of the correlation by comparing the cross-correlation coefficient with the threshold. Specifically, the analysis unit 120 compares the absolute value of the cross-correlation coefficient with a threshold. The threshold value is, for example, 0.75, but is not limited thereto. A plurality of time-series data in which the absolute value of the cross-correlation coefficient is larger than 0.75 is set as a group having a large correlation.

The correction unit 130 corrects the attribute information based on the result of the analysis by the analysis unit 120. Specifically, the correction unit 130 refers to the attribute information DB 164 stored in the storage unit 160, and uses the plurality of pieces of attribute information read from the storage unit 160 and the analysis result obtained from the analysis unit 120. Modify attribute information. The correction unit 130 reflects the result of the correction in the attribute information DB 164.

In the present embodiment, the correction unit 130 classifies, for each group classified by the analysis unit 120, a plurality of time-series data included in the group into a minority group and a majority group based on corresponding attribute information. . The correction unit 130 corrects the attribute information of the time-series data belonging to the minority into the attribute information of the time-series data belonging to the majority.

Specifically, when the number of time-series data belonging to the minority is equal to or less than half of the number of time-series data belonging to the majority, the correction unit 130 deletes the attribute information of the time-series data belonging to the minority. Modify to the attribute information of the time series data belonging to the faction. For example, when the number of time-series data belonging to the minority is equal to or less than 1/4 of the number of time-series data belonging to the majority, the correction unit 130 determines that the total number of time-series data belonging to the group is equal to or less than 20%. , The attribute information of the time-series data belonging to the minority is corrected. The specific operation of the correction will be described later in detail with an example.

The detection unit 140 detects an abnormality occurring in a plurality of elements based on the corrected attribute information and the plurality of time-series data. For example, the detection unit 140 detects an abnormality for each abnormality to be detected by using a predetermined abnormality detection model or rule. The detection unit 140 refers to the time-series data DB 162, the attribute information DB 164, and the abnormality detection model DB 166 stored in the storage unit 160, and reads necessary data to detect an abnormality. The detection unit 140 outputs the result of the detection of the abnormality to the output unit 150.

異常 Abnormalities to be detected include, for example, abnormalities that occur in a manufacturing apparatus such as a mounting machine, a molding machine, a coating machine, a welding machine or a press machine. Alternatively, the abnormality to be detected may include an abnormality that occurs in an air conditioner such as an air conditioner, or an abnormality that occurs in a space (specifically, a room) in which air conditioning is controlled by the air conditioner. Further, the abnormality to be detected may include an abnormality occurring in the compressor. The specific operation of the abnormality detection will be described later in detail with an example.

The analysis unit 120, the correction unit 130, and the detection unit 140 are each realized by, for example, software executed by a processor. Alternatively, at least one of the analysis unit 120, the correction unit 130, and the detection unit 140 may be realized by hardware such as an electronic circuit including a plurality of circuit elements.

The output unit 150 outputs information indicating the detected abnormality and the time when the abnormality occurs. In the present embodiment, output unit 150 further outputs information indicating the cause of the abnormality or information indicating a countermeasure for the abnormality that is predetermined according to the detected abnormality. Specifically, the output unit 150 refers to the countermeasure information DB 168 stored in the storage unit 160, determines a countermeasure corresponding to the detected abnormality, and outputs information indicating the determined countermeasure.

The output unit 150 is realized by, for example, a display for displaying information, but is not limited thereto. For example, the output unit 150 may be realized by a speaker that outputs information as sound. Alternatively, the output unit 150 may be realized by an output interface or a communication interface for outputting information to an external display, a speaker, or the like.

The storage unit 160 is realized by, for example, a semiconductor memory or a HDD (Hard Disk Drive). In the present embodiment, the storage unit 160 stores various data and information necessary for executing the abnormality detection method in a database. Specifically, as shown in FIG. 2, the storage unit 160 stores a time-series data DB 162, an attribute information DB 164, an abnormality detection model DB 166, and a countermeasure information DB 168.

The time series data DB 162 includes a plurality of time series data acquired by the acquisition unit 110. The time series data DB 162 is generated and updated when the acquisition unit 110 acquires the time series data. A specific example of the time-series data DB 162 will be described later with reference to FIG.

The attribute information DB 164 includes a plurality of pieces of attribute information acquired by the acquisition unit 110. The plurality of pieces of attribute information respectively correspond to the plurality of pieces of time-series data. The attribute information DB 164 is generated and updated when the acquisition unit 110 acquires the attribute information. The attribute information DB 164 is updated by the correction unit 130. A specific example of the attribute information DB 164 will be described later with reference to FIG.

The anomaly detection model DB 166 includes a model or a rule used by the detection unit 140 for anomaly detection. In the abnormality detection model DB 166, available models or rules are associated with each abnormality to be detected.

(5) The measure information DB 168 includes a plurality of measure plans. In the countermeasure information DB 168, one or more predetermined countermeasures are associated with each detected abnormality. The one or more countermeasures include, for example, stopping the device in which the abnormality is detected, instructing maintenance of the device, and presenting an alternative device that can be used in place of the device.

Next, the time-series data DB 162 and the attribute information DB 164 stored in the storage unit 160 will be described.

FIG. 3 is a diagram illustrating an example of a plurality of time-series data acquired by the abnormality detection device 100 according to the present embodiment. Specifically, FIG. 3 shows a part of the time-series data DB 162 stored in the storage unit 160.

時 As shown in FIG. 3, in the time-series data DB 162, a plurality of time-series data are managed in association with each time. The time is represented by, for example, date and time (date and time), but may not include the date.

(3) In the time-series data DB 162 shown in FIG. 3, the time-series data is shown as data for one column. An identifier (specifically, data ID) is attached to the time-series data. The data ID is composed of a combination of “Data-” and a two-digit numerical value, such as “Data-01” in the example shown in FIG. The configuration of the data ID is not particularly limited.

One piece of time-series data includes a numerical value of a parameter (for example, power consumption) for each fixed period. The numerical value may be, for example, an instantaneous value of the parameter, or a cumulative value or an average value of the parameter within a period. The period is, for example, one second or more and several ten minutes or less, but is not limited thereto. In the example shown in FIG. 3, the time-series data indicates a numerical value every 10 minutes.

異なる Different attribute information is assigned to each of the plurality of time-series data. FIG. 4 is a diagram illustrating an example of attribute information acquired by the abnormality detection device 100 according to the present embodiment. Specifically, FIG. 4 illustrates the attribute information DB 164 stored in the storage unit 160.

属性 As shown in FIG. 4, in the attribute information DB 164, attribute information is associated with each data ID representing time-series data. In the present embodiment, the attribute information is classified into a plurality of items. The items are an installation room, a device name, and a data type. The installation room and the device name indicate the room in which the device to which the sensor or the measuring device for measuring the numerical value of the corresponding time-series data is mounted is installed, and the name of the device. The type of data corresponds to the content of the parameter of the device.

The attribute information DB 164 is generated by, for example, input by a person such as an administrator of the factory 1, a designer, an operator of each device, or a maintenance worker. For this reason, an erroneous input due to a human error may occur, and the correspondence between the attribute information and the time-series data may include an error.

Even if the attribute information is input correctly, if the configuration of the manufacturing line is changed or the installation position of the sensor or measuring instrument is changed, the attribute information is corrected. Need to do. If the attribute information is not corrected, an error occurs in the correspondence between the attribute information and the time-series data.

(4) In FIG. 4, the attribute information in which an error has occurred and needs to be corrected is shown in bold letters, and the background is shaded with dots. The error is detected by the analysis by the analysis unit 120, and is corrected by the correction unit 130.

Next, an operation (an abnormality detection method) of the abnormality detection device 100 according to the present embodiment will be described.

FIG. 5 is a flowchart showing the operation of the abnormality detection device 100 according to the present embodiment.

As shown in FIG. 5, in the abnormality detection device 100, the acquisition unit 110 first acquires the time-series data and the attribute information (S10). The acquisition unit 110 acquires a sensor value (that is, a numerical value of each parameter) from, for example, a sensor or a measuring device provided in each device of the factory 1 every 10 minutes, and stores the acquired sensor value in the storage unit 160, for example. Acquire a plurality of time-series data for a certain period such as a day or a week. In addition, the acquisition unit 110 acquires attribute information corresponding to a plurality of time-series data by receiving an input from a manager of the factory 1 or the like. Either the acquisition of the time-series data or the acquisition of the attribute information may be performed first, or may be performed simultaneously.

Next, the analysis unit 120 analyzes the cross-correlation of the plurality of time-series data (S11). In the present embodiment, analysis unit 120 selects two time-series data from time-series data DB 162 stored in storage unit 160, and calculates a cross-correlation coefficient between the selected two time-series data. The analysis unit 120 repeats the selection of two time series data for which the cross-correlation coefficient has not been calculated from the N time series data stored in the storage unit 160, so that N (N−1) / 2 Is calculated. Furthermore, the analysis unit 120 classifies the plurality of time-series data into a plurality of groups by comparing the calculated cross-correlation coefficient with a threshold. For example, the analysis unit 120 determines that two time-series data in which the absolute value of the cross-correlation coefficient is greater than 0.75 belong to the same group, and classifies the N time-series data into a plurality of groups.

(4) In the following, for example, a case in which 23 pieces of time-series data Data-01 to Data-23 are classified will be described.

FIG. 6 is a diagram showing an example of the result of the cross-correlation analysis by the abnormality detection device 100 according to the present embodiment. Specifically, FIG. 6 schematically shows a state in which 20 time-series data Data-01 to Data-20 are classified into four groups A to D. The remaining Data-21 to Data-23 are not shown in FIG. 6 because they are classified into single groups E to G, respectively.

As shown in FIG. 6, for example, all five time-series data Data-01, Data-02, Data-03, Data-04, and Data-05 have an absolute value of a cross-correlation coefficient of 0. As a result of calculating a value larger than 75, the values are classified into one group A. Similarly, for all of the five time-series data Data-07, Data-11, Data-13, Data-14, and Data-18, a value in which the absolute value of each cross-correlation coefficient is greater than 0.75 is calculated. As a result, they are classified into one group B. The same applies to groups C and D.

FIG. 6 shows not only a group having a large correlation but also a correlation (medium correlation or small correlation) between the groups. The difference in the correlation strength is determined by using a plurality of thresholds. The details will be described later as a modification of the present embodiment.

After the groups are classified, the correction unit 130 corrects the attribute information as shown in FIG. 5 (S12). In the present embodiment, the correction unit 130 corrects the device name in the attribute information.

Hereinafter, a specific example of the device name correcting process will be described with reference to FIGS.

FIG. 7 is a flowchart showing a process of correcting attribute information in the operation of abnormality detection apparatus 100 according to the present embodiment.

FIG. 8 is a diagram showing an example of the correction of the attribute information by the abnormality detection device 100 according to the present embodiment. In FIG. 8, a grouping result obtained by the cross-correlation analysis and various types of information associated when performing the attribute information correction process are added to the attribute information DB 164.

FIG. 9 is a diagram showing an example of attribute information before correction for each group by the abnormality detection device 100 according to the present embodiment. Specifically, FIG. 9 corresponds to the grouping result of FIG. 6, and shows, instead of the data ID, the device name and the type of data associated with the data ID.

First, as shown in FIG. 7, the correction unit 130 sets a correction flag (temporary) based on the grouping result obtained by the cross-correlation analysis (S20). Specifically, the correction unit 130 classifies the time-series data into a minority group and a majority group based on the attribute information for each group. Here, the time-series data is classified into a minority group and a majority group based on the device name which is the attribute information to be corrected.

As shown in FIGS. 8 and 9, in the group A, all the device names of the five time-series data are “mounting machines”, and all of them are classified as majority. In the group B, the device name of the four time-series data Data-06, Data-09, Data-16 and Data-19 is “molding machine”, whereas the device name of one time-series data Data-10 is Is the “mounting machine”. That is, in the group B, the four time-series data Data-06, Data-09, Data-16, and Data-19 are classified into the majority, and one time-series data Data-10 is classified into the minority. Similarly, in the group C, the time-series data Data-13 is classified as a minority, and in the group D, the time-series data Data-08 is classified as a minority.

The correction unit 130 sets the correction flag (provisional) of each of the time-series data Data-08, Data-10, and Data-13 classified as the minority to “TRUE”, and sets the remaining 20 time-series data. Each correction flag (temporary) is set to “FALSE”. The correction flag (temporary) is information indicating whether or not the device name of the associated time-series data is to be corrected. When the correction flag (temporary) is “TRUE”, a process of correcting the device name of the time-series data associated with the correction flag (temporary) is performed.

Note that, in the present embodiment, the correction unit 130 sets the correction flag (provisional) for the minority only when the ratio of the time-series data classified to the minority to the total number of the time-series data in the group is sufficiently small. Is set to “TRUE”. For example, when the ratio of the time series data is 20% or less, the correction flag (temporary) of the minority is set to “TRUE”. As shown in FIGS. 8 and 9, the proportion of the minority in all of the groups B to D is 20% or less. For this reason, the correction unit 130 sets the correction flag (temporary) of each of the time series data Data-08, Data-10, and Data-13, which are the minorities of the groups B to D, to “TRUE”.

Next, as shown in FIG. 7, the correction unit 130 provisionally sets the candidate information (S21). The candidate information is, specifically, a device name that becomes a correction candidate. In the present embodiment, the device name that is a correction candidate is the device name of the time-series data belonging to the majority of the group. Therefore, as shown in FIG. 8, the device name of the correction candidate of the time-series data Data-13 belonging to the group B is “applicator” which is the device name of the time-series data belonging to the majority of the group B. Similarly, the device name of the correction candidate of the time-series data Data-10 belonging to the group C is “molding machine”. The device name of the correction candidate of the time-series data Data-08 belonging to the group D is “press machine”.

Next, as shown in FIG. 7, the following processing (S23 to S27) is performed on the attribute information whose correction flag (temporary) is “TRUE” (“TRUE” in S22). The process (S23 to S27) is not performed on the attribute information whose correction flag (temporary) is "FALSE" ("FALSE" in S22). Hereinafter, the case of the time series data Data-08 will be described.

First, the correction unit 130 sets a duplication flag (S23). Specifically, the correction unit 130 sets “TRUE” when the set of the installation room, the device name, and the data type after correction overlaps with the set of the installation room, the device name, and the data type of another data ID. If no duplicate data exists, "FALSE" is set. For example, in the time-series data Data-08, the installation room, the device name, and the data type after the correction are “RoomB”, “press machine”, and “operation time”, respectively, and do not overlap with other data. Therefore, the correction unit 130 sets the overlap flag to “FALSE”.

Next, the correction unit 130 extracts a representative value of the time-series data (S24). The representative value is, for example, at least one of a minimum value, a maximum value, and an average value of numerical values included in the time-series data. Specifically, the correction unit 130 reads the minimum value, the maximum value, and the average value of the time-series data Data-08 with reference to the time-series data DB 162.

Next, the correction unit 130 extracts an allowable representative value (S25). The permissible representative value is a value determined based on, for example, the specifications of the device or a past operation history, and is stored in the storage unit 160. The permissible representative value is at least one of a minimum value and a maximum value of a range permitted as the operation of the apparatus, and an assumed average value. The correction unit 130 associates the allowable representative value read from the storage unit 160 with the data ID. When the allowable representative value is unknown, the correction unit 130 associates that the allowable representative value is unknown with the data ID.

Next, the correction unit 130 determines whether or not inconsistency has occurred between the allowable representative value and the representative value of the time-series data (S26). For example, if the minimum value of the time series data is equal to or larger than the minimum value of the allowable representative value, no inconsistency occurs. If the maximum value of the time series data is equal to or less than the maximum value of the allowable representative value, no inconsistency occurs. If the average value of the time-series data is equal to the average value of the allowable representative values (for example, the difference is within ± 10% to 20%), no inconsistency occurs.

If there is no inconsistency (No in S26), the correction unit 130 corrects the device name in the attribute information DB 164 to a device name of a correction candidate (S27). By setting the correction flag to “TRUE”, the correction of the attribute information of the time-series data is completed.

After that, the processing from the setting of the duplication flag (S23) to the correction to the candidate information (S27) is repeated for the time-series data in which one of the correction flag (temporary) and the correction flag is “FALSE”.

If there is a contradiction between the allowable representative value and the representative value of the time-series data (Yes in S26), the correction unit 130 sets another candidate information and executes the above-described processing (S23 to S27). You may. For example, when the minimum value of the time-series data is smaller than the minimum value of the allowable representative value, a contradiction occurs. Alternatively, when the maximum value of the time-series data is larger than the maximum value of the allowable representative values, a contradiction occurs. If the average value of the time-series data is not equal to the average value of the allowable representative values (for example, the difference is within ± 10% to 20%), a contradiction occurs. Alternatively, the correction unit 130 corrects the attribute information by causing the output unit 150 to output information notifying that the correction has not been correctly performed, and causing the administrator of the factory 1 to input the attribute information. Is also good.

{When the duplication flag is set to “TRUE”, it means that there is at least one other piece of data having the same attribute information as the attribute information of the data ID for which “TRUE” is set. For this reason, the correction unit 130 may set another candidate information and execute the above-described processing (S23 to S27). Alternatively, information notifying that the correction was not correctly performed may be output from the output unit 150, and the information may be corrected by a manager of the factory 1 or the like.

After the attribute information has been corrected as described above, as shown in FIG. 5, the detection unit 140 detects an abnormality (S13). The detection of the abnormality may be performed immediately after the attribute information is corrected, or may be performed at a predetermined timing. Further, the detection of the abnormality may be periodically repeated.

Specifically, the detection unit 140 reads data from each of the time-series data DB 162, the attribute information DB 164, and the abnormality detection model DB 166 of the storage unit 160 for each abnormality to be detected, and detects an abnormality based on the read data. The presence / absence and the time when an abnormality has occurred are detected. Hereinafter, a case where the detection unit 140 detects an abnormality of the molding machine will be described as an example.

For example, the detection unit 140 reads and uses the abnormality detection model of the molding machine from the abnormality detection model DB 166. In the abnormality detection model of the molding machine, it is determined that no abnormality has occurred in the molding machine when the ratio between the number of parts supplied to the molding machine and the power consumption of the molding machine falls within a predetermined range. When the ratio deviates from the predetermined range, it is determined that an abnormality has occurred in the molding machine. In the abnormality detection model of the molding machine, for example, when the number of inputs [pieces] / power consumption [kWh] is in the range of 1.2 or more and 2.0 or less, it is determined that no abnormality has occurred, and When [unit] / power consumption [kWh] is less than 1.2 or greater than 2.0, it is determined that an abnormality has occurred.

Specifically, the detection unit 140 specifies the time-series data corresponding to each of the power consumption and the number of inputs of the target molding machine by referring to the attribute information DB 164. As shown in FIG. 8, it is specified that the power consumption of the molding machine is time-series data Data-06, and the number of injections of the molding machine is time-series data Data-10.

Here, if the attribute information is not corrected, the attribute information DB 164 before correction shown in FIG. 4 does not include time-series data corresponding to the number of injections of the molding machine. Therefore, although it actually exists, the time series data corresponding to the number of injections of the molding machine is not specified, and the abnormality detection process cannot be performed.

On the other hand, in the present embodiment, since the attribute information has been corrected in step S12, the time-series data Data-10 corresponding to the number of injections of the molding machine is specified. For this reason, the abnormality detection process can be performed with high accuracy.

FIG. 10 is a diagram illustrating an example of an abnormality detection process performed by the abnormality detection device 100 according to the present embodiment. In FIG. 10, the horizontal axis represents time. The vertical axis of FIG. 10A represents the power consumption [kWh] and the number of inputs [pieces]. That is, FIG. 10A is a two-dimensional graph of time-series data Data-06 indicating power consumption and time-series data Data-10 indicating the number of inputs. The vertical axis in FIG. 10B represents the number of inputs / power consumption [pcs / kWh].

よう As shown in FIG. 10, in the period T1 from 11:10 to 11:50, the number of inputs is lower than the power consumption. The number of inputs / power consumption in this period T1 is a value lower than 1.2 (> 0) as shown in FIG. For this reason, the detecting unit 140 determines that an abnormality indicating that the productivity has decreased has occurred in the molding machine in the period T1.

Also, in the period T2 from 13:00 to 13:40, power is consumed in the molding machine even though no components are loaded. The number of inputs / power consumption in this period T2 is 0, which is lower than 1.2, as shown in FIG. For this reason, the detecting unit 140 determines that an abnormality indicating that useless power consumption occurs has occurred in the molding machine in the period T2.

The type of the abnormality is predetermined in the abnormality detection model of the molding machine based on the value of the number of inputs / power consumption. For example, when the number of inputs / power consumption is 0 and the power consumption is greater than 0, the type of abnormality is determined to be useless power consumption. When the number of inputs / power consumption is larger than 0 and smaller than 1.2, it is determined that the type of abnormality is productivity reduction. When the number of inputs / power consumption is greater than 2.0, the type of abnormality is defined as excessive supply of components or a sensor failure.

(5) After the abnormality is detected by the detection unit 140 in this way, as shown in FIG. 5, the output unit 150 outputs information indicating the detected abnormality and the occurrence time of the abnormality (S14). In addition, the output unit 150 determines a countermeasure against the abnormality by referring to the countermeasure information DB 168 stored in the storage unit 160 based on the detected abnormality, and outputs information indicating the determined countermeasure. I do.

FIG. 11 is a diagram illustrating an example of information output by the abnormality detection device 100 according to the present embodiment. As illustrated in FIG. 11, the output unit 150 generates an image representing the detected abnormality, the occurrence time of the abnormality, and a countermeasure for the abnormality, and displays the generated image on a display or the like, so that the administrator of the factory 1 or the like. To present.

(5) Here, during the period T1 from 11:10 to 11:50, the occurrence of an abnormality such as a decrease in productivity has occurred, and the review of the production conditions of the molding machine has been displayed as a countermeasure against the abnormality. Similarly, in the period T2 from 13:00 to 13:40, the occurrence of the abnormality of wasteful power consumption and the turning off of the power of the molding machine are displayed as a measure against the abnormality. The number of output countermeasures may be plural for each abnormality.

In the present embodiment, the processing up to the correction of the attribute information is performed as a preceding step of the abnormality detection method, and the processing after the detection of the abnormality is performed as a subsequent step. If the attribute information has been corrected at least once and it has been confirmed that the attribute information has been correctly corrected, the cross-correlation analysis (S11) and the attribute information correction (S12) may not be performed. Good. In this case, acquisition of time-series data (S10), detection of abnormality (S13), and output of information (S14) may be periodically repeated.

Detection of an abnormality and output of information may be performed in real time in parallel with manufacture of a product. Accordingly, when an abnormality is detected, a response such as maintenance can be promptly performed, so that a decrease in production efficiency can be suppressed.

異常 Furthermore, the detection of the abnormality and the output of the information may be performed every period such as one week or one month. This makes it possible to specify a device or the like in which an abnormality is likely to occur when a product is manufactured for a long period of time. For this reason, it can be used for formulation of a future production plan, such as a change in the design of a production line or replacement of a production device.

As described above, the abnormality detection method and the abnormality detection device according to the present embodiment are capable of suppressing not only a short-term decrease in production efficiency with emphasis on real-time performance but also a decrease in production efficiency from a medium- to long-term perspective. Can help.

(Modification)
Subsequently, a modified example of the abnormality detection device according to the first embodiment will be described. In this modified example, a cross-correlation coefficient and a delay are calculated in the cross-correlation analysis, and the attribute information is corrected and an abnormality is detected based on the calculated cross-correlation coefficient and the delay. In this modification, the grouping of the time-series data is performed using a plurality of thresholds. The following description focuses on differences from the first embodiment, and description of common points is omitted or simplified.

In the present modification, the analysis unit 120 selects two time-series data from a plurality of time-series data, and calculates a cross-correlation coefficient and a delay of the selected time-series data. By calculating the delay, the relationship between groups having relatively small correlation can be determined. Specifically, the analysis unit 120 determines the line configuration of each manufacturing apparatus based on the delay.

{Circle around (7)} The analysis unit 120 compares the absolute value of the cross-correlation coefficient with a plurality of thresholds. For example, the plurality of threshold values include not only 0.75 but also 0.6 and 0.4. This makes it possible to classify the groups according to the strength of the correlation. For example, by using three thresholds, a plurality of time-series data can be classified into a large correlation group, a medium correlation group, a small correlation group, and no correlation. Further, when a plurality of groups having a large correlation are formed, the strength of the correlation between the plurality of groups can be determined.

For example, when the absolute value of the cross-correlation coefficient is larger than the smallest threshold (for example, 0.4) among the plurality of thresholds, the analysis unit 120 determines that there is a small correlation between the two time-series data. . In the present modification, the analysis unit 120 determines that the manufacturing apparatuses corresponding to the time-series data included in the group determined to have a small correlation form the same manufacturing line.

Here, the classification result based on a plurality of thresholds of the 20 time-series data shown in FIGS. 6 and 9 will be described. For example, the absolute value of the cross-correlation coefficient between the time-series data classified into group A and the time-series data classified into group B is larger than 0.65 and 0.75 or less. For this reason, it can be determined that the group A and the group B have a moderate correlation. In other words, when the threshold value used for classifying the group is set to 0.65 instead of 0.75, ten pieces of time-series data are classified into one group including the group A and the group B. The same applies to the groups A and C, and the same applies to the groups C and D.

{Also, for example, the absolute value of the cross-correlation coefficient between the time-series data classified into group B and the time-series data classified into group C is larger than 0.40 and 0.65 or less. Therefore, it is determined that group A and group C have a small correlation. Further, the absolute value of the cross-correlation coefficient between the time-series data classified into group A and the time-series data classified into group D is 0.40 or less, and there is almost no correlation between group A and group D. It is determined that there is not.

Thus, for example, each of the groups A to D shown in FIG. 6 has a small correlation or a moderate correlation with each other. Therefore, the analysis unit 120 determines that the mounting machine, the molding machine, the coating machine, and the press machine constitute one manufacturing line.

Furthermore, the analysis unit 120 determines the order of each manufacturing apparatus based on the delay. Specifically, the analysis unit 120 determines that the delay of the second time-series data with respect to the first time-series data is a positive value and the absolute value thereof is larger, the manufacturing time corresponding to the second time-series data is larger. The apparatus determines that the apparatus is located downstream of the manufacturing line from the manufacturing apparatus corresponding to the first time-series data. Similarly, as the delay of the second time-series data with respect to the first time-series data is a negative value and the absolute value thereof is larger, the analysis unit 120 determines that the manufacturing apparatus corresponding to the second time-series data has a larger value. , Is located upstream of the manufacturing line with respect to the manufacturing apparatus corresponding to the first time-series data.

Hereinafter, an example will be described in which the configuration of the production line is determined based on the classification result shown in FIG. The determination of the configuration of the production line is performed by the analysis unit 120 after the attribute information is corrected as described in the first embodiment. Therefore, the time-series data included in each of the groups A to D corresponds to each of the mounting machine, the coating machine, the molding machine, and the press machine.

As shown in FIG. 9, for example, as a result of the calculation by the analysis unit 120, a delay of one minute was obtained for the groups A and C. Therefore, it can be seen that the “molding machine” corresponding to the time-series data included in the group C is located downstream of the manufacturing line, rather than the “mounting machine” corresponding to the time-series data included in the group A. . Similarly, a delay of 30 seconds was calculated for groups B and A. For this reason, it can be seen that the “mounting machine” is located downstream of the manufacturing line with respect to the “coating machine”. In addition, a delay of 3 minutes was calculated for group D and group C. For this reason, it can be seen that the “molding machine” is located downstream of the manufacturing line with respect to the “press machine”. Note that no correlation was obtained between Group D and any of Groups A and B.

From the above results, the configuration of the production line as shown in FIG. 12 is obtained. FIG. 12 is a diagram illustrating a configuration example of a manufacturing apparatus determined by a cross-correlation analysis by the abnormality detection apparatus 100 according to the present modification.

As described above, according to the present modification, the configuration of the manufacturing line can be determined by the cross-correlation analysis by the analysis unit 120. Therefore, the detection unit 140 can detect the abnormality with higher accuracy.

In the following, as an example, a model for detecting an abnormality of a coating machine will be described. For example, in the abnormality detection model of the coating machine, when the ratio between the number of parts produced by the coating machine and the number of parts supplied to the mounting machine after the delay time of Δt is within a predetermined range, It is determined that no abnormality has occurred in the coating machine. When the ratio deviates from the predetermined range, it is determined that an abnormality has occurred. For example, the abnormality detection model of the coating machine is such that when the production number [pieces] of the coating machine / the number [pieces] of the mounting machine after 30 seconds is included in the range of 0.8 or more and 1.2 or less, the abnormality If it is determined that no occurrence has occurred, and the number of coating machines produced [pieces] / the number of loading machines after 30 seconds [pieces] is less than 0.8 or greater than 1.2, an abnormality occurs. It is determined that there is.

FIG. 13 is a diagram illustrating an example of an abnormality detection process performed by the abnormality detection device 100 according to the present modification. In FIG. 13, the horizontal axis represents time. The vertical axis indicates the number of production machines [pcs] and the number of mounting machines [pcs].

よう As shown in FIG. 13, the production number of the applicator became zero at 13:12, and the production of the applicator was stopped. Thereafter, at 13:13, the number of mounted machines becomes zero. Since the production number of coating machines / the number of mounting machines is 0 or a value that cannot be calculated, the detection unit 140 determines that an abnormality has been detected.

At this time, the detection unit 140 determines the period from when the production number of the coating machine becomes 0 to when the number of mounting machines becomes 0, and the delay between the coating machine and the mounting machine calculated by the analysis unit 120. Compare with time. When the period from when the production number of the coating machine becomes 0 to when the number of the mounting machines becomes 0 is equal to the delay time (for example, the difference is twice or less), the detection unit 140 detects the production line. It is determined that an abnormality has occurred in the device located upstream and no abnormality has occurred in the device located downstream. Thus, in the examples shown in FIGS. 12 and 13, it is determined that an abnormality has occurred in the coating machine and no abnormality has occurred in the mounting machine.

出力 Based on the detection result of such an abnormality, the output unit 150 outputs the detected abnormality, the occurrence time of the abnormality, and a measure.

FIG. 14 is a diagram illustrating an example of information output by the abnormality detection device 100 according to the present modification. As illustrated in FIG. 14, the output unit 150 generates an image indicating the detected abnormality, the occurrence time of the abnormality, and a countermeasure for the abnormality, and displays the image on a display or the like, so that the administrator of the factory 1 or the like. To present.

Here, at 13:12, the occurrence of the abnormality that the production of the coating machine has stopped has occurred, and the maintenance of the coating machine is displayed as a countermeasure against the abnormality. Similarly, at 13:13, the occurrence of an abnormality indicating that the production of the mounting machine has stopped and the maintenance of the coating machine are displayed as a countermeasure against the abnormality. That is, although the production of the mounting machine is stopped, the cause is the stoppage of the coating machine. Therefore, the output unit 150 displays not the maintenance of the mounting machine but the maintenance of the coating machine.

If the analysis of the production line according to the present modification is not performed and the configuration of the production line is unknown, it remains unknown that the stop of the production of the mounting machine is caused by the stop of the coating machine. According to this modification, not only the occurrence of an abnormality but also the cause thereof can be detected with high accuracy, and a highly effective countermeasure can be proposed.

(Embodiment 2)
Next, a second embodiment will be described. The following description focuses on differences from the first embodiment, and description of common points is omitted or simplified.

FIG. 15 is a diagram illustrating a configuration of a plurality of compressors A to C that are targets of the abnormality detection device 200 according to the present embodiment. Each of the plurality of compressors A to C is an example of a plurality of elements involved in manufacturing a product. The plurality of compressors A to C are devices that compress and output a gas such as air. The outputs from each of the plurality of compressors A to C are integrated. Specifically, the output of the compressor A, the output of the compressor B, and the output of the compressor C are integrated and output.

As shown in FIG. 15, the abnormality detection device 200 acquires pressure data, which is time-series data of the output pressure of each of the plurality of compressors A to C, and flow rate data, which is time-series data of the flow rate. Further, the abnormality detection device 200 acquires power data, which is time-series data of power consumption, by acquiring sensor values of power consumption in each compressor from sensors attached to the plurality of compressors A to C. Further, the abnormality detection device 200 acquires integrated pressure data, which is time series data of the integrated pressure of the plurality of compressors A to C, and integrated flow rate data, which is time series data of the flow rate. The abnormality detection device 200 detects an abnormality occurring in the plurality of compressors A to C based on the acquired plurality of time-series data.

FIG. 16 is a block diagram showing a configuration of an abnormality detection device 200 according to the present embodiment. As illustrated in FIG. 16, the abnormality detection device 200 includes an acquisition unit 210, an analysis unit 220, a correction unit 230, a detection unit 240, and an output unit 150. Further, the abnormality detection device 200 includes a storage unit 260 for storing various data and information.

The acquisition unit 210 acquires a plurality of time-series data obtained from each of the plurality of compressors A to C and attribute information of each of the plurality of time-series data. Specifically, the acquiring unit 210 acquires pressure data and flow rate data from each of the plurality of compressors A to C, and integrated pressure data and integrated flow rate data. For example, a pressure gauge and a flow meter are attached to each output line of the plurality of compressors A to C and the output line after integration. The acquisition unit 210 acquires a plurality of pressure data, a plurality of flow data, an integrated pressure data, and a pressure sensor value and a flow sensor value by periodically acquiring a pressure sensor value and a flow sensor value from a pressure gauge and a flow meter attached to each output line. Obtain integrated flow data. Further, the acquisition unit 210 acquires a plurality of power data by periodically acquiring power values from sensors attached to the plurality of compressors A to C.

The plurality of time-series data and attribute information acquired by the acquisition unit 210 are stored in the storage unit 260. In the storage unit 260, a plurality of pressure data, a plurality of flow data, a plurality of power data, integrated pressure data, and integrated flow data are managed in the time-series data DB 262, and attribute information is managed in the attribute information DB 264.

The analysis unit 220 analyzes the cross-correlation of the plurality of time-series data acquired by the acquisition unit 210. In the present embodiment, analysis section 220 performs multiple regression analysis. Specifically, the analysis unit 220 creates a plurality of combinations of pressure data and flow rate data from each of the plurality of compressors A to C, and candidates for attribute information corresponding to each of the pressure data and flow rate data. The analysis unit 220 determines the likelihood of the created combination by performing multiple regression analysis, and determines an optimal solution (optimal combination).

For example, for each combination created, the analysis unit 220 performs a multiple regression analysis using a plurality of pressure data and a plurality of flow data constituting the combination as independent variables, and using the integrated pressure data and the integrated flow data as dependent variables. The analysis unit 220 calculates a determination coefficient (that is, an R value) by multiple regression analysis for each combination. The analysis unit 220 determines the combination corresponding to the largest determination coefficient among the plurality of calculated determination coefficients as the optimal combination. The analysis unit 220 outputs information indicating the optimal combination to the correction unit 230.

The correction unit 230 corrects the attribute information based on the result of the analysis by the analysis unit 220. Specifically, the correction unit 230 refers to the attribute information DB 264 stored in the storage unit 260, and stores a plurality of pieces of attribute information read from the storage unit 260 and information indicating an optimal combination obtained from the analysis unit 220. To correct attribute information. The correction unit 230 reflects the result of the correction in the attribute information DB 264. The specific operation of the correction by the correction unit 230 will be described later using an example.

The detection unit 240 detects an abnormality occurring in the compressors A to C based on the corrected attribute information and the pressure data, the flow rate data, the power data, the integrated pressure data, and the integrated flow rate data. For example, the detection unit 240 detects an abnormality of the compressors A to C based on an abnormality detection model regarding the power efficiency of the compressor. The detection unit 240 refers to the time-series data DB 262, the attribute information DB 264, and the abnormality detection model DB 166 stored in the storage unit 260, and reads necessary data to detect an abnormality. The detection unit 240 outputs a detection result of the abnormality to the output unit 150. The specific operation of the abnormality detection will be described later in detail with an example.

The storage unit 260 is realized by, for example, a semiconductor memory or an HDD. In the present embodiment, the storage unit 260 stores various data and information necessary for executing the abnormality detection method in a database. Specifically, as shown in FIG. 16, the storage unit 260 stores a time-series data DB (database) 262, an attribute information DB 264, an abnormality detection model DB 166, and a countermeasure information DB 168.

The time series data DB 262 includes a plurality of pressure data, a plurality of flow rate data, integrated pressure data, and integrated flow rate data acquired by the acquisition unit 210. In addition, the time-series data DB 262 includes power data for each compressor acquired by the acquisition unit 210. The time-series data DB 262 includes, for each data ID assigned to the time-series data, a numerical value for each time (specifically, a pressure value, a flow rate value, an electric power value, etc.), similarly to the time-series data DB 162 illustrated in FIG. ).

The attribute information DB 264 includes a plurality of pieces of attribute information acquired by the acquisition unit 210. The plurality of pieces of attribute information respectively correspond to the plurality of pieces of time-series data. The attribute information DB 264 is generated and updated when the acquisition unit 210 acquires the attribute information. The attribute information DB 264 is updated by the correction unit 230.

FIG. 17 is a diagram illustrating an example of attribute information acquired by the abnormality detection device 200 according to the present embodiment. Specifically, FIG. 17 illustrates the attribute information DB 264 stored in the storage unit 260.

As shown in FIG. 17, in the attribute information DB 264, attribute information is associated with each data ID representing time-series data. In the present embodiment, the data ID is represented by P1 to P3, V1 to V3, Pt, Vt, and U1 to U3. The attribute information is classified into a plurality of items. The items are an installation room, a device name, and a data type. The installation room and the device name indicate the room where the compressor to which the sensor for measuring the numerical value of the corresponding time-series data is mounted is installed, and the name of the compressor. The type of data is the pressure or flow rate of the compressor.

In FIG. 17, attribute information in which an error has occurred and needs to be corrected is shown in bold letters and the background is shaded with dots. The error is detected by the analysis of the analysis unit 220 and corrected by the correction unit 230.

Next, an operation (an abnormality detection method) of the abnormality detection device 200 according to the present embodiment will be described.

FIG. 18 is a flowchart showing the operation of the abnormality detection device 200 according to the present embodiment. As shown in FIG. 18, in the abnormality detection device 200, first, the acquisition unit 210 acquires time-series data and attribute information (S30). The acquisition unit 210 includes, for example, a pressure gauge and a flow meter provided at each output of the plurality of compressors A to C and an output after integration, and sensors provided at the plurality of compressors A to C (specifically, For example, by acquiring a sensor value from a power meter every 10 minutes and storing the sensor value in the storage unit 260, a plurality of time-series data for a certain period such as one day or one week is acquired. Further, the acquiring unit 210 acquires attribute information corresponding to a plurality of time-series data by receiving inputs from managers of the plurality of compressors A to C and the like. Either the acquisition of the time-series data or the acquisition of the attribute information may be performed first, or may be performed simultaneously.

Next, the analysis unit 220 creates a plurality of combinations based on the time-series data DB 262 (S31). For example, as shown in FIG. 19, the analysis unit 220 combines each of the three pressure data P1 to P3 and each of the three flow rate data V1 to V3 so as not to overlap, thereby providing six combinations. Create

Here, FIG. 19 is a diagram illustrating an example of a combination created by the abnormality detection device 200 according to the present embodiment. Further, it is assumed that the integrated pressure data, the integrated flow rate data, and the power data have no error and need not be corrected. It is also assumed that there is no error between the pressure data and the flow rate data.

において In FIG. 19, the combination (P, V) indicates that they correspond to the same compressor. For example, in the combination Pattern-1, the pressure data P1 and the flow rate data V1 correspond to the first compressor, the pressure data P2 and the flow rate data V2 correspond to the second compressor, and the pressure data P3 and the flow rate data V3. Represents that it corresponds to the third compressor. The same applies to Pattern-2 to Pattern-6 in the combination.

Next, as shown in FIG. 18, the analysis unit 220 performs a multiple regression analysis for each created combination (S32). Specifically, for each combination, the analysis unit 220 sets the integrated output (Pt, Vt) as a dependent variable Y, and separates each time-series data (P1, V1), (P2, V2), and (P3, V3) independently. Multiple regression analysis is performed using the variables X1 to X3. Thereby, the analysis unit 220 calculates the determination coefficient (that is, the R value) for each combination.

Next, the analysis unit 220 selects an optimal solution, that is, an optimal combination (S33). Specifically, the analysis unit 220 selects a combination of the largest determination coefficient among the plurality of calculated determination coefficients as an optimal combination (that is, a correct combination). In the example shown in FIG. 19, the combination Pattern-1 is selected as the optimum combination.

Next, the correction unit 230 corrects the attribute information based on the analysis result (S34). Specifically, the correction unit 230 extracts a group in which the device name (specifically, the name of the compressor) is different for each pair of the pressure data and the flow rate data based on the optimum combination.

In the example shown in FIG. 17, in the set (P1, V1), the device name of both the pressure data and the flow rate data is “compressor A”. In the set (P2, V2), the device name of both the pressure data and the flow rate data is “compressor B”. On the other hand, in the set (P3, V3), the device name of the pressure data is “compressor B”, whereas the device name of the flow rate data is “compressor C”. For this reason, the correction unit 230 determines that at least one of the device names in the set (P3, V3) is incorrect and needs to be corrected.

The correction unit 230 determines whether or not inconsistency occurs when one device name of the pressure data and the flow rate data is corrected to the other device name in the set (P3, V3), for example. Specifically, even if the device name of the pressure data P3 is corrected to “compressor C” which is the device name of the flow rate data V3, no inconsistency occurs. On the other hand, when the device name of the flow rate data V3 is corrected to “compressor B” which is the device name of the pressure data P3, the pressure data of the compressor B already exists and contradicts. Therefore, the correction unit 230 corrects the device name of the pressure data P3 to “compressor C” so that no inconsistency occurs.

If the device name of the flow rate data V3 is "compressor A", both the pressure data and the flow rate data are incorrect. In this case, inconsistency occurs even if any device name is matched. If there is no case where no inconsistency occurs, the correction unit 230 corrects both the pressure data and the flow rate data to the name of the compressor that does not exist in the attribute information DB 264 (for example, “compressor C”), It may be determined whether or not a contradiction occurs. Alternatively, the correction unit 230 corrects the attribute information by causing the output unit 150 to output information indicating that the correction has not been correctly performed, and causing the manager of the factory 1 to input the attribute information. Is also good.

After the attribute information has been corrected as described above, the detection unit 240 detects an abnormality as shown in FIG. 18 (S35). The detection of the abnormality may be performed immediately after the attribute information is corrected, or may be performed at a predetermined timing. Further, the detection of the abnormality may be periodically repeated.

Specifically, the detection unit 240 reads data from each of the time-series data DB 262, the attribute information DB 264, and the abnormality detection model DB 166 of the storage unit 260 for each abnormality to be detected, and detects an abnormality based on the read data. The presence / absence and the time when an abnormality has occurred are detected.

では In the present embodiment, detection section 240 uses an abnormality detection model based on power efficiency A (k) expressed by the following equation (1).

(1) U (k, t) = A (k) V (k, t) P (k, t)

In equation (1), A (k) is the power efficiency of the k-th compressor. U (k, t) is the power consumption of the k-th compressor at time t. V (k, t) is the flow rate of the k-th compressor at time t. P (k, t) is the pressure of the k-th compressor at time t. For example, the compressors A to C are the first to third compressors, respectively.

As shown in equation (1), the power efficiency A (k) is the reciprocal of the ratio of the product of the pressure P (k, t) and the flow rate V (k, t) to the input power U (k, t). is there. Specifically, the detection unit 240 calculates the power efficiency of each compressor based on the equation (1), and when the calculated power efficiency exceeds a predetermined threshold, determines that an abnormality has occurred in the compressor. . When the calculated power efficiency is equal to or smaller than the threshold, the detection unit 240 determines that no abnormality has occurred in the compressor.

U (k, t), V (k, t) and P (k, t) are all obtained by referring to the time-series data DB 262. Here, if the attribute information is not corrected, the pressure data of the compressor C does not exist in the example shown in FIG. Therefore, the detection unit 240 cannot calculate the power efficiency A (3) for the compressor C, and cannot detect an abnormality.

On the other hand, in the present embodiment, since the attribute information is corrected, the power efficiency A (3) of the compressor C can be calculated, and it is possible to determine whether the compressor C is abnormal.

FIG. 20 is a diagram illustrating an example of an abnormality detection process performed by the abnormality detection device 200 according to the present embodiment. In FIG. 20, the horizontal axis represents the pressure P [MPa]. The vertical axis represents the power U [kWh].

The power efficiency A (k) corresponds to the slope of the plot for each compressor in FIG. Each plot corresponds to pressure data and power data at a predetermined time. In the example shown in FIG. 20, it can be seen that the slope of the compressor A increases during the period from 11:20 to 12:40. That is, the power efficiency of the compressor A is degraded. For this reason, the detection unit 240 determines that an abnormality has occurred in the compressor A during the period.

{Circle around (4)} As described above, after the detection unit 240 detects an abnormality, as illustrated in FIG. 16, the output unit 150 outputs information indicating the detected abnormality and the occurrence time of the abnormality (S36). Further, based on the detected abnormality, the output unit 150 refers to the measure information DB 168 stored in the storage unit 260 to determine a measure for the abnormality, and outputs information indicating the determined measure. I do.

FIG. 21 is a diagram illustrating an example of information output by the abnormality detection device 200 according to the present embodiment. As shown in FIG. 21, the output unit 150 generates an image representing the detected abnormality, the occurrence time of the abnormality, and a countermeasure for the abnormality, and displays the image on a display or the like to manage the compressors A to C. To others.

Here, during the period from 11:20 to 12:40, the occurrence of an abnormality such as a decrease in the operation efficiency of the compressor A and the operation of the compressors B and C are displayed as a countermeasure against the abnormality. That is, since the operation efficiency of the compressor A is reduced, a countermeasure for recommending the operation of the compressors B and C, which are alternative devices, is displayed.

(Embodiment 3)
Next, a third embodiment will be described. The following description focuses on differences from the first embodiment, and description of common points is omitted or simplified.

FIG. 22 is a diagram showing the configuration of the factory 2 to which the abnormality detection device 300 according to the present embodiment is applied. The factory 2 differs from the factory 1 according to the first embodiment in the devices arranged in each room. Specifically, in each of the three rooms A to C, a plurality of devices related to the manufacture of a product are arranged. Specifically, as shown in FIG. 22, in the room A, a mounting machine A1, a coating machine A1, a mounting machine A2, a coating machine A2, and an air conditioner A are arranged. In the room B, a molding machine B, a coating machine B, and an air conditioner B are arranged. In the room C, a press C and an air conditioner C are arranged.

部屋 Note that the rooms A to C do not have to be physically separated rooms. That is, in the factory 2, the rooms A to C may be included in one large room, or may be a space virtually divided in association with the air conditioner. That is, the rooms A to C are an example of a space in which each of the air conditioners A to C is arranged.

The abnormality detection device 300 acquires a plurality of time-series data obtained from each device arranged in each room and generates a plurality of time-series data based on the acquired plurality of time-series data, as in the first embodiment. Detect abnormalities. The abnormality detection device 300 is communicably connected to a sensor or a measuring device provided in each device by wire or wirelessly.

FIG. 23 is a block diagram showing a configuration of an abnormality detection device 300 according to the present embodiment. As shown in FIG. 23, the abnormality detection device 300 is different from the abnormality detection device 100 according to the first embodiment in that the analysis unit 320, the correction unit 130, and the detection unit 140 are replaced with an analysis unit 320 and a correction unit. The difference is that a point 330 and a detecting section 340 are provided.

The analysis unit 320 analyzes the cross-correlation of the plurality of time-series data acquired by the acquisition unit 110. In the present embodiment, analysis section 320 performs multiple regression analysis.

機器 Products related to products generally generate heat by operating. When the generated heat is trapped, the operation efficiency of each device decreases, and the production efficiency decreases. For this reason, an air conditioner for releasing generated heat is provided in each room. That is, there is a correlation between the operation state of each device and the operation state of the air conditioner. Specifically, there is a strong correlation between the power consumption of the air conditioner and the parameters of the device provided in the same space as the space where the air conditioner is provided. Therefore, in the present embodiment, analysis unit 320 performs multiple regression analysis using time-series data of power consumption of air conditioners A to C as dependent variables and time-series data of parameters of other devices as independent variables. Based on the result of the multiple regression analysis, an optimal combination of the air conditioners A to C (that is, the rooms A to C) and each device is determined.

The correction unit 330 corrects the attribute information based on the result of the multiple regression analysis. Specifically, the correction unit 330 refers to the attribute information DB 164 stored in the storage unit 160, and stores the plurality of pieces of attribute information read from the storage unit 160 and the result of the multiple regression analysis obtained from the analysis unit 320. The attribute information is corrected using the indicated information. For example, the correction unit 330 determines whether there is an error in the correspondence between the installation room and the time-series data in the attribute information DB 164, and if there is an error, corrects the erroneous installation room to a correct installation room. The correction unit 330 reflects the result of the correction in the attribute information DB 264. The specific operation of the correction by the correction unit 330 will be described later with an example.

The detection unit 340 detects an abnormality occurring in each device based on the information indicating the corrected installation room and a plurality of time-series data. Further, the detection unit 340 determines a device that can perform an alternative operation instead of the device in which the abnormality has occurred. Specifically, based on the information indicating the corrected installation room, the detection unit 340 determines a device that is installed in the same room as the device in which the abnormality has occurred and that is of the same type as a device that can perform an alternative operation. The detection unit 340 outputs the detection result of the abnormality and the determination result of the device that can perform the alternative operation to the output unit 150. The specific operation of the abnormality detection will be described later in detail with an example.

Next, an operation (an abnormality detection method) of the abnormality detection device 300 according to the present embodiment will be described.

FIG. 24 is a flowchart showing the operation of the abnormality detection device 300 according to the present embodiment. As shown in FIG. 24, in the abnormality detection device 300, first, the acquisition unit 110 acquires time-series data and attribute information (S40). The specific acquisition process is the same as step S10 according to the first embodiment.

Next, the analysis unit 320 performs a multiple regression analysis using the plurality of time-series data (S41). Specifically, analysis unit 320 performs multiple regression analysis using the time series data of the power consumption of each of air conditioners A to C as a dependent variable and the remaining time series data as independent variables.

{For example, the analysis unit 320 calculates a coefficient for each independent variable by performing multiple regression analysis using the time-series data Data-21 of the power consumption of the air conditioner A as a dependent variable and the remaining time-series data as independent variables. Similarly, a coefficient for each independent variable is calculated using the time series data Data-22 of the power consumption of the air conditioner B and the time series data Data-23 of the power consumption of the air conditioner C as dependent variables.

Next, the analysis unit 320 determines an optimal combination of each device and a room based on the result of the multiple regression analysis (S42). Specifically, the analysis unit 320 determines, for each time-series data (independent variable), the dependent variable when the calculated coefficient is the largest, and determines the dependent variable in the room where the air conditioner corresponding to the determined dependent variable is located. It is determined that a device corresponding to the time-series data is arranged.

FIG. 25 is a diagram illustrating an example of a result of a multiple regression analysis performed by the abnormality detection device according to the present embodiment. FIG. 25 shows only four data (Data-01, Data-06, Data-07 and Data-20) out of the 20 time-series data Data-01 to Data-20 excluding the air conditioners A to C. The calculated coefficients are shown.

For example, for the time-series data Data-01 (independent variable X1), the coefficient at the time-series data Data-21 is the largest. For this reason, it is determined that the device (specifically, the mounting device) corresponding to the time-series data Data-01 is located in the same space as the air conditioner A. In FIG. 25, cells that represent the combination with the largest coefficient are shaded with dots.

Here, for example, it is assumed that time-series data and a room (or an air conditioner) are associated with each other as follows. Specifically, the time-series data located in the same room as the air conditioner A includes time-series data Data-01, Data-02, Data-03, Data-04, Data-05, Data-07, Data-11, Data-11. -13, Data-14 and Data-18. The time-series data located in the same room as the air conditioner B are time-series data Data-06, Data-09, Data-10, Data-16, and Data-19. The time-series data located in the same room as the air conditioner C is time-series data Data-08, Data-12, Data-15, Data-17, and Data-20.

Note that the analysis unit 320 may determine a plurality of dependent variables whose calculated coefficients exceed a threshold value for each time-series data constituting the independent variable. That is, the analysis unit 320 may determine a plurality of rooms (air conditioners) corresponding to the time-series data. For example, when the spaces where the air conditioning is controlled by a plurality of air conditioners overlap, the coefficient of the time-series data of the device located in the overlapped space has a large value in each of the plurality of air conditioners.

Next, as shown in FIG. 24, the correction unit 330 corrects the attribute information based on the combination determined in this way (S43).

FIG. 26 is a diagram illustrating an example of the correction of the attribute information by the abnormality detection device according to the present embodiment. The correction unit 330 detects an error of the installation room in the attribute information DB 164 based on the combination of the room and the time-series data described above, and corrects the detected error.

For example, among the time series data that should correspond to the same room as the air conditioner A, the installation room corresponding to five time series data Data-07, Data-11, Data-13, Data-14, and Data-18 is “RoomB”. "And is different from the installation room" RoomA "corresponding to the time-series data of the air conditioner A. Therefore, the correction unit 330 corrects the installation room corresponding to the five time-series data Data-07, Data-11, Data-13, Data-14, and Data-18 to the same “RoomA” as the air conditioner A. The same applies to time-series data corresponding to the same room as the air conditioner B and time-series data corresponding to the same room as the air conditioner C.

After the attribute information has been corrected as described above, as shown in FIG. 24, the detection unit 340 detects an abnormality (S44). The detection of the abnormality may be performed immediately after the attribute information is corrected, or may be performed at a predetermined timing. Further, the detection of the abnormality may be periodically repeated.

More specifically, the detection unit 340 reads data from each of the time-series data DB 162, the attribute information DB 164, and the abnormality detection model DB 166 of the storage unit 160 for each abnormality to be detected, and detects an abnormality based on the read data. The presence / absence and the time when an abnormality has occurred are detected.

In the present embodiment, the detection unit 340 uses an abnormality detection model of a mounting machine and a coating machine. For example, in the abnormality detection model of the mounting machine, when the production number of the components produced by the mounting machine is equal to or greater than the threshold, it is determined that no abnormality has occurred in the mounting machine, and when the production number falls below the threshold, , It is determined that an abnormality has occurred in the mounting machine. In addition, in the abnormality detection model of the coating machine, when the number of components to be supplied to the coating machine is equal to or greater than the threshold value, it is determined that no abnormality has occurred in the coating machine, and when the number of input components falls below the threshold value. Then, it is determined that an abnormality has occurred in the coating machine.

検出 Furthermore, when a plurality of mounting machines are present in the same room, the detection unit 340 determines at least one of the plurality of mounting machines as a device capable of performing an alternative operation of another mounting machine. In addition, when a plurality of applicators exist in the same room, the detection unit 340 determines at least one of the plurality of applicators as a device capable of performing an alternative operation to another applicator.

FIG. 27 is a diagram illustrating an example of an abnormality detection process performed by the abnormality detection device according to the present embodiment. In FIG. 27, the horizontal axis represents time. The vertical axis indicates the number of production machines [pieces] and the number of application machines [pieces]. Here, as shown in FIG. 22, it is assumed that components produced by the mounting machine A1 are supplied to the coating machine A1 and components produced by the mounting machine A2 are supplied to the coating machine A2.

よう As shown in FIG. 27, at 12:00, an abnormality occurred in the mounting machine A2, and the production was stopped. Since the production number of the mounting machine A2 falls below the threshold value and becomes zero, the detection unit 340 detects that an abnormality has occurred in the mounting machine A2. As a result of the analysis by the analysis unit 320, it has been found that the mounting machine A2 and the mounting machine A1 are provided in the same room A together with the coating machine A1 and the coating machine A2. For this reason, when an abnormality has occurred in the mounting machine A2, the detection unit 340 determines the mounting machine A1 as a device capable of performing an alternative operation.

At this time, the supply of components produced by the mounting machine A2 to the coating machine A2 is stopped by stopping the mounting machine A2. Instead, a part of the components produced by the mounting machine A1 is supplied to the coating machine A2. For example, by operating the mounting machine A1 at the maximum capacity, it is possible to produce 1.5 times the number of components in a normal state. Therefore, the coating machine A2 can be supplied with half the number of components from the normal state from the mounting machine A1. The applicator A2 can lower the operation capability as compared with the normal operation and reduce power consumption.

At # 13: 10, the operation of the mounting machine A2 is restored. Thereby, each device can be returned to the normal operation.

As described above, after the abnormality is detected by the detection unit 340, as illustrated in FIG. 24, the output unit 150 outputs information indicating the detected abnormality and the time of occurrence of the abnormality, and The information indicating the measure is output (S45).

FIG. 28 is a diagram showing an example of information output by the abnormality detection device according to the present embodiment. As shown in FIG. 28, as a countermeasure, the maintenance of the abnormally stopped mounting machine A2 and the alternative operation performed by the mounting machine A1 are displayed on a display or the like, as a countermeasure.

For example, by presenting the information shown in FIG. 28 to the manager of the factory 2 in real time, the manager can perform processing for the abnormality in real time. Specifically, the administrator can suppress a decrease in production efficiency by causing the mounting machine A1 to perform an alternative operation. In addition, by performing maintenance on the mounting machine A2 during the period when the mounting machine A1 is performing the alternative operation, it is possible to quickly perform a recovery operation while suppressing a decrease in production efficiency.

As described above, according to the present embodiment, the room in which each device is installed can be correctly determined by analyzing the time-series data, so that not only can abnormalities be detected with high accuracy, but also alternative operations can be performed. Can propose new equipment.

In the following, a case where the correspondence between the rooms provided with the respective devices cannot be correctly determined will be described with reference to FIGS. 29 and 30.

FIG. 29 is a diagram illustrating an example of an abnormality detection process performed by the abnormality detection device according to the comparative example. FIG. 30 is a diagram illustrating an example of information output by the abnormality detection device according to the comparative example. In the abnormality detection device according to the comparative example, since the room is not determined by the multiple regression analysis, it is not possible to determine that the coating machine A1, the coating machine A2, the mounting machine A1, and the mounting machine A2 are provided in the same room. .

Therefore, as shown in FIG. 30, even if information indicating that the mounting machine A2 has stopped abnormally at 12:00 is output, as a countermeasure, it is not possible to propose a device that can perform an alternative operation. Therefore, as a measure, it is output that maintenance of the mounting machine A2 is performed.

(4) Also, since the mounting machine A2 has stopped abnormally, the components produced by the mounting machine A2 are not supplied to the coating machine A2. Therefore, as shown in FIG. 29, shortly after the stop of the mounting machine A2, the number of components supplied to the coating machine A2 becomes zero. Therefore, it is erroneously detected that an abnormality has occurred in the coating machine A2. As a result, as shown in FIG. 30, the fact that the coating machine A2 has stopped due to an abnormality and that maintenance of the coating machine A2 is to be performed as a countermeasure against the abnormality are output.

(4) Therefore, the maintenance of the coating machine A2 in which no abnormality has occurred originally takes time, and the time required for recovery may be long. As described above, according to the abnormality detection device according to the comparative example, the abnormality cannot be accurately detected, and the production efficiency of the manufacturing system may decrease.

(Modification)
Next, a modified example of the abnormality detection device according to the third embodiment will be described. In the present modification, the detection unit 340 detects an abnormality based on a thermal model of each room by the air conditioner. The process of correcting the attribute information except for the difference in the abnormality detection method is the same as in the third embodiment. The following description focuses on differences from the third embodiment, and description of common points is omitted or simplified.

FIG. 31 is a diagram showing a configuration of a thermal model that is an object of the abnormality detection device 300 according to the present modification. Based on the thermal model shown in FIG. 31, heat transfer in the space where the air conditioner is provided is represented by the following equation (2).

In the equation (2), _Iac is the power consumption of the air conditioner. P _k is the power consumption of each device provided in the same room as the air conditioner. The term obtained by multiplying the total value (ΣP _k ) of the power consumption of each device by the coefficient K corresponds to the total heat generation of the devices provided in the same room as the air conditioner.

R corresponds to the thermal resistance of the room provided with the air conditioner. _Vr is the room temperature of the room provided with the air conditioner. V _o is the outside air temperature. Further, when the heat capacity of the room air conditioner is provided with Q, the heat capacity Q is represented by the product of the time derivative value of the coefficient C and V _r -V _o.

Detection unit 340, the dependent variable _{I ac} of the formula (2), and .SIGMA.P _{_k,} and _(V r -V _{_o),} as independent variables and the time differential value of _(V r -V _o), a multiple regression Perform analysis. As a result, an optimal combination of the coefficients K, 1 / R and C is obtained. In this modification, the acquisition unit 110 acquires time-series data of power consumption of each air conditioner, time-series data of power consumption of each device, time-series data of outside temperature, and time-series data of room temperature. I do.

FIG. 32 is a diagram illustrating an example of an abnormality detection process performed by the abnormality detection device according to the present modification. In FIG. 32, the horizontal axis represents time. The vertical axis represents the values of K, R, and C coefficients. Each of K, R, and C becomes a value close to 1 when no abnormality occurs.

The detection unit 340 compares each of K, R, and C with a threshold, and determines that an abnormality has occurred in the thermal environment of the modeled room when at least one of K, R, and C is below the threshold. I do. The threshold value is, for example, 0.800, but is not limited to this.

例 In the example shown in FIG. 32, the thermal resistance R decreases during the period from 14:00 to 15:00. During a period in which the thermal resistance of the room is decreasing, it is estimated that the door or window provided in the room is in an open state. For example, the countermeasure information DB 168 is associated with a countermeasure for checking a window and a door of a room for an abnormality such as a decrease in thermal resistance.

Based on the above detection results, as shown in FIG. 33, the output unit 150 outputs information indicating the detected abnormality, the occurrence time of the abnormality, and a measure for the abnormality. I do. FIG. 33 is a diagram illustrating an example of information output by the abnormality detection device according to the present modification.

(Embodiment 4)
Next, a fourth embodiment will be described.

In the first to third embodiments, an example has been described in which the abnormality detection method is executed by the

abnormality detection device

100, 200, or 300 realized by one computer device. On the other hand, the abnormality detection method according to the present embodiment is executed by distributed processing by a plurality of devices.

FIG. 34 is a block diagram showing a configuration of an abnormality detection system 400 that executes the abnormality detection method according to the present embodiment. As shown in FIG. 34, the abnormality detection system 400 includes an information processing device 401, a data storage device 402, and an information presentation device 403. In the example illustrated in FIG. 34, each device included in the abnormality detection system 400 is provided between the network 404 and each manufacturing device, that is, inside the network 404. The network 404 is a communication network for connecting to an external device such as the Internet, for example.

The abnormality detection method according to the present embodiment may be executed by abnormality detection system 410 shown in FIG. FIG. 35 is a block diagram showing a configuration of an abnormality detection system 410 that executes the abnormality detection method according to the present embodiment.

As shown in FIG. 35, the abnormality detection system 410 includes an information processing device 401, a data storage device 402, and an information presentation device 403. Each device included in the abnormality detection system 410 is provided outside the network 404.

As described above, each device included in the

abnormality detection system

400 or 410 may be provided inside or outside the network 404. For example, only the information processing device 401 may be provided inside the network 404, and the data storage device 402 and the information presentation device 403 may be provided outside the network 404. Alternatively, only the data storage device 402 may be provided inside the network 404, and the information processing device 401 and the information presentation device 403 may be provided outside the network 404. Alternatively, only the information presentation device 403 may be provided inside the network 404, and the information processing device 401 and the data storage device 402 may be provided outside the network 404.

FIG. 36 is a block diagram showing a configuration of an abnormality detection system 400 according to the present embodiment. The configuration of the abnormality detection system 410 is the same as that of the abnormality detection system 400.

36, as shown in FIG. 36, the information processing apparatus 401 includes an acquisition unit 110, an analysis unit 120, a correction unit 130, a detection unit 140, and an output unit 150. The information processing device 401 is, for example, a computer device. The information processing device 401 is realized by a nonvolatile memory storing a program, a volatile memory serving as a temporary storage area for executing the program, an input / output port, a processor executing the program, and the like.

The data storage device 402 is a storage device that stores the time-series data DB 162, the attribute information DB 164, the abnormality detection model DB 166, and the countermeasure information DB 168. The data storage device 402 is realized by an HDD, a semiconductor memory, or the like.

The information presenting device 403 is a device that presents information output from the output unit 150 to a user such as a factory manager. The information presentation device 403 is, for example, a display or a speaker.

As described above, in this embodiment, the abnormality detection method is distributed by a plurality of devices. The processing amount in each device is reduced, and processing can be performed in a short period of time by distributed processing.

Note that each of the information processing device 401 and the data storage device 402 may not be a single device, and may be implemented by a plurality of devices.

(Other embodiments)
As described above, the abnormality detection method, the information processing apparatus, and the abnormality detection system according to one or more aspects have been described based on the embodiments, but the present disclosure is not limited to these embodiments. Unless departing from the gist of the present disclosure, various modifications conceivable by those skilled in the art are applied to the present embodiment, and forms configured by combining components in different embodiments are also included in the scope of the present disclosure. It is.

For example, a correction candidate may be presented to a user, and the user (administrator) may make a correction. For example, the abnormality detection device may include an input interface that receives a correction instruction from a user.

(4) For example, the detection unit may determine the type of the abnormality by clustering the detected abnormality. By determining the type of the abnormality, the accuracy of estimating the cause of the abnormality and selecting a measure for the abnormality can be further improved.

分析 Also, for example, the analysis may be an analysis method other than the statistical analysis. Any analysis method may be used as long as the mutual relationship between a plurality of time-series data can be determined.

The communication method between the devices described in the above embodiment is not particularly limited. When wireless communication is performed between devices, the wireless communication method (communication standard) is, for example, ZigBee (registered trademark), Bluetooth (registered trademark), or short-range wireless communication such as wireless LAN (Local Area Network). is there. Alternatively, the wireless communication method (communication standard) may be communication via a wide area communication network such as the Internet. Wired communication may be performed between the devices instead of wireless communication. Specifically, the wired communication is power line communication (PLC) or communication using a wired LAN.

In addition, in the above embodiment, another processing unit may execute the process executed by the specific processing unit. Further, the order of the plurality of processes may be changed, or the plurality of processes may be executed in parallel. The distribution of the components included in the abnormality detection system to a plurality of devices is an example. For example, components included in one device may be included in another device. Further, the abnormality detection system may be realized as a single device.

For example, the processing described in the above embodiment may be realized by centralized processing using a single device (system), or may be realized by distributed processing using a plurality of devices. Good. In addition, the number of processors that execute the program may be one or more. That is, centralized processing or distributed processing may be performed.

Further, in the above embodiment, all or a part of the components such as the control unit may be configured by dedicated hardware, or may be realized by executing a software program suitable for each component. Is also good. Each component may be realized by a program execution unit such as a CPU (Central Processing Unit) or a processor reading and executing a software program recorded on a recording medium such as an HDD (Hard Disk Drive) or a semiconductor memory. Good.

The components such as the control unit may be configured by one or a plurality of electronic circuits. Each of the one or more electronic circuits may be a general-purpose circuit or a dedicated circuit.

The one or more electronic circuits may include, for example, a semiconductor device, an IC (Integrated Circuit), an LSI (Large Scale Integration), and the like. The IC or LSI may be integrated on one chip, or may be integrated on a plurality of chips. Here, the term is referred to as an IC or an LSI, but the term varies depending on the degree of integration, and may be called a system LSI, a VLSI (Very Large Scale Integration), or a ULSI (Ultra Large Scale Integration). Further, an FPGA (Field Programmable Gate Array) programmed after manufacturing the LSI can be used for the same purpose.

In addition, general or specific aspects of the present disclosure may be realized by a system, an apparatus, a method, an integrated circuit, or a computer program. Alternatively, the present invention may be realized by a non-transitory computer-readable recording medium such as an optical disk, an HDD, or a semiconductor memory in which the computer program is stored. Further, the present invention may be realized by any combination of a system, an apparatus, a method, an integrated circuit, a computer program, and a recording medium.

In addition, in each of the above embodiments, various changes, replacements, additions, omissions, and the like can be made within the scope of the claims or equivalents thereof.

(4) The present disclosure can be used as an abnormality detection method that can detect an abnormality with high accuracy, and can be used, for example, for manufacturing management and maintenance in a factory or the like.

1, 2

factory

100, 200, 300

abnormality detection device

110, 210

acquisition unit

120, 220, 320

analysis unit

130, 230, 330

correction unit

140, 240, 340 detection unit 150

output unit

160, 260

storage unit

162, 262 Series data DB
164,264 Attribute information DB
166 Anomaly detection model DB
168 Measure information DB
400, 410 abnormality detection system 401 information processing device 402 data storage device 403 information presentation device 404 network

Claims

A plurality of time-series data obtained from a plurality of elements related to the manufacture of the product, and a step of acquiring each attribute information of the plurality of time-series data,
Analyzing the interrelationship of the plurality of time-series data,
Correcting the attribute information based on the result of the analysis,
Based on the corrected attribute information and the plurality of time-series data, detecting an abnormality occurring in the plurality of elements,
A method of outputting information indicating the detected abnormality and the time of occurrence of the abnormality.
The abnormality detection method according to claim 1, wherein the analysis is a statistical analysis.
The abnormality detection method according to claim 2, wherein the statistical analysis is a cross-correlation analysis.
In the step of performing the analysis,
Calculating a cross-correlation coefficient of the plurality of time-series data,
Classifying the plurality of time-series data into a plurality of groups based on the calculated cross-correlation coefficient,
In the correcting step, for each classified group, a plurality of time-series data included in the group is classified into a minority and a majority based on corresponding attribute information, and the time-series data belonging to the minority 4. The abnormality detection method according to claim 3, wherein the attribute information is corrected to attribute information of time-series data belonging to the majority.
In the correcting step, when the number of the time-series data belonging to the minority is equal to or less than half the number of the time-series data belonging to the majority, the attribute information of the time-series data belonging to the minority is changed to the majority. The abnormality detection method according to claim 4, wherein the attribute information is corrected to attribute information of time-series data belonging to a group.
The abnormality detection method according to claim 2, wherein the statistical analysis is a multiple regression analysis.
The plurality of elements includes a plurality of compressors with an output from each integrated;
In the obtaining step, a plurality of pressure data as time series data of pressure from each of the plurality of compressors and a plurality of flow rate data as time series data of flow rate, and integrated time series data of pressure after integration. The pressure data and the integrated flow rate data that is the time series data of the flow rate are obtained as the plurality of time series data,
In the step of performing the analysis,
A plurality of combinations of the plurality of pressure data and the plurality of flow rate data, and a plurality of attribute information candidates corresponding to each of the plurality of pressure data and the plurality of flow rate data,
The multiple regression analysis is performed for each created combination, using a plurality of pressure data and a plurality of flow rate data constituting the combination as independent variables, and using the integrated pressure data and the integrated flow rate data as dependent variables. Anomaly detection method.
The plurality of elements are:
Air conditioning equipment arranged in each of the plurality of spaces;
Each including a plurality of manufacturing devices arranged in any one of the plurality of spaces,
In the acquiring step, time-series data of power consumption of each of the plurality of air conditioners and time-series data of each parameter of the plurality of manufacturing apparatuses are obtained as the plurality of time-series data,
The abnormality detection method according to claim 6, wherein, in the step of performing the analysis, a multiple regression analysis is performed using time-series data of the parameter as an independent variable and time-series data of the power consumption as a dependent variable.
The abnormality detection method according to any one of claims 1 to 8, wherein in the outputting step, information indicating a cause of the detected abnormality is further output.
The abnormality detection method according to any one of claims 1 to 9, wherein in the outputting step, information indicating a measure for the abnormality, which is determined according to the detected abnormality, is further output.
A plurality of time-series data obtained from a plurality of elements related to the manufacture of a product, and an acquisition unit that obtains attribute information of each of the plurality of time-series data,
An analysis unit that analyzes the mutual relationship between the plurality of time-series data,
A correcting unit that corrects the attribute information based on a result of the analysis,
Based on the corrected attribute information and the plurality of time-series data, a detection unit that detects an abnormality that occurs in the plurality of elements,
An information processing apparatus comprising: an output unit that outputs information indicating a detected abnormality and a time at which the abnormality has occurred.
An information processing apparatus according to claim 11,
A data storage device for storing a plurality of time-series data and a plurality of attribute information obtained by the obtaining unit,
A presentation device that presents information output from the output unit.