Abnormality cause estimation device, learning device, precision diagnosis system, and abnormality cause estimation method

Abstract

The present invention is provided with: a sensor data acquisition unit (10) that acquires a plurality of time-series sensor data collected by a plurality of sensors (300) provided in a plurality of device components constituting a target device; an abnormality detection unit (30) that detects, on the basis of the plurality of pieces of sensor data, a plurality of abnormality detection sensors in which an abnormality has occurred; an abnormality detection order estimation unit (40) that estimates, for the plurality of abnormality detection sensors, an abnormality detection order in which abnormality has been detected to occur; an abnormality propagation path tracking unit (50) that estimates an abnormality propagation order in which the abnormality propagates on the basis of abnormality detection sensor information relating to the plurality of abnormality detection sensors and an estimation structure indicating a dependency relationship between the device components; and an abnormality cause estimation unit (60) that estimates an abnormality cause on the basis of the abnormality detection order and the abnormality propagation order.

Description

Technical Field

[0001] The present disclosure relates to an abnormality cause estimation device, a learning device, a precision diagnosis system, and an abnormality cause estimation method. Background Art

[0002] In equipment such as a unit or a factory, multiple elements (hereinafter referred to as "equipment structural elements") such as multiple machines that constitute the equipment operate in conjunction, and sensor data (variables) collected by sensors attached to and installed on these equipment structural elements also have certain correlations. Therefore, when an abnormality occurs in one of the multiple equipment structural elements that constitute the equipment and the abnormality is detected, the influence of the abnormality propagates and the abnormality is detected in multiple sensor data. In such a case, it is not easy to identify (accurately diagnose) the equipment structural element that is the source of the abnormality.

[0003] Therefore, conventionally, there is known a technology that can estimate the device component causing the abnormal operation state by using sensor data collected by sensors of the device components installed in the device when multiple device components in the device are in an abnormal operation state.

[0004] For example, Patent Document 1 discloses an abnormality diagnosis system that estimates a part that is the cause of a state change in a device based on state changes detected by multiple detection units set according to each part (machine) of the device based on changes in the relationship between multiple operating data related to the part being the target, and relationship information between detection units that stores the propagation relationship of the influence between the parts of the device corresponding to the detection units.

[0005] Prior art literature

[0006] Patent Literature

[0007] Patent Document 1: International Publication No. 2017 / 159016 Summary of the invention

[0008] Problem that the invention aims to solve

[0009] In the prior art disclosed in Patent Document 1, an operator or the like is required to grasp the propagation relationship of influences between a plurality of equipment components in advance and to prepare information equivalent to inter-detection unit relationship information defining the propagation relationship based on the grasped propagation relationship.

[0010] On the other hand, for example, in a complex facility or a large-scale facility that has feedback control, it is difficult for an operator or the like to understand the propagation relationship of influences between machines constituting the facility.

[0011] Therefore, in the conventional technology, the propagation relationship of the influence between a plurality of device components cannot be understood, or even if it can be understood, the accuracy is low, and thus there is a problem that the cause of an abnormality occurring in the device may not be estimated.

[0012] The present disclosure has been made to solve the above problems, and an object of the present disclosure is to provide an abnormality cause estimation device that can estimate the cause of an abnormality occurring in a device regardless of the complexity of the device or the scale of the device.

[0013] Means used to solve problems

[0014] The abnormality cause estimation device disclosed in the present invention comprises: a sensor data acquisition unit, which acquires a plurality of time series of sensor data collected by a plurality of sensors provided in a plurality of device structural elements constituting an object device; an abnormality detection unit, which detects a plurality of abnormality detection sensors having abnormalities among the plurality of sensors based on the plurality of sensor data acquired by the sensor data acquisition unit; an abnormality detection sequence estimation unit, which estimates the abnormality detection sequence of the plurality of abnormality detection sensors, based on the detection times of the plurality of abnormality detection sensors detected by the abnormality detection unit; an abnormality propagation path tracking unit, which estimates the abnormality propagation sequence of the abnormality propagation based on the abnormality detection sensor information related to the plurality of abnormality detection sensors detected by the abnormality detection unit and the estimated structure showing the dependency relationship between the device structural elements; and an abnormality cause estimation unit, which estimates the cause of the abnormality based on the abnormality detection sequence estimated by the abnormality detection sequence estimation unit and the abnormality propagation sequence estimated by the abnormality propagation path tracking unit.

[0015] Effects of the Invention

[0016] According to the present disclosure, since it is configured as described above, the abnormality cause estimation device can estimate the cause of abnormality occurring in a device regardless of the complexity of the device or the scale of the device. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a diagram showing a configuration example of a precision diagnosis system including the abnormality cause estimation device according to the first embodiment.

[0018] Figure 2 This is a diagram showing a configuration example of the abnormality cause estimation device according to the first embodiment.

[0019] Figure 3 This is a diagram for explaining the structure of sensor data in the first embodiment.

[0020] Figure 4 This is a diagram for explaining a specific example of the abnormality propagation order estimation process performed by the abnormality propagation path tracking unit in the first embodiment.

[0021] Figure 5 This is another diagram for explaining a specific example of the abnormality propagation order estimation process performed by the abnormality propagation path tracking unit in the first embodiment.

[0022] Figure 6 This is another diagram for explaining a specific example of the abnormality propagation order estimation process performed by the abnormality propagation path tracking unit in the first embodiment.

[0023] Figure 7 This is a diagram for explaining a specific example of abnormality cause estimation processing performed by the abnormality cause estimation unit in the first embodiment.

[0024] Figure 8 This is a diagram for explaining a screen example of an abnormality cause estimation result screen that the abnormality cause estimation result output unit causes the display device to display in the first embodiment.

[0025] Fig. 9 This is a diagram for explaining another screen example of the abnormality cause estimation result screen that the abnormality cause estimation result output unit causes the display device to display in the first embodiment.

[0026] Fig. 10A is a diagram showing an example of the contents of the abnormality detection order estimation result, Fig. 10B is a diagram showing an example of the contents of the abnormality propagation order estimation result, Fig. 10C This is a diagram showing an example of the content of the abnormality cause order estimation result.

[0027] Fig.11 This is a diagram showing a configuration example of a learning device according to Embodiment 1.

[0028] Fig.12 This is a diagram showing the concept of an example of a learning process in which the relationship structure learning unit learns the relationship structure in the first embodiment.

[0029] Fig.13 This is a flowchart for explaining the operation of the abnormality cause estimation device according to the first embodiment.

[0030] Fig.14 This is a flowchart for explaining the operation of the learning device according to the first embodiment.

[0031] Fig.15 It is used to illustrate Fig.14 This is a flowchart of the details of the processing of step ST23.

[0032] Fig.16A and Fig. 16B This is a diagram showing an example of the hardware configuration of the abnormality cause estimation device 100 according to the first embodiment.

[0033] Fig.17This is a diagram showing a configuration example of a precision diagnosis system in which the abnormality cause estimation device and the learning device in the first embodiment include a common sensor data acquisition unit and a common data storage unit.

[0034] Fig.18 This is a diagram showing a configuration example of a precision diagnosis system in which a learning device learns a correlation structure for each operating state of a target device in Implementation 1, and an abnormality cause estimation device estimates the cause of the abnormality based on the correlation structure corresponding to the operating state of the target device learned by the learning device.

[0035] Fig.19 This is a diagram showing a configuration example of an abnormality cause estimation device including a correlation structure correction unit in the first embodiment.

[0036] Fig. 20 This is a diagram for explaining the concept of an example of a process of correcting a related structure by a related structure correcting unit in the abnormality cause estimation device including the related structure correcting unit in the first embodiment.

[0037] Fig.21 This is a diagram showing a configuration example of an abnormality cause estimation device including a relationship change estimation unit in the first embodiment.

[0038] Fig. 22 This is a diagram for explaining the concept of an example of relationship change order estimation processing performed by the relationship change estimation unit based on the learned relationship structure and the relationship structure at the time of abnormality when the abnormality cause estimation device configured as the first embodiment includes a relationship change estimation unit.

[0039] Fig.23 This is a diagram showing a configuration example of an abnormality cause estimation device including an abnormality cause device estimation unit in the first embodiment and having a configuration for estimating the cause of abnormality for each device.

[0040] Fig.24 This is a conceptual diagram for illustrating an example of machine unit abnormal cause estimation processing in which the abnormal cause machine estimation unit estimates the abnormal cause on a machine basis based on machine-attached sensor information, abnormal detection order estimation results, and abnormal propagation order estimation results, when the abnormal cause estimation device configured as implementation mode 1 includes an abnormal cause machine estimation unit.

[0041] Fig.25 This is a diagram showing a screen example of an abnormality cause device estimation result screen that the abnormality cause estimation result output unit causes the display device to display in the first embodiment.

[0042] Fig.26This is a diagram showing a configuration example of an abnormality cause estimation device that includes a correlation structure map output unit and is configured to output correlation structure map display information to a display device in the first embodiment.

[0043] Fig. 27 This is a diagram for explaining an example of a diagram screen displayed on a display device when the correlation structure diagram output unit outputs the correlation structure diagram display information in the case where the abnormality cause estimation device is configured to include a correlation structure diagram output unit in the first embodiment.

[0044] Fig.28 This is a diagram showing an example of the content of the association structure.

[0045] Fig.29 This is a diagram showing an example of the content of abnormality detection sensor information.

[0046] Fig.30 This is a diagram showing an example of the content of the common cause order estimation result.

[0047] Fig.31 This is a flowchart for explaining an example of the operation of the abnormality cause estimation device in the case where the abnormality cause estimation device is configured to include a correlation structure map output unit in the first embodiment.

[0048] Fig.32 This is a diagram showing a configuration example of a learning device including a learning sensor pair generating unit in the first embodiment.

[0049] Fig.33 This is a diagram for explaining the concept of an example of a method in which the learning sensor pair generating unit generates sensor pair information based on device design information when the learning device is configured to include a learning sensor pair generating unit in Embodiment 1. DETAILED DESCRIPTION

[0050] Hereinafter, in order to explain the present disclosure in more detail, the mode for implementing the present disclosure will be described with reference to the drawings.

[0051] Implementation method 1.

[0052] The abnormality cause estimation device of the first embodiment is used for all devices such as power generation units or FA (Factory Automation) systems, in which some abnormalities appear in the sensor data collected in the device. In addition, the sensor data is collected by sensors provided in a plurality of elements constituting the device (hereinafter referred to as "device structural elements"). In the first embodiment, the device structural element is assumed to be a machine, for example. One or more sensors 300 are provided in one machine. In the following first embodiment, for convenience, as an example, it is assumed that one sensor is provided in one machine.

[0053] For example, the abnormality cause estimation device monitors the sensor data collected in the equipment that is the monitoring object, that is, the object of detecting the occurrence of abnormality (hereinafter referred to as "target equipment"), and detects multiple sensors that detect the occurrence of abnormality (hereinafter referred to as "abnormality detection sensors") based on the sensor data. In addition, as described above, in the equipment, the equipment structural elements (here, machines) that constitute the equipment act in conjunction, and the sensor data collected by the sensors attached to these machines and installed in the machines also have certain correlations. In the case where an abnormality occurs in a device structural element among multiple device structural elements and the abnormality is detected, the impact of the abnormality is propagated, and the abnormality is detected in multiple sensor data, in other words, in multiple sensors.

[0054] Then, when the abnormality cause estimation device detects multiple abnormality detection sensors, it estimates the abnormality cause based on the sensor data collected by the multiple abnormality detection sensors, and prompts the operator such as the on-site maintenance personnel of the target equipment with information related to the estimation result. For example, the abnormality cause estimation device prompts the operator with information related to the estimated result of the abnormality cause by causing the display device to display the information related to the estimated result of the abnormality cause. The abnormality cause estimation device prompts the operator with information related to the estimated result of the abnormality cause, for example, in a manner that enables the operator to grasp the sensors installed in the machine that becomes the abnormality cause or the order in which the operator should inspect. As a result, the abnormality cause estimation device can reduce unnecessary inspection work performed by the operator and reduce the operator's load. In addition, the abnormality cause estimation device can estimate the abnormality cause with a quantitative indicator that does not rely on human subjectivity and prompt the basis for the estimation.

[0055] Figure 1 1 is a diagram showing a configuration example of a precision diagnosis system 1000 including the abnormality cause estimation device 100 according to the first embodiment.

[0056] The precise diagnosis system 1000 includes an abnormality cause estimation device 100, a learning device 200, a sensor 300, and a display device 400. In the first embodiment, the sensor 300 and the display device 400 are provided in the precise diagnosis system 1000, but this is only an example. The precise diagnosis system 1000 does not necessarily have to include the sensor 300 and the display device 400, and the sensor 300 and the display device 400 may be provided in a system connected to the precise diagnosis system 1000 outside the precise diagnosis system 1000.

[0057] In addition, Figure 1 In the figure, for the sake of simplicity of explanation, only one sensor 300 is shown, but there may be a plurality of sensors 300. The abnormality cause estimation device 100 is connected to the plurality of sensors 300. In addition, there may be a plurality of display devices 400.

[0058] The abnormality cause estimation device 100 is connected to the learning device 200 , the sensor 300 , and the display device 400 .

[0059] The abnormality cause estimation apparatus 100 estimates the cause of an abnormality occurring in a target device (not shown).

[0060] Specifically, the abnormality cause estimation device 100 detects the sensor 300 (hereinafter referred to as the "abnormality detection sensor") where an abnormality has occurred based on the sensor data obtained from the sensor 300 and the learned association structure generated by the learning device 200, tracks the propagation path of the abnormality between the abnormality detection sensors, and thereby estimates the cause of the abnormality occurring in the target device.

[0061] In the first embodiment, the abnormality cause estimation device 100 "estimates the cause of the abnormality" means estimating the abnormality cause score indicating the degree of possibility of the source of the abnormality and the abnormality cause order based on the abnormality cause score in units of sensors 300, and generating information related to the abnormality cause score and the abnormality cause order.

[0062] Then, the abnormality cause estimation device 100 displays information related to the estimated abnormality cause on the display device 400 .

[0063] The details of the abnormality cause estimation device 100 and related structures will be described later.

[0064] The learning device 200 estimates the association structure using sensor data collected by the sensor 300 provided in the target device during normal operation of the target device. In the first embodiment, the estimation of the association structure performed by the learning device 200 is also referred to as "learning". That is, the "learned association structure" used by the abnormality cause estimation device 100 when estimating the cause of the abnormality occurring in the target device can be said to be the "estimated structure" as the association structure estimated by the learning device 200. The association structure is information showing the dependency relationship of multiple device structural elements constituting the target device. The association structure shows the dependency relationship of the device structural elements by showing the dependency relationship of the sensors provided in the device structural elements. The association structure is, for example, information showing the dependency relationship of multiple device structural elements constituting the target device in a matrix. The association structure can also be information showing the dependency relationship of multiple device structural elements constituting the target device in the form of JSON (JavaScript (registered trademark) Object Notification), which is a typical description method of words. In the first embodiment, as an example, the association structure constituting the target device is information showing the dependency relationship of multiple device structural elements in a matrix.

[0065] In addition, the normal operation of the target device is specifically the normal operation of the multiple device components (here, machines) constituting the target device. Therefore, the sensor data collected by the sensor 300 during the normal operation of the target device is specifically the sensor data collected by the sensor 300 installed in each machine during the normal operation of the multiple machines constituting the target device.

[0066] The details of the learning device 200 will be described later.

[0067] The sensor 300 is provided in a plurality of device components (here, devices) constituting the target device.

[0068] The sensor 300 outputs the sensor data to the abnormality cause estimation device 100 .

[0069] The sensor data is, for example, time series data of sensor measurement values ​​obtained at predetermined intervals for a predetermined amount of time by the sensor 300 provided in each device as a device component constituting the target device. The sensor data, for example, represents at least one sensor measurement value of the opening, deviation, rotation speed, conductivity, flow rate, pressure, temperature, concentration, or water level. In addition, this is only an example, and the sensor data may also include control values ​​such as command values ​​or reference values ​​obtained at predetermined intervals for a predetermined amount of time by a plurality of sensors 300.

[0070] In the following first embodiment, the sensor data is assumed to be at least one time series data of sensor measurement values ​​such as opening, deviation, rotation speed, conductivity, flow rate, pressure, temperature, concentration or water level obtained by multiple sensors 300 at specified intervals for a specified time period.

[0071] The display device 400 is, for example, a display included in a PC (Personal Computer) installed in a management room or the like where an operator performs work. The display device 400 may be, for example, a touch panel display included in a tablet terminal carried by an operator.

[0072] First, the abnormality cause estimation device 100 according to the first embodiment will be described.

[0073] Figure 2 1 is a diagram showing a configuration example of the abnormality cause estimation device 100 according to the first embodiment.

[0074] In addition, Figure 2 In the figure, the illustration of the learning device 200 is omitted.

[0075] and Figure 1 Likewise, in Figure 2In the figure, for the sake of simplicity, only one sensor 300 is shown, but this is just an example. A plurality of sensors 300 may be connected to the abnormality cause estimation device 100. In the first embodiment, it is assumed that a plurality of sensors 300 are connected to the abnormality cause estimation device 100. In the following first embodiment, the plurality of sensors 300 are also referred to as sensors 300.

[0076] The abnormality cause estimation device 100 includes a sensor data acquisition unit 10 , a data storage unit 20 , an abnormality detection unit 30 , an abnormality detection order estimation unit 40 , an abnormality propagation path tracking unit 50 , an abnormality cause estimation unit 60 , and an abnormality cause estimation result output unit 70 .

[0077] The sensor data acquisition unit 10 acquires sensor data from the sensor 300 .

[0078] As described above, in embodiment 1, the sensor data is time series data of sensor measurement values ​​(for example, at least one sensor measurement value of opening, deviation, rotation speed, conductivity, flow rate, pressure, temperature, concentration or water level) obtained at specified intervals for a specified time period from sensors 300 installed in multiple machines as structural elements of the equipment.

[0079] For example, when the number of sensors 300 is represented by n, the sensors 300 are represented by X1, X2, X3, X4, ..., Xn. In addition, when sensor data is collected at each of time 1, time 2, ..., time t, the sensor data is represented by a two-dimensional data frame whose number of behavior times is t and whose number of columns is n is sensors. The sensor data of the first sensor X1 at time 1 is represented by X11, the sensor data of the second sensor X2 at time 1 is represented by X21, and the sensor data of the first sensor X1 at time 2 is represented by X12 (refer to Figure 3 ) In addition, in the following first embodiment, the sensor data acquired by the sensor data acquisition unit 10 is referred to as “sensor data D1”.

[0080] The sensor data acquisition unit 10 stores the acquired sensor data D1 in the data storage unit 20 .

[0081] The abnormality detection unit 30 performs abnormality detection processing on the sensor data D1 stored in the data storage unit 20 by the sensor data acquisition unit 10 .

[0082] Specifically, the abnormality detection unit 30 performs abnormality detection processing using a well-known univariate abnormality detection method on the sensor data D1 as time series data stored in the data storage unit 20 by the sensor data acquisition unit 10, and detects a plurality of abnormality detection sensors in the sensor 300. In the first embodiment, the occurrence of an abnormality in the sensor 300 means that the value of the sensor data collected by the sensor 300 is abnormal. That is, the abnormality detection sensor refers to the sensor 300 whose value of the sensor data collected by the sensor 300 is abnormal. In addition, in the first embodiment, the occurrence of an abnormality in the sensor 300 means that the abnormality occurs in the device provided with the sensor 300.

[0083] As well-known univariate anomaly detection methods, Discord (non-patent literature: KEOGH, Eamonn; LIN, Jessica; FU, Ada. Hot sax: Efficiently finding the most unusual time series subsequence. In: Data mining, fifth IEEE international conference on. IEEE, 2005.) or Hotelling's T^2 theory are cited.

[0084] In addition, in the following first embodiment, a plurality of abnormality detection sensors detected by the abnormality detection unit 30 are also simply referred to as “abnormality detection sensors”.

[0085] The abnormality detection unit 30 stores information related to the abnormality detection sensor (hereinafter referred to as “abnormality detection sensor information”) and information related to the detection time when the abnormality detection sensor detects the abnormality (hereinafter referred to as “abnormality detection time information”) in the data storage unit 20 .

[0086] In the following embodiment 1, the abnormality detection sensor information is referred to as "abnormality detection sensor information D3", and the abnormality detection time information is referred to as "abnormality detection time information D4". The abnormality detection sensor information D3 is information indicating the abnormality detection sensor. The information indicating the abnormality detection sensor is information that can identify the abnormality detection sensor, such as an ID assigned to each sensor 300. The abnormality detection time information D4 is information that associates information that can identify the abnormality detection sensor with the time when the abnormality detection sensor is detected.

[0087] In addition, the abnormality detection unit 30 may also summarize the abnormality detection sensor information D3 and the abnormality detection time information D4 as information corresponding to the information indicating the abnormality detection sensor and the time when the abnormality detection sensor is detected (hereinafter referred to as "abnormality detection result"). In this case, the abnormality detection unit 30 stores the abnormality detection result in the data storage unit 20.

[0088] Here, the abnormality detection unit 30 acquires the sensor data D1 from the sensor data acquisition unit 10 via the data storage unit 20 , but this is just an example. The abnormality detection unit 30 may acquire the sensor data D1 directly from the sensor data acquisition unit 10 .

[0089] The abnormality detection sequence estimation unit 40 obtains the abnormality detection sensor information D3 and the abnormality detection time information D4 stored in the data storage unit 20 by the abnormality detection unit 30, and performs abnormality detection sequence estimation processing. In this abnormality detection sequence estimation processing, for the abnormality detection sensor, the order in which the abnormalities are detected is estimated, more specifically, the order in which the abnormalities are detected in the sensor data D1 collected by the abnormality detection sensor (hereinafter referred to as "abnormality detection sequence").

[0090] Specifically, the abnormality detection order estimation unit 40 assigns the abnormality detection order to the abnormality detection sensors in order of abnormality detection time from earliest to latest, based on the abnormality detection sensor information D3 and the abnormality detection time information D4 .

[0091] Specifically, the abnormality detection order estimation unit 40 assigns the abnormality detection order on to the sensor Xn detected as the abnormality detection sensor. Here, the abnormality detection order on is a real value. For example, the abnormality detection order estimation unit 40 assigns the abnormality detection order on in ascending order, starting from the sensor Xn with the earliest corresponding abnormality detection time. For example, when there are multiple sensors Xn with the same corresponding abnormality detection time, the abnormality detection order estimation unit 40 assigns the same abnormality detection order on to the multiple sensors Xn with the same abnormality detection time. For example, the abnormality detection order estimation unit 40 may also assign the abnormality detection order "0" to the sensor Xn with the earliest corresponding abnormality detection time, and then assign the elapsed time from the time corresponding to the abnormality detection order "0" to the other sensors Xn as the abnormality detection order on.

[0092] The abnormality detection order estimation unit 40 stores the result obtained by assigning the abnormality detection order (hereinafter referred to as "abnormality detection order estimation result") in the data storage unit 20. In the following embodiment 1, the abnormality detection order estimation result is set as "abnormality detection order estimation result D5". The abnormality detection order estimation result D5 is information obtained by matching information indicating the abnormality detection sensor, the abnormality detection time, and the information indicating the abnormality detection order.

[0093] Here, the abnormality detection sequence estimation unit 40 obtains the abnormality detection sensor information D3 and the abnormality detection time information D4 from the abnormality detection unit 30 via the data storage unit 20, but this is just an example. The abnormality detection sequence estimation unit 40 may also obtain the abnormality detection sensor information D3 and the abnormality detection time information D4 directly from the abnormality detection unit 30.

[0094] The abnormality propagation path tracking unit 50 obtains the abnormality detection sensor information D3 and the associated structure stored by the abnormality detection unit 30 from the data storage unit 20, and performs an abnormality propagation order estimation process to estimate the order of abnormality propagation (hereinafter referred to as "abnormality propagation order") based on the acquired abnormality detection sensor information D3 and the associated structure. In the abnormality propagation order estimation process, the abnormality propagation path tracking unit 50 estimates the abnormality propagation order for the sensor 300. In addition, for a certain sensor 300, the abnormality detection order assigned by the abnormality detection order estimation unit 40 and the abnormality propagation order estimated by the abnormality propagation path tracking unit 50 are not necessarily the same order. For example, due to the problem of time resolution, even if the same order is assigned to the abnormality detection order, the order relationship of the abnormality propagation order may have a situation of being in the first and last order. In addition, for example, due to the problem of the accuracy of the abnormality detection time, sometimes the order opposite to the abnormality propagation order may be assigned to the abnormality detection order. In addition, for example, due to the problem of the accuracy of the abnormality detection time, even if the same order is assigned to the abnormality detection order, the order relationship of the abnormality propagation order may have a situation of being in the first and last order. Even if it is relatively easy to detect whether an abnormality has occurred, it is difficult to accurately detect the time of its occurrence, and therefore, there may be a problem of accuracy in the time of abnormality detection.

[0095] The association structure is generated based on at least one statistic between a plurality of sensor data through learning in the learning device 200, and is stored in the data storage unit 20. In the following embodiment 1, the association structure is referred to as "association structure D2". As described above, in embodiment 1, the association structure D2 is information showing the dependency relationship between a plurality of devices as a plurality of device structural elements constituting the target device in a matrix.

[0096] In detail, the abnormality propagation path tracking unit 50 determines the dependency relationship between the abnormality detection sensors based on the abnormality detection sensor information D3 and the association structure D2, and tracks the abnormal propagation path of the abnormality along the direction of the dependency relationship. Then, the abnormality propagation path tracking unit 50 assigns the abnormality propagation order from the abnormality detection sensor located upstream of the abnormality propagation, which is considered to be the source of the abnormality, toward the abnormality detection sensor located downstream, such as "1" (the first), "2" (the second), "3" (the third), etc. The abnormality propagation path tracking unit 50 generates an estimation result of the abnormality propagation order (hereinafter referred to as "abnormality propagation order estimation result") and stores it in the data storage unit 20. In the following embodiment 1, the abnormality propagation order estimation result is set as "abnormality propagation order estimation result D6".

[0097] Here, the abnormality propagation order estimation process performed by the abnormality propagation path tracking unit 50 will be described with reference to the drawings and specific examples.

[0098] Figure 4 , Figure 5 as well as Figure 6 This is a diagram for explaining a specific example of the abnormality propagation order estimation process performed by the abnormality propagation path tracking unit 50 in the first embodiment.

[0099] 〈Generation of the Influence Communication Relationship Matrix〉

[0100] In the abnormality propagation order estimation process, the abnormality propagation path tracking unit 50 first converts the association structure D2 into an influence propagation relationship matrix that indicates the presence or absence of dependency between sensor data through a matrix based on the association structure D2 stored in the data storage unit 20. In the first embodiment, the influence propagation relationship matrix is ​​referred to as "influence propagation relationship matrix D9".

[0101] Figure 4 This is a diagram for explaining the concept of an example of processing in which the abnormality propagation path tracking unit 50 converts the correlation structure D2 into the influence propagation relationship matrix D9 in the first embodiment.

[0102] Here, the association structure D2 is a three-dimensional array in which the first dimension is the type m of statistical indicators, the second dimension is the number n of sensors, and the third dimension is the number n of sensors. In addition, the statistical indicator is an indicator that describes the dependency relationship between sensor data. The details of the statistical indicator will be described later.

[0103] Let A(k) be the two-dimensional matrix corresponding to the k-th statistical index, and let a(k)ij be the statistic describing the dependency relationship from the i-th sensor data to the j-th sensor data in the k-th statistical index. In addition, the i-th sensor data is the sensor data collected by the i-th sensor Xi, and the j-th sensor data is the sensor data collected by the j-th sensor Xj.

[0104] The abnormal propagation path tracking unit 50 performs preprocessing of the statistics used for tracking the selected abnormal propagation path (hereinafter referred to as "tracking statistics") for the above-mentioned association structure D2, and generates a preprocessed association structure. In the following embodiment 1, the preprocessed association structure is assumed to be "preprocessed association structure D8". The preprocessed association structure D8 has the same data structure as the association structure D2.

[0105] The abnormal propagation path tracking unit 50 may select only the statistic with a large dependency as the statistic for tracking, for example. In this case, the abnormal propagation path tracking unit 50 sets a threshold value (hereinafter referred to as "statistic selection threshold value") according to each type of statistical index, and selects the statistic a(k)ij whose absolute value |a(k)ij| of the statistic is larger than the statistic selection threshold value as the statistic for tracking. For example, when the absolute value of the statistic is larger than the statistic selection threshold value, the abnormal propagation path tracking unit 50 substitutes the statistic a(k)ij, which is an element of the association structure D2, into the element b(k)ij of the preprocessed association structure D8, and when the absolute value of the statistic is smaller than the statistic selection threshold value, the abnormal propagation path tracking unit 50 substitutes "0" indicating no dependency into the element b(k)ij of the preprocessed association structure D8. In addition, the statistic selection threshold value may be manually set in advance by an operator or the like operating an input device such as a mouse or keyboard (not shown), or may be automatically determined by the abnormal propagation path tracking unit 50 based on sensor data. For example, the abnormality propagation path tracking unit 50 can determine a relative statistic selection threshold value based on the average value (median value or quantile, etc.) of the statistic.

[0106] The abnormality propagation path tracking unit 50 performs a conversion process on the preprocessed correlation structure D8 to convert it into an influence propagation relationship matrix D9, and generates the influence propagation relationship matrix D9.

[0107] The abnormal propagation path tracking unit 50 determines the dependency relationship between the sensor data based on one or more statistical indicators including at least one directed statistical indicator, and converts the preprocessed association structure D8 into the influence propagation relationship matrix D9. In this case, as a method for determining the dependency relationship between the sensor data, the abnormal propagation path tracking unit 50 may, for example, determine that the sensor data have a dependency relationship when at least one of the statistics b(1)ij, b(2)ij, ..., b(m)ij is not "0", or may determine that the sensor data have a dependency relationship when all of the statistics b(1)ij, b(2)ij, ..., b(m)ij are not "0". In detail, for example, when i=1 and j=2, it is also possible to determine that the sensor data collected by sensor X1 and the sensor data collected by sensor X2 have a dependency relationship when at least one of the statistics b(1)12, b(2)12,..., b(m)12 is not "0", and it is also possible to determine that the sensor data collected by sensor X1 and the sensor data collected by sensor X2 have a dependency relationship when all the statistical indicators b(1)12, b(2)12,..., b(m)12 are not "0".

[0108] The statistical indicator used by the abnormal propagation path tracking unit 50 to determine the dependency relationship between sensor data can also be manually selected by an operator or the like from m types of statistical indicators. To give a specific example, for example, the abnormal propagation path tracking unit 50 displays a setting screen of the types of statistical indicators, such as a check box for each type of statistical indicator, on the display device 400. The operator or the like operates an input device such as a mouse or a keyboard to select a statistical indicator from the setting screen. The abnormal propagation path tracking unit 50 accepts the statistical indicator selected by the operator or the like as a statistical indicator for determining the dependency relationship between sensor data.

[0109] In implementation mode 1, as an example, Figure 4 As shown, the influence propagation relationship matrix D9 becomes a two-dimensional matrix with the first dimension being the number n of sensors 300 and the second dimension being the number n of sensors 300 .

[0110] Here, the elements of the association structure D2 and the preprocessed association structure D8 are real values, and the elements of the influence propagation relationship matrix D9 are Boolean values. For example, in the association structure D2 and the preprocessed association structure D8, the greater the absolute value of the element |a(k)ij|, |b(k)ij|, the greater the dependency relationship, and in the influence propagation relationship matrix D9, when the element cij is "1", it means that there is a dependency relationship.

[0111] For example, the abnormal propagation path tracking unit 50 substitutes "1" into cij, which is an element of the influence propagation relationship matrix D9, when there is a dependency relationship from the i-th sensor data to the j-th sensor data; and substitutes "0" into cij, which is an element of the influence propagation relationship matrix D9, when there is no dependency relationship from the i-th sensor data to the j-th sensor data.

[0112] 〈Estimation of the order of anomaly propagation〉

[0113] In the abnormality propagation order estimation process, after generating the influence propagation relationship matrix D9 , the abnormality propagation path tracking unit 50 estimates the abnormality propagation order based on the abnormality detection sensor information D3 acquired from the data storage unit 20 and the generated influence propagation relationship matrix D9 .

[0114] Figure 5 and Figure 6 This is a diagram for explaining the concept of an example of a process in which the abnormality propagation path tracking unit 50 estimates the abnormality propagation order based on the abnormality detection sensor information D3 and the influence propagation relationship matrix D9 in the first embodiment.

[0115] Here, as an example, the number of sensors 300 is set to 6, and the sensor 300 is represented by sensor Xn (n=1, . . . , 6). Also, as an example, it is assumed that sensors X1, X2, X4, and X6 among sensors X1, X2, X3, X4, X5, and X6 are abnormality detection sensors. Also, as an example, the influence propagation relationship matrix D9 is a two-dimensional matrix of 6 rows×6 rows that represents the dependency relationship between sensor data related to sensor Xn.

[0116] First, the abnormal propagation path tracking unit 50 converts the influence propagation relationship matrix D9 into an influence propagation graph D10. Figure 5 As shown, the influence propagation graph D10 is a directed graph in which sensors X1, X2, X3, X4, X5, and X6 are represented as nodes, and the dependency relationship between sensor data related to sensors X1, X2, X3, X4, X5, and X6 is represented as edges. Sensors X1, X2, X3, X4, X5, and X6 correspond to nodes N51, N52, N53, N54, N55, and N56, respectively. For example, when there is a dependency relationship in one direction from sensor X2 to sensor X1, the dependency relationship is represented by an edge of a one-way arrow from node N52 to node N51 in the influence propagation graph D10.

[0117] After converting the influence propagation relationship matrix D9 into the influence propagation graph D10 , the abnormality propagation path tracking unit 50 then converts the influence propagation graph D10 into the abnormality propagation graph D11 based on the abnormality detection sensor information D3 .

[0118] For example, Figure 5 As shown, the abnormality propagation path tracing unit 50 selects only the dependency relationship related to the abnormality detection sensor Xn from among the dependency relationships represented by the influence propagation graph D10, and converts it into the abnormality propagation graph D11. In this case, the abnormality propagation path tracing unit 50 selects only the nodes corresponding to the sensors X1, X2, X3, X4, X5, and X6 and the edges between the nodes corresponding to the sensors X1, X2, X3, X4, X5, and X6 represented by the influence propagation graph D10, and the edges between the nodes corresponding to the abnormality detection sensors X1, X2, X4, and X6. For example, Figure 5 In the example shown, sensor X3 is not included in the abnormality detection sensors X1, X2, X4, and X6. Therefore, the abnormality propagation path tracking unit 50 does not select the node N53 corresponding to the sensor X3. As a result, in the abnormality propagation graph D11, the node N53 corresponding to the sensor X3 is deleted, and the edge between the node N53 and the node N54 associated with the node N53 is also deleted.

[0119] Then, the anomaly propagation path tracing unit 50 estimates the anomaly propagation order based on the anomaly propagation graph D11 .

[0120] For example, the abnormality propagation path tracking unit 50 assigns the abnormality propagation order on to the abnormality detection sensor Xn. Here, the abnormality propagation order on is a real value. The abnormality propagation path tracking unit 50 assigns the abnormality propagation order on in ascending order, for example, from the abnormality detection sensor Xn located upstream of the abnormality propagation, which is considered to be the source of the abnormality.

[0121] At once Figure 5 In the example shown, the abnormality propagation path tracking unit 50 first sets the node that does not lead out a one-way arrow from other nodes as the node located at the most upstream of the abnormality propagation, and assigns the smallest abnormality propagation order on to the abnormality detection sensor Xn corresponding to the node. Figure 5 , nodes N52, N54, and N56 are nodes from which no one-way arrows are drawn from other nodes. Therefore, the abnormal propagation path tracking unit 50 assigns abnormal propagation orders o2, o4, and o6 to the abnormal detection sensors X2, X4, and X6 corresponding to the nodes N52, N54, and N56, respectively. At this time, o2=o4=o6. For example, the abnormal propagation path tracking unit 50 sets o2=o4=o6=1. That is, the abnormal propagation path tracking unit 50 sets the abnormal propagation order of the abnormal detection sensors X2, X4, and X6 to "1" (the first).

[0122] Next, the abnormality propagation path tracking unit 50 assigns an abnormality propagation order on that is greater than the assigned abnormality propagation order on to the abnormality detection sensor Xn corresponding to the node located at the end point of the one-way arrow starting from the node assigned the smallest abnormality propagation order on (here "1"). In addition, the abnormality propagation path tracking unit 50 assigns an abnormality propagation order on to the abnormality detection sensor Xn corresponding to the node located at the end point of the two-way arrow starting from the node assigned the smallest abnormality propagation order on (here "1"). Figure 5 , node N51 is the node at the end point of the one-way arrow from node N52. Therefore, the abnormal propagation path tracking unit 50 assigns the abnormal propagation order o1 to the abnormal detection sensor X1 corresponding to the node N51, that is, the abnormal propagation order o1 greater than "1" (the first). For example, the abnormal propagation path tracking unit 50 sets the abnormal propagation order o1=2, and the abnormal propagation path tracking unit 50 sets the abnormal propagation order of the abnormal detection sensor X1 to "2" (the second). In addition, Figure 5 , there is a node N56 at the end point of the double-headed arrow from the node N52, but the exception propagation order o6=1 has been assigned to the node N56.

[0123] Thereafter, the abnormality propagation path tracking unit 50 repeats the above-described allocation of the abnormality propagation order on in the same manner until the abnormality propagation order on is allocated to the abnormality detection sensors Xn corresponding to all the nodes on the abnormality propagation graph D11.

[0124] However, there may be multiple exception propagation orders assigned by the selection method of the node to be traced, that is, the path of exception propagation. Figure 5 In the example shown, the bidirectional arrow between the node N52 and the node N56 is a unidirectional arrow from the node N56 to the node N52. In this case, there are two candidates for the abnormality propagation order o1 assigned to the abnormality detection sensor X1 corresponding to the node N51. Specifically, as candidates for the abnormality propagation order o1 assigned to the abnormality detection sensor X1, the abnormality propagation order o1 assigned based on the path from the node N54 directly to the node N51 and the abnormality propagation order o1 assigned based on the path from the node N56 to the node N51 via the node N52 are cited as candidates for the abnormality propagation order o1 assigned to the abnormality detection sensor X1. In this case, the abnormality propagation path tracking unit 50, for example, assigns the candidate with a later order among the candidates for the abnormality propagation order o1 to the abnormality propagation order o1. In addition, this is just an example, and the abnormality propagation path tracking unit 50, for example, may also assign the candidate with a higher order to the abnormality propagation order o1 of the node N51.

[0125] The abnormality propagation path tracking unit 50 generates an estimation result of the abnormality propagation order on (hereinafter referred to as "abnormality propagation order estimation result") and stores it in the data storage unit 20. In the following embodiment 1, the abnormality propagation order estimation result is referred to as "abnormality propagation order estimation result D6".

[0126] The abnormality propagation order estimation result D6 is, for example, information representing the abnormality detection sensor Xn (such as Figure 5 ), indicating whether the abnormality detection sensor Xn is included in the abnormality detection sensor flag fn (as shown in D6A in FIG. Figure 5 D6B in FIG. 5 ), and the abnormal propagation order on estimated by the abnormal propagation path tracking unit 50 (as shown in FIG. Figure 5 In addition, Figure 5 In the example shown, the sensor Xn included in the abnormality propagation order estimation result D6 is only the abnormality detection sensor Xn. In addition, the abnormality detection sensor flag fn is a Boolean value. Figure 5 In the example shown, the abnormality detection sensor Xn set in the abnormality propagation order estimation result D6, specifically the abnormality detection sensors X1, X2, X4, and X6 are all abnormality detection sensors. Therefore, the abnormality propagation path tracking unit 50 assigns (True) to the abnormality detection sensor flag fn of the abnormality detection sensors X1, X2, X4, and X6, for example.

[0127] In addition, when using Figure 5 In the estimation of the abnormality propagation order based on the abnormality detection sensor information D3 and the influence propagation relationship matrix D9 performed by the abnormality propagation path tracking unit 50, the abnormality propagation path tracking unit 50 only selects the dependency relationship between the sensor data related to the abnormality detection sensor Xn when converting the influence propagation graph D10 into the abnormality propagation graph D11, but this is just an example.

[0128] For example, Figure 6 As shown, when the abnormality propagation path tracking unit 50 converts the influence propagation graph D10 into the abnormality propagation graph D11, when at least one of the two different sensors Xn is the abnormality detection sensor Xn, the dependency relationship between the sensor data related to the sensor Xn can also be selected.

[0129] In this case, the abnormality propagation path tracking unit 50 selects the nodes N51, N52, N53, N54, N55, N56 corresponding to the sensors X1, X2, X3, X4, X5, X6 represented by the influence propagation graph D10 and the edges between the nodes N51, N52, N53, N54, N55, N56 corresponding to the sensors X1, X2, X3, X4, X5, X6, and the nodes connected to the edges, where at least one of the nodes connected is the edge of the node corresponding to the abnormality detection sensors X1, X2, X4, X6 and the node connected to the edge. For example, the node N53 corresponding to the sensor X3 and the node N54 corresponding to the sensor X4 are connected by the edge of the bidirectional arrow. Here, the sensor X3 is not included in the abnormality detection sensor Xn, but the sensor X4 is included in the abnormality detection sensor Xn. Therefore, in the abnormality propagation graph D11, the edge between the node N53 and the node N45 is not deleted.

[0130] In this case, for example, the abnormality propagation path tracking unit 50 may generate the abnormality propagation graph D11 in such a manner that the nodes N53 and N55 corresponding to the sensors X3 and X5 that are not included in the abnormality detection sensor Xn are known among the nodes N51, N52, N53, N54, and N55 corresponding to the sensors X1, X2, X3, X4, X5, and X6 represented by the abnormality propagation graph D11. Figure 6 In the anomaly propagation graph D11 shown, nodes N51, N52, N54, and N56 are represented by solid-line circles, and nodes N53 and N55 are represented by dotted-line circles.

[0131] Furthermore, in this case, the abnormality propagation path tracking unit 50 estimates the abnormality propagation order on based on the abnormality propagation graph D11 including the nodes N53 and N55 corresponding to the sensors X3 and X5 that are not included in the abnormality detection sensor Xn among the sensors Xn.

[0132] For example, in Figure 6 , nodes N53 and N54 do not have one-way arrows drawn from other nodes. Therefore, the abnormal propagation path tracking unit 50 assigns abnormal propagation orders o3 and o4 to sensors X3 and X4, respectively. At this time, o3=o4. For example, the abnormal propagation path tracking unit 50 sets o3=o4=1. That is, the abnormal propagation path tracking unit 50 sets the abnormal propagation orders o3 and o4 of sensors X3 and X4 to "1" (the first), respectively.

[0133] Next, the abnormality propagation path tracking unit 50 assigns abnormality propagation orders o1 and o5, which are greater than the abnormality propagation order o4, to the sensors X1 and X5 corresponding to the nodes N51 and N55 located at the end point of the one-way arrow from the node N54 to which the smallest abnormality propagation order (here, "1") is assigned, respectively. For example, the abnormality propagation path tracking unit 50 sets o1=o5=2. That is, the abnormality propagation path tracking unit 50 sets the abnormality propagation order of the sensors X1 and X5 to "2" (the second one). In addition, the abnormality propagation path tracking unit 50 assigns abnormality propagation orders o2 and o6, which are greater than the abnormality propagation order o5, to the sensors X2 and X6 corresponding to the nodes N52 and N56 located at the end point of the one-way arrow from the node N55, respectively. For example, the abnormality propagation path tracking unit 50 sets o2=o6=3. That is, the abnormality propagation path tracking unit 50 sets the abnormality propagation order of the sensors X2 and X6 to "3" (the third one).

[0134] In addition, although the abnormality propagation order o1, in other words, "2" (the second) has been assigned to X1 corresponding to the node N51, since the node N51 is the node located at the end point of the one-way arrow starting from the node N52, the abnormality propagation path tracking unit 50 can finally reallocate an order greater than the abnormality propagation order o2, in other words, "3" (the third) to the sensor X1 corresponding to the node N51 as the abnormality propagation order o1. For example, the abnormality propagation path tracking unit 50 can also reallocate the abnormality propagation order o1 to "4" (the fourth).

[0135] In addition, Figure 6 In the example shown, nodes N51, N52, N53, N54, N55, and N56 on the abnormality propagation graph D11 include nodes N53 and N55 corresponding to sensors X3 and X5 that are not included in the abnormality detection sensor Xn. Therefore, the abnormality propagation path tracking unit 50 may weight the abnormality propagation orders o3 and o5 assigned to sensors X3 and X5 after assigning abnormality propagation orders o1, o2, o3, o4, o5, and o6 to sensors X1, X2, X3, X4, X5, and X6, respectively. For example, Figure 6 In the example shown, weights b are added to the abnormality propagation orders o3 and o5 corresponding to the sensors X3 and X5 included in the abnormality propagation order estimation result D6. The weight b is a real value of 0 or more.

[0136] Then, the anomaly propagation path tracking unit 50 stores the anomaly propagation order estimation result D6 in the data storage unit 20 .

[0137] As mentioned above, in Figure 6In the example shown, the sensors Xn (specifically, the sensors X1, X2, X3, X4, X5, and X6) set in the abnormality propagation order estimation result D6 include sensors X3 and X5 that are not included in the abnormality detection sensors Xn (specifically, the abnormality detection sensors X1, X2, X4, and X6). The abnormality propagation path tracking unit 50 assigns (False) to the abnormality detection sensor flags fn of the sensors X3 and X5.

[0138] Here, the abnormality propagation path tracking unit 50 obtains the abnormality detection sensor information D3 from the abnormality detection unit 30 via the data storage unit 20 , but this is just an example. The abnormality propagation path tracking unit 50 may directly obtain the abnormality detection sensor information D3 from the abnormality detection unit 30 .

[0139] Return to Figure 2 A description is given of a configuration example of the abnormality cause estimation device 100 shown.

[0140] The abnormality cause estimation unit 60 obtains the abnormality detection sequence estimation result D5 output by the abnormality detection sequence estimation unit 40 and the abnormality propagation sequence estimation result D6 output by the abnormality propagation path tracking unit 50 from the data storage unit 20, and performs abnormality cause estimation processing. In this abnormality cause estimation processing, the cause of the abnormality is estimated based on the abnormality detection sequence estimated by the abnormality detection sequence estimation unit 40 and the abnormality propagation sequence estimated by the abnormality propagation path tracking unit 50.

[0141] Specifically, the abnormality cause estimation unit 60 calculates an abnormality cause score indicating the degree of possibility of the abnormality source based on the abnormality detection order estimation result D5 and the abnormality propagation order estimation result D6, and assigns an order based on the calculated abnormality cause score (hereinafter referred to as "abnormality cause order"). The higher the degree of possibility of the abnormality source, the smaller the abnormality cause score becomes.

[0142] First, the abnormality cause estimation unit 60 obtains the abnormality detection order estimation result D5 and the abnormality propagation order estimation result D6 from the data storage unit 20. The abnormality cause estimation unit 60 calculates the corresponding abnormality cause score according to the abnormality detection order included in the abnormality detection order estimation result D5 and the abnormality propagation order included in the abnormality propagation order estimation result D6 for each sensor 300. Here, the abnormality detection order, the abnormality propagation order, and the abnormality cause score are all real values. In addition, the abnormality cause estimation unit 60 calculates the corresponding abnormality cause score for all sensors 300 included in the abnormality detection order estimation result D5 or the abnormality propagation order estimation result D6.

[0143] The abnormality cause estimation unit 60 calculates a representative value as the abnormality cause score using, for example, a weighted average of the abnormality detection order and the abnormality propagation order. In addition, the abnormality cause estimation unit 60 may also calculate, for example, a representative value such as a minimum value or a maximum value as the abnormality cause score. That is, the abnormality cause estimation unit 60 may also calculate, for example, the smaller or larger of the abnormality detection order and the abnormality propagation order as the abnormality cause score.

[0144] In addition, when only one of the abnormality detection order and the abnormality propagation order is set, the abnormality cause estimation unit 60 may also consider setting only the order of the set one to weight the abnormality cause score. For example, when only one of the abnormality detection order and the abnormality propagation order is set, the abnormality cause estimation unit 60 adds a weight b to the abnormality cause score. In addition, the weight b is a real value greater than 0. For example, when the abnormality propagation path tracking unit 50 converts the impact propagation graph D10 into the abnormality propagation graph D11 and also includes the node corresponding to the sensor Xn not included in the abnormality detection sensor Xn and allocates the abnormality propagation order, it may occur that the abnormality detection order is not set but the abnormality propagation order is set.

[0145] After calculating the abnormality cause score for each sensor 300, the abnormality cause estimation unit 60 assigns an abnormality cause order based on the abnormality cause score to the sensor 300 based on the calculated abnormality cause score. The abnormality cause order is a real value.

[0146] The abnormality cause estimation unit 60 assigns the abnormality cause order in ascending order, for example, starting from the sensor 300 with the smallest corresponding abnormality cause score. When there are a plurality of sensors 300 with the same corresponding abnormality cause score, the abnormality cause estimation unit 60 assigns the same abnormality cause order to the plurality of sensors 300.

[0147] After assigning the abnormal cause order to each sensor 300, the abnormal cause estimation unit 60 generates information related to the abnormal cause order assigned to each sensor 300 (hereinafter referred to as "abnormal cause order estimation result") and stores it in the data storage unit 20. In the following embodiment 1, the abnormal cause order estimation result is set to "abnormal cause order estimation result D7".

[0148] The abnormality cause order estimation result D7 is information obtained by associating information indicating the sensor 300, an abnormality detection sensor flag indicating whether the sensor 300 is an abnormality detection sensor included in the abnormality detection order estimation result D5, an abnormality cause score, and an abnormality cause order.

[0149] In addition, the abnormality detection sensor flag is a Boolean value. For example, when the sensor 300 is an abnormality detection sensor, the abnormality cause estimation unit 60 assigns (True) to the abnormality detection sensor flag corresponding to the sensor 300, and when the sensor 300 is not an abnormality detection sensor, the abnormality cause estimation unit 60 assigns (False) to the abnormality detection sensor flag corresponding to the sensor 300. For example, when the information related to the sensor 300 is included in the abnormality detection order estimation result D5, the abnormality cause estimation unit 60 may determine that the sensor 300 is the abnormality detection sensor Xi.

[0150] The abnormality cause estimation process performed by the abnormality cause estimation unit 60 as described above will be described with reference to the drawings and specific examples.

[0151] Figure 7 This is a diagram for explaining a specific example of the abnormality cause estimation process performed by the abnormality cause estimation unit 60 in the first embodiment.

[0152] exist Figure 7 In FIG. 1 , as an example, the number of sensors 300 is set to n (n=1 to 6), and the number n of sensors 300 is represented as sensors Xn. In addition, sensors X1, X2, X4, and X6 among the sensors Xn are assumed to be abnormality detection sensors.

[0153] In addition, Figure 7 In the example shown, the abnormality propagation order estimation result D6 used by the abnormality cause estimation unit 60 in the abnormality cause estimation process is as follows: Figure 5 As shown, only the information on the abnormality propagation order corresponding to the abnormality detection sensor Xi is recorded.

[0154] At once Figure 7 In the illustrated example, the abnormality cause estimation unit 60 obtains the abnormality detection order estimation result D5 and the abnormality propagation order estimation result D6 related to the abnormality detection sensors X1 , X2 , X4 , and X6 from the data storage unit 20 .

[0155] Here, in the abnormality detection order estimation result D5, the abnormality detection orders o1, o2, o4, o6 corresponding to the abnormality detection sensors X1, X2, X4, X6 ( Figure 7 In addition, in the abnormality propagation order estimation result D6, the abnormality propagation order o1, o2, o4, o6 ( Figure 7 ) are “2” (the second one), “1” (the first one), “1” (the first one), and “1” (the first one).

[0156] Here, the abnormality cause estimation unit 60 uses the average to calculate the abnormality cause score sn (n=1, 2, 4, 6) corresponding to the abnormality detection sensor Xn. In this case, the abnormality cause estimation unit 60 calculates the abnormality cause scores s1, s2, s4, and s6 corresponding to the abnormality detection sensors X1, X2, X4, and X6 as "2", "2", "1", and "2.5", respectively. Based on the calculated abnormality cause scores s1, s2, s4, and s6, the abnormality cause estimation unit 60 assigns the abnormality cause orders o1, o2, o4, and o6 corresponding to the abnormality detection sensors X1, X2, X4, and X6 as "2" (the second), "2" (the second), "1" (the first), and "3" (the third), respectively.

[0157] Then, the abnormality cause estimation unit 60 generates an abnormality cause order estimation result D7.

[0158] Specifically, the abnormality cause estimation unit 60 generates information indicating abnormality detection sensors X1, X2, X4, and X6 ( Figure 7 ), indicating whether the abnormality detection sensors X1, X2, X4, X6 are included in the abnormality detection order estimation result D5, the abnormality detection sensor flags f1, f2, f4, f6 ( Figure 7 As shown in D7B), abnormal cause scores s1, s2, s4, s6 ( Figure 7 D7C in the figure), and the order of abnormal causes o1, o2, o4, o6 ( Figure 7 The abnormal cause sequence estimation result D7 corresponding to D7D in FIG.

[0159] Here, the sensor Xn included in the abnormality detection order estimation result D5 is equal to the abnormality detection sensor Xn. Therefore, the abnormality detection sensor flags f1, f2, f4, and f6 corresponding to the abnormality detection sensors X1, X2, X4, and X6 are all (True).

[0160] Then, the abnormality cause estimation unit 60 stores the generated abnormality cause order estimation result D7 in the data storage unit 20 .

[0161] In addition, here, the abnormality cause estimation unit 60 obtains the abnormality detection order estimation result D5 from the abnormality detection order estimation unit 40 via the data storage unit 20, and obtains the abnormality propagation order estimation result D6 from the abnormality propagation path tracking unit 50 via the data storage unit 20, but this is just an example. The abnormality cause estimation unit 60 may also directly obtain the abnormality detection order estimation result D5 and the abnormality propagation order estimation result D6 from the abnormality detection order estimation unit 40 and the abnormality propagation path tracking unit 50, respectively.

[0162] Return to Figure 2A description is given of a configuration example of the abnormality cause estimation device 100 shown.

[0163] The abnormal cause estimation result output unit 70 obtains the abnormal cause sequence estimation result D7 output by the abnormal cause estimation unit 60, the abnormal detection sequence estimation result D5 output by the abnormal detection sequence estimation unit 40, and the abnormal propagation sequence estimation result D6 output by the abnormal propagation path tracking unit 50 from the data storage unit 20, and outputs information related to the estimation result of the cause of the abnormality estimated by the abnormal cause estimation unit 60.

[0164] In detail, the abnormal cause estimation result output unit 70 outputs information (hereinafter referred to as "abnormal cause estimation result display information") for displaying the following screen (hereinafter referred to as "abnormal cause estimation result screen") to the display device 400 based on the abnormal cause sequence estimation result D7, the abnormal detection sequence estimation result D5 and the abnormal propagation sequence estimation result D6. The screen shows information related to the estimation result of the cause of the abnormality estimated by the abnormal cause estimation unit 60.

[0165] In the first embodiment, the abnormality cause estimation result output unit 70 is provided in the abnormality cause estimation device 100, but this is only an example. The abnormality cause estimation result output unit 70 may also be provided in a device such as a display (not shown) connected to the abnormality cause estimation device 100 via a wired or wireless signal line.

[0166] here, Figure 8 and Fig. 9 This is a diagram for explaining a screen example of the abnormality cause estimation result screen which the abnormality cause estimation result output unit 70 causes the display device 400 to display in the first embodiment.

[0167] As an example, Figure 8 and Fig. 9 An example of the abnormality cause estimation result screen is shown when the number of sensors 300 is set to 6 (sensor Xn. n=1 to 6).

[0168] exist Figure 8 and Fig. 9 In the figure, the abnormality cause estimation result screens are represented by "D12-1" and "D12-2" respectively.

[0169] For example, Figure 8 and Fig. 9 As shown, the abnormal cause estimation result screen has 11 display frames, namely, display frame D12A, display frame D12B, display frame D12C, display frame D12D, display frame D12E, display frame D12F, display frame D12G, display frame D12H, display frame D12I, display frame D12J and display frame D12K.

[0170] The abnormal cause estimation result output unit 70 displays, for example, an abnormal cause estimation result list that summarizes information related to the estimation result of the abnormal cause on the abnormal cause estimation result screen. The abnormal cause estimation result list is, for example, a list that displays information indicating sensor Xn, information indicating an abnormal detection sensor flag, abnormal detection time, abnormal detection order, abnormal propagation order, abnormal cause score, and abnormal cause order for each sensor Xn. Figure 8 and Fig. 9 In the abnormality cause estimation result screen shown, the abnormality cause estimation result lists are indicated by "D12-1a" and "D12-2a", respectively.

[0171] Figure 8 The screen example of the abnormality cause estimation result screen shown is as follows: Fig. 10A The content shown, the content of the abnormal propagation order estimation result D6 is Fig. 10B The contents shown, the contents of abnormal cause sequence estimation result D7 are Fig. 10C An example of a screen with the content shown.

[0172] The abnormal cause estimation result output unit 70 outputs the abnormal cause estimation result display information to the display device 400, and the abnormal cause estimation result display information displays the sensor Xn information indicating the abnormal cause sequence estimation result D7 in the display frame D12A, displays the abnormal detection sensor flag information indicating the abnormal cause sequence estimation result D7 in the display frame D12B, displays the abnormal detection time information indicating the abnormal detection sequence estimation result D5 in the display frame D12C, displays the abnormal detection sequence of the abnormal detection sequence estimation result D5 in the display frame D12D, and displays the abnormal propagation sequence of the abnormal propagation sequence estimation result D6 in the display frame D12C. The display box D12E displays the abnormal cause score of the abnormal cause sequence estimation result D7 in the display box D12F, displays the abnormal cause sequence of the abnormal cause sequence estimation result D7 in the display box D12G, displays the sorting buttons for re-sorting the arrangement order of the abnormal cause estimation results in ascending order based on the abnormal detection sequence estimation result D5, the abnormal propagation sequence of the abnormal propagation sequence estimation result D6, and the abnormal cause sequence of the abnormal cause sequence estimation result D7 in the display boxes D12I, D12J, and D12K, respectively, and displays a check box for accepting an instruction to display only the abnormal detection sensor in the display box D12H. As a result, the display device 400 displays Figure 8 The abnormality cause estimation result screen is shown.

[0173] For example, the abnormal cause estimation result output unit 70 displays "True" in the display box D12B when the abnormal detection sensor flag of the abnormal cause sequence estimation result D7 is set to (True), and displays "False" in the display box D12B when the abnormal detection sensor flag of the abnormal cause sequence estimation result D7 is set to (False).

[0174] Furthermore, the abnormality cause estimation result output unit 70 displays a blank column when, for example, the abnormality detection time or the abnormality detection order of the abnormality detection order estimation result D5 has no set value.

[0175] In addition, the abnormality cause estimation result output unit 70 displays information indicating the sensor Xn, information indicating the abnormality detection sensor flag, abnormality detection time, abnormality detection order, abnormality propagation order, abnormality cause score, and abnormality cause order, which are associated with the sensor Xn, in the order of the ID assigned to the sensor Xn, for example, in the initial state of the abnormality cause estimation result list. Here, the initial state of the abnormality cause estimation result list refers to the state of the abnormality cause estimation result list when the abnormality cause estimation result output unit 70 causes the display device 400 to display the abnormality cause estimation result list for the first time, for example, after the power is turned on.

[0176] Figure 8 The example of the abnormality cause estimation result list shown is an example of the abnormality cause estimation result list in the initial state.

[0177] In the initial state of the abnormality cause estimation result list, the abnormality cause estimation result output unit 70 displays the abnormality cause estimation result list as follows: Figure 8 As shown, the state in which the instruction to reorder the data is not performed and only the instruction of the abnormality detection sensor is displayed is set.

[0178] Operators such as confirmation Figure 8 The abnormality cause estimation result screen is shown. Thus, the operator can grasp the information related to the estimation result of the abnormality cause. For example, the operator can determine the machine that is the cause of the abnormality based on the information of the sensor Xn that is the cause of the abnormality. In addition, the operator can grasp the order in which the abnormal machine should be inspected. Therefore, the operator can reduce unnecessary inspection work and the operator's load is reduced.

[0179] In addition, Figure 8 When the abnormality cause estimation result list is displayed, the operator can instruct to reorder the information shown in the abnormality cause estimation result list.

[0180] For example, the operator can instruct the reordering of information by operating an input device such as a mouse or keyboard (not shown) and pressing the sorting buttons of the display boxes D12I, D12J, and D12K. For example, when receiving an instruction to reorder information, the abnormal cause estimation result output unit 70 displays the sorting button that has been input, or in other words, has been pressed, in black. Then, the abnormal cause estimation result output unit 70 outputs abnormal cause estimation result display information to the display device 400, for example, and the abnormal cause estimation result display information displays the abnormal cause estimation result list in which the information to be displayed has been reordered according to the input instruction. As a result, the abnormal cause estimation result list displayed by the display device 400 is updated to the abnormal cause estimation result list after the displayed information has been reordered.

[0181] In addition, for example, the operator or the like operates the input device and presses the check box displayed in the display box D12H, thereby inputting an instruction to display only the information related to the abnormality detection sensor in the abnormality estimation result list. For example, when an instruction to display only the information related to the abnormality detection sensor in the abnormality estimation result list is input, the abnormality estimation result output unit 70 displays the above check box as checked. Then, for example, the abnormality estimation result output unit 70 outputs the abnormality estimation result display information for displaying only the information related to the abnormality detection sensor in the abnormality estimation result list according to the input instruction to the display device 400. As a result, the abnormality estimation result list displayed by the display device 400 is updated to the abnormality estimation result list that only displays the information related to the abnormality detection sensor.

[0182] Fig. 9 : is a diagram showing an example of the following abnormality cause estimation result screen: Figure 8 When the abnormal cause estimation result screen shown is displayed, the operator presses the sort button in the display box D12K of the abnormal cause estimation result list, and the abnormal cause estimation result output unit 70 that receives the instruction re-sorts the information in the abnormal cause estimation result list in ascending order of the abnormal cause. Thereafter, the operator presses the check box in the display box D12H, and the abnormal cause estimation result output unit 70 that receives the instruction displays the abnormal cause estimation result list that only displays the rows corresponding to the abnormal detection sensors whose abnormal detection sensor flag is (True).

[0183] exist Fig. 9The abnormal cause estimation result screen shown shows the following overview of the abnormal cause estimation results: only the rows corresponding to sensors X1, X2, X4, and X6 (i.e., abnormal detection sensors) whose abnormal detection sensor flag is (True) are displayed, and based on the abnormal cause order, the rows corresponding to sensors X1, X2, X4, and X6 are reordered in ascending order of the abnormal cause order to rows corresponding to sensor X4, rows corresponding to sensor X2, rows corresponding to sensor X6, and rows corresponding to sensor X1.

[0184] exist Fig. 9 In the abnormality cause estimation result screen shown in FIG. 1 , a check mark is displayed in the check box of the display box D12H. Thus, the operator can recognize that only the abnormality detection sensor is displayed in the abnormality cause estimation result screen. Fig. 9 In the abnormality cause estimation result screen shown, the sort button of the display box D12K is filled in. This allows the operator to recognize that the rows of the abnormality cause estimation result list in the abnormality cause estimation result screen are re-sorted in ascending order of the abnormality cause sequence.

[0185] In addition, when the operator presses the same sort button again, for example, when any one of the sort buttons displayed in the display boxes D12H, D12J, and D12K is pressed, that is, when the display of the abnormal cause estimated result list has been re-sorted, the display of the abnormal cause estimated result list can be returned to the state before the display of the abnormal cause estimated result list was re-sorted. When the abnormal cause estimated result output unit 70 receives the intention that the same sort button has been pressed again, it displays the filled sort button in the unfilled state. Then, the abnormal cause estimated result output unit 70 outputs the abnormal cause estimated result display information for displaying the abnormal cause estimated result list before the re-sorting to the display device 400. As a result, the abnormal cause estimated result list displayed by the display device 400 is updated to the abnormal cause estimated result list before the re-sorting (refer to Figure 8 ).

[0186] In addition, when the operator presses the check box again, for example, when the check box of the display box D12H is pressed, that is, when the list of abnormal cause estimation results showing only the information related to the abnormal detection sensor is displayed, the display of the list of abnormal cause estimation results can be returned to the state before indicating that only the information related to the abnormal detection sensor is displayed. When the abnormal cause estimation result output unit 70 receives the fact that the check box has been pressed again, it displays the unchecked check box. Then, the abnormal cause estimation result output unit 70 outputs the abnormal cause estimation result display information for displaying the list of abnormal cause estimation results before switching to displaying only the information related to the abnormal detection sensor to the display device 400. As a result, the list of abnormal cause estimation results displayed by the display device 400 is updated to the list of abnormal cause estimation results before switching to displaying only the information related to the abnormal detection sensor (refer to Figure 8 ).

[0187] In addition, here, the initial state of the abnormal cause estimation result screen is Figure 8 For example, the abnormal cause estimation result output unit 70 may also re-sort the information related to the sensor 300 displayed in the abnormal cause estimation result list in the initial state of the abnormal cause estimation result screen in advance based on any one of the abnormal detection order, abnormal propagation order, and abnormal cause order, and may also display the abnormal cause estimation result list that only displays the abnormal detection sensor.

[0188] Return to Figure 2 A description is given of a configuration example of the abnormality cause estimation device 100 shown.

[0189] The data storage unit 20 stores various information.

[0190] In detail, the data storage unit 20 stores, for example, an association structure D2 generated by the learning device 200, sensor data D1 acquired by the sensor data acquisition unit 10, abnormality detection sensor information D3 and abnormality detection time information D4 output by the abnormality detection unit 30, an abnormality detection sequence estimation result D5 output by the abnormality detection sequence estimation unit 40, an abnormality propagation sequence estimation result D6 output by the abnormality propagation path tracking unit 50, and an abnormality cause sequence estimation result D7 output by the abnormality cause estimation unit 60.

[0191] In addition, in Embodiment 2, Figure 2 As shown, the data storage unit 20 is provided in the abnormality cause estimation device 100, but this is only an example. The data storage unit 20 may be provided outside the abnormality cause estimation device 100 and in a place where the abnormality cause estimation device 100 can refer to.

[0192] A configuration example of the learning device 200 according to the first embodiment will be described.

[0193] Fig.11 This is a diagram showing a configuration example of a learning device 200 according to the first embodiment.

[0194] The learning device 200 uses sensor data collected by the sensor 300 provided in the target device during normal operation of the target device for learning. Specifically, the learning device 200 estimates the association structure D2 using sensor data collected by the sensor 300 provided in the target device during normal operation of the target device. In addition, specifically, the normal operation of the target device refers to the normal operation of multiple machines constituting the target device. Therefore, specifically, the sensor data collected by the sensor 300 during normal operation of the target device refers to the sensor data collected by the sensor 300 provided in each machine during normal operation of the multiple machines constituting the target device.

[0195] The learning device 200 stores the learned correlation structure D2 in the data storage unit 20 of the abnormality cause estimation device 100 .

[0196] In addition, Fig.11 In the figure, for the sake of simplicity of explanation, only the data storage unit 20 is shown among the components of the abnormality cause estimation device 100.

[0197] The learning device 200 includes a learning sensor data acquisition unit 210 , a learning data storage unit 220 , a learning preprocessing unit 230 , and a correlation structure learning unit 240 .

[0198] The learning sensor data acquisition unit 210 acquires learning data for learning the association structure D2. The learning data includes sensor data acquired from a plurality of sensors 300.

[0199] In addition, here, not all the learning data acquired by the learning sensor data acquisition unit 210 is used for learning the association structure D2. The learning preprocessing unit 230 described later acquires the learning data actually used for learning the association structure D2 based on the learning data acquired by the learning sensor data acquisition unit 210. Therefore, more accurately speaking, the learning data acquired by the learning sensor data acquisition unit 210 is a learning data candidate. The details of the learning preprocessing unit 230 will be described later.

[0200] The learning sensor data acquisition unit 210 stores the acquired learning data candidates in the learning data storage unit 220 .

[0201] Here, the candidate data for learning is information of the same type as the sensor data D1 acquired by the abnormality cause estimation device 100 from the sensor 300, and more specifically, information of the same content as the sensor data D1 (at least one measured value or control value among opening, deviation, rotation speed, conductivity, flow rate, pressure, temperature, concentration, or water level), and is sensor data collected by the sensor 300 during normal operation of multiple devices as multiple device components of the target device. That is, the candidate data for learning includes sensor measured values ​​or control values ​​during normal operation of the target device. In addition, the candidate data for learning is information of the same type as the sensor data D1 acquired from the same sensor 300 as the sensor 300 from which the abnormality cause estimation device 100 acquires the sensor data D1.

[0202] The learning data candidates are prepared in advance by, for example, a manager or the like, and stored in a location that can be referenced by the learning device 200. For example, the learning data candidates may also be stored in the data storage unit 20 of the abnormality cause estimation device 100. In this case, the learning sensor data acquisition unit 210 may acquire the learning data candidates from the data storage unit 20 of the abnormality cause estimation device 100.

[0203] In the first embodiment, the learning data candidate acquired by the learning sensor data acquisition unit 210 is referred to as “learning data candidate D21 ”.

[0204] In addition, Fig.11 In the figure, the learning data candidates obtained by the learning sensor data acquisition unit 210 from a place other than the abnormal cause estimation device 100 are set as "learning data candidate D21", and the learning data candidates obtained by the learning sensor data acquisition unit 210 from the abnormal cause estimation device 100 are set as "learning data candidate D22", and they are distinguished and illustrated. This is for easy understanding, and the contents of "learning data candidate D21" and "learning data candidate D22" are the same.

[0205] Therefore, in the following description, both “learning data candidate D21” and “learning data candidate D22” are described as “learning data candidate D21”.

[0206] The learning preprocessing unit 230 obtains the learning data candidate D21 obtained by the learning sensor data obtaining unit 210 from the learning data storage unit 220 and preprocesses the learning data candidate D21. In addition, the learning preprocessing unit 230 obtains the learning data candidate D21 stored at the prescribed time from the learning data storage unit 220 at every prescribed time.

[0207] Specifically, the learning preprocessing unit 230 performs data conversion or selection on the learning data candidates D21 to obtain learning data that is actually used for learning the relationship structure D2.

[0208] For example, the learning preprocessing unit 230 converts the learning data candidate D21 into a first-order difference sequence and uses the converted learning data candidate D21 as learning data. In this case, the sensor data included in the learning data candidate D21 is converted into data indicating a change amount.

[0209] In addition, for example, the learning preprocessing unit 230 may also select only sensor data with a large variance from the sensor data included in the learning data candidate D21, and set the selected sensor data as the learning data. In this case, the learning preprocessing unit 230 sets a threshold value for the variance value, and selects sensor data with a variance value larger than the threshold value from the sensor data included in the learning data candidate D21 as the learning data. In addition, the threshold value for the variance value is set manually, for example, by an operator or the like. The operator or the like operates an input device such as a mouse or a keyboard in advance to input the threshold value for the variance value, and sets the threshold value for the variance value. In addition, for example, when the measurement error of the sensor 300 is known, the learning preprocessing unit 230 may also set the square of the measurement error as the threshold value for the variance value.

[0210] The learning preprocessing unit 230 stores the acquired learning data in the learning data storage unit 220. In the first embodiment, the learning data acquired by the learning preprocessing unit 230 is referred to as "learning data D23".

[0211] In the following first embodiment, as an example, it is assumed that the preprocessing performed by the learning preprocessing unit 230 on the learning data candidate D21 is selected, and the learning preprocessing unit 230 obtains sensor data whose variance value is larger than a threshold value from among the sensor data included in the learning data candidate D21 as learning data D23.

[0212] Here, the learning preprocessing unit 230 acquires the learning data candidate D21 from the learning sensor data acquiring unit 210 via the learning data storage unit 220 , but this is just an example. The learning preprocessing unit 230 may directly acquire the learning data candidate D21 from the learning sensor data acquiring unit 210 .

[0213] The relationship structure learning unit 240 acquires the learning data D23 output by the learning preprocessing unit 230 from the learning data storage unit 220 , and learns the relationship structure D2 based on the learning data D23 .

[0214] Specifically, the relationship structure learning unit 240 calculates at least one statistic indicating the relationship between two different sensor data for the plurality of sensor data included in the learning data D23 , and learns the relationship structure D2 based on the calculated statistic.

[0215] The association structure learning unit 240 uses, for example, correlation or cross-correlation, or waveform-based statistical indicators such as Granger Causality, Transfer entropy, CCM (Convergent cross mapping), DTW (Dynamic Time Warping), as indicators when calculating statistics representing the relationship between sensor data (hereinafter referred to as "statistical indicators"). In addition, the association structure learning unit 240 can also use distribution-based statistical indicators such as KL (Kullback Leibler) divergence or HI (Histogram Intersection) as statistical indicators. Statistical indicators are classified into undirected and directed types. Undirected statistical indicators refer to statistical indicators such as correlation that cannot identify the direction of dependency, and directed statistical indicators refer to statistical indicators such as Granger causality that can identify the direction of dependency.

[0216] In implementation mode 1, as an example, the association structure learning unit 240 calculates one or more statistics including at least one directional statistical index representing the relationship between two different sensor data for multiple sensor data included in the learning data D23, and learns the association structure D2 based on the calculated statistics.

[0217] The learning process of the relationship structure learning unit 240 learning the relationship structure D2 will be described with reference to a specific example using the drawings.

[0218] Fig.12 This is a diagram showing the concept of an example of a learning process of the relationship structure learning unit 240 learning the relationship structure D2 in the first embodiment.

[0219] like Fig.12 As shown, the learning data D23 is a two-dimensional data frame in which the number of rows is t and the number of columns is n (represented by sensors Xn (n=1, 2, 3, 4, ..., n) of the sensors 300. The association structure D2 is initialized so that all elements defined by the association structure D2 are "0". Here, the association structure learning unit 240 uses m types of statistical indicators when learning the association structure D2.

[0220] 〈Data selection for learning〉

[0221] The association structure learning unit 240 selects sensor data collected by two different sensors Xn from the sensor data collected by the sensor Xn included in the learning data D23. For example, the association structure learning unit 240 selects the i-th sensor data collected by the i-th sensor Xi and the j-th sensor data collected by the j-th sensor Xj. Hereinafter, the i-th sensor data collected by the i-th sensor Xi will also be referred to as “i-th sensor data”, and the j-th sensor data collected by the j-th sensor Xj will also be referred to as “j-th sensor data”.

[0222] 〈Statistical calculation〉

[0223] Next, the association structure learning unit 240 uses m types of statistical indicators to calculate the statistics a(1)ij, a(2)ij, ..., a(m)ij from the i-th sensor data to the j-th sensor data as the statistics from the sensor Xi to the sensor Xj, and calculates the statistics a(1)ij, a(2)ij, ..., a(m)ij from the i-th sensor data to the j-th sensor data as the statistics from the sensor Xi to the sensor Xj. Then, the association structure learning unit 240 obtains information D24A related to the statistics between the sensor Xi and the sensor Xj, which is composed of the statistics a(k)ij from the sensor Xi to the sensor Xj and the statistics a(k)ji from the sensor Xj to the sensor Xi (hereinafter referred to as "inter-sensor statistical information"). Here, the association structure learning unit 240 converts the inter-sensor statistical information D24A as needed so that the greater the absolute value |a(k)ij| of the statistics, the greater the dependency relationship between the sensor data. For example, the p-value corresponding to the Granger Causality statistic a(k)ij takes a value between 0 and 1, and the smaller the p-value is, the less likely it is that there is no dependency between the i-th sensor data and the j-th sensor data. In this case, the association structure learning unit 240 converts the p-value in a manner that indicates that the larger the statistic, the greater the dependency, and sets the (1-p-value) as the statistic a(k)ij in the inter-sensor statistical information. On the other hand, assuming that the association structure learning unit 240 calculates the statistic using correlation as a non-directional statistical indicator, for example, the correlation coefficient ρ corresponding to the related statistic a(k)ij takes a value between -1 and 1, and the larger the absolute value of ρ is, the greater the dependency between the i-th sensor data and the j-th sensor data is. In this case, the association structure learning unit 240 does not convert the inter-sensor statistical information D24A.

[0224] Furthermore, the association structure learning unit 240 creates pairs of two different sensor data in all combinations of sensor data included in the learning data D23 , calculates statistics for all two different sensor data, and acquires inter-sensor statistical information D24A.

[0225] 〈Associative structure learning〉

[0226] The association structure learning unit 240 uses the statistics corresponding to all pairs of sensor data set in the inter-sensor statistical information D24A as elements to learn the association structure D2. For example, the statistic |a(k)ij| from the i-th sensor Xi to the j-th sensor Xj calculated using the k-th statistical index among the m types of statistical indexes is substituted into the k-th element of the first dimension, the i-th element of the second dimension, and the j-th element of the third dimension of the association structure D2.

[0227] After learning the correlation structure D2 as described above, the correlation structure learning unit 240 stores the learned correlation structure D2 in the data storage unit 20 of the abnormality cause estimation device 100 .

[0228] For example, the association structure learning unit 240 may store the association structure D2 in the learning data storage unit 220. In this case, in the abnormality cause estimation device 100, the abnormality propagation path tracking unit 50 downloads the association structure D2 to be used from the learning data storage unit 220 to the data storage unit 20, for example, each time the abnormality propagation order estimation process is performed.

[0229] Here, the relation structure learning unit 240 obtains the learning data D23 from the learning preprocessing unit 230 via the learning data storage unit 220 , but this is just an example. The relation structure learning unit 240 may directly obtain the learning data D23 from the learning preprocessing unit 230 .

[0230] Return to Fig.11 Description of the configuration example of the learning device 200 shown.

[0231] The learning data storage unit 220 stores various information related to the learning performed by the learning device 200 .

[0232] Specifically, the learning data storage unit 220 stores, for example, the learning data candidate D21 acquired by the learning sensor data acquisition unit 210 and the learning data D23 output by the learning preprocessing unit 230. The learning data storage unit 220 may also store the relationship structure D2 learned by the relationship structure learning unit 240.

[0233] In addition, here, the learning data storage unit 220 is provided in the learning device 200, but this is just an example, and the learning data storage unit 220 may be provided outside the learning device 200 and in a place that the learning device 200 can refer to.

[0234] In the first embodiment, the learning device 200 includes the learning preprocessing unit 230, but this is only an example, and the learning device 200 does not necessarily need to include the learning preprocessing unit 230. When the learning device 200 does not include the learning preprocessing unit 230, the association structure learning unit 240, for example, uses all the learning data candidates D21 obtained by the learning sensor data acquisition unit 210 as the learning data D23 actually used for learning the association structure D2, and learns the association structure D2 using the learning data D23 obtained by the learning sensor data acquisition unit 210. That is, in the learning device 200, the association structure learning unit 240 uses the multiple learning data candidates obtained by the learning sensor data acquisition unit 210 as multiple learning data, calculates at least one statistic between the multiple learning data based on the multiple learning data, and learns the estimated structure (association structure D2) showing the dependency relationship between the device structural elements based on the calculated statistic.

[0235] The operations of the abnormality cause estimation device 100 and the learning device 200 according to the first embodiment will be described.

[0236] First, the operation of the abnormality cause estimation device 100 according to the first embodiment will be described.

[0237] Fig.13 This is a flowchart for explaining the operation of the abnormality cause estimation device 100 according to the first embodiment.

[0238] The sensor data acquisition unit 10 acquires the sensor data D1 from the sensor 300 (step ST1 ).

[0239] The sensor data acquisition unit 10 stores the acquired sensor data D1 in the data storage unit 20 .

[0240] The abnormality detection unit 30 performs abnormality detection processing on the sensor data D1 stored in the data storage unit 20 by the sensor data acquisition unit 10 in step ST1 (step ST2 ).

[0241] The abnormality detection unit 30 stores the abnormality detection sensor information D3 and the abnormality detection time information D4 in the data storage unit 20 .

[0242] If the abnormality detection unit 30 detects an abnormality detection sensor in step ST2, the operation of the abnormality cause estimation device 100 proceeds to step ST3. If the abnormality detection unit 30 does not detect an abnormality detection sensor in step ST2, the abnormality cause estimation device 100 ends. Fig.13 The process is shown in the flowchart.

[0243] For example, when the abnormality detection unit 30 detects an abnormality detection sensor in step ST2, it notifies the abnormality detection order estimation unit 40 of this fact, and the operation of the abnormality cause estimation device 100 proceeds to step ST3. On the other hand, when the abnormality detection unit 30 does not detect an abnormality detection sensor in step ST2, it notifies this fact to the control unit (not shown) of the abnormality cause estimation device 100, and the control unit ends the processing of the abnormality cause estimation device 100.

[0244] In step ST3, the abnormality detection sequence estimation unit 40 obtains the abnormality detection sensor information D3 and the abnormality detection time information D4 stored in the data storage unit 20 by the abnormality detection unit 30 in step ST2, and performs abnormality detection sequence estimation processing. In this abnormality detection sequence estimation processing, for the abnormality detection sensor, the abnormality detection sequence in which the abnormality is detected, more specifically, the abnormality detection sequence in which the abnormality is detected in the sensor data D1 collected by the abnormality detection sensor is estimated (step ST3).

[0245] The abnormality detection order estimation unit 40 stores the abnormality detection order estimation result D5 in the data storage unit 20 .

[0246] The abnormality propagation path tracking unit 50 obtains the abnormality detection sensor information D3 and the associated structure D2 stored by the abnormality detection unit 30 in step ST3 from the data storage unit 20, and performs abnormality propagation order estimation processing. In this abnormality propagation order estimation processing, the abnormality propagation order is estimated based on the acquired abnormality detection sensor information D3 and the associated structure D2 (step ST4).

[0247] The anomaly propagation path tracing unit 50 outputs the anomaly propagation order estimation result D6 to the data storage unit 20 .

[0248] The abnormality cause estimation unit 60 obtains the abnormality detection sequence estimation result D5 output by the abnormality detection sequence estimation unit 40 in step ST3 and the abnormality propagation sequence estimation result D6 output by the abnormality propagation path tracking unit 50 in step ST4 from the data storage unit 20, and performs abnormality cause estimation processing. In this abnormality cause estimation processing, the cause of the abnormality is estimated based on the abnormality detection sequence estimated by the abnormality detection sequence estimation unit 40 and the abnormality propagation sequence estimated by the abnormality propagation path tracking unit 50 (step ST5).

[0249] The abnormality cause estimation unit 60 generates an abnormality cause order estimation result D7 , and stores the generated abnormality cause order estimation result D7 in the data storage unit 20 .

[0250] The abnormal cause estimation result output unit 70 obtains from the data storage unit 20 the abnormal cause sequence estimation result D7 output by the abnormal cause estimation unit 60 in step ST5, the abnormal detection sequence estimation result D5 output by the abnormal detection sequence estimation unit 40 in step ST3, and the abnormal propagation sequence estimation result D6 output by the abnormal propagation path tracking unit 50 in step ST4, and outputs information related to the estimation result of the abnormal cause estimated by the abnormal cause estimation unit 60 (step ST6).

[0251] Specifically, the abnormality cause estimation result output unit 70 outputs abnormality cause estimation result display information for displaying the abnormality cause estimation result screen to the display device 400. Thus, the abnormality cause estimation result output unit 70 presents information on the estimation result of the cause of the abnormality to the operator.

[0252] In addition, when the abnormality cause estimation device 100 does not include the abnormality cause estimation result output unit 70, the abnormality cause estimation device 100 Fig.13 The process of step ST6 can be omitted in the operation of the abnormality cause estimation device 100 shown in the flowchart of FIG. The process of step ST6 is performed by a device external to the abnormality cause estimation device 100, for example.

[0253] In this way, the abnormality cause estimation device 100 detects a plurality of abnormality detection sensors based on a plurality of time series of sensor data collected by a plurality of sensors 300 provided in the target device, and estimates the abnormality detection order in which the abnormality is detected for the plurality of abnormality detection sensors based on the detection times of the plurality of abnormality detection sensors. The abnormality cause estimation device 100 estimates the abnormality propagation order in which the abnormality propagates based on the abnormality detection sensor information related to the plurality of abnormality detection sensors and the estimated structure (association structure) showing the dependency relationship between the plurality of constituent machines constituting the target device, and estimates the cause of the abnormality based on the estimated abnormality detection order and the abnormality propagation order.

[0254] Since the abnormality cause estimation apparatus 100 estimates the cause of the abnormality occurring in the target device based on the order in which the abnormality occurs and the order in which the abnormality propagates, it is possible to estimate the cause of the abnormality more appropriately based on a plurality of criteria.

[0255] Furthermore, since the abnormality cause estimation device 100 estimates the order of abnormality propagation based on the estimation structure (correlation structure), there is no need to wait until sufficient sensor data is collected in order to construct the correlation structure D2 for diagnosing the target device, and the cause of the abnormality can be estimated in advance.

[0256] That is, the abnormality cause estimation apparatus 100 can estimate the cause of the abnormality occurring in the target device regardless of the complexity of the target device or the scale of the target device.

[0257] Furthermore, the abnormality cause estimation device 100 outputs the estimation result of the cause of the abnormality.

[0258] Therefore, the abnormality cause estimation device 100 improves the interpretability and explanation of the estimation result of the abnormality cause for the operator. The abnormality cause estimation device 100 can reduce the unnecessary inspection work of the operator and reduce the load of the operator. In addition, the abnormality cause estimation device 100 can estimate the cause of the abnormality with a quantitative index that does not rely on human subjectivity and provide the basis for the estimation. The operator can determine the inspection order of the equipment with less labor.

[0259] Furthermore, the abnormality cause estimation device 100 detects the abnormality detection sensor using a univariate abnormality detection method such as Hotelling's theory or Discord.

[0260] Therefore, the abnormality cause estimation device 100 can more appropriately detect an abnormality caused by a change in one sensor data D1 alone. For example, an abnormality caused by a change in one sensor data D1 is an abnormality detected in sensor data D1 collected by a single sensor 300 that is not related to other sensors 300 .

[0261] Furthermore, the abnormality cause estimation device 100 detects the abnormality detection sensor using a multivariate abnormality detection method such as Graphical Lasso.

[0262] Therefore, the abnormality cause estimation device 100 can more appropriately detect abnormalities in the relationship changes between the plurality of sensor data D1. The abnormality in the relationship changes between the plurality of sensor data D1 refers to, for example, an abnormality that occurs in the sensor data D1 collected by the sensor 300 of the upstream machine when two equipment structural elements (here, machines) are in a control relationship, and is an abnormality that also appears in the sensor data D1 collected by the sensor 300 of the downstream machine. For example, when an abnormality occurs in the valve opening detected by a certain valve that controls the flow rate, an abnormality also occurs in the flow rate measured by the flow meter that measures the flow rate.

[0263] Next, the operation of the learning device 200 according to the first embodiment will be described.

[0264] Fig.14 This is a flowchart for explaining the operation of the learning device 200 according to the first embodiment.

[0265] The learning sensor data acquisition unit 210 acquires the learning data candidate D21 for learning the association structure D2 (step ST21). Specifically, the learning sensor data acquisition unit 210 acquires the learning data candidate, which includes a plurality of time series of sensor data collected by a plurality of sensors installed in the target device during normal operation of the target device.

[0266] The learning sensor data acquisition unit 210 stores the acquired learning data candidates in the learning data storage unit 220 .

[0267] The learning preprocessing unit 230 acquires the learning data candidate D21 acquired by the learning sensor data acquiring unit 210 in step ST21 from the learning data storage unit 220 , preprocesses the learning data candidate D21 , and acquires learning data D23 (step ST22 ).

[0268] The learning preprocessing unit 230 stores the acquired learning data D23 in the learning data storage unit 220 .

[0269] The relationship structure learning unit 240 acquires the learning data D23 output by the learning preprocessing unit 230 from the learning data storage unit 220 , and learns the relationship structure D2 based on the learning data D23 (step ST23 ).

[0270] After learning the correlation structure D2 , the correlation structure learning unit 240 stores the learned correlation structure D2 in the data storage unit 20 of the abnormality cause estimation device 100 .

[0271] here, Fig.15 It is used to illustrate Fig.14 This is a flowchart of the details of the processing of step ST23.

[0272] The association structure learning unit 240 is based on Fig.14 The learning data D23 acquired by the learning preprocessing unit 230 in step ST22 selects sensor data collected by two different sensors 300 from the sensor data included in the learning data D23.

[0273] That is, the correlation structure learning unit 240 acquires a pair of two different sensor data based on the learning data D23 (step ST231 ).

[0274] Furthermore, the association structure learning unit 240 defines all combinations of a plurality of sensor data included in the learning data D23 as pairs of sensor data.

[0275] The correlation structure learning unit 240 extracts the pair of sensor data generated in step ST231 , and calculates at least one statistic between two different sensor data (step ST232 ).

[0276] The correlation structure learning unit 240 calculates statistics for all two different sensor data, and acquires inter-sensor statistical information D24A.

[0277] Then, the association structure learning unit 240 learns the association structure D2 using the statistics corresponding to all pairs of sensor data set in the inter-sensor statistical information D24A as elements (step ST233 ).

[0278] In addition, when the learning device 200 does not include the learning preprocessing unit 230, the learning device 200 Fig.14 In the operation of the learning device 200 shown in the flowchart of FIG. 1 , the process of step ST22 can be omitted.

[0279] In this way, the learning device 200 obtains a plurality of time series sensor data collected by a plurality of sensors 300 installed in the target device during normal operation of the target device as learning data candidates, and obtains a plurality of learning data for learning based on the plurality of learning data candidates. The learning device 200 calculates at least one statistic between the plurality of sensor data included in the learning data based on the obtained learning data, and learns the estimated structure (correlation structure D2) based on the calculated statistic.

[0280] For example, the comprehensiveness of the association structure taught manually depends on the connection relationship of the device structural elements, in other words, the sensor 300, which is known to the person. In contrast, the learning device 200 can extract the association between the sensor data D1 in a comprehensive manner, and as a result, can provide the association structure D2 that suppresses the omission of the connection relationship of the sensor 300. The learning device 200 provides the abnormality cause estimation device 100 with the estimation structure (association structure) when tracking the sensor 300 that is the source of the abnormality, so that the abnormality cause estimation device 100 can more appropriately track the sensor 300 that is the source of the abnormality, and can improve the estimation accuracy of the abnormality cause.

[0281] Furthermore, even if an association structure that only includes qualitative information such as connection information is manually taught, the learning device 200 can quantitatively determine the magnitude of the association or the direction of influence between the sensor data D1, and can more appropriately generate and provide an estimated structure (association structure) that can estimate the cause of the abnormality.

[0282] Furthermore, the learning device 200 calculates statistics using waveform-based statistical indicators such as correlation, Granger Causality, or DTW.

[0283] Therefore, the learning device 200 can provide an estimation structure (correlation structure D2) that can track the propagation of abnormality based on such a dependency relationship on the waveform and can more appropriately estimate the cause of the abnormality.

[0284] Furthermore, the learning device 200 calculates statistics using a statistical index based on distribution, such as KL divergence or HI.

[0285] Therefore, the learning device 200 can provide an estimation structure (correlation structure D2) that can track the propagation of anomalies based on such a dependency relationship in distribution and can more appropriately estimate the cause of the anomaly.

[0286] Furthermore, the learning device 200 calculates statistics using waveform-based statistical indicators and distribution-based statistical indicators.

[0287] Therefore, the learning device 200 can provide an estimation structure (correlation structure D2) that can track the propagation of abnormality based on such a dependency relationship in waveform or distribution and can more appropriately estimate the cause of abnormality.

[0288] Fig.16A and Fig. 16B This is a diagram showing an example of the hardware configuration of the abnormality cause estimation device 100 according to the first embodiment.

[0289] In the first embodiment, the functions of the sensor data acquisition unit 10, the abnormality detection unit 30, the abnormality detection order estimation unit 40, the abnormality propagation path tracking unit 50, the abnormality cause estimation unit 60, the abnormality cause estimation result output unit 70, and the control unit (not shown) are implemented by the processing circuit 1601. That is, the abnormality cause estimation device 100 includes the processing circuit 1601 for estimating the cause of the abnormality occurring in the target device using the estimation structure (correlation structure D2) indicating the dependency relationship between the plurality of device components constituting the target device.

[0290] The processing circuit 1601 may be as follows Fig.16A As shown in the figure, it is a dedicated hardware, or it can be Fig. 16B Shown is a processor 1604 that executes programs stored in memory.

[0291] When the processing circuit 1601 is dedicated hardware, the processing circuit 1601 corresponds to, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or a combination thereof.

[0292] In the case where the processing circuit is a processor 1604, the functions of the sensor data acquisition unit 10, the abnormality detection unit 30, the abnormality detection sequence estimation unit 40, the abnormality propagation path tracking unit 50, the abnormality cause estimation unit 60, the abnormality cause estimation result output unit 70, and the control unit not shown in the figure are implemented by software, firmware, or a combination of software and firmware. The software or firmware is described in the form of a program and stored in the memory 1605. The processor 1604 executes the functions of the sensor data acquisition unit 10, the abnormality detection unit 30, the abnormality detection sequence estimation unit 40, the abnormality propagation path tracking unit 50, the abnormality cause estimation unit 60, the abnormality cause estimation result output unit 70, and the control unit not shown in the figure by reading and executing the program stored in the memory 1605. That is, the abnormality cause estimation device 100 has a memory 1605 for storing a program, and the program performs the above-mentioned results when executed by the processor 1604. Fig.13 Steps ST1 to ST6 of the present invention. In addition, the program stored in the memory 1605 can also be said to be a program that causes the computer to execute the processing procedures or methods of the sensor data acquisition unit 10, the abnormality detection unit 30, the abnormality detection order estimation unit 40, the abnormality propagation path tracking unit 50, the abnormality cause estimation unit 60, the abnormality cause estimation result output unit 70, and the control unit not shown. Here, the memory 1605 corresponds to, for example, RAM, ROM (Read Only Memory), flash memory, EPROM (Erasable Programmable ROM), EEPROM (registered trademark. Omitted below) (Electrically Erasable Programmable Read-Only Memory), etc., non-volatile or volatile semiconductor memory, magnetic disk, floppy disk, optical disk, high-density disk, mini disk, DVD (Digital Versatile Disc), etc.

[0293] In addition, the functions of the sensor data acquisition unit 10, the abnormality detection unit 30, the abnormality detection order estimation unit 40, the abnormality propagation path tracking unit 50, the abnormality cause estimation unit 60, the abnormality cause estimation result output unit 70, and the control unit not shown in the figure may be partially realized by dedicated hardware and partially realized by software or firmware. For example, the functions of the sensor data acquisition unit 10 and the abnormality cause estimation result output unit 70 can be realized by the processing circuit 1601 as dedicated hardware, and the functions of the abnormality detection order estimation unit 40, the abnormality propagation path tracking unit 50, the abnormality cause estimation unit 60, and the control unit not shown in the figure can be realized by the processor 1604 reading and executing the program stored in the memory 1605.

[0294] The data storage unit 20 is composed of an auxiliary storage device (not shown).

[0295] Furthermore, the abnormality cause estimation device 100 includes devices such as the sensor 300 and the display device 400 , and an input interface device 1602 and an output interface device 1603 that perform wired or wireless communication.

[0296] The hardware configuration example of the learning device 200 according to the first embodiment is also as follows Fig.16A and Fig. 16B shown.

[0297] In the first embodiment, the functions of the learning sensor data acquisition unit 210, the learning preprocessing unit 230, and the association structure learning unit 240 are implemented by the processing circuit 1601. That is, the learning device 200 includes the processing circuit 1601 for performing the following control: based on a plurality of time series of sensor data collected by a plurality of sensors 300 provided in the target device when the target device is operating normally, the estimated structure (association structure D2) indicating the dependency relationship between a plurality of device structural elements constituting the target device is learned.

[0298] The processing circuit 1601 may be as follows Fig.16A As shown in the figure, it is a dedicated hardware, or it can be Fig. 16B Shown is a processor 1604 that executes programs stored in memory.

[0299] When the processing circuit 1601 is dedicated hardware, the processing circuit 1601 corresponds to, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or a combination thereof.

[0300] When the processing circuit is a processor 1604, the functions of the learning sensor data acquisition unit 210, the learning preprocessing unit 230, and the associated structure learning unit 240 are implemented by software, firmware, or a combination of software and firmware. The software or firmware is described in the form of a program and stored in the memory 1605. The processor 1604 reads and executes the program stored in the memory 1605 to execute the functions of the learning sensor data acquisition unit 210, the learning preprocessing unit 230, and the associated structure learning unit 240. That is, the learning device 200 has a memory 1605 for storing a program, and the program is executed by the processor 1604. Fig.14Steps ST21 to ST23 of the present invention. In addition, the program stored in the memory 1605 can also be said to be a program that causes the computer to execute the processing procedures or methods of the learning sensor data acquisition unit 210, the learning preprocessing unit 230, and the associated structure learning unit 240. Here, the memory 1605 corresponds to, for example, RAM, ROM (Read Only Memory), flash memory, EPROM (Erasable Programmable ROM), EEPROM (Electrically Erasable Programmable Read-Only Memory) and other non-volatile or volatile semiconductor memories, magnetic disks, floppy disks, optical disks, high-density disks, mini disks, DVDs (Digital Versatile Discs), etc.

[0301] In addition, the functions of the learning sensor data acquisition unit 210, the learning preprocessing unit 230, and the association structure learning unit 240 may be partially implemented by dedicated hardware and partially implemented by software or firmware. For example, the learning sensor data acquisition unit 210 may be implemented by the processing circuit 1601 as dedicated hardware, and the learning preprocessing unit 230 and the association structure learning unit 240 may be implemented by the processor 1604 reading and executing the program stored in the memory 1605.

[0302] The learning data storage unit 220 is composed of an auxiliary storage device (not shown).

[0303] Furthermore, the learning device 200 includes devices such as the abnormality cause estimation device 100 , and an input interface device 1602 and an output interface device 1603 that perform wired or wireless communication.

[0304] In the above embodiment 1, for convenience, it is assumed that one sensor 300 is provided in one machine, but this is only an example, and one machine may be provided with a plurality of sensors 300. For example, in the above embodiment 1, the equipment components may be a plurality of parts of the machine, and each part may be provided with a sensor 300. In this case, the abnormality cause estimation device 100 presents information related to the estimation result of the abnormality cause to the operator in a manner that enables the operator to understand the sensors provided in the parts that become the abnormality cause or the order in which the operator should inspect.

[0305] 〈Variation Example〉

[0306] In the above embodiment 1, the abnormality cause estimation device 100 and the learning device 200 are respectively provided with the sensor data acquisition unit 10 and the learning sensor data acquisition unit 210, but this is only an example. In addition, in the above embodiment 1, the abnormality cause estimation device 100 and the learning device 200 are respectively provided with the data storage unit 20 and the learning data storage unit 220, but this is only an example.

[0307] For example, the abnormality cause estimation device 100 and the learning device 200 may include a common sensor data acquisition unit and a common data storage unit and may access each other.

[0308] Fig.17 This is a diagram showing a configuration example of a precise diagnosis system 1000 in which the abnormality cause estimation device 100 and the learning device 200 in the first embodiment include a sensor data acquisition unit 310 and a data storage unit 320 that are shared.

[0309] In addition, Fig.17 Although the diagram is omitted for simplicity of explanation, the abnormality cause estimation device 100 includes, in addition to the data storage unit 320, an abnormality detection unit 30, an abnormality detection sequence estimation unit 40, an abnormality propagation path tracking unit 50, an abnormality cause estimation unit 60, an abnormality cause estimation result output unit 70, and a control unit. However, the abnormality cause estimation device 100 does not necessarily have the abnormality cause estimation result output unit 70. In addition, Fig.17 Although not shown in the figure for the sake of simplicity, the learning device 200 includes a learning preprocessing unit 230 and a correlation structure learning unit 240 in addition to the sensor data acquisition unit 310 .

[0310] exist Fig.17 In the structural example of the precision diagnosis system 1000 shown, the abnormality cause estimation device 100 has a data storage unit 320, and the learning device 200 has a sensor data acquisition unit 310, but this is just an example. In the precision diagnosis system 1000, it is also possible that the learning device 200 has a data storage unit 320, and the abnormality cause estimation device 100 has a sensor data acquisition unit 310.

[0311] In addition, in the precise diagnosis system 1000 , the sensor data acquisition unit 310 and the data storage unit 320 may be provided in either the abnormality cause estimation device 100 or the learning device 200 .

[0312] 〈Variation Example〉

[0313] In the above-mentioned first embodiment, the abnormality cause estimation device 100 is configured to estimate the cause of abnormality based on one correlation structure D2, but this is only an example.

[0314] For example, the abnormality cause estimation device 100 may be configured to estimate the abnormality cause based on the correlation structure corresponding to the operating state of the target device. In this case, the learning device 200 learns the correlation structure for each operating state of the target device.

[0315] Fig.18 This is a diagram showing a structural example of a precision diagnosis system 1000 in which, in implementation mode 1, a learning device 200 learns an associated structure according to each operating state of an object device, and an abnormality cause estimation device 100 estimates the cause of an abnormality based on the associated structure corresponding to the operating state of the object device learned by the learning device 200.

[0316] In addition, Fig.18 Although the illustration is omitted for simplicity of explanation, the abnormality cause estimation device 100 includes a sensor data acquisition unit 10, an abnormality detection unit 30, an abnormality detection sequence estimation unit 40, an abnormality cause estimation unit 60, an abnormality cause estimation result output unit 70, and a control unit in addition to the data storage unit 20 and the abnormality propagation path tracking unit 50. However, the abnormality cause estimation device 100 does not necessarily have the abnormality cause estimation result output unit 70. In addition, Fig.18 Although not shown in the figure for simplicity of explanation, the learning device 200 includes a learning sensor data acquisition unit 210 and a learning pre-processing unit 230 in addition to a learning data storage unit 220 and a correlation structure learning unit 240 .

[0317] like Fig.18 As shown, in the learning device 200, the association structure learning unit 240 obtains the device operating state information D31 indicating the operating state of the target device corresponding to the learning data candidate D21. For example, an operator operates an input device such as a mouse or a keyboard to input the device operating state information D31, and the association structure learning unit 240 receives the input device operating state information D31, thereby obtaining the device operating state information D31. For example, the association structure learning unit 240 may obtain a control signal or the like in the target device, estimate the operating state based on the obtained control signal or the like, and thereby obtain the device operating state information D31.

[0318] Then, after learning the association structure D2 based on the learning data D23 output from the learning preprocessing unit 230, the association structure learning unit 240 assigns the acquired device operating state information D31 to the learned association structure D2 as an association structure D32. The association structure learning unit 240 stores the association structure D32 assigned with the device operating state information D31 in the data storage unit 20 of the abnormality cause estimation device 100. The association structure learning unit 240 may also store the association structure D32 in the learning data storage unit 220.

[0319] The learning device 200 performs the above-described learning based on various operating states of the target equipment, and learns the association structure D32 corresponding to the various operating states.

[0320] In addition, in this case, when using Fig.14 In the operation of the learning device 200 illustrated in the flowchart, the association structure learning unit 240 obtains the device operating state information D31 in advance before performing the process of step ST23, and in step ST23, performs the process of generating and storing the association structure D32. The learning device 200 repeats the process according to the operating state of the target device. Fig.14 The actions are shown in the flowchart.

[0321] In the abnormality cause estimation device 100, the abnormality propagation path tracking unit 50 obtains the device operating state information D31. Then, when performing the abnormality propagation order estimation process, the abnormality propagation path tracking unit 50 estimates the abnormality propagation order based on the acquired abnormality detection sensor information D3, the device operating state information D31, and the association structure D32 stored in the data storage unit 20 by the learning device 200. Specifically, the abnormality propagation path tracking unit 50 selects the association structure D32 corresponding to the operating state of the target device, and estimates the abnormality propagation order using the selected association structure D32.

[0322] For example, in the learning device 200, the association structure learning unit 240 may pre-store the association structure D32 in the learning data storage unit 220, and in the abnormality cause estimation device 100, the abnormality propagation path tracking unit 50 may download the association structure D32 to be used from the learning data storage unit 220 to the data storage unit 20 each time the abnormality propagation order estimation process is performed.

[0323] In addition, in this case, when using Fig.13 In the operation of the abnormality cause estimation device 100 illustrated in the flowchart, the abnormality propagation path tracking unit 50 estimates the abnormality propagation order in step ST4 based on the abnormality detection sensor information D3, the equipment operation status information D31, and the association structure D32 stored in the data storage unit 20 by the learning device 200.

[0324] In this way, the abnormality cause estimation device 100 can be structured as follows: the abnormality propagation path tracking unit 50 estimates the abnormality propagation order based on the abnormality detection sensor information D3, the equipment operation status information D31 representing the operation status of the object equipment, and the estimated structure (association structure D32) that shows the dependency relationship between multiple object structural elements constituting the object equipment according to the operation status of the object equipment.

[0325] By adopting such a configuration, the abnormality cause estimation device 100 can estimate the abnormality cause in detail based on the correlation structure D32 that can cope with the change in the dependency relationship between the sensors 300 accompanying the change in the operating state of the target equipment and has improved reliability.

[0326] In the learning device 200, after the association structure learning unit 240 calculates at least one statistic between multiple sensor data based on the learning data and learns the estimated structure (association structure D2) based on the calculated statistic, it generates an association structure D32 obtained by assigning equipment operating status information D31 to the association structure D2. As a result, the reliability of the association structure provided to the abnormal cause estimation device 100 is improved, and the association structure D32 that can make a detailed estimate of the cause of the abnormality can be provided to the abnormal cause estimation device 100.

[0327] 〈Variation Example〉

[0328] In the above-mentioned first embodiment, the abnormality cause estimation device 100 includes the data storage unit 20 , but this is only an example.

[0329] For example, a single or multiple network storage devices (not shown) configured on a communication network may store various data, and in the abnormality cause estimation device 100, the abnormality detection unit 30, the abnormality detection sequence estimation unit 40, the abnormality propagation path tracking unit 50, the abnormality cause estimation unit 60, and the abnormality cause estimation result output unit 70 access the network storage device.

[0330] 〈Variation Example〉

[0331] In the above first embodiment, in the abnormality cause estimation device 100 , the abnormality detection unit 30 performs abnormality detection processing on the sensor data D1 using a well-known univariate abnormality detection method to detect an abnormality detection sensor in the sensor 300 . However, this is only an example.

[0332] For example, the abnormality detection unit 30 may also use a known multivariate abnormality detection method to detect the abnormality detection sensor. For example, as a known multivariate abnormality detection method, a method such as Graphical Lasso is cited. In addition, the abnormality detection unit 30 may also use a single variable abnormality detection method and a multivariate abnormality detection method at the same time to perform abnormality detection processing.

[0333] In this way, the abnormality cause estimation device 100 detects the abnormality detection sensor by using a single-variable abnormality detection method, a multi-variable abnormality detection method, or both methods. Even if the way in which the abnormality occurring in the target device appears in the sensor data differs depending on the type of the abnormality, it is possible to appropriately detect the abnormality detection sensor that detects the occurrence of various types of abnormalities and to make a detailed estimate of the cause of the abnormality.

[0334] 〈Variation Example〉

[0335] In the above-mentioned first embodiment, the abnormality cause estimation device 100 may be configured to include a correlation structure correction unit 330 that corrects the correlation structure D2 stored in the data storage unit 20 .

[0336] Fig.19 This is a diagram showing a configuration example of the abnormality cause estimation device 100 including the correlation structure correction unit 330 in the first embodiment.

[0337] In addition, Fig.19 Although the illustration is omitted for simplicity of explanation, the abnormality cause estimation device 100 includes a sensor data acquisition unit 10, an abnormality detection unit 30, an abnormality detection sequence estimation unit 40, an abnormality cause estimation unit 60, an abnormality cause estimation result output unit 70, and a control unit in addition to the associated structure correction unit 330, the data storage unit 20, and the abnormality propagation path tracking unit 50. However, the abnormality cause estimation device 100 does not necessarily have the abnormality cause estimation result output unit 70. In addition, Fig.19 Although not shown in the figure for the sake of simplicity, the abnormality cause estimation device 100 is connected to the learning device 200 , the sensor 300 , and the display device 400 .

[0338] The associated structure correction unit 330 obtains information D33 related to sensor pairs having a dependency relationship (hereinafter referred to as "dependent pair information") and information D34 related to sensor pairs having no dependency relationship (hereinafter referred to as "non-dependent pair information") . The dependent pair information D33 and the non-dependent pair information D34 may be manually generated by an operator based on the operator's technical knowledge, or may be generated by the associated structure correction unit 330 based on information indicating a physical connection relationship such as design information of the device. For example, the dependent pair information D33 and the non-dependent pair information D34 may be generated by combining manual generation based on the operator's technical knowledge with generation of information indicating a physical connection relationship based on design information of the device by the associated structure correction unit 330.

[0339] The association structure correction unit 330 obtains the association structure D2 from the data storage unit 20, corrects the dependency relationship between the sensor data with respect to the association structure D2 based on the dependent pair information D33 and the non-dependent pair information D34, and stores the corrected association structure D35 in the data storage unit 20. In addition, the corrected association structure D35 has the same data structure as the association structure D2.

[0340] When the corrected association structure D35 is stored in the data storage unit 20 , the abnormality propagation path tracking unit 50 uses the association structure D35 to perform abnormality propagation order estimation processing.

[0341] The details of the process of correcting the related structure D2 by the related structure correction unit 330 will be described using a specific example.

[0342] Fig. 20 This is a diagram for explaining the concept of an example of a process of correcting the correlation structure D2 by the correlation structure correcting unit 330 in the abnormality cause estimation device 100 including the correlation structure correcting unit 330 in the first embodiment.

[0343] Here, as an example, the number of sensors 300 is set to 3, and the sensors 300 are represented by sensors Xn (n=1, 2, 3). In addition, in the dependent pair information D33, it is defined that the i-th sensor Xi from the sensor Xn has a dependency relationship with the j-th sensor Xj, and in the non-dependent pair information D34, it is defined that the i-th sensor Xi from the sensor Xn has no dependency relationship with the j-th sensor Xj. i and j are both 1, 2, or 3.

[0344] At this time, for example, among the sensors X1 , X2 , and X3 , there is a dependency relationship between sensor data from sensor X1 to sensor X2 and from sensor X3 to sensor X2 , and there is no dependency relationship between sensor data from sensor X2 to sensor X1 .

[0345] Assume that the association structure D2 is a three-dimensional array with the first dimension being the statistical index, the second dimension being the number of sensors Xn, and the third dimension being the number of sensors Xn. The number of types of statistical indexes is two.

[0346] The association structure correction unit 330 corrects the statistic a(k)ij of the association structure D2 to the statistic a'(k)ij based on the dependent pair information D33 and the non-dependent pair information D34. Here, the statistic a(k)ij and the statistic a'(k)ij are real values. For example, the association structure correction unit 330 corrects the statistic a(k)ij of the association structure D2 corresponding to the pair of sensors Xi and Xj included in the dependent pair information D33 to the statistic a'(k)ij that is larger than the upper limit value of the statistic and shows the case of a dependency relationship according to each statistical index. In addition, this is just an example. For example, the association structure correction unit 330 may also correct the statistic a(k)ij of the association structure D2 corresponding to the pair of sensors Xi and Xj included in the dependent pair information D33 to the statistic a'(k)ij that is larger than the threshold value set for selecting the dependency relationship.

[0347] exist Fig. 20 In the example, the association structure correction unit 330 corrects the statistics a(1)12 and a(2)12 corresponding to the sensor data from the sensor X1 to the sensor X2 having a dependency relationship to the statistics a'(1)12 and a'(2)12, which are upper limit values ​​of the statistical index, based on the dependency pair information D33. In addition, the association structure correction unit 330 corrects the statistics a(1)32 and a(2)32 corresponding to the sensor data from the sensor X3 to the sensor X2 having a dependency relationship to the statistics a'(1)32 and a'(2)32, which are upper limit values ​​of the statistical index, based on the dependency pair information D33.

[0348] Furthermore, the correlation structure correction unit 330 corrects the statistic a(k)ij of the correlation structure D2 corresponding to the pair of sensors Xi and Xj included in the independent pair information D34 to a statistic a′(k)ij indicating the lack of dependency for each statistical index.

[0349] exist Fig. 20 In the example, the association structure correction unit 330 corrects the statistics a(1)21 and a(2)21 corresponding to the sensor data from the sensor X2 to the sensor X1 that have no dependency relationship to statistics a'(1)21 and a'(2)21 indicating no dependency relationship, respectively, based on the non-dependency pair information D34.

[0350] like Fig.19 As shown, when the abnormality cause estimation device 100 is configured to include the associated structure correction unit 330, Fig.13In the operation of the abnormality cause estimation device 100 illustrated in the flowchart, the association structure correction unit 330 corrects the dependency relationship between the sensor data for the association structure D2 based on the dependent pair information D33 and the non-dependent pair information D34 before performing the processing of step ST4, and stores the corrected association structure D35 in the data storage unit 20. The abnormality propagation path tracking unit 50 estimates the abnormality propagation order based on the abnormality detection sensor information D3, the equipment operation state information D31, and the association structure D35 corrected by the association structure correction unit 330 in step ST4.

[0351] In this way, the abnormality cause estimation device 100 can improve the reliability of the associated structure D2 and make a detailed estimate of the abnormality cause by having an associated structure correction unit 330, wherein the associated structure correction unit 330 corrects the dependency relationship between sensor data for the associated structure D2 (estimated structure) based on dependent pair information D33 related to pairs of sensors 300 with a dependency relationship among the sensors 300 and non-dependent pair information D34 related to pairs of sensors 300 without a dependency relationship among the sensors 300.

[0352] 〈Variation Example〉

[0353] In the above embodiment 1, the abnormality cause estimation device 100 may also be configured to include a relationship change estimation unit 340, which estimates the change in the relationship between the sensor data by comparing the learned association structure D2 with the association structure D36 when the abnormality occurs. The change in the relationship between the sensor data is assumed to be, for example, the destruction of the relationship between the sensor data. The relationship change estimation unit 340 estimates the location where the destruction of the relationship between the sensor data is large among the elements of the association structure D2 as the location where the relationship between the sensor data has changed.

[0354] Fig.21 This is a diagram showing a configuration example of the abnormality cause estimation device 100 including the relationship change estimation unit 340 in the first embodiment.

[0355] In addition, Fig.21 Although the illustration is omitted for simplicity of explanation, the abnormality cause estimation device 100 includes a sensor data acquisition unit 10, an abnormality detection unit 30, an abnormality detection sequence estimation unit 40, an abnormality propagation path tracking unit 50, and a control unit in addition to the relationship change estimation unit 340, the data storage unit 20, the abnormality cause estimation unit 60, and the abnormality cause estimation result output unit 70. However, the abnormality cause estimation device 100 does not necessarily have the abnormality cause estimation result output unit 70. In addition, Fig.21Although not shown in the figure for simplicity of explanation, the learning device 200 includes a learning sensor data acquisition unit 210 in addition to a learning data storage unit 220 , a learning preprocessing unit 230 , and a correlation structure learning unit 240 .

[0356] The relationship change estimation unit 340 obtains the association structure D2, the association structure D36 at the time of abnormality, and the abnormality detection sensor information D3 from the data storage unit 20. Then, the relationship change estimation unit 340 performs a relationship change order estimation process, in which the relationship structure D2 is compared with the association structure D36 based on the association structure D2, the association structure D36, and the abnormality detection sensor information D3, and the order of change of the relationship between the sensor data is estimated (hereinafter referred to as "relationship change order").

[0357] The association structure D36 at the time of abnormality is acquired through the following processing. In the learning device 200, the learning sensor data acquisition unit 210 acquires the sensor data D1 when the abnormality occurs (including the period when the abnormality is detected) from the data storage unit 20 of the abnormality cause estimation device 100, and stores it in the learning data storage unit 220. The learning preprocessing unit 230 acquires the sensor data D1 when the abnormality occurs from the learning data storage unit 220, preprocesses the acquired sensor data D1 when the abnormality occurs, and outputs the preprocessed sensor data D38 when the abnormality occurs to the association structure learning unit 240. In addition, the learning preprocessing unit 230 may also output the preprocessed sensor data D38 when the abnormality occurs to the association structure learning unit 240 via the learning data storage unit 220. The association structure learning unit 240 learns the association structure D36 at the time of abnormality based on the preprocessed sensor data D38 when the abnormality occurs output from the learning preprocessing unit 230. The association structure learning unit 240 may learn the association structure D36 in the same manner as the manner in which the association structure D2 is learned. The association structure learning unit 240 learns the association structure D36 at the time of abnormality based on the sensor data D1 at the time of abnormality occurrence.

[0358] The correlation structure learning unit 240 stores the learned abnormal correlation structure D36 in the data storage unit 20 of the abnormality cause estimation device 100 .

[0359] The details of the relationship change order estimation process performed by the relationship change estimation unit 340 will be described.

[0360] First, the relationship change estimation unit 340 obtains the association structure D2 and the association structure D36 when abnormal from the data storage unit 20, and obtains the association structure (hereinafter referred to as the "association structure for relationship change estimation") and the association structure when abnormal (hereinafter referred to as the "association structure for relationship change estimation") from the association structure D2 and the association structure D36 when abnormal. The association structure (hereinafter referred to as the "association structure for relationship change estimation") is an association structure corresponding to a non-directional statistical index selected for estimating relationship changes. Here, the association structure for relationship change estimation and the association structure for relationship change estimation when abnormal are two-dimensional matrices A(k) and A'(k) obtained by extracting only the parts corresponding to the kth statistical index from the association structure D2 and the association structure D36 when abnormal, respectively. The relationship change estimation unit 340 selects a statistical index that is meaningful when comparing statistics as the statistical index to be extracted. For example, the relationship change estimation unit 340 sets the statistical index calculated based on the p-value of the hypothesis test as a statistical index that is not meaningful when comparing statistics, and selects other statistical indexes as statistical indexes that are meaningful when comparing statistics. The relationship change estimation unit 340 can estimate the destruction of the relationship, in other words, the relationship change, by comparing the magnitude relationship of the correlation coefficient, which is a statistic corresponding to the correlation.

[0361] Next, the relationship change estimation unit 340 calculates the change d(k)ij based on the element a(k)ij of the association structure for relationship change estimation and the element a'(k)ij of the abnormal time association structure for relationship change estimation. Then, the relationship change estimation unit 340 generates information, i.e., the association structure change, which uses the calculated change d(k)ij as an element and represents the change d(k)ij in a matrix.

[0362] Here, the elements a(k)ij, a'(k)ij and the change d(k)ij are all real values. The relationship change estimation unit 340 may calculate the change d(k)ij of the element as the change amount of the association structure using the absolute value of the difference between the absolute value of the element a(k)ij of the association structure for relationship change estimation and the absolute value of the element a'(k)ij of the association structure for relationship change estimation when abnormality occurs, or may calculate the change d(k)ij of the element as the change amount of the association structure using the absolute value of the difference between the element a(k)ij of the association structure for relationship change estimation and the element a'(k)ij of the association structure for relationship change estimation when abnormality occurs.

[0363] Next, the relationship change estimation unit 340 obtains the abnormality detection sensor information D3 from the data storage unit 20, and calculates the relationship change degree for each sensor 300, more specifically, each abnormality detection sensor, based on the calculated element of the associated structure change, namely the change d(k)ij and the abnormality detection sensor information D3.

[0364] Here, both the change amount d(k)ij and the relationship change degree are real values. The relationship change estimation unit 340 calculates the relationship change degree corresponding to the i-th sensor 300, that is, the abnormality detection sensor Xi, among the n sensors 300, using, for example, the average of the elements in the i-th row of the association structure change amount excluding the i-th column and corresponding to the sensor 300 (i.e., the abnormality detection sensor) included in the abnormality detection sensor information D3. In addition, this is just an example, and the relationship change estimation unit 340 may also calculate the relationship change degree corresponding to the i-th sensor 300, that is, the abnormality detection sensor Xi, among the n sensors 300, using, for example, the average of the elements in the i-th row of the association structure change amount excluding the i-th column.

[0365] Next, the relationship change estimation unit 340 assigns the corresponding relationship change order to the abnormality detection sensor based on the calculated relationship change degree corresponding to the abnormality detection sensor. Here, the relationship change degree and the relationship change order are both real values. The relationship change estimation unit 340 assigns the relationship change order in a manner such that the corresponding relationship change order becomes ascending order, for example, from the abnormality detection sensor with a larger relationship change degree value. However, when the relationship change degrees corresponding to multiple abnormality detection sensors are equal, the relationship change estimation unit 340 assigns the same relationship change order to the multiple abnormality detection sensors.

[0366] After assigning the relationship change order, the relationship change estimation unit 340 generates a relationship change order estimation result D37. The relationship change order estimation result is information in which information indicating the abnormality detection sensor, the relationship change degree, and the relationship change order are associated with each other.

[0367] The relationship change estimation unit 340 stores the relationship change order estimation result D37 in the data storage unit 20 .

[0368] The relationship change order estimation process performed by the relationship change estimation unit 340 as described above will be described with reference to the drawings and specific examples.

[0369] Fig. 22 This is a diagram for explaining the concept of an example of relationship change order estimation processing performed by the relationship change estimation unit 340 based on the learned relationship structure D2 and the abnormal relationship structure D36 when the abnormality cause estimation device 100 of the first embodiment is configured to include the relationship change estimation unit 340.

[0370] Here, as an example, the number of sensors 300 is set to 4, and the sensors 300 are represented by sensors Xn (n=1, 2, 3, 4). Also, as an example, it is assumed that sensors X1, X2, and X3 among sensors X1, X2, X3, and X4 are abnormality detection sensors.

[0371] First, the relationship change estimation unit 340 obtains the association structure D2 and the abnormal association structure D36 from the data storage unit 20, and obtains the relationship change estimation association structure D2A and the abnormal association structure D36A from the association structure D2 and the abnormal association structure D36. Fig. 22 In the figure, the associated structure D2 and the associated structure D36 in the abnormal state are omitted.

[0372] exist Fig. 22 In the example, the relationship change estimation association structure D2A is represented by a two-dimensional matrix obtained by extracting only the part corresponding to the k-th statistical index from the association structure D2. The relationship change estimation association structure D2A is represented by a two-dimensional matrix with the first dimension set to the number of sensors Xn4 and the second dimension set to the number of sensors Xn4. The statistical index is an undirected statistical index, and the type of the statistical index is correlation. In addition, as Fig. 22 As shown, the abnormal-time correlation structure D36A for estimating a relationship change and the correlation structure change amount D39 have the same data structure as the correlation structure D2A for estimating a relationship change.

[0373] exist Fig. 22 As an example, the element of the association structure change amount D39, that is, the change amount d(k)ij, is the absolute value of the difference between the absolute value of the element a(k)ij of the association structure D2A for relationship change estimation and the absolute value of the element a'(k)ij of the abnormal time association structure D36A for relationship change estimation. For example, the relationship change estimation unit 340 calculates the change amount between the sensor X1 and the sensor X2 as ||a(k)12|-|a'(k)12||=d(k)12 based on the element a(k)12 in the first row and second column of the association structure D2A for relationship change estimation and the element a'(k)12 in the first row and second column of the abnormal time association structure D36A for relationship change estimation. Here, |·| represents an absolute value.

[0374] In addition, Fig. 22In the example, the relationship change estimation unit 340 calculates the relationship change degree dn corresponding to the abnormality detection sensor Xn using the average of the elements corresponding to the sensor 300 (i.e., the abnormality detection sensor) included in the abnormality detection sensor information D3 in the n-th row excluding the n-th column of the association structure change amount D39. Therefore, the relationship change estimation unit 340, for example, sets the relationship change degree d1 corresponding to the abnormality detection sensor X1 to the average of the elements d(k)nn (where n=1, 2, 3) corresponding to the abnormality detection sensors X1, X2, and X3 in the first row excluding the first column of the association structure change amount D39. Specifically, the relationship change estimation unit 340 calculates the relationship change degree d1 of the abnormality detection sensor X1 as the average of the element d(k)12 in the first row and the second column and the element d(k)13 in the first row and the third column of the association structure change amount D39 corresponding to the abnormality detection sensors X2 and X3. The relationship change estimation unit 340 similarly calculates the relationship change degrees d2 and d3 corresponding to the sensors X2 and X3, respectively. The relationship change estimation unit 340 assigns corresponding relationship change orders o1, o2, and o3 to the sensors X1, X2, and X3, respectively, based on the calculated relationship change degrees d1, d2, and d3.

[0375] Then, the relationship change estimation unit 340 stores the relationship change order estimation result D37 in the data storage unit 20 .

[0376] The abnormality cause estimation unit 60 obtains the abnormality detection order estimation result D5, the abnormality propagation order estimation result D6, and the relationship change order estimation result D37 from the data storage unit 20, and performs abnormality cause estimation processing that takes into account the relationship change order based on the abnormality detection order estimation result D5, the abnormality propagation order estimation result D6, and the relationship change order estimation result D37. In detail, performing abnormality cause estimation processing that takes into account the relationship change order based on the abnormality detection order estimation result D5, the abnormality propagation order estimation result D6, and the relationship change order estimation result D37 means that the abnormality cause estimation unit 60 calculates the corresponding abnormality cause score based on the abnormality detection order included in the abnormality detection order estimation result D5, the abnormality propagation order included in the abnormality propagation order estimation result D6, and the relationship change order included in the relationship change order estimation result D37, and estimates the abnormality cause order based on the calculated abnormality cause score.

[0377] The abnormality cause estimation unit 60 may calculate the abnormality cause score from the abnormality detection order, the abnormality propagation order, and the relationship change order in the same way as calculating the abnormality cause score from the abnormality detection order and the abnormality propagation order.

[0378] The abnormality cause estimation unit 60 stores the abnormality cause order estimation result D40 taking the relationship change order into consideration in the data storage unit 20 .

[0379] The abnormal cause estimation result output unit 70 obtains the abnormal cause sequence estimation result D40, the abnormal detection sequence estimation result D5, the abnormal propagation sequence estimation result D6, and the relationship change sequence estimation result D37 from the data storage unit 20, and outputs information related to the estimation result of the abnormal cause estimated by the abnormal cause estimation unit 60. Specifically, the abnormal cause estimation result output unit 70 outputs abnormal cause estimation result display information for displaying an abnormal cause estimation result screen to the display device 400 based on the abnormal cause sequence estimation result D40, the abnormal detection sequence estimation result D5, the abnormal propagation sequence estimation result D6, and the relationship change sequence estimation result D37. The abnormal cause estimation result screen shows information related to the estimation result of the abnormal cause estimated by the abnormal cause estimation unit 60.

[0380] In addition, in the abnormality cause estimation device 100, Fig.21 When the relationship change estimation unit 340 is provided as shown in FIG. Fig.13 In the operation of the abnormality cause estimation device 100 described in the flowchart, the relationship change estimation unit 340 estimates the change in the relationship between the sensor data before performing the process of step ST5 and stores the relationship change order estimation result D37 in the data storage unit 20.

[0381] In step ST5, the abnormality cause estimation unit 60 obtains the abnormality detection order estimation result D5, the abnormality propagation order estimation result D6 and the relationship change order estimation result D37 from the data storage unit 20, and performs abnormality cause estimation processing taking into account the relationship change order based on the abnormality detection order estimation result D5, the abnormality propagation order estimation result D6 and the relationship change order estimation result D37.

[0382] Thus, the abnormality cause estimation device 100 includes a relationship change estimation unit 340, which compares the association structure D2 (estimated structure) with the association structure D36 when the abnormality occurs, and the abnormality detection sensor information D3, and estimates the change in the relationship between the sensor data. The abnormality cause estimation unit 60 is configured to estimate the cause of the abnormality by considering the change in the relationship between the sensor data estimated by the relationship change estimation unit 340 based on the abnormality detection order estimated by the abnormality detection order estimation unit 40 and the abnormality propagation order estimated by the abnormality propagation path tracking unit 50. As a result, the reliability of the abnormality cause order estimation result D7 is improved, and the abnormality cause can be estimated in detail. By including the relationship change estimation unit 340 in the abnormality cause estimation device 100, a reference for estimating the abnormality cause in the abnormality cause estimation device 100 is added.

[0383] 〈Variation Example〉

[0384] In the above embodiment 1, the abnormality cause estimation device 100 "estimates the cause of the abnormality" means estimating the abnormality cause score indicating the degree of possibility of the abnormality source and the abnormality cause order based on the abnormality cause score for each sensor 300, and generating information related to the abnormality cause score and the abnormality cause order. In addition, the abnormality cause estimation device 100 "estimates the cause of the abnormality" may also include the following: estimating the abnormality cause score and the abnormality cause order based on the abnormality cause score for each device.

[0385] Fig.23 This is a diagram showing a configuration example of the abnormality cause estimation device 100 which includes the abnormality cause device estimation unit 350 in the first embodiment and has a configuration for estimating the cause of abnormality for each device.

[0386] In addition, Fig.23 Although the illustration is omitted for simplicity of explanation, the abnormality cause estimation device 100 includes a sensor data acquisition unit 10, an abnormality detection unit 30, an abnormality detection sequence estimation unit 40, an abnormality propagation path tracking unit 50, an abnormality cause estimation unit 60, and a control unit in addition to the abnormality cause machine estimation unit 350, the data storage unit 20, and the abnormality cause estimation result output unit 70. However, the abnormality cause estimation device 100 does not necessarily have the abnormality cause estimation result output unit 70. In addition, Fig.23 Although not shown in the figure for the sake of simplicity, the abnormality cause estimation device 100 is connected to the learning device 200 .

[0387] The abnormality cause machine estimation unit 350 obtains the machine-attached sensor information D41. In addition, the abnormality cause machine estimation unit 350 obtains the abnormality detection order estimation result D5 and the abnormality propagation order estimation result D6 from the data storage unit 20. The abnormality cause machine estimation unit 350 performs a machine-based abnormality cause estimation process in which the cause of the abnormality is estimated on a machine-by-machine basis based on the machine-attached sensor information D41, the abnormality detection order estimation result D5, and the abnormality propagation order estimation result D6.

[0388] The machine-attached sensor information D41 is table data indicating which machine the sensor 300 is installed in. For example, an operator operates an input device such as a mouse or a keyboard to input the machine-attached sensor information D41, and the abnormal cause machine estimation unit 350 receives the input machine-attached sensor information D41 to obtain the machine-attached sensor information D41.

[0389] For example, in the device-attached sensor information D41 , information indicating a device and information indicating the sensor 300 provided in the device are associated with each other.

[0390] The machine-based abnormality cause estimation process performed by the abnormality cause machine estimation unit 350 will be described in detail.

[0391] First, the abnormality-causing device estimation unit 350 obtains the device-attached sensor information D41 , the abnormality detection order estimation result D5 , and the abnormality propagation order estimation result D6 .

[0392] The abnormality cause device estimation unit 350 converts the abnormality detection order estimation result D5 into a device-based abnormality detection order estimation result (hereinafter referred to as “device abnormality detection order estimation result”) D42 based on the device-attached sensor information D41 and the abnormality detection order estimation result D5.

[0393] Specifically, the abnormality-cause machine estimation unit 350 matches the information of the sensor 300 corresponding to the machine attached sensor information D41 with the information of the sensor 300 included in the abnormality detection order estimation result D5 for a certain machine U among U machines (U is an integer). Then, the abnormality-cause machine estimation unit 350 obtains the abnormality detection order corresponding to the information of the matched sensor 300 in the abnormality detection order estimation result D5, and calculates the abnormality detection order total value. Here, the abnormality detection order total value is a real value. The abnormality-cause machine estimation unit 350, for example, uses the weighted average of the abnormality detection order corresponding to the information of the matched sensor 300 to set the representative value as the abnormality detection order total value. In addition, the abnormality-cause machine estimation unit 350 may also use the representative value such as the minimum value or the maximum value of the abnormality detection order corresponding to the information of the matched sensor 300 as the abnormality detection order total value.

[0394] Then, the abnormality cause machine estimation unit 350 assigns the abnormality detection order of the machine unit (hereinafter referred to as "machine abnormality detection order") to the machine U based on the calculated total value of the abnormality detection order. Here, the machine abnormality detection order ouU is a real value. The abnormality cause machine estimation unit 350 assigns the machine abnormality detection order to the machine u in a manner such that the corresponding machine abnormality detection order becomes ascending order, starting from the abnormality detection order total value with a smaller value, for example. However, when the abnormality detection order total values ​​calculated for multiple machines are equal, the abnormality cause machine estimation unit 350 assigns the same machine abnormality detection order to the multiple machines.

[0395] The abnormality cause device estimation unit 350 generates a device abnormality detection order estimation result D42 which is information corresponding to information representing the device, the abnormality detection order total value, and the device abnormality detection order for each device, and stores the device abnormality detection order estimation result D42 in the data storage unit 20 .

[0396] Furthermore, the abnormality cause machine estimation unit 350 converts the abnormality propagation order estimation result D6 into the abnormality propagation order estimation result of the machine unit (hereinafter referred to as “machine abnormality propagation order estimation result”) D43 based on the machine attached sensor information D41 and the abnormality propagation order estimation result D6.

[0397] Specifically, when U machines are represented by machine U (U=1, 2, . . . , U), the abnormality-causing machine estimation unit 350 matches the information indicating the sensor 300 corresponding to the machine attached sensor information D41 with the information indicating the sensor 300 included in the abnormality propagation order estimation result D6 for a certain machine U. Then, the abnormality-causing machine estimation unit 350 obtains the abnormality propagation order corresponding to the information indicating the matched sensor 300 in the abnormality propagation order estimation result D6, and calculates the abnormality propagation order total value. Here, the abnormality propagation order total value is a real value. The abnormality-causing machine estimation unit 350 uses, for example, a weighted average of the abnormality propagation order corresponding to the information indicating the matched sensor 300, and sets the representative value as the abnormality propagation order total value. In addition, the abnormality-causing machine estimation unit 350 may also set, for example, a representative value such as the minimum value or the maximum value of the abnormality propagation order corresponding to the information indicating the matched sensor 300 as the abnormality propagation order total value.

[0398] Then, the abnormality cause machine estimation unit 350 assigns the abnormality propagation order of the machine unit (hereinafter referred to as "machine abnormality propagation order") to the machine U based on the calculated abnormality propagation order total value. Here, the abnormality detection order is a real value. The abnormality cause machine estimation unit 350 assigns the machine abnormality propagation order to the machine U in a manner such that the corresponding machine abnormality propagation order is in ascending order, starting from the abnormality propagation order total value with a smaller value. However, when the abnormality propagation order total values ​​calculated for multiple machines are equal, the abnormality cause machine estimation unit 350 assigns the same machine abnormality propagation order to the multiple machines.

[0399] The abnormality cause machine estimation unit 350 generates a machine abnormality propagation order estimation result D43, which is information obtained by associating information indicating the machine, the abnormality detection machine flag, the abnormality propagation order total value, and the machine abnormality propagation order, for each machine, and stores the machine abnormality propagation order estimation result D43 in the data storage unit 20. The abnormality detection machine flag indicates, for each machine, whether there is an abnormality detection sensor in the sensor 300 provided in the machine. The abnormality detection machine flag is a Boolean value.

[0400] In addition, as described above, in the above embodiment 1, a plurality of sensors 300 may be provided in the machine. For example, if there is only one abnormality detection sensor among the plurality of sensors 300 provided in a certain machine U, the abnormality cause machine estimation unit 350 sets (True) the abnormality detection machine flag corresponding to the certain machine U in the machine abnormality propagation order estimation result D43. That is, for example, in the case where a plurality of sensors 300 are provided in a certain machine U and there is one or more abnormality detection sensors among the plurality of sensors 300, the abnormality cause machine estimation unit 350 sets (True) the abnormality detection machine flag corresponding to the certain machine U. On the other hand, for example, in the case where a plurality of sensors 300 are provided in a certain machine U and there is not one abnormality detection sensor among the plurality of sensors 300, the abnormality cause machine estimation unit 350 sets (False) the abnormality detection machine flag.

[0401] Furthermore, the abnormality-causing device estimation unit 350 estimates the cause of the abnormality for each device based on the generated device abnormality detection order estimation result D42 and device abnormality propagation order estimation result D43.

[0402] Specifically, the abnormality cause machine estimation unit 350 calculates a machine abnormality cause score for each machine based on the machine abnormality detection order set in the machine abnormality detection order estimation result D42 and the machine abnormality propagation order set in the machine abnormality propagation order estimation result D43. Here, the machine abnormality cause score is a real value.

[0403] The abnormal cause machine estimation unit 350 uses the weighted average of the machine abnormality detection order and the machine abnormality propagation order for each machine, for example, and uses the representative value as the machine abnormality cause score. In addition, the abnormal cause machine estimation unit 350 may also use the representative value such as the maximum value or minimum value of the machine abnormality detection order and the machine abnormality propagation order as the machine abnormality cause score for each machine. In addition, in the case where only one of the machine abnormality detection order and the machine abnormality propagation order is set, the abnormal cause machine estimation unit 350 may directly use the value of the set one as the machine abnormality cause score. In this case, the abnormal cause machine estimation unit 350 may also consider the case where only one of the orders is set to weight the machine abnormality cause score.

[0404] Then, the abnormal cause machine estimation unit 350 assigns the abnormal cause order of each machine (hereinafter referred to as "machine abnormal cause order") based on the calculated machine abnormal cause score for each machine. Here, the machine abnormal cause order is a real value. For example, the abnormal cause machine estimation unit 350 assigns the machine abnormal cause order in ascending order, starting from the machine with the smaller value of the corresponding machine abnormal cause score, for machine U. However, when the machine abnormal cause scores calculated for multiple machines U are equal, the abnormal cause machine estimation unit 350 assigns the same machine abnormal cause order to the multiple machines U.

[0405] The abnormality cause machine estimation unit 350 generates a machine abnormality cause sequence estimation result D44, which is information obtained by associating information indicating the machine, the abnormality detection machine flag, the machine abnormality cause score, and the machine abnormality cause sequence, for each machine, and stores the machine abnormality cause sequence estimation result D44 in the data storage unit 20. In addition, the abnormality cause machine estimation unit 350 may set, for the machine U, the value of the abnormality detection machine flag set corresponding to the information indicating the machine U in the machine abnormality propagation sequence estimation result D43, in the abnormality detection machine flag corresponding to the machine abnormality cause sequence estimation result D44.

[0406] The machine-based abnormality cause estimation process performed by the abnormality cause machine estimation unit 350 as described above will be described with reference to the drawings and specific examples.

[0407] Fig.24 This is a conceptual diagram for illustrating an example of a machine unit abnormal cause estimation process in which the abnormal cause estimation device 100 of embodiment 1 is configured to include an abnormal cause machine estimation unit 350, and the abnormal cause machine estimation unit 350 estimates the abnormal cause on a machine basis based on machine-attached sensor information D41, abnormal detection order estimation results D5, and abnormal propagation order estimation results D6.

[0408] Here, as an example, the number of sensors 300 is set to 6, and the sensors 300 are represented by sensors Xn (n=1 to 6). In addition, as an example, the number of devices is 3, and the devices are represented by devices U (U=1, 2, 3). In addition, sensors X1 and X2 are provided in device 1, sensors X3, X4, and X5 are provided in device 2, and sensor X6 is provided in device 3.

[0409] The abnormality-causing device estimation unit 350 determines the abnormality detection order on ( Fig.24 The abnormality detection order total value suU (shown in D5C in FIG. 1 ) and the machine-attached sensor information D41 are calculated by, for example, using an average. Fig.24 ). For example, in machine 1, the sensors Xn added to machine 1 are sensor X1 and sensor X2, and therefore, the abnormal cause machine estimation unit 350 sets the abnormal detection order total value su1 corresponding to machine 1 to the average value of the abnormal detection order o1 and the abnormal detection order o2. In addition, for example, in machine 2, the sensors Xn added to machine 2 are sensor X3, sensor X4, and sensor X5. However, among sensors X3, sensor X4, and sensor X5, the abnormal detection sensor, in other words, the sensor Xn included in the abnormal detection order estimation result D5 is only sensor X4. Therefore, the abnormal cause machine estimation unit 350 directly sets the abnormal detection order o4 corresponding to sensor X4 to the abnormal detection order total value su2 corresponding to machine 2. Similarly, the abnormal cause machine estimation unit 350 directly sets the abnormal detection order o6 corresponding to sensor X6 to the abnormal detection order total value su6 corresponding to machine 3.

[0410] Next, the abnormality cause machine estimation unit 350 allocates a machine abnormality detection order ouU to each machine based on the abnormality detection order total value suU calculated for each machine. Specifically, the abnormality cause machine estimation unit 350 allocates corresponding machine abnormality detection orders ou1, ou2, and ou3 to machines 1, 2, and 3, respectively, based on the calculated abnormality detection order total values ​​su1, su2, and su3.

[0411] Then, the abnormality cause device estimation unit 350 generates information indicating devices 1, 2, and 3 ( Fig.24 As shown in D42A in FIG), the total value of abnormal detection sequence su1, su2, su3 ( Fig.24 D42B in the figure), and the machine abnormality detection sequence ou1, ou2, ou3 ( Fig.24 The machine abnormality detection order estimation result D42 obtained by corresponding to the machine abnormality detection order estimation result D42 is stored in the data storage unit 20.

[0412] In addition, the abnormality-causing device estimation unit 350 determines the abnormality propagation order on ( Fig.24 The abnormality propagation order total value suU (shown in D6C) and the device-attached sensor information D41 are calculated for each device by using, for example, an average. Fig.24). For example, in machine 1, the sensors Xn added to machine 1 are sensors X1 and X2, and therefore, the abnormality cause machine estimation unit 350 sets the abnormality propagation order total value su1 corresponding to machine 1 to the average value of the abnormality propagation order o1 and the abnormality propagation order o2. In addition, for example, in machine 2, the sensors Xi added to machine 2 are sensors X3, X4, and X5, and therefore, the abnormality cause machine estimation unit 350 sets the abnormality propagation order total value su2 corresponding to machine 2 to the average value of the abnormality propagation order o3, the abnormality propagation order o4, and the abnormality propagation order o5. Similarly, the abnormality cause machine estimation unit 350 directly sets the abnormality propagation order o6 corresponding to sensor X6 to the abnormality propagation order total value su6 corresponding to machine 3.

[0413] Next, the abnormality cause machine estimation unit 350 assigns the machine abnormality propagation order ouU based on the abnormality propagation order total value suU calculated for each machine. Specifically, the abnormality cause machine estimation unit 350 assigns the corresponding machine abnormality propagation order ou1, ou2, ou3 to machines 1, 2, and 3 respectively based on the calculated abnormality propagation order total values ​​su1, su2, and su3.

[0414] Then, the abnormality cause device estimation unit 350 generates information indicating devices 1, 2, and 3 ( Fig.24 Indicated by D43A), abnormal detection machine flag ( Fig.24 In D43B), the total values ​​of abnormal propagation order su1, su2, su3 ( Fig.24 D43C), and the machine exception propagation order ou1, ou2, ou3 ( Fig.24 The machine abnormality propagation order estimation result D43 obtained by corresponding to the machine abnormality propagation order estimation result D43 is stored in the data storage unit 20.

[0415] Furthermore, the abnormality cause machine estimation unit 350 calculates the machine abnormality cause score suU using the average for each machine based on the machine abnormality detection order ouU set in the machine abnormality detection order estimation result D42 and the machine abnormality propagation order ouU set in the machine abnormality propagation order estimation result D43. For example, the abnormality cause machine estimation unit 350 calculates the average of the machine abnormality detection order ou1 and the machine abnormality propagation order ou1 of machine 1 as the machine abnormality cause score su1 for machine 1.

[0416] Then, the abnormality cause device estimation unit 350 assigns a device abnormality cause order ouU to each device based on the device abnormality cause score suU. For example, the abnormality cause device estimation unit 350 assigns corresponding device abnormality cause orders ou1, ou2, and ou3 to devices 1, 2, and 3, respectively, based on the calculated device abnormality cause scores su1, su2, and su3.

[0417] The abnormality cause device estimation unit 350 generates information indicating devices 1, 2, and 3 ( Fig.24 D44A in the figure), abnormal detection machine flag ( Fig.24 As shown in D44B), machine abnormality cause scores su1, su2, su3 ( Fig.24 D44C in the figure), and the order of machine abnormality causes ou1, ou2, ou3 ( Fig.24 The information obtained by corresponding to the machine abnormality cause sequence estimation result D44) is stored in the data storage unit 20.

[0418] As described above, when the abnormal cause machine estimation unit 350 stores the machine abnormality detection sequence estimation result D42, the machine abnormality propagation sequence estimation result D43, and the machine abnormality cause sequence estimation result D44 in the data storage unit 20, the abnormal cause estimation result output unit 70 obtains the machine abnormality detection sequence estimation result D42, the machine abnormality propagation sequence estimation result D43, and the machine abnormality cause sequence estimation result D44 output by the abnormal cause machine estimation unit 350 via the data storage unit 20, and outputs information related to the estimation result of the cause of the abnormality of the machine unit based on the machine abnormality detection sequence estimation result D42, the machine abnormality propagation sequence estimation result D43, and the machine abnormality cause sequence estimation result D44.

[0419] In detail, the abnormal cause estimation result output unit 70 outputs information (hereinafter referred to as "abnormal cause machine estimation result display information") for displaying the following screen (hereinafter referred to as "abnormal cause machine estimation result screen") to the display device 400 based on the machine abnormality detection sequence estimation result D42, the machine abnormality propagation sequence estimation result D43 and the machine abnormality cause sequence estimation result D44, which shows information related to the estimation result of the cause of the abnormality of each machine estimated by the abnormal cause machine estimation unit 350.

[0420] Fig.25 This is a diagram showing a screen example of the abnormality cause device estimation result screen that the abnormality cause estimation result output unit 70 causes the display device 400 to display in the first embodiment.

[0421] As an example, Fig.25An example of the abnormality-causing device estimation result screen is shown when the number of devices is set to 3 (device U. U=1 to 3).

[0422] exist Fig.25 In the display of the abnormal cause machine estimation result, "D48-1" is displayed.

[0423] For example, Fig.25 As shown, the abnormal cause machine estimation result screen has 10 display frames, namely, display frame D48A, display frame D48B, display frame D48C, display frame D48D, display frame D48E, display frame D48F, display frame D48G, display frame D48H, display frame D48I and display frame D48J.

[0424] The abnormal cause estimation result output unit 70 displays a list of abnormal cause machine estimation results, for example, in the abnormal cause machine estimation result screen, which summarizes the information related to the abnormal cause estimation result by machine. The abnormal cause machine estimation result list is, for example, a list that displays information indicating the machine U, information indicating the abnormal detection machine flag, the machine abnormality detection order, the machine abnormality propagation order, the machine abnormality cause score, and the machine abnormality cause order for each machine U. Fig.25 In the abnormality cause device estimation result screen shown, the abnormality cause device estimation result list is indicated by "D48-1a".

[0425] The abnormal cause estimation result output unit 70 outputs the abnormal cause machine estimation result display information to the display device 400, and the abnormal cause machine estimation result display information displays the machine U information indicating the machine abnormal cause sequence estimation result D44 in the display frame D48A, displays the abnormal detection machine flag information indicating the machine abnormality propagation sequence estimation result D43 in the display frame D48B, displays the machine abnormality detection sequence of the machine abnormality detection sequence estimation result D42 in the display frame D48C, displays the machine abnormality propagation sequence of the machine abnormality propagation sequence estimation result D43 in the display frame D48D, and displays the machine abnormality cause sequence estimation result D44 in the display frame D48C. The machine abnormality cause score is displayed in display box D48E, the machine abnormality cause sequence of the machine abnormality cause sequence estimation result D44 is displayed in display box D48F, the machine abnormality detection sequence based on the machine abnormality detection sequence estimation result D42, the machine abnormality propagation sequence of the machine abnormality propagation sequence estimation result D43, and the machine abnormality cause sequence of the machine abnormality cause sequence estimation result D44, and the sorting buttons for re-sorting the arrangement order of the abnormality cause machine estimation results in ascending order are displayed in display boxes D48H, D48I, and D48J, respectively, and a check box for accepting instructions to display only abnormality detection machines is displayed in display box D48G. As a result, the display device 400 displays Fig.25The abnormality cause machine estimation result screen is shown.

[0426] Fig.25 The abnormal cause machine estimation result screen shown is the one that has been used Figure 8 The error cause estimation result screen described above is set to display the result by device. The sorting buttons and check boxes have the same functions as the ones used. Figure 8 The functions of the sort button and check box described are the same, so duplicate descriptions are omitted.

[0427] In addition, in this case, when using Fig.13 In the operation of the abnormality cause estimation device 100 illustrated in the flowchart, the abnormality cause machine estimation unit 350 performs machine unit abnormality cause estimation processing before or after the processing of step ST5. In the machine unit abnormality cause estimation processing, the cause of the abnormality is estimated on a machine basis based on the machine-attached sensor information D41, the abnormality detection sequence estimation result D5, and the abnormality propagation sequence estimation result D6.

[0428] Furthermore, for example, the abnormality cause estimation result output unit 70 may be able to select whether to output information on the estimation result of the abnormality cause in sensor units as described in the first embodiment or to output information on the estimation result of the abnormality cause in equipment units.

[0429] In this way, the abnormality cause estimation device 100 can be configured to include an abnormality cause machine estimation unit 350, which estimates the abnormality cause for each machine based on the machine-attached sensor information, the abnormality detection order estimated by the abnormality detection order estimation unit 40, and the abnormality propagation order estimated by the abnormality propagation path tracking unit 50. As a result, the abnormality cause estimation device 100 can enable the operator to efficiently identify the machine that is the cause of the abnormality. In addition, the abnormality cause estimation device 100 can enable the operator to efficiently grasp the order of inspection for the machine that should have an abnormality.

[0430] 〈Variation Example〉

[0431] In the above embodiment 1, the abnormality cause estimation device 100 can also be a structure with an associated structure diagram output unit 360, which outputs information for displaying a diagram related to the associated structure D2 stored in the data storage unit 20 to the display device 400 (hereinafter referred to as "information for displaying the associated structure diagram").

[0432] Fig.26 This is a diagram showing a configuration example of the abnormality cause estimation device 100 which includes the correlation structure map output unit 360 and is configured to output the correlation structure map display information to the display device 400 in the first embodiment.

[0433] In addition, Fig.26 Although the illustration is omitted for simplicity of explanation, the abnormality cause estimation device 100 includes a sensor data acquisition unit 10, an abnormality detection unit 30, an abnormality detection sequence estimation unit 40, an abnormality propagation path tracking unit 50, an abnormality cause estimation unit 60, an abnormality cause estimation result output unit 70, and a control unit in addition to the associated structure diagram output unit 360 and the data storage unit 20. However, the abnormality cause estimation device 100 does not necessarily have the abnormality cause estimation result output unit 70. In addition, Fig.26 Although not shown in the figure for the sake of simplicity, the abnormality cause estimation device 100 is connected to the learning device 200 .

[0434] The associated structure graph output unit 360 outputs associated structure graph display information for displaying a graph associated with the associated structure D2 to the display device 400 based on the associated structure D2, the abnormality detection sensor information D3, and the abnormality cause order estimation result D7 stored in the data storage unit 20. The graph associated with the associated structure D2 is, for example, a graph that associates the associated structure D2, the abnormality detection sensor, and the estimation result of the cause of the abnormality.

[0435] The display device 400 displays a screen (hereinafter referred to as a “diagram screen”) in which a diagram related to the related structure D2 is displayed, based on the related structure diagram display information output from the related structure diagram output unit 360 .

[0436] Furthermore, the correlation structure map output unit 360 may be provided in a device such as the display device 400 that is external to the abnormality cause estimation device 100 and is connected to the abnormality cause estimation device 100 via a wired or wireless signal line.

[0437] When the abnormality cause estimation device 100 is configured to include the correlation structure map output unit 360 , an example of a diagram screen displayed on the display device 400 by the correlation structure map output unit 360 outputting the correlation structure map display information will be described with reference to the drawings.

[0438] Fig. 27 This is a diagram for explaining an example of a diagram screen displayed on the display device 400 by the correlation structure diagram output unit 360 outputting the correlation structure diagram display information when the abnormality cause estimation device 100 is configured to include the correlation structure diagram output unit 360 in the first embodiment.

[0439] Here, as an example, the number of sensors 300 is set to 6, and the sensors 300 are represented by sensors Xn (n=1 to 6).

[0440] also, Fig. 27The example of the screen shown in the figure is that the content of the association structure D2 stored in the data storage unit 20 is Fig.28 The content shown, the content of the abnormality detection sensor information D3 is Fig.29 The content shown, abnormal cause sequence estimation result D7 is Fig.30 In the case of the content shown, this is an example of a diagram screen displayed based on the relationship structure diagram display information output by the relationship structure diagram output unit 360.

[0441] In addition, when outputting information for displaying the association structure diagram, the association structure diagram output unit 360 will determine whether each element of the association structure D2 is an element with a dependency relationship or an element without a dependency relationship according to each statistical indicator, and convert the association structure D2 into an association structure that uses information indicating the presence or absence of a dependency relationship as an element (hereinafter referred to as a "association structure after dependency determination").

[0442] For example, the association structure graph output unit 360 determines, for each statistical index, each element of the association structure D2 as an element having a dependency relationship when it is greater than a preset threshold value for dependency selection, and determines it as an element having no dependency relationship when it is less than the threshold value for dependency selection. Then, the association structure graph output unit 360 generates a post-dependency determination association structure represented by a matrix in which elements having a dependency relationship are set to "1" and elements having no dependency relationship are set to "0", for example.

[0443] exist Fig.28 In the figure, the dependency determination degree association structure of each statistical indicator is shown in conjunction with the association structure D2 ( Fig.28 ).

[0444] exist Fig. 27 In the diagram screen, for example, a relationship structure graph ( Fig. 27 ).

[0445] In the diagram screen, in addition to the associated structure diagram, a screen (hereinafter referred to as an "indicator specification screen") displaying a check box (hereinafter referred to as an "indicator specification check box") is displayed. Fig. 27 ), this check box is used to accept the designation of the type of statistical index for the object of displaying the corresponding element in the association structure diagram. Here, there are three types of statistical indexes, so Fig. 27In the index designation screen shown, an index designation check box corresponding to "statistical index 1" for accepting the designation of the first statistical index, an index designation check box corresponding to "statistical index 2" for accepting the designation of the second statistical index, and an index designation check box corresponding to "statistical index 3" for accepting the designation of the third statistical index are displayed. For example, the operator specifies the statistical index as the object of displaying the correlation (edge) between the corresponding sensors Xn by checking the index designation check box. Fig. 27 The indicator specification check box corresponding to "Statistical indicator 1" and the indicator specification check box corresponding to "Statistical indicator 2" are checked in the indicator specification screen, in other words, the first statistical indicator and the second statistical indicator are specified. Therefore, only the edges corresponding to the statistical indicators checked in the indicator specification check boxes, i.e., the first statistical indicator and the second statistical indicator, are displayed on the association structure diagram.

[0446] For example, Fig. 27 As shown in FIG. 1 , the edges corresponding to each statistical index are displayed with different line types so that the edges corresponding to the statistical index can be known. Fig. 27 In the illustrated association structure diagram, the edge corresponding to the first statistical indicator is displayed with a solid arrow (see, for example, D45G), and the edge corresponding to the second statistical indicator is displayed with a dotted arrow (see, for example, D45H). This is just an example, and for example, the edges corresponding to the statistical indicators may be displayed with different arrow colors.

[0447] In addition, on the graph screen, a screen (hereinafter referred to as the "node condition specification screen") displaying check boxes (hereinafter referred to as the "node condition specification check boxes") for specifying display conditions related to nodes (hereinafter referred to as the "node display conditions") is displayed. Fig. 27 The node display conditions are preset. Fig. 27 In the node condition specification screen shown, three conditions are set as node display conditions: "Display only abnormality detection sensors", "Emphasis on abnormality detection sensors", and "Display abnormality causes in order". For example, the operator specifies the node display condition by checking the node condition specification check box. Fig. 27 The state in which the node condition specifying check box corresponding to “emphasize abnormality detection sensor” and the node condition specifying check box corresponding to “display abnormality cause order” are checked in the node condition specifying screen is shown.

[0448] Therefore, in the association structure graph, the nodes ( Fig. 27In addition, here, the abnormal detection sensor is emphasized by filling the node corresponding to the abnormal detection sensor, but the method of emphasizing the abnormal detection sensor is not limited to this, and the node corresponding to the abnormal detection sensor can also be emphasized by other methods.

[0449] In addition, in the association structure diagram, the abnormality causes are sequentially displayed on the nodes corresponding to the sensor Xn. Fig. 27 In the association structure diagram shown, the order of abnormal causes is displayed as "rank 1", "rank 2", "rank 3", "rank 4" or "rank 5". In addition, here, the order of abnormal causes is displayed as "rank △", but the display method of the order of abnormal causes is not limited to this, as long as it is displayed in a way that the order of abnormal causes is known.

[0450] In addition, here, three conditions are set in the node display conditions: "only display abnormality detection sensors", "emphasize abnormality detection sensors", and "display abnormality causes in order", but this is just an example, and other conditions can also be set in the node display conditions.

[0451] In the association structure diagram, for example, a sensor name for identifying the sensor Xn is displayed at each node for the operator who has checked the diagram screen. Fig. 27 In the display, sensor names "X1", "X2", "X3", "X4", "X5", and "X6" are displayed.

[0452] For example, an edge is displayed when the statistic of an element of the association structure D2 is greater than a threshold value for dependency selection set according to each statistical index. The statistic being greater than the threshold value means that there is a dependency relationship between the sensors Xn. In addition, the association structure diagram output unit 360 can determine the dependency relationship between the sensors Xn based on the dependency determination post-association structure.

[0453] When the dependency between sensors Xn is in one direction, the directed graph shows the edge of a unidirectional arrow, and when the dependency between sensors Xn is in two directions, the directed graph shows the edge of a bidirectional arrow. For example, here, there is a dependency from sensor X2 to sensor X1 in one direction, so Fig. 27 As shown, in the directed graph, the node representing sensor X2 ( Fig. 27 ) to the node representing sensor X1 ( Fig. 27 The side of the one-way arrow (shown as D45A in Fig. 27 In addition, for example, here, sensor X2 and sensor X6 have a dependency relationship in both directions, so, as shown in FIG. Fig. 27As shown, in the directed graph, between the node representing sensor X2 and the node representing sensor X6 ( Fig. 27 The edge with a double-headed arrow (as shown in D45F in Fig. 27 ).

[0454] Fig.31 This is a flowchart for explaining an example of the operation of the abnormality cause estimation device 100 in the case where the abnormality cause estimation device 100 is configured to include the correlation structure map output unit 360 in the first embodiment.

[0455] The abnormality cause estimation device 100 is used in addition to Fig.13 In addition to the actions described in the flowchart, Fig.31 In addition, assume that Fig.31 The actions shown in the flowchart are performed Fig.13 The processing of steps ST1 to ST5 is then performed at least once. Fig.31 The actions shown in the flowchart can also be performed in Fig.13 The process is performed after the process of step ST5, before the process of step ST6, after the process of step ST6, or in parallel with the process of step ST6.

[0456] The relationship structure diagram output unit 360 receives an instruction to display the relationship structure diagram (step ST31 ).

[0457] For example, the operator operates an input device such as a mouse or a keyboard to call up an input screen for displaying the associated structure diagram on the display device 400. The operator inputs a display instruction for the associated structure diagram from the input screen for displaying the associated structure diagram. The associated structure diagram output unit 360 receives the display instruction for the associated structure diagram input by the operator.

[0458] The associated structure diagram output unit 360 outputs associated structure diagram display information for displaying a diagram associated with the associated structure D2 to the display device 400 based on the associated structure D2, the abnormality detection sensor information D3, and the abnormality cause sequence estimation result D7 stored in the data storage unit 20 (step ST32). Fig. 27 The picture shown.

[0459] As described above, since the abnormality cause estimation device 100 is configured to include the correlation structure map output unit 360 , the abnormality cause estimation device 100 can improve the descriptiveness of information related to the estimation result of the cause of the abnormality.

[0460] 〈Variation Example〉

[0461] In the above embodiment 1, the learning device 200 may also be configured to include a learning sensor pair generation unit 370, which generates pairs of sensors 300 from a plurality of sensors 300 based on information defining the connection relationship of a plurality of machines constituting the target device and a plurality of sensors 300 provided in the plurality of machines (hereinafter referred to as "device design information"). In this case, the association structure learning unit 240 acquires learning sensor data according to each pair of sensors 300 generated by the learning sensor pair generation unit 370, and learns the association structure D2. In addition, for example, after the operator or the like generates the device design information based on the design drawing, the operator operates an input device such as a mouse or a keyboard to input the generated device design information, and the learning device 200 acquires the device design information by accepting the input device design information.

[0462] Fig.32 This is a diagram showing a configuration example of a learning device 200 including a learning sensor pair generation unit 370 in the first embodiment.

[0463] In addition, Fig.32 Although the illustration is omitted for simplicity of explanation, the learning device 200 includes a learning sensor pair generation unit 370, a correlation structure learning unit 240, and a learning data storage unit 220, as well as a learning sensor data acquisition unit 210 and a learning preprocessing unit 230. Fig.32 Although not shown in the figure for the sake of simplicity, the learning device 200 is connected to the abnormality cause estimation device 100 .

[0464] The learning sensor pair generation unit 370 acquires the device design information, and based on the acquired device design information, generates a pair of sensors 300 from among the plurality of sensors 300 to be used when the association structure learning unit 240 learns the association structure D2.

[0465] Specifically, the learning sensor pair generator 370 determines a combination of two sensors 300 based on the device design information, and generates information (hereinafter referred to as “sensor pair information”) D47 in which the combination is listed.

[0466] The learning sensor pair generation unit 370 outputs the generated sensor pair information D47 to the correlation structure learning unit 240 .

[0467] When the sensor pair information D47 is output from the learning sensor pair generation unit 370, the association structure learning unit 240 calculates a statistic representing the relationship between two different sensor data for a plurality of sensor data included in the learning data D23 based on the pair of sensors 300 set in the sensor pair information D47, and learns the association structure D2 based on the calculated statistic.

[0468] The process of generating the sensor pair information D47 by the learning sensor pair generating unit 370 will be described with reference to specific examples using the drawings.

[0469] Fig.33 This is a diagram for explaining the concept of an example of a method in which the learning sensor pair generating unit 370 generates the sensor pair information D47 based on the device design information D46 when the learning device 200 is configured to include the learning sensor pair generating unit 370 in the first embodiment.

[0470] Here, as an example, the number of sensors 300 is set to 8, and the sensors 300 are represented by sensors Xn (n=1 to 8).

[0471] Here, as an example, the target device is composed of five devices (device D46A, device D46B, device D46C, device D46D, and device D46E).

[0472] Furthermore, the device D46A is provided with a sensor X1 and a sensor X2 , the device D46B is provided with a sensor X3 , the device D46C is provided with a sensor X4 and a sensor X5 , the device D46D is provided with a sensor X6 and a sensor X7 , and the device D46E is provided with a sensor X8 .

[0473] Furthermore, machine D46A is in a connection relationship with machine D46B, machine D46B is in a connection relationship with machine D46A, machine D46C, and machine D46D, machine D46C is in a connection relationship with machine D46B and machine D46E, machine D46D is in a connection relationship with machine D46B, and machine D46E is in a connection relationship with machine D46C. In addition, in terms of design, machines in a connection relationship are connected by an undirected line. In this case, the device design information D46 becomes Fig.33 In addition, Fig.33 In the example, the device design information D46 is shown in a block diagram, but this is only an example. The device design information D46 may be information that allows knowing the connection relationship between the plurality of devices constituting the target device and the plurality of sensors Xn installed in the plurality of devices.

[0474] The learning sensor pair generator 370 generates a pair of two different sensors Xn based on the device design information D46. Specifically, the learning sensor pair generator 370 generates a pair of a sensor Xn attached to a certain device and a sensor Xn attached to a device connected to the device.

[0475] For example, in Fig.33In the example shown, according to the device design information D46, the device D46A and the device D46B are in a connection relationship. In this case, the sensor X1 and the sensor X2 provided in the device D46A are also in a connection relationship with the sensor X3 provided in the device D46B. Then, the learning sensor pair generation unit 370 generates a pair of the sensor X1 and the sensor X3, and a pair of the sensor X2 and the sensor X3.

[0476] Furthermore, when two or more sensors Xn are provided in one device, the learning sensor pair generation unit 370 generates a pair of different sensors Xn among the two or more sensors Xn provided in the same device.

[0477] For example, in Fig.33 In the example shown, the sensor X1 and the sensor X2 are provided in the device D46A according to the equipment design information D46. Then, the learning sensor pair generation unit 370 generates a pair of the sensor X1 and the sensor X2.

[0478] In this way, the learning sensor pair generation unit 370 generates, as sensor pairs, for example, a pair consisting of two sensors installed in different machines and connected to each other, and a pair consisting of two different sensors installed in one machine.

[0479] In addition, the sensor pair generated by the learning sensor pair generating unit 370 as described above is only an example. The learning sensor pair generating unit 370 may, for example, only generate a pair consisting of two sensors that are installed in different machines and are connected to each other, or only generate a pair consisting of two different sensors installed in one machine.

[0480] As described above, when the learning device 200 is configured to include the learning sensor pair generation unit 370, Fig.14 In the operation of the learning device 200 described in the flowchart of FIG23 , before the processing of step ST231 is performed, the learning sensor pair generation unit 370 generates sensor pair information D47 and outputs it to the association structure learning unit 240. In step ST231, the association structure learning unit 240 obtains two different sensor data pairs based on the learning data D23 and the sensor pair information D47. In addition, the association structure learning unit 240 sets all combinations of multiple sensor data included in the learning data D23 as pairs of sensor data based on the sensor pair information D47.

[0481] In this way, the learning device 200 includes a learning sensor pair generation unit 370, which generates pairs of sensors 300 from a plurality of sensors 300 based on the device design information, and the association structure learning unit 240 is configured to acquire learning sensor data according to each pair of sensors 300 generated by the learning sensor pair generation unit 370, and learn the association structure D2. Thus, the learning device 200 can suppress the possibility of detecting dependencies between sensors 300 with low correlation in design, and can learn the association structure D2 with improved reliability. As a result, the learning device 200 can provide the abnormality cause estimation device 100 with the association structure D2 that can estimate the abnormality cause in detail.

[0482] As described above, the abnormality cause estimation device 100 of embodiment 1 is configured to include: a sensor data acquisition unit 10, which acquires a plurality of time series of sensor data collected by a plurality of sensors 300 provided in a plurality of device structural elements constituting the target device; an abnormality detection unit 30, which detects a plurality of abnormality detection sensors in which abnormalities occur among the plurality of sensors 300 based on the plurality of sensor data acquired by the sensor data acquisition unit 10; an abnormality detection order estimation unit 40, which estimates the abnormality detection order in which abnormalities are detected for the plurality of abnormality detection sensors based on the detection times at which the abnormality detection unit 30 detects the plurality of abnormality detection sensors; an abnormality propagation path tracking unit 50, which estimates the abnormality propagation order of abnormality propagation based on abnormality detection sensor information D3 related to the plurality of abnormality detection sensors detected by the abnormality detection unit 30 and an estimated structure (association structure D2) showing the dependency relationship between the device structural elements; and an abnormality cause estimation unit 60, which estimates the cause of the abnormality based on the abnormality detection order estimated by the abnormality detection order estimation unit 40 and the abnormality propagation order estimated by the abnormality propagation path tracking unit 50.

[0483] Therefore, the abnormality cause estimation device 100 can estimate the cause of abnormality occurring in the equipment regardless of the complexity of the equipment or the scale of the equipment.

[0484] Furthermore, the abnormality cause estimation device 100 is configured to include an abnormality cause estimation result output unit 70 that outputs information on an estimation result of the abnormality cause estimated by the abnormality cause estimation unit 60 .

[0485] Therefore, the abnormality cause estimation device 100 improves the interpretability and explanation of the abnormality cause estimation result for the operator. The abnormality cause estimation device 100 can reduce the operator's unnecessary inspection work and reduce the operator's load. In addition, the abnormality cause estimation device 100 can estimate the cause of the abnormality with a quantitative index that does not rely on human subjectivity and provide the basis for the estimation. The operator can determine the inspection order of the equipment with less labor.

[0486] Furthermore, the abnormality cause estimation device 100 can be configured to detect the abnormality detection sensor using a univariate abnormality detection method.

[0487] Therefore, the abnormality cause estimation apparatus 100 can more appropriately detect an abnormality caused by a single change in one sensor data D1.

[0488] Furthermore, the abnormality cause estimation device 100 can be configured to detect the abnormality detection sensor using a multivariate abnormality detection method.

[0489] Therefore, the abnormality cause estimation apparatus 100 can more appropriately detect an abnormality in a change in the relationship between the plurality of sensor data D1.

[0490] Furthermore, the abnormality cause estimation device 100 can be configured to detect the abnormality detection sensor using a univariate abnormality detection method or a multivariate abnormality detection method.

[0491] Therefore, the abnormality cause estimation apparatus 100 can more appropriately detect an abnormality in which a single sensor data D1 changes independently or an abnormality in which the relationship between a plurality of sensor data D1 changes.

[0492] In addition, in the abnormality cause estimation device 100, the abnormality propagation path tracking unit 50 can be configured to estimate the abnormality propagation order based on the abnormality detection sensor information D3, the equipment operation status information D31 indicating the operation status of the target equipment, and the estimation structure (association structure D32) showing the dependency relationship between equipment structural elements according to the operation status of the target equipment.

[0493] Therefore, the abnormality cause estimation device 100 can cope with the change in the dependency relationship between the sensors 300 accompanying the change in the operating state of the target equipment, and can estimate the abnormality cause in detail based on the correlation structure D32 with improved reliability.

[0494] In addition, the abnormality cause estimation device 100 can have an associated structure correction unit 330, which corrects the dependency relationship between sensor data in the estimation structure (associated structure D2) based on dependent pair information D33 related to pairs of sensors with a dependency relationship among multiple sensors 300, and non-dependent pair information D34 related to pairs of sensors 300 without a dependency relationship among multiple sensors 300.

[0495] Therefore, the abnormality cause estimation device 100 can improve the reliability of the estimation structure and can estimate the abnormality cause in detail.

[0496] In addition, the abnormality cause estimation device 100 can include a relationship change estimation unit 340, which compares the estimated structure with the estimated structure when the abnormality occurs based on the estimated structure (association structure D2), the estimated structure when the abnormality occurs (association structure D36) and the abnormality detection sensor information D3, and estimates the change in the relationship between the sensor data. The abnormality cause estimation unit 60 is constructed to estimate the cause of the abnormality based on the abnormality detection order estimated by the abnormality detection order estimation unit 40 and the abnormality propagation order estimated by the abnormality propagation path tracking unit 50, taking into account the change in the relationship between the sensor data estimated by the relationship change estimation unit 340.

[0497] Therefore, in the abnormality cause estimation device 100, the reliability of the abnormality cause order estimation result D7 is improved, and the abnormality cause can be estimated in detail.

[0498] In addition, the abnormality cause estimation device 100 can be configured to include an abnormality cause machine estimation unit 350, which estimates the cause of the abnormality on a machine basis based on the machine-attached sensor information D41 obtained by corresponding the machine installed in the target device with the sensor 300 installed in the machine, the abnormality detection sequence estimated by the abnormality detection sequence estimation unit 40, and the abnormality propagation sequence estimated by the abnormality propagation path tracking unit 50.

[0499] Therefore, the abnormality cause estimation device 100 can enable the operator to efficiently identify the device that is the cause of the abnormality. In addition, the abnormality cause estimation device 100 can enable the operator to efficiently understand the order in which the abnormal device should be inspected.

[0500] In addition, the abnormal cause estimation device 100 can be configured to have an associated structure diagram output unit 360, which outputs associated structure diagram display information based on the estimated structure (associated structure D2), the abnormality detection sensor information D3, and information related to the estimated result of the abnormal cause estimated by the abnormal cause estimation unit 60. The associated structure diagram display information is used to display a diagram that associates the estimated structure, the abnormality detection sensor, and the estimated result of the abnormal cause.

[0501] Therefore, the abnormality cause estimation device 100 can improve the descriptiveness of information related to the estimation result of the abnormality cause.

[0502] In addition, as described above, the learning device 200 of embodiment 1 is configured to include: a learning sensor data acquisition unit 210, which acquires multiple time series sensor data collected by multiple sensors 300 installed in the target device when the target device is operating normally as learning data candidates; and an associated structure learning unit 240, which uses the multiple learning data candidates acquired by the learning sensor data acquisition unit 210 as multiple learning data, calculates at least one statistic between the multiple learning data based on the learning data, and learns an estimated structure (associated structure D2) showing the dependency relationship between device structural elements based on the calculated statistic.

[0503] Therefore, the learning device 200 can extract the correlation between the sensor data D1 in an all-inclusive manner, and as a result, can provide the correlation structure D2 that suppresses the omission of the connection relationship of the sensor 300. The learning device 200 provides the abnormality cause estimation device 100 with the estimation structure (correlation structure D2) when tracking the sensor 300 that is the source of the abnormality, so that the abnormality cause estimation device 100 can more appropriately track the sensor 300 that is the source of the abnormality, and can improve the estimation accuracy of the abnormality cause.

[0504] In addition, the learning device 200 includes a learning preprocessing unit 230, which obtains multiple learning data for learning based on multiple learning data candidates obtained by the learning sensor data acquisition unit 210. The association structure learning unit 240 can be configured to calculate at least one statistic between the multiple learning data based on the learning data obtained by the learning preprocessing unit 230, and learn the estimated structure (association structure D2) based on the calculated statistic.

[0505] Therefore, the learning device 200 can extract the correlation between the sensor data D1 in an all-inclusive manner, and as a result, can provide the correlation structure D2 that suppresses the omission of the connection relationship of the sensor 300. The learning device 200 provides the abnormality cause estimation device 100 with the estimation structure (correlation structure D2) when tracking the sensor 300 that is the source of the abnormality, so that the abnormality cause estimation device 100 can more appropriately track the sensor 300 that is the source of the abnormality, and can improve the estimation accuracy of the abnormality cause.

[0506] In addition, in the learning device 200, the learning preprocessing unit 230 can select multiple learning data candidates whose variance values ​​are smaller than a selected threshold from the multiple learning data candidates acquired by the learning sensor data acquisition unit 210, and acquire the selected multiple learning data candidates as multiple learning data.

[0507] Therefore, the learning device 200 extracts the correlation between the sensor data D1 in an all-inclusive manner, and as a result, can provide the correlation structure D2 that suppresses the omission of the connection relationship of the sensor 300. The learning device 200 provides the abnormality cause estimation device 100 with the estimation structure (correlation structure) when tracking the sensor 300 that is the source of the abnormality, so that the abnormality cause estimation device 100 can more appropriately track the sensor 300 that is the source of the abnormality, and can improve the estimation accuracy of the abnormality cause.

[0508] Furthermore, in the learning device 200 , the correlation structure learning unit 240 may be configured to calculate statistics using statistical indicators based on waveforms.

[0509] Therefore, the learning device 200 can provide an estimation structure (correlation structure D2) that can track the propagation of abnormality based on such a dependency relationship on the waveform and can more appropriately estimate the cause of the abnormality.

[0510] Furthermore, the learning device 200 can be configured to calculate statistics using a statistical index based on distribution.

[0511] Therefore, the learning device 200 can provide an estimation structure (correlation structure D2) that can track the propagation of anomalies based on such a dependency relationship in distribution and can more appropriately estimate the cause of the anomaly.

[0512] Furthermore, the learning device 200 can be configured to calculate statistics using a statistical index based on a waveform and a statistical index based on a distribution.

[0513] Therefore, the learning device 200 can provide an estimation structure (correlation structure D2) that can track the propagation of abnormality based on such a dependency relationship in waveform or distribution and can more appropriately estimate the cause of abnormality.

[0514] In addition, the learning device 200 can be configured to include a learning sensor pair generating unit 370, which generates pairs of sensors 300 from multiple sensors 300 based on device design information D46 that defines the connection relationship between multiple machines constituting the target device and multiple sensors 300 installed in the multiple machines. The associated structure learning unit 240 obtains learning data based on the pairs of sensors 300 generated by the learning sensor pair generating unit 370, and learns the estimated structure (associated structure D2).

[0515] Therefore, the learning device 200 is designed to suppress the possibility of detecting dependencies between sensors 300 with low correlation, and learn the correlation structure D2 with improved reliability. As a result, the learning device 200 can provide the abnormality cause estimation device 100 with the correlation structure D2 that can estimate the abnormality cause in detail.

[0516] In addition, the present disclosure can modify any constituent elements of the embodiments or omit any constituent elements of the embodiments.

[0517] Industrial Applicability

[0518] The abnormality cause estimation device and the abnormality cause estimation apparatus of the present disclosure can estimate the cause of an abnormality occurring in a device regardless of the complexity of the device or the scale of the device.

[0519] Description of Reference Numerals

[0520] 1000 precision diagnosis system, 100 abnormality cause estimation device, 10, 310 sensor data acquisition unit, 20, 320 data storage unit, 30 abnormality detection unit, 40 abnormality detection order estimation unit, 50 abnormality propagation path tracking unit, 60 abnormality cause estimation unit, 70 abnormality cause estimation result output unit, 330 associated structure correction unit, 340 relationship change estimation unit, 350 abnormality cause machine estimation unit, 360 associated structure map output unit, 200 learning device, 210 sensor data acquisition unit for learning, 220 data storage unit for learning, 230 preprocessing unit for learning, 240 associated structure learning unit, 370 sensor pair generation unit for learning, 300 sensor, 400 display device, 1601 processing circuit, 1602 input interface device, 1603 output interface device, 1604 processor, 1605 memory.

Claims

1. An abnormality cause estimation device, wherein: The abnormality cause estimation device comprises: a sensor data acquisition unit that acquires a plurality of time-series sensor data collected by a plurality of sensors provided in a plurality of device components constituting the target device; an abnormality detection unit that detects a plurality of abnormality detection sensors in which abnormalities occur among the plurality of sensors based on the plurality of sensor data acquired by the sensor data acquisition unit; an abnormality detection order estimating unit that estimates an abnormality detection order in which the abnormality is detected as occurring for the plurality of abnormality detection sensors based on the detection times of the plurality of abnormality detection sensors detected by the abnormality detecting unit; an abnormality propagation path tracking unit that estimates an abnormality propagation order of the abnormality propagation based on abnormality detection sensor information related to the plurality of abnormality detection sensors detected by the abnormality detection unit and an estimated structure showing a dependency relationship between the device structural elements; as well as An abnormality cause estimation unit estimates a cause of the abnormality based on the abnormality detection order estimated by the abnormality detection order estimation unit and the abnormality propagation order estimated by the abnormality propagation path tracking unit.

2. The abnormality cause estimation device according to claim 1, characterized in that: The estimation structure is represented by a matrix.

3. The abnormality cause estimation device according to claim 1 or 2, characterized in that: The abnormality cause estimation device includes an abnormality cause estimation result output unit that outputs information on an estimation result of the cause of the abnormality estimated by the abnormality cause estimation unit.

4. The abnormality cause estimation device according to any one of claims 1 to 3, characterized in that: The abnormality detection unit detects the abnormality detection sensor using a univariate abnormality detection method.

5. The abnormality cause estimation device according to any one of claims 1 to 3, characterized in that: The abnormality detection unit detects the abnormality detection sensor using a multivariate abnormality detection method.

6. The abnormality cause estimation device according to any one of claims 1 to 3, characterized in that: The abnormality detection unit detects the abnormality detection sensor using a univariate abnormality detection method and a multivariate abnormality detection method.

7. The abnormality cause estimation device according to any one of claims 1 to 6, characterized in that: The abnormality propagation path tracking unit estimates the abnormality propagation order based on the abnormality detection sensor information, the device operating state information indicating the operating state of the target device, and the estimated structure indicating the dependency relationship between the device components according to the operating state of the target device.

8. The abnormality cause estimation device according to any one of claims 1 to 7, characterized in that: The abnormal cause estimation device includes an associated structure correction unit, which corrects the dependency relationship between the sensor data for the estimation structure based on dependent pair information related to the pairs of sensors having a dependency relationship among the multiple sensors and non-dependent pair information related to the pairs of sensors having no dependency relationship among the multiple sensors.

9. The abnormality cause estimation device according to any one of claims 1 to 8, characterized in that: The abnormality cause estimation device includes a relationship change estimation unit, which estimates a change in the relationship between sensor data by comparing the estimated structure with the estimated structure when the abnormality occurs based on the estimated structure, the estimated structure when the abnormality occurs, and the abnormality detection sensor information. The abnormality cause estimation unit estimates the cause of the abnormality based on the abnormality detection order estimated by the abnormality detection order estimation unit and the abnormality propagation order estimated by the abnormality propagation path tracking unit, taking into account the change in the relationship between the sensor data estimated by the relationship change estimation unit.

10. The abnormality cause estimation device according to any one of claims 1 to 9, characterized in that: The abnormality cause estimation device includes an abnormality cause machine estimation unit, which estimates the cause of the abnormality in units of the machine based on machine-attached sensor information obtained by corresponding the machine installed in the target device with the sensor installed in the machine, the abnormality detection sequence estimated by the abnormality detection sequence estimation unit, and the abnormality propagation sequence estimated by the abnormality propagation path tracking unit.

11. The abnormality cause estimation device according to any one of claims 1 to 10, characterized in that: The abnormality cause estimation device includes an associated structure diagram output unit, which outputs associated structure diagram display information based on the estimated structure, the abnormality detection sensor information, and information related to the estimated result of the cause of the abnormality estimated by the abnormality cause estimation unit. The associated structure diagram display information is used to display a diagram that associates the estimated structure, the abnormality detection sensor and the estimated result of the cause of the abnormality.

12. A learning device, wherein: The learning device comprises: a learning sensor data acquisition unit that acquires, as learning data candidates, a plurality of time-series sensor data collected by a plurality of sensors installed in the target device during normal operation of the target device; as well as An associated structure learning unit, which takes the multiple learning data candidates obtained by the learning sensor data acquisition unit as multiple learning data, calculates at least one statistic between the multiple learning data based on the multiple learning data, and learns an estimated structure showing the dependency relationship between the device structural elements based on the calculated statistic.

13. The learning device according to claim 12, characterized in that: The learning device includes a learning preprocessing unit that acquires a plurality of the learning data for learning based on the plurality of the learning data candidates acquired by the learning sensor data acquiring unit. The relational structure learning unit calculates at least one of the statistics between a plurality of the learning data based on the learning data acquired by the learning preprocessing unit, and learns the estimated structure based on the calculated statistics.

14. The learning device according to claim 13, characterized in that: The learning preprocessing unit selects a plurality of the learning data candidates having variance values ​​smaller than a selection threshold value from among the plurality of the learning data candidates acquired by the learning sensor data acquiring unit, and acquires the selected plurality of the learning data candidates as a plurality of the learning data.

15. The learning device according to any one of claims 12 to 14, characterized in that: The association structure learning unit calculates the statistic using a waveform-based statistical index.

16. The learning device according to any one of claims 12 to 14, characterized in that: The association structure learning unit calculates the statistic using a statistical index based on distribution.

17. The learning device according to any one of claims 12 to 14, characterized in that: The association structure learning unit calculates the statistic using a waveform-based statistical index and a distribution-based statistical index.

18. The learning device according to any one of claims 12 to 17, characterized in that: The learning device includes a learning sensor pair generating unit, which generates the sensor pair from the plurality of sensors based on device design information, wherein the device design information defines the connection relationship between the plurality of machines constituting the target device and the plurality of sensors provided in the plurality of machines. The associated structure learning unit acquires the learning data based on the sensor pair generated by the learning sensor pair generating unit, and learns the estimated structure.

19. A precision diagnostic system, wherein: The precision diagnosis system comprises: The abnormality cause estimation device according to any one of claims 1 to 11; and A learning device as claimed in any one of claims 12 to 18.

20. A method for estimating abnormal causes, wherein: The abnormality cause estimation method comprises the following steps: The sensor data acquisition unit acquires a plurality of time-series sensor data collected by a plurality of sensors provided in a plurality of device components constituting the target device; The abnormality detection unit detects a plurality of abnormality detection sensors in which abnormality occurs among the plurality of sensors based on the plurality of sensor data acquired by the sensor data acquisition unit; The abnormality detection order estimation unit estimates the abnormality detection order in which the abnormality is detected as occurring for the plurality of abnormality detection sensors based on the detection times of the plurality of abnormality detection sensors detected by the abnormality detection unit; an abnormality propagation path tracking unit estimates an abnormality propagation order of the abnormality propagation based on abnormality detection sensor information related to the plurality of abnormality detection sensors detected by the abnormality detection unit and an estimated structure showing a dependency relationship between the device structural elements; as well as The abnormality cause estimation unit estimates the cause of the abnormality based on the abnormality detection order estimated by the abnormality detection order estimation unit and the abnormality propagation order estimated by the abnormality propagation path tracking unit.

Images