Abnormality diagnosing apparatus, abnormality diagnosing method, and program

Abstract

An abnormality diagnosing apparatus includes: an abnormality determining section that determines presence or absence of abnormality for each of state quantities acquired from a device; and a cause inferring section that uses a cause correspondence table in which causes of abnormality modes of the device determined by fault tree analysis are corresponded to state quantities that become abnormal when the causes occur, and infers a cause of abnormality of the device from the state quantity determined by the abnormality determining section as having abnormality.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to an abnormality diagnosing apparatus, an abnormality diagnosing method, and a program.

[0002] This application claims priority based on Japanese Patent Application No. 2020-130788 filed on July 31, 2020, the contents of which are incorporated herein. BACKGROUND

[0003] It is known that in large refrigerators, equipment such as submersible pumps, complete equipment, and the like, an observation value is detected with an observation system, and an abnormality diagnosis that determines a cause of an abnormality is performed, and the soundness is monitored.

[0004] For example, in Patent Literature 1, a state estimation model of the complete equipment is created from real data observed in the observation system, and a part including an observation value in which a deviation based on a prediction of the state estimation model is added is regarded as an abnormality to determine a site of a behavior in which an abnormality occurs, that is, an abnormality cause.

[0005] PRIOR ART DOCUMENTS

[0006] PATENT LITERATURE

[0007] Patent Literature 1: JP Patent No. 6246755 SUMMARY

[0008] PROBLEMS TO BE SOLVED BY THE INVENTION

[0009] However, the technology described in Patent Literature 1 does not systematically show a method of corresponding an abnormality cause to an observation value in which an abnormality occurs. Therefore, in the technology described in Patent Literature 1, even if an observation value in which an abnormality occurs can be determined, in order to correctly diagnose an abnormality cause, a large number of independent observation systems corresponding to the abnormality cause must be prepared. For this reason, a technology that performs an abnormality diagnosis with good accuracy using an existing observation system is sought.

[0010] The present disclosure is made in view of such a problem, and provides an abnormality diagnosing apparatus, an abnormality diagnosing method, and a program that can easily and accurately estimate a cause of an abnormality of equipment.

[0011] MEANS FOR SOLVING THE PROBLEM

[0012] According to one embodiment of the present disclosure, an abnormality diagnosing apparatus includes: an abnormality determination section that respectively determines presence or absence of an abnormality for a state quantity acquired from equipment; and a cause estimation section that uses a cause correspondence table that establishes a correspondence between a cause of an abnormal mode of the equipment determined by a fault tree analysis and the state quantity that becomes abnormal if the cause is generated, and estimates a cause of an abnormality of the equipment from a state quantity determined by the abnormality determination section as having an abnormality.

[0013] According to one embodiment of the present disclosure, an abnormality diagnosis method includes: determining whether or not each of state quantities acquired from a device is abnormal; and inferring a cause of an abnormality of the device from the state quantity determined to be abnormal, using a cause correspondence table that is established using causes of abnormality modes of the device determined by fault tree analysis and the state quantities that become abnormal when the causes are generated.

[0014] According to one embodiment of the present disclosure, a program causes a computer of an abnormality diagnosis apparatus to execute: determining whether or not each of state quantities acquired from a device is abnormal; and inferring a cause of an abnormality of the device from the state quantity determined to be abnormal, using a cause correspondence table that is established using causes of abnormality modes of the device determined by fault tree analysis and the state quantities that become abnormal when the causes are generated.

[0015] Effects of Invention

[0016] According to the abnormality diagnosis apparatus, the abnormality diagnosis method, and the program according to the present disclosure, a cause of an abnormality of a device can be inferred easily and with good precision. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 is a diagram showing the overall structure of an apparatus according to a first embodiment of the present disclosure.

[0018] Figure 2 is a first diagram for explaining a cause correspondence table according to the first embodiment of the present disclosure.

[0019] Figure 3 is a second diagram for explaining the cause correspondence table according to the first embodiment of the present disclosure.

[0020] Figure 4 is a third diagram for explaining the cause correspondence table according to the first embodiment of the present disclosure.

[0021] Figure 5 is a flowchart showing an example of a process of an abnormality diagnosis apparatus according to the first embodiment of the present disclosure.

[0022] Figure 6 is a first diagram for explaining a cause correspondence table according to a second embodiment of the present disclosure.

[0023] Figure 7 is a second diagram for explaining the cause correspondence table according to the second embodiment of the present disclosure.

[0024] Figure 8 is a third diagram for explaining the cause correspondence table according to the second embodiment of the present disclosure.

[0025] Figure 9is a flowchart showing an example of the process of the abnormality diagnosing apparatus according to the second embodiment of the present disclosure.

[0026] Figure 10 is a diagram showing a functional structure of the abnormality diagnosing apparatus according to the third embodiment of the present disclosure.

[0027] Figure 11 is a first diagram for explaining a cause correspondence table according to the third embodiment of the present disclosure.

[0028] Figure 12 is a second diagram for explaining a cause correspondence table according to the third embodiment of the present disclosure.

[0029] Figure 13 is a flowchart showing an example of the process of the abnormality diagnosing apparatus according to the third embodiment of the present disclosure.

[0030] Figure 14 is a diagram showing an overall structure of the equipment according to the fourth embodiment of the present disclosure.

[0031] Figure 15 is a diagram showing a function of the observation system restriction determining section in the fourth embodiment of the present disclosure.

[0032] Figure 16 is a diagram showing a function of the model estimating section in the fourth embodiment of the present disclosure.

[0033] Figure 17 is a diagram showing an observation model in the fourth embodiment of the present disclosure.

[0034] Figure 18 is a diagram explaining a full-scale error.

[0035] Figure 19 is a diagram explaining an indicated value error.

[0036] Figure 20 is a diagram explaining an Allan variance.

[0037] Figure 21 is a diagram explaining a relationship between the Allan variance and a time window.

[0038] Figure 22 is a diagram showing a function of the model estimating section in the fourth embodiment of the present disclosure.

[0039] Figure 23 is a diagram showing a function of the matching determining section in the fourth embodiment of the present disclosure.

[0040] Figure 24 is a diagram showing a function of the matching determining section in the fourth embodiment of the present disclosure.

[0041] Figure 25 This diagram illustrates the function of the matching determination unit in the fourth embodiment of this disclosure.

[0042] Figure 26 This diagram illustrates the function of the matching determination unit in the fourth embodiment of this disclosure.

[0043] Figure 27 This is a flowchart of the estimation method of the fourth embodiment of this disclosure.

[0044] Figure 28 This diagram illustrates an example of the hardware structure of the computer included in the anomaly diagnosis device and the estimation device according to each embodiment. Detailed Implementation

[0045] The embodiments disclosed herein are described below using accompanying drawings. Throughout the drawings, identical or equivalent structures are labeled with the same reference numerals, and common descriptions are omitted.

[0046] <First Implementation>

[0047] (Overall structure)

[0048] Figure 1 This is a diagram showing the overall structure of the equipment according to the first embodiment of this disclosure.

[0049] like Figure 1 As shown, equipment 9 includes device 1, anomaly diagnosis device 5, and a first observation system 3.

[0050] Equipment 1 is, for example, a large refrigeration unit, a submersible pump, etc.

[0051] The anomaly diagnostic device 5 detects the presence or absence of anomalies in the equipment 1 and diagnoses the causes of such anomalies. Furthermore, the anomaly diagnostic device 5 is connected to the first observation system 3 via wired or wireless communication.

[0052] The first observation system 3 is a system for observing the state of the device 1, and includes, for example, multiple sensors (first sensor 31, second sensor 32, third sensor 33, ...).

[0053] The first sensor 31 is, for example, located inside the device 1, and can be a pressure sensor that measures the internal pressure of the device 1 as an observed value A1. The second sensor 32 is, for example, located at the inlet or outlet of the device 1, and can be a thermometer that measures the inlet or outlet temperature of the device 1 as an observed value A2. The third sensor 33 is, for example, located in the device 1, and can be a flow meter that measures the flow rate of liquids, gases, etc., flowing through the device 1 as an observed value A3. Furthermore, the observed values ​​A1, A2, A3, ... are examples of the state variables of the device 1 in this embodiment.

[0054] (Functional configuration of abnormality diagnosing apparatus)

[0055] The abnormality diagnosing apparatus 5 is an apparatus for detecting presence or absence of an abnormality of the device 1 and inferring a cause of the abnormality. As shown in Figure 1 , the abnormality diagnosing apparatus 5 is provided with an observation value acquiring section 51, an abnormality determining section 52, and a cause inferring section 53.

[0056] The observation value acquiring section 51 acquires observation values Al, A2, A3 observed by the first observation system 3 of the device 1 as a state quantity of the device 1.

[0057] The abnormality determining section 52 determines presence or absence of an abnormality with respect to the state quantity acquired from the device 1, respectively. In the present embodiment, the abnormality determining section 52 determines, with respect to the observation values Al, A2, A3 acquired by the observation value acquiring section 51, that it is normal if it is within the constraint Rl, and that it is abnormal if it is outside the constraint Rl.

[0058] The cause inferring section 53 uses a cause correspondence table TB1 in which causes of an abnormal mode of the device 1 determined by a fault tree analysis and state quantities that become abnormal if the causes occur are established in correspondence, and infers a cause of an abnormality of the device 1 from the state quantity determined by the abnormality determining section 52 as abnormal.

[0059] The notification section 54 is a display apparatus, a sound output apparatus, or the like for notifying a determination result of the abnormality determining section 52 and an inference result of the cause inferring section 53. Further, the notification section 54 notifies data to an apparatus different from the abnormality diagnosing apparatus 5, such as a computer, a smartphone, a tablet, or the like operated by an operator who performs an operation of the device 1, monitoring, or the like.

[0060] (Cause correspondence table)

[0061] Figure 2 is a first drawing for explaining the cause correspondence table related to the first embodiment of the present disclosure.

[0062] Figure 3 is a second drawing for explaining the cause correspondence table related to the first embodiment of the present disclosure.

[0063] Figure 4 is a third drawing for explaining the cause correspondence table related to the first embodiment of the present disclosure.

[0064] In the present embodiment, the cause correspondence table TB1 is set to be created in advance by a designer or the like of the abnormality diagnosing apparatus 5 before the start of the operation of the device 1. As for the creation method of this cause correspondence table TB1, reference is made to Figures 2-4 for an explanation.

[0065] First, in the design stage of the abnormality diagnosing apparatus 5, an operator who performs design, development, maintenance, or the like of the abnormality diagnosing apparatus 5Figure 2

[0066] An abnormal pattern represents an undesirable failure, accident, or the like of the device 1. A factor represents an event related to generation of the abnormal pattern. The factor can be further divided into an upper factor and a lower factor that causes the upper factor, and is configured in a hierarchical manner. For example, as the abnormal pattern, an event of "the condenser pressure becoming high pressure and becoming unable to operate" can be set. At this time, as the factor 1, an event of "failure of the pressure sensor (first sensor 31)" can be set, and as the factor 3, an event of "the pressure observation value (observation value Al) actually becoming high" or the like can be set. Further, with respect to the factor 3, "malfunction of the cooling tower" and "malfunction of the circulating pump" or the like can be further set as lower factors 3-1 and 3-2.

[0067] In addition, in Figure 2 An example in which one abnormal pattern is set is shown, but is not limited thereto. A plurality of abnormal patterns can be set in correspondence with the structure and characteristics of the device 1, and the tree shown in Figure 2 is constituted for each abnormal pattern. Further, an example in which the factor is divided into two hierarchical layers is shown in Figure 2 , but is not limited thereto. The factor can be divided into three or more hierarchical layers in correspondence with the structure and characteristics of the device 1.

[0068] Further, the work personnel establishes correspondence between a plurality of factors obtained by the fault tree analysis and observation values that become abnormal in a case where the factor is generated (values other than the constraint R1). In the present embodiment, as shown in Figure 2 , the work personnel establishes correspondence between the observation values Al, A2, A3 that can be observed by the first observation system 3 of the device 1 and each factor. For example, if it is known from the product knowledge of the device 1 that the observation value Al and the observation value A3 become abnormal in a case where the factor 3-1 is generated, the factor 3-1 and the observation value Al and the observation value A3 are established in correspondence.

[0069] When the work personnel completes establishment of correspondence between the fault tree analysis result and the observation value, the factor correspondence table TB1 shown in Figure 3 is created. The factor correspondence table TB1 represents which factor of the abnormal pattern will be generated for each combination pattern of the determination results of the observation values Al, A2, A3.

[0070] For example, as shown in Figure 4 ​As shown, only observation A1 from device 1 is considered abnormal. In this case, the three factors that would arise when observation A1 is abnormal are considered: "Factor 1-1", "Factor 2", and "Factor 3-1". However, "Factor 2" and "Factor 3-1" are factors that arise when both observation A1 and observation A3 are abnormal. Figure 4 In the example, since observation A3 is normal, it is excluded from the candidate factors. Therefore, in this example, it can be presumed that "factor 1-1 (factor 1)" is the factor causing the abnormal pattern. The technician creates a correspondence between the factors that each combination of the judgment results of observations A1, A2, and A3 will generate. Figure 3 The factor correspondence table TBI is shown.

[0071] (Process flow of abnormal diagnosis methods)

[0072] Figure 5 This is a flowchart illustrating an example of the processing of the anomaly diagnosis device according to the first embodiment of this disclosure.

[0073] The following is for reference Figure 5 This section will explain the details of the abnormal diagnosis process of the abnormal diagnosis device 5 involved in this embodiment.

[0074] like Figure 5 As shown, the observation acquisition unit 51 of the abnormality diagnosis device 5 acquires observation values ​​A1, A2, and A3 from the first sensor 31, the second sensor 32, and the third sensor 33 of the device 1, respectively (step S101).

[0075] Next, the anomaly determination unit 52 of the anomaly diagnosis device 5 determines whether there is an anomaly in the observed values ​​A1, A2, and A3 respectively (step S102).

[0076] Next, the cause estimation unit 53 of the anomaly diagnosis device 5 determines whether there are any observations that are judged to be abnormal (step S103).

[0077] For example, if the observed value A1 is within the threshold (upper or lower limit) set as a constraint R1, the anomaly determination unit 52 determines that the observed value A1 is normal. On the other hand, if the observed value A1 exceeds the threshold, the anomaly determination unit 52 determines that the observed value A1 is abnormal.

[0078] Furthermore, the anomaly determination unit 52 can determine that the observed value A1 is normal if the change in the observed value A1 is within a threshold set as a constraint R1. In this case, the anomaly determination unit 52 determines that the observed value A1 is abnormal if the change in the observed value A1 exceeds the threshold.

[0079] Furthermore, the anomaly determination unit 52 can also use known anomaly detection algorithms such as k-nearest neighbor method and One Class SVM (Support Vector Machine) to determine whether the observation value A1 indicates a normal pattern within constraint R1 or an anomaly pattern outside constraint R1.

[0080] The anomaly determination unit 52 also determines whether the observed values ​​A2 and A3 are normal or abnormal.

[0081] Next, the anomaly determination unit 52 determines whether all observations are normal (step S103).

[0082] If the anomaly determination unit 52 determines that all observed values ​​A1, A2, and A3 are normal (step S103: Yes), it determines that the operating status of the equipment 1 is normal (step S104).

[0083] On the other hand, if the anomaly determination unit 52 determines that at least one of the observed values ​​A1, A2, and A3 is abnormal (step S103: No), it determines that the operating state of device 1 is abnormal (anomaly mode is generated). In this case, the cause estimation unit 53 determines the cause based on the cause correspondence table TB1 ( Figure 3 To infer the cause of the abnormal pattern (step S105).

[0084] Specifically, the cause estimation unit 53 selects a combination pattern from the cause correspondence table TB1 and the determination result of the observation value from the anomaly determination unit 52 to establish a corresponding cause. For example, such as Figure 4 As in the example, the determination result of observation A1 is "abnormal", and the determination results of observations A2 and A3 are "normal". In this case, the cause estimation unit 53 refers to the cause correspondence table TB1 and establishes "cause 1-1" corresponding to "mode 4" representing the combination of the determination results as the cause of the abnormal mode of device 1.

[0085] In addition, sometimes such as Figure 3 As with "Mode 5", there exists a mode where the cause is not narrowed down to a single factor. In this case, the cause estimation unit 53 may not determine only one cause for the abnormal mode, but may estimate both "Cause 2" and "Cause 3-1" corresponding to "Mode 5" as candidates for causes of the abnormal mode.

[0086] Further, in other embodiments, an additional observation system (not shown) for identifying "factor 2" and "factor 3-1" can also be provided for "pattern 5" in which the factor is narrowed down to one from the factor-to-correspondence table TB1. In this case, the factor inference unit 53 can further infer whether the factor of the abnormal pattern observed when "pattern 5" is observed is "factor 2" or "factor 3-1" based on the observation value obtained from the additional observation system. At this time, the additional observation system can be provided only at a portion corresponding to the factor of "pattern 5" that cannot be identified based on the observation value of the first observation system 3. Therefore, the additional observation system does not need to be provided for the factor narrowed down to one from the observation value of the first observation system 3. For this reason, even in the case where the additional observation system is provided for identification of the factor, the increase in the number of observation systems can be minimized.

[0087] Further, in other embodiments, the factor inference unit 53 can further perform a checking process for narrowing down the factor and further infer whether the factor of the abnormal pattern observed when "pattern 5" is observed is "factor 2" or "factor 3-1". In this way, since the additional observation system for factor identification is not required, the cost involved in the abnormal diagnosis of the device 1 can be minimized.

[0088] Next, the notification unit 54 notifies the result of the determination by the abnormality determination unit 52 (step S106). Further, the notification unit 54 further notifies the factor inferred by the factor inference unit 53 in the case where it is determined by the abnormality determination unit 52 that the operation state of the device 1 is abnormal.

[0089] The abnormality diagnosis apparatus 5 detects the presence or absence of an abnormality in the device 1 in real time by repeatedly performing the series of processes illustrated in FIG. 8 in the operation of the device 1 and infers the factor of the abnormality when the abnormality occurs. Figure 5 The abnormality diagnosis apparatus 5 detects the presence or absence of an abnormality in the device 1 in real time by repeatedly performing the series of processes illustrated in FIG. 8 in the operation of the device 1 and infers the factor of the abnormality when the abnormality occurs.

[0090] (EFFECTS)

[0091] As described above, in the abnormality diagnosis apparatus 5 according to the present embodiment, in the case where the abnormality determination unit 52 detects an abnormality in at least one of the state quantities (observation values Al, A2, A3) obtained from the device 1, the factor inference unit 53 infers the factor of the abnormal pattern of the device 1 based on the factor-to-correspondence table TB1 in which the factor of the abnormal pattern and the state quantity that becomes abnormal when the factor occurs are associated with each other.

[0092] In this way, the abnormality diagnosis apparatus 5 can easily and accurately infer the factor of the abnormality of the device 1 based on the factor-to-correspondence table TB1.

[0093] <2nd Embodiment>

[0094] Next, the abnormality diagnosis device 5 according to the second embodiment of the present disclosure will be described. Common elements with the first embodiment are denoted by the same reference numerals and detailed description will be omitted.

[0095] (Factor correspondence table)

[0096] Figure 6 is a first diagram for explaining the factor correspondence table according to the second embodiment of the present disclosure.

[0097] Figure 7 is a second diagram for explaining the factor correspondence table according to the second embodiment of the present disclosure.

[0098] Figure 8 is a third diagram for explaining the factor correspondence table according to the second embodiment of the present disclosure.

[0099] In the present embodiment, the factor correspondence table TB2 is also created in advance by the designer or the like of the abnormality diagnosis device 5 before the start of the operation of the device 1. The method of creating the factor correspondence table TB2 according to the present embodiment will be described with reference to Figures 6-8

[0100] First, as shown in Figure 6 , at the design stage of the abnormality diagnosis device 5, the fault tree analysis of the device 1 is performed by the operator who performs the design, development, maintenance, or the like of the abnormality diagnosis device 5. This is the same as in the first embodiment. Further, in the present embodiment, a plurality of factors obtained by the fault tree analysis are each associated with the abnormality degree of the value that each observation value Al, A2, A3 can take in the case where the factor has occurred.

[0101] The abnormality degree is a value that discretizes each observation value into a plurality of levels corresponding to the difference from the constraint R1 set for each observation value. For example, the abnormality degree can be discretized into any one of 2 to 7 levels with respect to each observation value as shown in the discretization table of Figure 7 . For example, in the case of being discretized into 3 levels, the abnormality degree is represented in 3 stages of "negative" in which the observation value becomes a value more on the negative side than the constraint R1, "normal" which is within the constraint R1, and "positive" in which the observation value becomes a value more on the positive side than the constraint R2. Further, in the case of being discretized into 5 levels, it is also possible to further divide "negative" / "positive" into 2 stages of "large negative" / "large positive" and "small negative" / "small positive" respectively, in correspondence with how far the observation value deviates from the constraint R1 to the negative side / positive side. In the case of being discretized into 7 levels, it is also possible to further divide "negative" / "positive" into 3 stages of "large negative" / "large positive", "medium negative" / "medium positive", and "small negative" / "small positive" respectively, in correspondence with how far the observation value deviates from the constraint R1 to the negative side / positive side, for example. ​

[0102] In the present embodiment, for example, in a case where it is difficult to narrow down the cause to one with the abnormality degree of only 2 levels ("normal" or "abnormal") (for example Figure 3 of the cause corresponds to "Pattern 5" of the cause correspondence table TB1), the abnormality degree of the observation value is subdivided into 3 levels to 7 levels. In addition, the discretization levels can also be different for each observation value. In the present embodiment, as shown in Figure 6 , an example in which each observation value is represented with the abnormality degree of 5 levels is explained.

[0103] When the work personnel completes the correspondence establishment of the fault tree analysis result and the abnormality degree of the observation value, the cause correspondence table TB2 as shown in Figure 8 is created. The cause correspondence table TB2 represents which cause of the abnormality pattern is generated for each combination pattern of the abnormality degrees of the observation values Al, A2, A3.

[0104] (Process flow of abnormality diagnosis method)

[0105] Figure 9 is a flowchart showing an example of the process of the abnormality diagnosis apparatus according to the second embodiment of the present disclosure.

[0106] Details of the abnormality diagnosis process of the abnormality diagnosis apparatus 5 according to the present embodiment will be explained below with reference to Figure 9 .

[0107] As shown in Figure 9 , the observation value acquisition unit 51 of the abnormality diagnosis apparatus 5 acquires the observation values Al, A2, A3 from the first sensor 31, the second sensor 32, the third sensor 33 of the device 1, respectively (step S111).

[0108] Next, the abnormality determination unit 52 of the abnormality diagnosis apparatus 5 determines the abnormality degree corresponding to the difference from the constraint R1 set for each observation value for the observation values Al, A2, A3, respectively (step S112).

[0109] For example, the abnormality determination unit 52 determines the observation value Al as any one of the abnormality degrees of 5 levels of "large negative", "small negative", "normal", "small positive", and "large positive". The abnormality determination unit 52, for example, determines that the observation value Al is normal in a case where the observation value Al is in a range of the first lower limit value or more and less than the first upper limit value set as the constraint R1. Further, the abnormality determination unit 52 determines that the observation value Al is "small negative" in a case where the observation value Al is less than the first lower limit value and is the second lower limit value or more, and determines that the observation value Al is "large negative" in a case where the observation value Al is less than the second lower limit value. The abnormality determination unit 52 determines that the observation value Al is "small positive" in a case where the observation value Al is the first upper limit value or more and less than the second upper limit value, and determines that the observation value Al is "large positive" in a case where the observation value Al is the second upper limit value or more.

[0110] In addition, in other embodiments, the abnormality determination section 52 can determine the degree of abnormality based on the amount of change in the observation value Al, and can use a known anomaly detection algorithm that utilizes a k-nearest neighbor method, One Class SVM (Support Vector Machine), or the like, to determine the degree of abnormality of the observation value Al.

[0111] Further, the abnormality determination section 52 determines the degree of abnormality for the observation values A2, A3 as well.

[0112] Next, the abnormality determination section 52 determines whether the degrees of abnormality of all of the observation values are normal (step S113).

[0113] The abnormality determination section 52 determines that the operation state of the device 1 is normal (step S114) when it is determined that the degrees of abnormality of all of the observation values Al, A2, A3 are normal (step S113: Yes).

[0114] On the other hand, the abnormality determination section 52 determines that the operation state of the device 1 is abnormal (an abnormal pattern occurs) when it is determined that at least one of the degrees of abnormality of the observation values Al, A2, A3 is other than normal (any one of large negative, small negative, small positive, and large positive) (step S113: No). In this case, the cause estimation section 53 estimates the cause of the abnormal pattern based on the cause correspondence table TB2 (step S115). Figure 8

[0115] Specifically, the cause estimation section 53 selects a cause that corresponds to the combination pattern of the degrees of abnormality of the observation values of the abnormality determination section 52 from the cause correspondence table TB2. For example, as in the example of Figure 8 the degree of abnormality of the observation value Al is “small positive”, the degree of abnormality of the observation value A2 is “normal”, and the degree of abnormality of the observation value A3 is “small negative”. In this case, the cause estimation section 53 refers to the cause correspondence table TB2 and estimates that “cause 2” that corresponds to the “pattern 2” that represents the combination of this determination result is the cause of the abnormal pattern of the device 1.

[0116] Next, the notification section 54 notifies the determination result of the abnormality determination section 52 (step S116). Further, the notification section 54 further notifies the cause estimated by the cause estimation section 53 when it is determined by the abnormality determination section 52 that the operation state of the device 1 is abnormal.

[0117] The abnormality diagnosis device 5 detects the presence or absence of an abnormality in the device 1 in real time and estimates the cause of the abnormality when an abnormality occurs by repeatedly performing the series of processes illustrated in Figure 9

[0118] ​​(Action and Effect)

[0119] As above, in the abnormality diagnosing apparatus 5 according to the present embodiment, the abnormality determining section 52 further determines the abnormality degree corresponding to the difference from the constraint Rl for each observation value. The cause inferring section 53 infers the cause of the abnormality pattern of the device 1 using the cause correspondence table TB2 that pre-establishes correspondence between the cause of the abnormality pattern and the abnormality degree of each observation value.

[0120] For example, in the first embodiment, sometimes the cause of the abnormality pattern cannot be narrowed down to one from the determination result (2 levels) indicating the presence or absence of abnormality of the observation value. However, the abnormality diagnosing apparatus 5 according to the present embodiment performs determination by not dividing each observation value into two levels of "normal" or "abnormal", but subdividing the abnormality degree into multiple stages, and narrows down the cause more finely.

[0121] <3rd Embodiment>

[0122] Next, the abnormality diagnosing apparatus 5 according to the third embodiment of the present disclosure will be described. Common elements with the first and second embodiments are denoted by the same reference numerals and detailed description will be omitted.

[0123] (Functional Configuration of Abnormality Diagnosing Apparatus)

[0124] Figure 10 is a diagram showing the functional configuration of the abnormality diagnosing apparatus according to the third embodiment of the present disclosure.

[0125] As shown in Figure 10 , the abnormality diagnosing apparatus 5 according to the present embodiment further includes an inferred value acquiring section 55.

[0126] The inferred value acquiring section 55 acquires the physical quantity inferred value PV1, PV2,... of the device 1 inferred on the basis of the observation values Al, A2, A3 acquired by the observation value acquiring section 51, as the state quantity of the device 1. In the present embodiment, the physical quantity inferred value PV is an internal state quantity of the device 1 that cannot be observed in the observation system, which has an influence on the relationship of the input and output conditions of the device 1. The physical quantity inferred value PV is, for example, the work quantity in the device 1 such as the end temperature difference of the heat exchanger possessed by the device 1, the refrigeration capacity, or the like.

[0127] (Cause Correspondence Table)

[0128] Figure 11 is a first diagram for explaining the cause correspondence table according to the third embodiment of the present disclosure.

[0129] Figure 12 is a second diagram for explaining the cause correspondence table according to the third embodiment of the present disclosure.

[0130] In this embodiment, the cause-and-effect table TB3 is also designed to be created in advance by the designer of the anomaly diagnosis device 5 before the operation of device 1 begins. (See reference) Figures 11-12 This section explains the method for creating the factor correspondence table TB3 involved in this embodiment.

[0131] First, such as Figure 11 As shown, during the design phase of the anomaly diagnosis device 5, the personnel responsible for the design, development, and maintenance of the anomaly diagnosis device 5 perform fault tree analysis on the equipment 1. This is the same as in the first and second embodiments. Furthermore, in this embodiment, a correspondence is established between the multiple factors obtained from the fault tree analysis and the degree of anomaly of the physical quantity estimation values ​​PV1, PV2, observed value A1, ... that can be obtained when these factors occur. Additionally, in Figure 11 The example shown categorizes the estimated physical quantities PV1, PV2, and state variable A1 into five levels of anomaly severity, but is not limited to this. If the causes of the anomalous pattern can be narrowed down to a single level, the anomaly severity of each physical quantity and observation can be set as follows: Figure 7 Any of the levels shown, from 2 to 7.

[0132] Once the operators have completed the mapping between the fault tree analysis results and the physical quantities and the degree of anomalies in the observed values, they will create... Figure 12 The cause-and-effect table TB3 is shown in the figure. The cause-and-effect table TB3 characterizes which cause of the anomalous pattern according to each combination of the anomaly degree of the estimated physical quantity PV1, the estimated physical quantity PV2, and the observed value A1.

[0133] (Process flow of abnormal diagnosis methods)

[0134] Figure 13 This is a flowchart illustrating an example of the processing of the anomaly diagnosis device according to the third embodiment of this disclosure.

[0135] The following is for reference Figure 13 This section will explain the details of the abnormal diagnosis process of the abnormal diagnosis device 5 involved in this embodiment.

[0136] like Figure 13 As shown, the observation acquisition unit 51 of the abnormality diagnosis device 5 acquires observation values ​​A1, A2, and A3 from the first sensor 31, the second sensor 32, and the third sensor 33 of the device 1, respectively (step S121).

[0137] Next, the estimated value acquisition unit 55 of the abnormality diagnosing apparatus 5 acquires the physical quantity estimated values PV1, PV2 of the device 1 estimated based on the observation values Al, A2, A3 (step S122). For example, the estimated value acquisition unit 55 acquires the physical quantity estimated values PV1, PV2 of the device 1 estimated using a known estimation algorithm using a Kalman filter, a hidden Markov model, or the like.

[0138] Next, the abnormality determination unit 52 of the abnormality diagnosing apparatus 5 determines whether or not there is an abnormality or the degree of abnormality in the observation values Al acquired by the observation value acquisition unit 51 and the physical quantity estimated values PV1, PV2 estimated by the estimated value acquisition unit 55, respectively (step S123). The details of the processing are the same as in the step S102 (of the first embodiment) or the step S112 (of the second embodiment). Figure 5 Figure 9

[0139] Next, the abnormality determination unit 52 determines whether or not all of the observation values and the physical quantities are normal (step S124).

[0140] The abnormality determination unit 52 determines that the operation state of the device 1 is normal when it is determined that all of the observation values Al and the physical quantity estimated values PV1, PV2 are normal (step S124: Yes) (step S125).

[0141] On the other hand, the abnormality determination unit 52 determines that the operation state of the device 1 is abnormal (abnormal pattern occurs) when it is determined that at least one of the observation values Al and the physical quantity estimated values PV1, PV2 is not normal (abnormal or any one of large negative, small negative, small positive, and large positive) (step S124: No). In this case, the cause estimation unit 53 estimates the cause of the abnormal pattern based on the cause correspondence table TB3 (of the first embodiment) (step S126). The details of the processing are the same as in the step S105 (of the first embodiment), the step S115 (of the second embodiment). Figure 12 Figure 5 Figure 9

[0142] Next, the notification unit 54 notifies the determination result of the abnormality determination unit 52 (step S127). In addition, the notification unit 54 further notifies the cause estimated by the cause estimation unit 53 when it is determined by the abnormality determination unit 52 that the operation state of the device 1 is abnormal.

[0143] The abnormality diagnosing apparatus 5 detects whether or not there is an abnormality in the device 1 in real time and estimates the cause of the abnormality when the abnormality occurs by repeatedly performing the series of processes illustrated in FIG. 10. Figure 13

[0144] (EFFECTS) ​​​​​​

[0145] As above, the abnormality diagnosing apparatus 5 according to the present embodiment further includes the estimated value acquisition unit 55 that estimates the physical quantity estimated values PV1, PV2 of the physical quantity of the device 1 that cannot be observed by the first observation system 3.

[0146] In this way, the abnormality diagnosing apparatus 5 can perform the determination of the presence or absence of abnormality and the estimation of the abnormality cause using the physical quantity estimated values PV1, PV2 that estimate the internal state quantity of the device 1 in addition to the observation value Al acquired from the first observation system 3 of the device 1. Thus, the abnormality diagnosing apparatus 5 can improve the detection accuracy for the abnormality pattern that is difficult to detect only by the change in the observation value.

[0147] <4th Embodiment>

[0148] Next, the abnormality diagnosing apparatus 5 according to the 4th embodiment of the present disclosure will be described. The same reference numerals are given to the constituent elements common to the 1st to 3rd embodiments, and detailed description will be omitted.

[0149] Figure 14 is a diagram showing the overall structure of the equipment according to the 4th embodiment of the present disclosure.

[0150] As shown in Figure 14 , the equipment 9 according to the present embodiment further includes the estimation apparatus 2. In addition, the equipment 9 according to the present embodiment includes the second observation system 4 in addition to the first observation system 3.

[0151] (Structure of the first observation system)

[0152] The first observation system 3 is a system for observing the state of the device 1.

[0153] The first observation system 3 observes the first observation value OB1.

[0154] For example, the first observation system 3 can be provided in the device 1.

[0155] For example, the first observation system 3 includes the first sensor 31.

[0156] For example, the first observation system 3 can observe the observation value Al as the first observation value OB1.

[0157] For example, the first sensor 31 can be provided in the inside of the device 1. At this time, the first sensor 31 can be a pressure sensor that measures the internal pressure of the device 1 as the observation value Al.

[0158] (Structure of the second observation system)

[0159] The second observation system 4 is a system for observing the state of the device 1.

[0160] The second observation system 4 observes a second observation value OB2.

[0161] The second observation system 4 is another observation system independent of the first observation system 3.

[0162] For example, the second observation system 4 can be provided in the device 1.

[0163] For example, the second observation system 4 can include a second sensor 41, a third sensor 42, and a fourth sensor 43.

[0164] For example, the second observation system 4 can observe observation values A2, A3, A4 as the second observation value OB2.

[0165] For example, the second sensor 41 can be provided at an outlet of the device 1. In this case, the second sensor 41 can be a thermometer that measures a temperature of a fluid such as a liquid or a gas flowing out of the device 1, that is, an outlet temperature, as the observation value A2.

[0166] Further, the third sensor 42 can be provided at an inlet of the device 1. In this case, the third sensor 42 can be a thermometer that measures a temperature of a fluid such as a liquid or a gas flowing into the device 1, that is, an inlet temperature, as the observation value A3.

[0167] Further, the fourth sensor 43 can be provided in the device 1. In this case, the fourth sensor 43 can be a flowmeter that measures a flow rate of a fluid such as a liquid or a gas flowing through the device 1 as the observation value A4.

[0168] (Structure of Estimation Device)

[0169] The estimation device 2 is a device for estimating parameters PR of each model of the device 1, the first observation system 3, and the second observation system 4.

[0170] Further, the estimation device 2 is a device for providing the anomaly diagnosis device 5 with a network model NWM including the estimated parameters PR.

[0171] The estimation device 2 includes an observation system constraint determination section 22, a model estimation section 23, and a matching determination section 24.

[0172] For example, the estimation device 2 can further include a physical quantity constraint determination section 25 and a model constraint determination section 26.

[0173] Further, the estimation device 2 can further include an acquisition section 21 and an output section 27.

[0174] (Structure of Acquisition Section)

[0175] The acquisition section 21 acquires the first observation value OB1.

[0176] For example, the acquisition unit 21 can acquire the observation value Al measured by the first sensor 31 as the first observation value OB1.

[0177] For example, the acquisition unit 21 can acquire the internal pressure of the device 1 measured by the first sensor 31 as the observation value Al.

[0178] The acquisition unit 21 acquires the second observation value OB2.

[0179] For example, the acquisition unit 21 can acquire the observation values A2, A3, A4 measured by the second sensor 41 as the second observation value OB2.

[0180] For example, the acquisition unit 21 can acquire the outlet temperature measured by the second sensor 41 as the observation value A2.

[0181] For example, the acquisition unit 21 can acquire the inlet temperature measured by the third sensor 42 as the observation value A3.

[0182] For example, the acquisition unit 21 can acquire the flow rate measured by the fourth sensor 43 as the observation value A4.

[0183] (Structure of observation system constraint determination unit)

[0184] The observation system constraint determination unit 22 determines the data within the first constraint R1, that is, the constraint-internal data DT2, based on the time-series data DT1 of each observation value of the first observation value OB1 observed by the first observation system 3 and the second observation value OB2 observed by the second observation system 4.

[0185] The observation system constraint determination unit 22 determines whether each observation value is data of an abnormal behavior as a single unit by determining the suitability of the constraint condition for each observation system unit.

[0186] For example, as shown in Figure 2 , the observation system constraint determination unit 22 can determine the constraint-internal data DT2 by the observation system constraints such as a constraint related to a maximum value and a minimum value, a constraint related to a moving average change rate, a constraint related to a spike generation frequency, a constraint related to a variance within a time window, and the like as the first constraint R1.

[0187] Thus, the observation system constraint determination unit 22 determines an abnormality based on the measurement principle (physical law) of each observation system itself, as a single unit of each observation value, including a value containing an abnormality.

[0188] For example, in Figure 15In the case of data DT3 in the time series data DT1 of the original observations observed by each observation system, data DT3 is taken as the peak of the observations. Data with a frequency higher than the specified threshold, or data with a variance of the observations within the time window higher than the specified threshold, are determined to be outside the constraint according to the first constraint R1.

[0189] Similarly, data DT4 in the time series data DT1 is considered outside the constraint if the observed value is greater than the maximum value specified above or less than the minimum value specified below, based on the first constraint R1.

[0190] Similarly, data DT5 in the time series data DT1 is considered as data whose moving average rate of change is greater than the specified threshold, and is determined to be outside the constraint according to the first constraint R1.

[0191] On the other hand, data that is not outside the constraints is judged as data inside the constraints, DT2.

[0192] For example, if an out-of-constraint determination occurs in any of the observation systems, the observation system constraint determination unit 22 can also determine the data of other observation systems at the same time as out of constraint, and not perform the subsequent processing in the estimation device 2.

[0193] (Structure of the model estimation section)

[0194] The model estimation unit 23 estimates the parameters PR of multiple models, including the model of each observation system, namely the observation model MLO, and the model in the equipment 1 equipped with the observation system, namely the physical model MLP, based on the constraint data DT2.

[0195] For example, the model estimation unit 23 may include a network model NWM, which includes multiple observation models MLO and multiple physical models MLP.

[0196] For example, each model of multiple observation models (MLO) and multiple physical models (MLP) can include parameter PR.

[0197] For example, model estimation unit 23 can be used as an example of network model NWM. Figure 3 The parameters PR of each model in the inferred observation model MLO and the physical model MLP in the network model NWMl shown are given.

[0198] like Figure 16 As shown, in the network model NWM1, the model estimation unit 23 can estimate the physical quantity estimation value PV1, which is one of the multiple physical quantity estimation values ​​PV, based on the observation value A1 via an observation model MLO.

[0199] On the other hand, in the network model NWM1, the model estimation unit 23 can also estimate the physical quantity estimation value PV1 based on the observation values ​​A2, A3 and A4 via multiple observation models MLO and multiple physical models MLP.

[0200] That is, in the network model NWM1, the model estimation unit 23 can estimate the common physical quantity estimate value PV1 from different systems.

[0201] like Figure 16 As shown, for example, in the model estimation unit 23, the parameters PR of the observation model MLO related to the observation value A2 can be estimated based on the observation value A2 and the observation value A2' before the step of observing the observation value A2 (the observation value A2' observed at the moment just before the observation value A2 is about to be observed).

[0202] Similarly, in the model estimation unit 23, the parameters PR of the observation model MLO related to the observation value A3 can be estimated based on the observation value A3 and the observation value A3' before the step of observing the observation value A3 (the observation value A3' observed at the moment just before the observation value A3 is about to be observed).

[0203] (Structure of the model)

[0204] For example, the observation model MLO and the physical model MLP can be represented by nonlinear polynomials based on theoretical and experimental formulas, respectively. In this case, the coefficients of the polynomials are equivalent to the parameter PR, which becomes the parameter used to represent the deviation of the actual device 1.

[0205] The physical model MLP is explained in detail here.

[0206] For example, each physical model (MLP) can be a model derived from pure physical laws known in physical phenomena.

[0207] For example, Figure 16 The physical model MLPA shown can be a model that can calculate the estimated value of physical quantity PVA representing the work done in equipment 1, such as refrigeration capacity, based on the estimated value of physical quantity PV2 representing outlet temperature, the estimated value of physical quantity PV3 representing inlet temperature, and the estimated value of physical quantity PV4 representing flow rate.

[0208] For example, Figure 16 The physical model MLPB shown can be a model that can calculate the estimated value of the physical quantity PVB representing the saturation temperature inside device 1 based on the estimated value of the physical quantity PV2 representing the outlet temperature and the estimated value of the physical quantity PVA representing the work done.

[0209] For example, Figure 16The physical model MLPC shown can be a model that calculates a physical quantity estimate PV1 representing the pressure (saturated vapor pressure) in the device 1 from a physical quantity estimate PVB representing the saturated temperature in the device 1.

[0210] Next, the observation model MLO is described in detail.

[0211] Each observation model MLO can be, for example, a model represented by Figure 17 a representative model as shown.

[0212] In Figure 17 , an error from a true value due to a characteristic of a sensor is characterized as an internal factor, and an error due to a value of an observation object is characterized as an external factor.

[0213] In addition to the internal factor and the external factor that can be modeled, an observation value is further affected by a random noise component.

[0214] In addition, the external factor requires any one of a premise that "it is small to an extent that can be ignored", "the external factor is measured with another sensor", and "only a data amount of noise can be handled".

[0215] The accuracy of an observation system is prescribed by any one of a full-scale error (±○% F.S.) as shown in Figure 18 or an indicated value error (±○% R.D.) as shown in Figure 19 , or an index that combines them, on the basis of considering all errors.

[0216] In addition, these error ranges are ultimately accuracies in a case of normal operation, and deviate from a range that causes a sensor to be installed out of position or to malfunction.

[0217] Further, in a case where the same value is continuously measured, the size of a deviation of a result of comparing values that are averaged with a time window to each other is called an Allan variance.

[0218] A typical Allan variance shows a variance as shown in Figure 20 In addition, Figure 20 σy(f) shown is a value associated with the size of the Allan variance.

[0219] As shown in Figure 20 , Figure 21 In a period in which the averaging time τ is short, the influence of noise is dominant, and the deviation between the average values slowly becomes small, but if the averaging time τ is lengthened, on the contrary, the influence of a long-term variation of an observation system parameter becomes apparent and the deviation slowly continuously becomes large.

[0220] Therefore, for example, ideally, the parameters PR of the observation model MLO can be estimated from recorded data in which only the Allan variance is minimized.

[0221] For example, the model estimation unit 23 can also use the most recent observation from the constraint-inclusive data DT2, which contains no anomalies and is considered as an individual observation in the observation system constraint determination unit 22, to accumulate the sum of the average value of the physical quantity estimation PV for the entire system and the sum of the squares of the errors of the physical quantity estimation PV for each system in the entire record for the physical quantity estimation value PV that can be estimated from each system. At this time, the accumulated value and the magnitude of the correction term for each parameter PR can be determined based on the magnitude of the penalty coefficient determined in advance, and the parameter PR can be estimated so that the value is minimized.

[0222] For example, the model estimation unit 23 can estimate the parameter PR for observation systems that can derive a common physical quantity estimate PV1 by transforming the values ​​of each model based on the network model NWM1, so that the sum of the penalty for the deviation between the physical quantity estimate PV1 estimated from each system and the penalty for the correction term of each parameter PR is minimized.

[0223] Furthermore, even when the number of observation systems is generally extremely small relative to the number of parameters PR to be estimated, the model estimation unit 23 can estimate the physical quantity PV1 by using multiple records.

[0224] For example, the observation model MLO may not be an observation model that is modeled exactly as it is, but rather an observation model that is simplified to a suitable range, taking into account the usage conditions of device 1 and the specifications of each sensor.

[0225] For example, in the observation model MLO associated with the observation A1, the penalty for the correction term of the bias, scaling factor (hereinafter also referred to as "SF") from the ideal state can be determined based on the sensor accuracy.

[0226] The model estimation unit 23 is configured as a reward function to determine the magnitude of the mean square error of the difference between the estimated values ​​PV1 of multiple physical quantities and the average value, which is the deviation between the estimated values ​​PV1 of physical quantities estimated from each system in the network model NWM1.

[0227] For example, such as Figure 22 As shown, when the estimated values ​​PV of each physical quantity can be estimated from three or more different systems in a network model NWM containing multiple models, the model estimation unit 23 can determine the reward function for the magnitude of the mean square error of the difference between the estimated values ​​PV of each physical quantity and the average value, as the deviation between the estimated values ​​PV of each physical quantity estimated from three or more systems.

[0228] and Figure 22As shown in the drawing, the model estimation unit 23 can perform estimation of the parameters PR so that the sum of the penalty for the deviation of the physical quantity estimation value PV estimated from each of the three or more systems from each other and the penalty for the correction term for each parameter PR becomes the minimum.

[0229] (Configuration of the matching determination unit)

[0230] The matching determination unit 24 determines the matching of the model based on the deviation of the first predicted observation value PA1 predicted from the second observation value OB2 based on the parameters PR of the plurality of models estimated by the model estimation unit 23 from the first observation value OB1.

[0231] Thus, the matching determination unit 24 can evaluate the size of the deviation by regressing a certain observation value from other observation values based on the estimated parameters PR.

[0232] Here, the first predicted observation value PA1 is a value derived by transformation based on the values of each model of the network model NWM1, and is a value equivalent to the first observation value OB1 in the network model NWM1.

[0233] For example, the matching determination unit 24 can compare the first predicted observation value PA1, which is a regression result predicted based on the parameters PR of each model estimated by the model estimation unit 23, and the first observation value OB1, which is a measured value.

[0234] For example, as the matching of the model, the matching determination unit 24 can determine whether the deviation of the first predicted observation value PA1 predicted from the second observation value OB2 from the first observation value OB1 is within the specification range of the first observation system 3.

[0235] For example, as shown in the drawing, as the matching of the model, the matching determination unit 24 can determine whether the deviation of the first predicted observation value PA1 predicted from the observation value A2, the observation value A3, and the observation value A4 from the observation value Al is within the specification range of the first observation system 3. Figure 23

[0236] For example, the matching determination unit 24 predicts the first predicted observation value PA1 from the observation value A2, the observation value A3, and the observation value A4 via the plurality of observation models MLO and the plurality of physical models MLP into which the parameters PR estimated by the model estimation unit 23 are introduced, and compares the predicted first predicted observation value PA1 and the observation value Al.

[0237] ​For example, the matching determination unit 24 can also obtain the difference between the first predicted observation value PA1 and the first observation value OB1 as the deviation of the first predicted observation value PA1 from the first observation value OB1. At this time, as the determination of the matching of the models, the matching determination unit 24 can determine that the plurality of observation models MLO and the plurality of physical models MLP into which the parameters PR are introduced match if the obtained difference is equal to or less than a predetermined threshold value.

[0238] For example, the matching determination unit 24 can also perform mutual regression and evaluate the matching of all the observation systems with respect to each other.

[0239] For example, the matching determination unit 24 can also perform evaluation in addition to the evaluation shown in FIG. 6, and predict the second predicted observation value PA2 from the observation value A1, the observation value A3, and the observation value A4 via the plurality of observation models MLO and the plurality of physical models MLP into which the parameters PR estimated by the model estimation unit 23 are introduced, and further compare the predicted second predicted observation value PA2 with the observation value A2. Figure 23 Figure 24 For example, the matching determination unit 24 can also perform evaluation in addition to the evaluation shown in FIG. 6, and predict the second predicted observation value PA2 from the observation value A1, the observation value A3, and the observation value A4 via the plurality of observation models MLO and the plurality of physical models MLP into which the parameters PR estimated by the model estimation unit 23 are introduced, and further compare the predicted second predicted observation value PA2 with the observation value A2.

[0240] Further, as with the evaluation related to the observation value A1 shown in FIG. 6 and the evaluation related to the observation value A2 shown in FIG. 7, the matching determination unit 24 can also perform evaluation related to the observation value A3 based on the observation value A1, the observation value A2, and the observation value A4, and can also perform evaluation related to the observation value A4 based on the observation value A1, the observation value A2, and the observation value A3. Figure 23 Figure 24 For example, as shown in FIG. 8, in a case where the same observation value is regressed by a plurality of systems independently, the matching determination unit 24 can also obtain the deviation for each system to which regression is performed, and take the minimum value of the obtained plurality of deviations as the evaluation result of the observation value.

[0241] For example, as shown in FIG. 8, in a case where the same observation value is regressed by a plurality of systems independently, the matching determination unit 24 can also obtain the deviation for each system to which regression is performed, and take the minimum value of the obtained plurality of deviations as the evaluation result of the observation value. Figure 25 Figure 26 For example, as shown in FIG. 8, in a case where the same observation value is regressed by a plurality of systems independently, the matching determination unit 24 can also obtain the deviation for each system to which regression is performed, and take the minimum value of the obtained plurality of deviations as the evaluation result of the observation value.

[0242] (Structure of Physical Quantity Constraint Determination Unit)

[0243] The physical quantity constraint determination unit 25 determines whether the physical quantity estimated value PV estimated based on the parameters PR is within the second constraint R2.

[0244] Here, the physical quantity estimated value PV is a physical quantity within the device 1 estimated by the physical model MLP.

[0245] For example, the physical quantity constraint determination unit 25 can also determine whether the physical quantity estimated value PV estimated based on the observation value of the parameters PR and the constraint-internal data DT2 when the matching is determined in the matching determination unit 24 is within the second constraint R2.

[0246] ​​​For example, the second constraint R2 can be a prescribed numerical range indicating a physical quantity within the design level (specification) of the device 1.

[0247] For example, the physical quantity constraint determination section 25 can determine normal if the physical quantity estimation value PV is within the second constraint R2, and abnormal if it is not. In the case of determining normal, in the determination of the physical quantity constraint determination section 25, it can be considered that the parameters PR introduced to each model are appropriate.

[0248] For example, in the case of determining that the physical quantity estimation value PV is abnormal, the physical quantity constraint determination section 25 determines that the device 1 is abnormal.

[0249] That is, if based on the assumption that the estimation of the physical quantity can be performed with respect to the constraint-in data DT2 as a record determined to be normal, if the physical quantity estimation value PV as the physical quantity is not included within the constraint of the design specification, it can be considered that the device 1 is abnormal.

[0250] For example, by the use of the first constraint Rl, in the case where each observation value of the constraint-in data DT2 is dispersed in time, the physical quantity constraint determination section 25 can perform time interpolation with respect to each dispersed observation value, and estimate the physical quantity estimation value PV based on the observation value subjected to time interpolation.

[0251] For example, the physical quantity constraint determination section 25 can determine whether the physical quantity estimation value PV estimated at each time step is within the second constraint R2.

[0252] For example, the physical quantity constraint determination section 25 compares each estimated physical quantity estimation value PV with the design level (specification), and performs determination of within the specification, above the specification, below the specification, and the like, whereby it is converted into discretized data of 2 to 7 stages.

[0253] (Configuration of the Model Constraint Determination Section)

[0254] The model constraint determination section 26 determines whether each parameter PR is within the third constraint R3.

[0255] For example, the model constraint determination section 26 can determine whether each parameter PR at the time of determining matching in the matching determination section 24 is within the third constraint R3.

[0256] For example, the third constraint R3 with respect to each parameter PR in the physical model MLP can be a prescribed numerical range indicating a parameter within the design level (specification) of the device 1.

[0257] For example, the third constraint R3 with respect to each parameter PR in the observation model MLO can be a prescribed numerical range indicating a parameter within the design level (specification) of each observation value.

[0258] For example, the model constraint determination unit 26 can determine as normal if each parameter PR is within the third constraint R3, and determine as abnormal if not. In the case of determining as normal, in the determination of the model constraint determination unit 26, each parameter PR introduced to each model can be regarded as appropriate.

[0259] For example, each parameter PR determined in the model constraint determination unit 26 can be each parameter estimated in a moving window or a batch time window in which mutual regression is performed for each.

[0260] For example, the model constraint determination unit 26 compares each parameter PR estimated and a design level (specification), and performs determination of within the specification, above the specification, below the specification, and the like, whereby, it is converted into discretized data of 2 to 7 stages.

[0261] For example, in the case of determining that the parameter PR of the physical model MLP is abnormal, the model constraint determination unit 26 can determine that the device 1 is abnormal.

[0262] That is, if based on the assumption that parameter estimation can be correctly performed, as long as the parameter PR of the physical model MLP is not included within the constraint of the design specification, it can be regarded as abnormality of the device 1.

[0263] For example, in the case of determining that the parameter PR of the observation model MLO is abnormal, the model constraint determination unit 26 can determine that the observation system is abnormal. Further, the model constraint determination unit 26 can determine that the sensor associated with the observation model MLO having the parameter PR determined as abnormal is abnormal.

[0264] That is, if based on the assumption that parameter estimation can be correctly performed, as long as the parameter PR of the observation model MLO is not included within the constraint of the design specification, it can be regarded as abnormality of the observation system.

[0265] (Structure of output unit)

[0266] The output unit 27 outputs the parameter PR determined as normal and the physical quantity estimated value PV estimated based on the parameter PR.

[0267] Further, the output unit 27 outputs the parameter PR determined as abnormal and the physical quantity estimated value PV estimated based on the parameter PR.

[0268] For example, the output unit 27 can output the meaning that the sensor associated with the observation model MLO having the parameter PR determined as abnormal is abnormal in the case where the parameter PR determined as abnormal is the parameter PR of the observation model MLO.

[0269] For example, the output unit 27 can output the meaning that the device 1 is abnormal in the case where the parameter PR determined as abnormal is the parameter PR of the physical model MLP.

[0270] (Action of Estimation Device)

[0271] The action of the estimation device 2 of this embodiment will be described.

[0272] The action of the estimation device 2 corresponds to the estimation method of this embodiment.

[0273] The action of the estimation device 2 can be implemented, for example, as shown in Figure 27

[0274] First, the acquisition unit 21 acquires the first observation value OB1 observed by the first observation system 3 and the second observation value OB2 observed by the second observation system 4 (ST01: acquisition step).

[0275] After the implementation of ST01, the observation system constraint determination unit 22 determines, based on the time series data DT1 of each observation value of the first observation value OB1 and the second observation value OB2, the in-constraint data DT2 that is data within the first constraint R1 (ST02: observation system constraint determination step).

[0276] After the implementation of ST02, the model estimation unit 23 estimates, based on the in-constraint data DT2, the parameters PR of a plurality of models including the observation model MLO that is a model of each observation system and the physical model MLP that is a model within the device 1 provided with the observation system (ST03: model estimation step).

[0277] After the implementation of ST03, the matching determination unit 24 determines the matching of the model based on the deviation of the first predicted observation value PA1 predicted from the second observation value OB2 from the first observation value OB1 according to the parameters PR of the plurality of models estimated by the model estimation unit 23 (ST04: matching determination step).

[0278] For example, in ST04, the matching determination unit 24 can determine that there is an abnormality in the observation model or the physical model in a case where the deviation is larger than a prescribed value (ST04A). At this time, the matching determination unit 24 can delete a record that exceeds the model even if optimized.

[0279] For example, after the implementation of ST04A, the matching determination unit 24 can determine the non-matching data among the in-constraint data DT2 and exclude it from the in-constraint data DT2 as abnormal data (ST04B).

[0280] For example, after the implementation of ST04, the physical quantity constraint determination unit 25 can determine whether the physical quantity estimated value PV estimated based on the parameters PR is within the second constraint R2 (ST05: physical quantity constraint determination step).

[0281] ​For example, in parallel with ST05, the model constraint determination unit 26 determines whether each parameter PR is within the third constraint R3 (ST06: model constraint determination step).

[0282] For example, after the execution of ST04 and ST05, the output unit 27 can output the parameter PR determined to be normal and the physical quantity estimated value PV estimated based on the parameter PR, and output the parameter PR determined to be abnormal and the physical quantity estimated value PV estimated based on the parameter PR (ST07: output step).

[0283] (Function of abnormality diagnosis device)

[0284] In the abnormality diagnosis device 5 according to the present embodiment, the estimated value acquisition unit 55 acquires the physical quantity estimated value PV of the device 1 from the estimation device 2, unlike the third embodiment. That is, the estimated value acquisition unit 55 of the abnormality diagnosis device 5 according to the present embodiment acquires the physical quantity estimated value PV estimated by the estimation device 2 as the state quantity of the device 1 in step S122 of Figure 13

[0285] Further, the abnormality determination unit 52 of the abnormality diagnosis device 5 according to the present embodiment can further determine the presence or absence of abnormality of the device 1 based on the determination results of the observation system constraint determination unit 22 and the physical quantity constraint determination unit 25 of the estimation device 2. For example, in step S123 of Figure 13 Figure 13

[0286] (Action effect)

[0287] As described above, in the abnormality diagnosis device 5 according to the present embodiment, the estimated value acquisition unit 55 acquires the physical quantity estimated value PV of the device 1 estimated by the estimation device 2 as the state quantity of the device 1.

[0288] Further, the estimation device 2 according to the present embodiment can estimate more correct physical quantity estimated values PV because it can estimate parameters PR of a plurality of models including the observation model MLO and the physical model MLP and perform optimization.

[0289] As such, by using correct physical quantity estimated values PV, the abnormality diagnosis device 5 can further improve the accuracy of estimating abnormality causes.​​​

[0290] Computer Hardware Architecture

[0291] Furthermore, in the above embodiments, programs for implementing various functions of the anomaly diagnosis device 5 and the estimation device 2 are recorded on a computer-readable recording medium. A computer system such as a microcomputer reads and executes the program recorded on the recording medium to perform various processes. Here, the various processing procedures of the computer system's CPU are stored in the form of a program on the computer-readable recording medium. By having the computer read and execute the program, the aforementioned various processes are performed. Furthermore, a computer-readable recording medium refers to a magnetic disk, optical disk, CD-ROM, DVD-ROM, semiconductor memory, etc. Additionally, the computer program can be distributed to a computer via a communication line, and the computer receiving the distribution can execute the program.

[0292] In the above embodiments, examples of the hardware structure of the computer that executes programs for implementing various functions of the anomaly diagnosis device 5 and the estimation device 2 will be described.

[0293] like Figure 20 As shown, the computer 29 of the anomaly diagnosis device 5 and the estimation device 2 respectively includes a CPU 291, a memory 292, a storage / reproduction device 293, an input / output interface (hereinafter referred to as "IO I / F") 294 and a communication interface (hereinafter referred to as "communication I / F") 295.

[0294] The memory 292 is a medium such as a random access memory (hereinafter referred to as "RAM") that temporarily stores data used in the programs executed in the abnormality diagnosis device 5 and the estimation device 2, respectively.

[0295] Storage / reproduction device 293 is a device for storing data to or reproducing data from external media such as CD-ROM, DVD, and flash memory.

[0296] IO I / F294 is an interface for inputting and outputting information, etc., between the anomaly diagnosis device 5 and the estimation device 2 and other devices.

[0297] The communication I / F295 is an interface for communication between the estimated device 2 and other devices via communication lines such as the Internet or dedicated communication lines.

[0298] <Other Implementation Methods>

[0299] The above describes embodiments of the present disclosure, but these embodiments are suggested as examples and are not intended to limit the scope of the disclosure. These embodiments can be implemented in other various forms, and various omissions, substitutions, and changes can be made without departing from the scope of the disclosure. These embodiments and variations thereof are included in the scope and spirit of the disclosure.

[0300] <Notes>

[0301] The abnormality diagnosing apparatus, the abnormality diagnosing method, and the program described in the above embodiments are grasped, for example, as follows.

[0302] (1) The abnormality diagnosing apparatus 5 of the first mode has: an abnormality determining section 52 that determines presence or absence of abnormality for each of state quantities acquired from the device 1; and a cause inferring section 53 that infers a cause of abnormality of the device 1 from the state quantity determined to have abnormality by the abnormality determining section 52, using a cause correspondence table in which a cause of an abnormality mode of the device 1 determined by fault tree analysis and the state quantity that becomes abnormal if the cause is generated are established in correspondence.

[0303] In this way, the abnormality diagnosing apparatus 5 can easily and with good precision infer the cause of abnormality of the device 1 based on the cause correspondence table.

[0304] (2) The abnormality diagnosing apparatus 5 of the second mode further has, on the basis of the abnormality diagnosing apparatus 5 of (1), an observation value acquiring section 51 that acquires an observation value observed by an observation system of the device 1 as the state quantity, and the abnormality determining section 52 compares the observation value acquired by the observation value acquiring section 51 and a normal value or a normal mode of each of the observation values to determine presence or absence of abnormality of the device 1.

[0305] In this way, the abnormality diagnosing apparatus 5 can easily and with good precision perform inference of the cause of abnormality, for example, by the observation value observed by the existing observation system of the device 1 and the cause correspondence table.

[0306] (3) The abnormality diagnosing apparatus 5 of the third mode further has, on the basis of the abnormality diagnosing apparatus 5 of (1) or (2), the abnormality determining section 52 that further determines an abnormality degree corresponding to a difference from a constraint set for each of the state quantities for each of the state quantities, and the cause inferring section 53 that infers the cause of abnormality of the device 1 using the cause correspondence table in which the cause and the abnormality degree of the state quantity that becomes abnormal if the cause is generated are further established in correspondence.

[0307] In this way, the abnormality diagnosing apparatus 5 can more finely narrow down the cause by determining the abnormality degree in a plurality of stages instead of dividing each observation value into two levels of “normal” or “abnormal”.

[0308] (4) The abnormality diagnosing apparatus 5 of the fourth aspect further includes, in addition to the abnormality diagnosing apparatus 5 of the second aspect, a physical value acquisition unit 55 that acquires a physical value PV1, PV2 of the device 1 estimated based on the observation value A1 acquired by the observation value acquisition unit 51, as the state quantity.

[0309] In this way, the abnormality diagnosing apparatus 5 can determine the presence or absence of an abnormality and estimate the cause of the abnormality using the physical value PV1, PV2 estimated for the internal state quantity of the device 1 in addition to the observation value A1 acquired from the first observation system 3 of the device 1. Thus, the abnormality diagnosing apparatus 5 can improve the detection accuracy for an abnormality pattern that is difficult to detect only from the change in the observation value.

[0310] (5) The abnormality diagnosing method of the fifth aspect includes the steps of determining the presence or absence of an abnormality for each of the state quantities acquired from the device 1, and estimating the cause of the abnormality of the device 1 from the state quantity determined to have an abnormality, using a cause correspondence table in which the cause of the abnormality pattern of the device 1 determined by fault tree analysis and the state quantity that becomes abnormal when the cause is generated are associated with each other.

[0311] (6) The program of the sixth aspect causes the computer 29 of the abnormality diagnosing apparatus 5 to execute the steps of determining the presence or absence of an abnormality for each of the state quantities acquired from the device 1, and estimating the cause of the abnormality of the device 1 from the state quantity determined to have an abnormality, using a cause correspondence table in which the cause of the abnormality pattern of the device 1 determined by fault tree analysis and the state quantity that becomes abnormal when the cause is generated are associated with each other.

[0312] Industrial applicability

[0313] According to the abnormality diagnosing apparatus, the abnormality diagnosing method, and the program according to the present disclosure, it is possible to easily and accurately estimate the cause of the abnormality of the device.

[0314] Explanation of reference signs

[0315] 2 Estimation apparatus

[0316] 21 Acquisition unit

[0317] 22 Observation system constraint determination unit

[0318] 23 Model estimation unit

[0319] 24 Matching determination unit

[0320] 25 Physical value constraint determination unit

[0321] 26 Model constraint determination unit

[0322] 27 output section

[0323] 29 computer

[0324] 3 first observation system

[0325] 31 first sensor

[0326] 32 second sensor

[0327] 33 third sensor

[0328] 4 second observation system

[0329] 41 second sensor

[0330] 42 third sensor

[0331] 43 fourth sensor

[0332] 5 abnormality diagnosing device

[0333] 51 observation value obtaining section

[0334] 52 abnormality determining section

[0335] 53 factor inferring section

[0336] 54 notifying section

[0337] 55 inferred value obtaining section

Claims

1. An abnormality diagnostic device, comprising: The anomaly detection unit determines whether there are any anomalies in the status variables obtained from the equipment; and The cause estimation unit establishes a corresponding cause mapping table using the causes of the abnormal mode of the device determined by fault tree analysis and the state quantities that become abnormal when the causes occur. Based on the state quantities that are determined to be abnormal by the abnormality determination unit, the cause of the device's abnormality is estimated. The anomaly determination unit further determines the degree of anomaly corresponding to the difference between the state quantity and the constraint set for each state quantity. The cause estimation unit uses a cause correspondence table to further establish the cause and the degree of abnormality of the state quantity that becomes abnormal when the cause occurs, in order to estimate the cause of the device's abnormality.

2. The abnormality diagnosis device according to claim 1, wherein, The abnormality diagnostic device also includes: The observation acquisition unit acquires the observation values ​​observed by the observation system of the device, and uses them as the state variables. The anomaly determination unit compares the observed value obtained by the observed value acquisition unit with the normal value or normal mode of the observed value to determine whether the device is abnormal.

3. The abnormality diagnosis device according to claim 1, wherein, The abnormality diagnostic device also includes: An observation acquisition unit acquires observations obtained by the observation system of the device; and The estimation value acquisition unit acquires an estimated value of the physical quantity of the device based on the observation value acquired by the observation value acquisition unit, and uses it as the state quantity.

4. An abnormality diagnosis method, comprising: Step 1: Determine whether there are any anomalies in the status variables obtained from the device; and Step 2 involves establishing a corresponding cause-and-effect table using the causes of the device's abnormal patterns determined through fault tree analysis and the state variables that become abnormal when the causes occur. The causes of the device's abnormality are then inferred based on the state variables that are determined to be abnormal. In the first step, the degree of abnormality corresponding to the difference between the state quantity and the constraint set for each state quantity is further determined. The second step uses a corresponding factor mapping table to further establish the factor and the degree of abnormality of the state quantity that becomes abnormal when the factor occurs, in order to presume the factor of the device's abnormality.

5. A non-transitory computer-readable recording medium storing a program that causes a computer of an anomaly diagnostic device to execute: Step 1: Determine whether there are any anomalies in the status variables obtained from the device; and Step 2 involves establishing a corresponding cause-and-effect table using the causes of the device's abnormal patterns determined through fault tree analysis and the state variables that become abnormal when the causes occur. The causes of the device's abnormality are then inferred based on the state variables that are determined to be abnormal. In the first step, the degree of abnormality corresponding to the difference between the state quantity and the constraint set for each state quantity is further determined. The second step further establishes a corresponding factor correspondence table for the factors and the degree of abnormality of the state quantity that becomes abnormal when the factors occur, in order to predetermine the factors of the device's abnormality.

Images