Complete plant monitoring device, complete plant monitoring method, and recording medium

Abstract

An acquisition unit acquires a bundle of detection values ​​for each of a plurality of sensor values ​​associated with a plant. A distance calculation unit calculates the Mahalanobis distance of the bundle of detection values ​​acquired by the acquisition unit, using a unit space formed by clustering the bundle of detection values ​​for each of the plurality of sensor values ​​as a reference. A determination unit determines whether the operating state of the plant is normal or abnormal based on whether the Mahalanobis distance is within a predetermined threshold. A trend determination unit determines a trend for at least one sensor value. An abnormality cause estimation unit estimates the cause of the abnormality based on the trend of each of the at least one sensor value and a fault location estimation database storing relationships between a plurality of abnormality causes that may occur in the plant and a plurality of sensor values ​​corresponding to each trend.

Description

Technical Field

[0001] The present invention relates to a plant monitoring device for monitoring the operating status of a plant, a plant monitoring method, and a computer-readable recording medium having a program recorded thereon.

[0002] This application claims priority based on Japanese Patent Application No. 2020-102395, filed in Japan on June 12, 2020, and incorporates the contents herein. Background Art

[0003] In various plants such as gas turbine power plants, nuclear power plants, and chemical plants, in order to monitor the normal operation of the plants, the status values ​​of various sensors in the plants, such as temperature and pressure, are acquired. The operating status of the plants is monitored based on these status values. However, in the event of an abnormality, the function of inferring the cause is required.

[0004] For example, the monitoring device disclosed in Patent Document 1 below acquires the status values ​​of each sensor in the plant online from the plant's computer and uses the Mahalanobis-Taguchi method (hereinafter referred to as the MT method) to determine if an abnormality exists. This monitoring device has the function of identifying the cause of the abnormality if it is determined to be abnormal.

[0005] In the MT method, a unit space is prepared in advance, which is composed of multiple state quantity bundles. This state quantity bundle is an aggregation of the state quantities of each of the multiple sensors. If the state quantity bundle is obtained from a complete set of equipment, the Mahalanobis distance (hereinafter referred to as MD distance) of the state quantity bundle is calculated based on the unit space. The normality of the operating status of the complete set of equipment is determined based on whether the Mahalanobis distance is within a pre-set threshold.

[0006] Previous technical literature

[0007] Patent Literature

[0008] Patent Document 1: Japanese Patent Application Laid-Open No. 2017-215863 Summary of the Invention

[0009] Technical issues to be solved by the invention

[0010] The MT method calculates the expected maximum SN ratio for each sensor based on the Mahalanobis distance, identifying sensors with increased Mahalanobis distances. With the conventional MT method, when focusing on a sensor, the Mahalanobis distance increases regardless of whether the sensor's value is high or low, making it impossible to distinguish between abnormally high and low values.

[0011] The object of the present invention is to provide a more reliable method for preparing a database for estimating a fault location and a method for estimating the cause of an abnormality by adding information on the causes of abnormalities that are likely to occur and those that are unlikely to occur under high / low values.

[0012] Means for solving technical problems

[0013] According to one embodiment, a complete plant monitoring device includes: an acquisition unit that acquires a bundle of detection values ​​of each sensor value among a plurality of sensor values ​​related to the complete plant; a distance calculation unit that calculates the Mahalanobis distance of the bundle of detection values ​​acquired by the acquisition unit based on a unit space formed by gathering the bundle of detection values ​​of each sensor value among the plurality of sensor values; a determination unit that determines whether the operating state of the complete plant is normal or abnormal based on whether the Mahalanobis distance is within a prescribed threshold value; and a determination unit that determines, for each sensor value, whether a high SN ratio is generated by a high value or a low value.

[0014] According to one method, a fault location inference database is implemented by adding information on whether a specific abnormality cause is more likely to occur due to a high value abnormality of the sensor or a low value abnormality, and whether the abnormality cause is less likely to occur due to a high value abnormality of the sensor value or a low value abnormality.

[0015] According to one method, as a database for distinguishing high-value anomalies / low-value anomalies of the SN ratio and inferring the fault location of the present invention, a more reliable abnormality cause can be inferred by combining information on whether a specific abnormality cause is likely to occur or not due to a high-value anomaly / low-value anomaly of the sensor value.

[0016] Effects of the Invention

[0017] According to at least one of the above aspects, the plant monitoring device can more reliably estimate the true cause of a failure by setting information on an event less likely to be caused by an abnormality in a sensor to a negative value. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a diagram for explaining the outline of the plant monitoring device according to the first embodiment.

[0019] Figure 2 This is a schematic block diagram showing the functional configuration of the plant monitoring device according to the first embodiment.

[0020] Figure 3 This is a diagram showing an example of the failure location estimation database according to the first embodiment.

[0021] Figure 4 This is a conceptual diagram showing the concept of Mahalanobis distance.

[0022] Figure 5 This is a flowchart showing a method for updating the failure location estimation database according to the first embodiment.

[0023] Figure 6 This is a flowchart showing the monitoring process of the plant according to the first embodiment.

[0024] Figure 7 This is a schematic block diagram showing the structure of a computer according to at least one embodiment. DETAILED DESCRIPTION

[0025] <First embodiment>

[0026] Figure 1 This is a diagram for explaining the outline of the plant monitoring device 20 according to the first embodiment.

[0027] The plant monitoring device 20 according to this embodiment is used to monitor the operating status of a plant 1, which has multiple evaluation items. The plant monitoring device 20 acquires detection values ​​representing status quantities for each evaluation item from detectors installed in various parts of the plant 1. The plant monitoring device 20 then uses the Mahalanobis-Taguchi method to determine whether the operating status of the plant 1 is normal or abnormal based on the acquired detection values.

[0028] Structure of complete equipment

[0029] The plant 1 according to this embodiment is a gas turbine combined power generation plant, comprising a gas turbine 10, a gas turbine generator 11, a waste heat recovery boiler 12, a steam turbine 13, a steam turbine generator 14, and a control device 40. In other embodiments, the plant 1 may also be a gas turbine power generation plant, a nuclear power generation plant, or a chemical plant.

[0030] The gas turbine 10 includes a compressor 101 , a combustor 102 , and a turbine 103 .

[0031] Compressor 101 compresses air drawn in through an air intake. Temperature sensors 101A and 101B are provided on compressor 101 as sensors for detecting the temperature within the compressor compartment (one of the evaluation parameters). For example, temperature sensor 101A can detect the temperature at the compressor compartment inlet (inlet air temperature), while temperature sensor 101B can detect the temperature at the compressor compartment outlet (outlet air temperature).

[0032] The combustor 102 mixes fuel F with compressed air introduced from the compressor 101 and burns the mixture to generate combustion gas. The combustor 102 is provided with a pressure sensor 102A as a detector for detecting the pressure of the fuel F (one of the evaluation parameters).

[0033] The turbine 103 is driven by combustion gas supplied from the combustor 102. Temperature sensors 103A and 103B are provided on the turbine 103 as sensors for detecting the temperature within the turbine chamber (one of the evaluation parameters). For example, the temperature sensor 103A can detect the temperature at the turbine chamber inlet (the inlet combustion gas temperature), while the temperature sensor 103B can detect the temperature at the turbine chamber outlet (the outlet combustion gas temperature).

[0034] The gas turbine generator 11 is connected to the rotor of the turbine 103 via the compressor 101 and generates electricity by utilizing the rotation of the rotor. The gas turbine generator 11 is provided with a thermometer 11A as a detector for detecting the temperature of the lubricating oil (one of the evaluation items).

[0035] The waste heat recovery boiler 12 generates steam by heating water using combustion gas (exhaust gas) discharged from the turbine 103. The waste heat recovery boiler 12 is provided with a liquid level gauge 12A as a detector for detecting the water level of the steam drum (one of the evaluation items).

[0036] The steam turbine 13 is driven by steam from the waste heat recovery boiler 12. A temperature sensor 13A is installed in the steam turbine 13 to detect the temperature within the turbine compartment (one of the evaluation parameters). The steam discharged from the steam turbine 13 is converted back into water by a condenser 132 and then supplied to the waste heat recovery boiler 12 via a water supply pump.

[0037] The steam turbine generator 14 is coupled to the rotor 131 of the steam turbine 13 and generates electricity by the rotation of the rotor 131. The steam turbine generator 14 is provided with a thermometer 14A as a detector for detecting the temperature of the lubricating oil (one of the evaluation items).

[0038] The above evaluation items are merely examples and are not limited to these. Other evaluation items for the plant 1 may include, for example, the output of the gas turbine generator 11, the pressure within the turbine 103 casing, the rotational speed and vibration of the turbine 103 or steam turbine 13 rotors, etc. In this case, detectors (not shown) that detect the state quantities of these evaluation items are installed in various parts of the plant 1.

[0039] The control device 40 is a device for controlling the operation of the plant 1. Furthermore, if the plant monitoring device 20 determines that the operating state of the plant 1 is abnormal, the control device 40 can control the operation of each component of the plant 1 based on a control signal from the plant monitoring device 20.

[0040] "structure"

[0041] Figure 2 This is a schematic block diagram showing the functional configuration of the plant monitoring device 20 according to the first embodiment.

[0042] The complete equipment monitoring device 20 includes a sensor value acquisition unit 201, a unit space storage unit 202, an MD distance calculation unit 203, a complete equipment abnormality determination unit 204, a maximum SN ratio calculation unit 205, an abnormal sensor extraction unit 206, a high value abnormality / low value abnormality determination unit 207, a fault location inference database 208, an abnormality cause inference unit 209, and an abnormality cause output display unit 210.

[0043] The sensor value acquisition unit 201 acquires detection values ​​from each of the multiple detectors installed in the plant 1. Each detector corresponds to a plurality of evaluation items. Specifically, the sensor value acquisition unit 201 acquires a bundle of detection values, which is an aggregation of the detection values ​​for each of the multiple evaluation items. The sensor value acquisition unit 201 acquires the bundle of detection values ​​at a predetermined acquisition cycle (e.g., one minute) and stores it in the unit space storage unit.

[0044] The unit space storage unit 202 stores combinations of bundles of detection values ​​acquired from normal plants as unit spaces of Mahalanobis distances.

[0045] The MD distance calculation unit 203 calculates the Mahalanobis distance representing the state of the plant 1 based on the unit space stored in the unit space storage unit 202, using the bundle of detection values ​​acquired by the sensor value acquisition unit 201 as factors. The Mahalanobis distance is a measure that indicates the magnitude of the difference between a reference sample and a newly acquired sample, expressed in terms of a unit space.

[0046] The plant abnormality determination unit 204 determines whether an abnormality has occurred in the plant 1 based on the Mahalanobis distance calculated by the MD distance calculation unit 203. Specifically, if the Mahalanobis distance is greater than or equal to a predetermined threshold, the plant abnormality determination unit 204 determines that an abnormality has occurred in the plant 1. The threshold is typically set to a value of 3 or greater.

[0047] If the plant abnormality determination unit 204 determines that an abnormality has occurred in the gas turbine T, the expected SN ratio calculation unit 205 calculates the expected SN ratio (Signal-Noise Ratio) in accordance with the Taguchi method based on the bundle of detection values ​​acquired by the sensor value acquisition unit 201. For example, the expected SN ratio calculation unit 205 calculates the expected SN ratio for the presence or absence of an item using an orthogonal table analysis. The larger the expected SN ratio, the higher the probability that an abnormality exists in the evaluation item associated with the detection value.

[0048] The abnormal sensor extraction unit 206 extracts at least one abnormal sensor as a sensor value that contributes significantly to an increase in the Mahalanobis distance based on the expected SN ratio calculated by the expected SN ratio calculation unit 205. For example, the abnormal sensor extraction unit 206 may extract a predetermined number of sensor values ​​ranked high in the expected SN ratio from among the plurality of sensor values ​​as abnormal sensors. Alternatively, the abnormal sensor extraction unit 206 may extract sensor values ​​whose expected SN ratios are greater than a predetermined threshold value from among the plurality of sensor values ​​as abnormal sensors.

[0049] The high-value anomaly / low-value anomaly determination unit 207 determines, for each of the multiple sensor values, whether the abnormality that has occurred is a high-value anomaly (an abnormality that occurs due to a high detection value of the sensor value) or a low-value anomaly (an abnormality that occurs due to a low detection value). That is, the high-value anomaly / low-value anomaly determination unit 207 determines whether the increase in the Mahalanobis distance has occurred due to an increase in the detection value or a decrease in the detection value. Specifically, the high-value anomaly / low-value anomaly determination unit 207 calculates the Mahalanobis distance when the value of the bundle of detection values ​​acquired by the sensor value acquisition unit 201 is increased or decreased for each sensor value, and determines whether it is a high-value anomaly or a low-value anomaly based on the increase or decrease in the Mahalanobis distance caused by the change in the value. In the case where the Mahalanobis distance increases due to an increase in the detection value, it can be determined that the sensor value is a high-value anomaly. In the case where the Mahalanobis distance increases due to a decrease in the detection value, it can be determined that the sensor value is a low-value anomaly. (Japanese Patent Application No. 2019-063575)

[0050] The fault location estimation database 208 is a fault location estimation database that indicates the relationship between evaluation items, abnormality causes, and high-value abnormalities / low-value abnormalities. Figure 3 : is a diagram showing an example of a fault location estimation database according to the first embodiment. Specifically, the fault location estimation database includes evaluation items ( Figure 3 The vertical axis) and each abnormal cause ( Figure 3The horizontal axis (the horizontal axis) stores the amount of information regarding the abnormality associated with the evaluation item observed when the abnormality cause occurs. As the value of the information amount increases, the more likely the same abnormality is observed in the associated evaluation item. The amount of information stored in the fault location estimation database 208 regarding the actual occurrence of high-value abnormalities or low-value abnormalities is expressed, for example, by the following equation (1).

[0051] [Formula 1]

[0052] I=log2[∑(x*w)+1] / log2(2)…(1)

[0053] Here, I represents the amount of information, x represents the number of events that occurred, and w represents a weight coefficient based on the reliability of the data.

[0054] For example, the weighting factor w can be larger when the cause of an actual abnormality has occurred and has been identified based on a report, etc., than when the cause of the abnormality has been identified based on FTA data (FT: Fault Tree) generated by maintenance personnel. Furthermore, the weighting factor w can be larger when the cause of the abnormality has been identified based on a method with higher accuracy than a report, such as offline analysis or simulation.

[0055] On the other hand, the cause that is less likely to occur when a sensor abnormality occurs is represented by, for example, the following equation (2), and is a negative value.

[0056] [Formula 2]

[0057] I=log2[∑(x*w) / {1-∑(x*w)}+1] / log2(2)…(2)

[0058] The weight w used for calculation of the amount of information related to a high-value anomaly or a low-value anomaly that did not actually occur may be greater than the weight w used for calculation of the amount of information related to a high-value anomaly or a low-value anomaly that actually occurred.

[0059] The abnormality cause inference unit 209 generates an M*2 row, N column matrix based on the fault location inference database. (Here, the M*2 portion is doubled to distinguish between high-value anomalies and low-value anomalies.) The fault location inference database associates information with M evaluation items and high-value anomalies and low-value anomalies. Therefore, the abnormality cause inference unit 209 generates an M*2 row, N column matrix by reading the information associated with each of the M evaluation items in the high-value anomaly / low-value anomaly determination unit 207. The abnormality cause inference unit 209 then multiplies a 1-row, M*2 column vector based on the expected SN ratio of each evaluation item by the generated M*2 row, N column matrix to obtain an N-row, 1-column vector based on the probability of the abnormal cause. The abnormality cause inference unit 209 then infers that the abnormality cause associated with the row with the largest element value in the obtained N-row, 1-column vector is the abnormality cause that occurred in the complete plant 1. That is, the abnormality cause estimation unit 209 calculates a weighted sum of the expected SN ratio of each evaluation item and the amount of information related to the abnormality of that item for each abnormality cause, and estimates the abnormality cause based on the weighted sum.

[0060] The abnormality cause output display unit 210 outputs the abnormality cause estimated by the abnormality cause estimation unit 209 in order of probability. Examples of output include display on a display, external data transmission, printing on paper, and audio output.

[0061] About the MT Law

[0062] Figure 4 This is a conceptual diagram showing the concept of Mahalanobis distance.

[0063] First, use Figure 4 The outline of the plant monitoring method based on the MT method is described.

[0064] like Figure 4 As shown, it is assumed that the sensor value acquisition unit 201 of the plant monitoring device 20 acquires the first and second detection values ​​of the plant 1 as a detection value bundle B. For example, the first detection value is "gas turbine output" and the second detection value is "boiler water level." In the MT method, the Mahalanobis distance D of the detection value bundle A acquired at a certain point in time is calculated, using a data set (a collection of multiple detection value bundles B) as a unit space S (reference data set).

[0065] The Mahalanobis distance D is a distance weighted by the dispersion or correlation of the detection values ​​within the unit space S, and increases with decreasing similarity with the data set within the unit space S. Here, the average Mahalanobis distance of the bundle B of detection values ​​constituting the unit space S is 1. When the operating state of the plant 1 is normal, the Mahalanobis distance D of the bundle A of detection values ​​is generally 4 or less. However, if the operating state of the plant 1 is abnormal, the value of the Mahalanobis distance D increases depending on the degree of abnormality.

[0066] Therefore, in the MT method, whether the operating status of the plant 1 is normal or abnormal is determined based on whether the Mahalanobis distance D is within a preset threshold value Dc. For example, if the Mahalanobis distance D1 of the detection value bundle A1 is below the threshold value Dc, the operating status of the plant 1 at the time the detection value bundle A1 was acquired is determined to be normal. Furthermore, if the Mahalanobis distance D2 of the detection value bundle A2 is greater than the threshold value Dc, the operating status of the plant 1 at the time the detection value bundle A2 was acquired is determined to be abnormal.

[0067] The threshold Dc is preferably set to a value greater than the maximum Mahalanobis distance among the Mahalanobis distances between the plurality of detection value bundles B constituting the unit space S. Furthermore, the threshold Dc is preferably set in consideration of the inherent characteristics of the plant 1. The threshold Dc can be changed by personnel via the plant monitoring device 20.

[0068] <Operation of Plant Monitoring Device 20>

[0069] Next, the operation of the plant monitoring device 20 will be described.

[0070] Before starting monitoring processing, the plant monitoring device 20 collects bundles of detection values ​​from the plant 1 while the plant 1 is operating normally, and accumulates the bundles of detection values ​​in the unit space storage unit 202. Furthermore, the plant monitoring device 20 may also obtain bundles of detection values ​​from other plants 1 having the same structure as the monitored plant 1 when operating normally, and record these bundles in the unit space storage unit 202.

[0071] (Monitoring Process of Plant 1)

[0072] When the unit space is recorded in the unit space storage unit 202 and the fault location estimation database is recorded in the fault location estimation database 208 , the plant monitoring device 20 executes the following monitoring process at a predetermined monitoring timing (for example, every hour).

[0073] Figure 6This is a flowchart showing the monitoring process of the plant 1 according to the first embodiment.

[0074] When the plant monitoring device 20 starts monitoring, the sensor value acquisition unit 201 acquires a bundle of detection values ​​from the plant 1 (step S31). The MD distance calculation unit 203 calculates the Mahalanobis distance based on the unit space stored in the unit space storage unit 202, using the bundle of detection values ​​acquired in step S31 as factors (step S32).

[0075] Next, the plant abnormality determination unit 204 determines whether an abnormality has occurred in the plant 1 based on the Mahalanobis distance calculated in step S32 (step S33). If the plant abnormality determination unit 204 determines that no abnormality has occurred in the plant 1 (step S33: No), the plant monitoring device 20 ends the monitoring process and waits for the next monitoring timing.

[0076] On the other hand, when the plant abnormality determination unit 204 determines that an abnormality has occurred in the plant 1 (step S33: Yes), the Wangda SN ratio calculation unit 205 calculates the Wangda SN ratio related to the Taguchi method for each evaluation item based on the bundle of detection values ​​obtained in step S31 and the Mahalanobis distance calculated in step S32 (step S34).

[0077] Next, the plant monitoring device 20 selects evaluation items one by one, and performs the following processing from step S36 to step S41 on each evaluation item (step S35 ).

[0078] First, the abnormally high / abnormally low value determination unit 207 increases the sensor value selected in step S35 by a predetermined amount in the bundle of detection values ​​acquired in step S31 (step S36). Next, the MD distance calculation unit 203 calculates the Mahalanobis distance based on the unit space stored in the unit space storage unit 202, using the bundle of detection values ​​modified in step S36 as factors (step S37).

[0079] The abnormal high / low value determination unit 207 determines whether the Mahalanobis distance has increased, decreased, or remained unchanged due to an increase in the detection value associated with the abnormal sensor (step S38). For example, the abnormal high / low value determination unit 207 may determine that the Mahalanobis distance has not changed when the difference in the Mahalanobis distance is less than a predetermined threshold.

[0080] If the Mahalanobis distance increases (step S38: Increase), the high-value anomaly / low-value anomaly determination unit 207 determines that the abnormal sensor extracted in step S35 has a high-value anomaly (step S39). On the other hand, if the Mahalanobis distance decreases (step S38: Decrease), the high-value anomaly / low-value anomaly determination unit 207 determines that the abnormal sensor extracted in step S35 has a low-value anomaly (step S40). If the Mahalanobis distance does not change (step S38: No Change), the high-value anomaly / low-value anomaly determination unit 207 determines that the abnormal sensor cannot be classified as an abnormal sensor extracted in step S35 (step S41).

[0081] The abnormal cause inference unit 209 uses the fault location inference database 208 to generate a matrix of M*2 rows and N columns (step S42). The abnormal cause inference unit 209 obtains a vector of N rows and 1 column with the probability of the abnormal cause as an element by multiplying the vector of 1 row and M*2 columns including the expected SN ratio of each evaluation item calculated in step S34 and whether it is a high-value abnormality or a low-value abnormality with the matrix of M*2 rows and N columns generated in step S42 (step S43). In addition, the items of the expected SN ratio that are considered unclassifiable are set to 0. Next, the abnormal cause inference unit 209 sorts each abnormal cause in descending order of the probability represented by the obtained vector (step S44). At this time, the abnormal cause inference unit 209 sets it to a negative number when it is less likely to occur than usual. Then, the abnormal cause output display unit 210 outputs the abnormal cause inferred by the abnormal cause inference unit 209 in the sorted order (step S45). For example, the error cause output and display unit 210 displays the error cause with the highest probability on the display. Upon receiving a user's instruction to display the next error cause, the error cause with the next highest probability is displayed on the display. Furthermore, for example, the error cause output and display unit 210 prints a list of the error causes in descending order of probability onto paper.

[0082] Function and Effect

[0083] Thus, according to the first embodiment, when the complete plant monitoring device 20 determines that an abnormality exists based on the Mahalanobis distance, it infers the cause of the abnormality based on the abnormality of each sensor value and the fault location inference database, wherein the fault location inference database stores the relationship between multiple abnormal causes that may occur in the complete plant 1 and the multiple sensor values ​​of each abnormality.

[0084] Thus, the plant monitoring device 20 can estimate the cause of the abnormality by classifying each sensor value as either high-value abnormality or low-value abnormality. Therefore, the plant monitoring device 20 can exclude events with low probability of occurrence from the estimated results of the abnormality cause.

[0085] Furthermore, the fault location estimation database according to the first embodiment stores information indicating the increased / decreased likelihood of the cause of an abnormality, by associating it with the cause and the high / low abnormal sensor values. Furthermore, the plant monitoring device 20 calculates, for each of the multiple sensor values, a value obtained by multiplying the information associated with the abnormality identified in the fault location estimation database by the expected SN ratio associated with that sensor value, and estimates the cause of the abnormality based on the sum of these values. This increases the probability of abnormal causes with a high information volume associated with sensor values ​​with a high expected SN ratio, while decreasing the probability of abnormal causes with a low information volume associated with sensor values ​​with a high expected SN ratio. Consequently, the plant monitoring device 20 can eliminate events with a low probability of occurrence from the estimated cause of the abnormality.

[0086] In addition, in other embodiments, this is not limited to the above. For example, the complete plant monitoring device 20 involved in other embodiments can also obtain an N-row, 1-column vector with the probability of the abnormal cause of the abnormality as an element by calculating the cosine similarity between a 1-row, M*2-column vector with the expected SN ratio of each sensor value as an element and an M*2-row, N-column matrix with the value of the fault location estimation database as an element. In addition, the cosine similarity is the value obtained by dividing the inner product of the vector (the weighted sum of each expected SN ratio and the amount of information related to the abnormal cause) by the product of the norm of each vector. For example, the complete plant monitoring device 20 involved in other embodiments can also obtain the weighted sum of the expected SN ratio of each sensor value and the amount of information related to the abnormal cause according to the abnormal cause without relying on matrix calculation.

[0087] Furthermore, the fault location estimation database according to the first embodiment stores positive information by associating the cause of an abnormality with sensor value anomalies that are highly likely to occur when the cause of the abnormality occurs. On the other hand, the fault location estimation database according to the first embodiment stores negative information by associating the cause of an abnormality with sensor value anomalies that are highly likely not to occur when the cause of the abnormality occurs. This allows the plant monitoring device 20 to actively reduce the probability of a highly likely abnormal cause not occurring. Consequently, the plant monitoring device 20 can eliminate events with a low probability of occurrence from the results of the abnormality cause estimation.

[0088] In other embodiments, this is not a limitation. For example, the fault location estimation database in other embodiments may store zero information in a manner that associates an abnormality cause with an abnormality sensor value that is highly likely not to occur when the abnormality cause occurs. In this case, while the probability of the abnormality cause is not significantly reduced compared to a case with negative information, by classifying abnormalities in each sensor value and inferring the abnormality cause, it is possible to eliminate events with a low probability of occurrence from the estimated abnormality cause.

[0089] Furthermore, the plant monitoring device 20 according to the first embodiment updates the fault location estimation database based on the bundle of detection values ​​when the cause of the abnormality occurs, increasing the amount of information associated with confirmed abnormalities for each sensor value and decreasing the amount of information associated with unconfirmed abnormalities. This allows the plant monitoring device 20 to automatically generate a fault location estimation database containing information related to the opposite direction. Furthermore, in other embodiments, this is not limiting, and negative information can also be manually entered by personnel.

[0090] Furthermore, the plant monitoring device 20 according to the first embodiment updates the information volume of at least one abnormal sensor having a large expected SN ratio among a plurality of sensor values. Thus, the plant monitoring device 20 can increase the clarity of the information volume of each sensor value in the fault location estimation database.

[0091] While one embodiment has been described in detail above with reference to the accompanying drawings, the specific structure is not limited to the above description and various design changes may be made. Specifically, in other embodiments, the order of the above processes may be appropriately changed. Furthermore, some processes may be performed simultaneously.

[0092] <Computer Architecture>

[0093] Figure 7 This is a schematic block diagram showing the structure of a computer according to at least one embodiment.

[0094] The computer 90 includes a processor 91 , a main memory 92 , a storage device 93 , and an interface 94 .

[0095] The plant monitoring device 20 is installed in a computer 90. The operations of each of the processing units are stored in the form of a program in a storage device 93. A processor 91 reads the program from the storage device 93, expands it in the main memory 92, and executes the processing described above according to the program. Furthermore, the processor 91 reserves storage areas corresponding to each of the storage units in the main memory 92 according to the program. Examples of the processor 91 include a CPU (Central Processing Unit), a GPU (Graphic Processing Unit), and a microprocessor.

[0096] The program can also be used to implement a part of the functions that the computer 90 performs. For example, the program can also perform functions by combining with other programs stored in a storage device or with other programs installed in other devices. In addition, in other embodiments, the computer 90 may also have a customized LSI (Large Scale Integrated Circuit) such as a PLD (Programmable Logic Device) in addition to the above structure, or a customized LSI (Large Scale Integrated Circuit) such as a PLD (Programmable Logic Device) instead of the above structure. Examples of PLDs include PAL (Programmable Array Logic), GAL (Generic Array Logic), CPLD (Complex Programmable Logi c Device), and FPGA (Field Programmable Gate Array). In this case, part or all of the functions implemented by the processor 91 can be implemented by the integrated circuit. Such an integrated circuit is also included in an example of a processor.

[0097] Examples of the storage device 93 include a HDD (Hard Disk Drive), an SSD (Solid State Drive), a magnetic disk, an optical magnetic disk, a CD-ROM (Compact Disc Read Only Memory), a DVD-ROM (Digital Versatile Disc Read Only Memory), and a semiconductor memory. The storage device 93 may be an internal medium directly connected to the bus of the computer 90, or an external medium connected to the computer 90 via an interface 94 or a communication line. Furthermore, when the program is distributed to the computer 90 via a communication line, the computer 90 that receives the distribution may expand the program in the main memory 92 to execute the above-mentioned processing. In at least one embodiment, the storage device 93 is a non-temporary tangible storage medium.

[0098] Furthermore, the program may be used to realize a part of the above functions. Furthermore, the program may be a so-called differential file (differential program) that realizes the above functions by combining with other programs stored in the storage device 93 .

[0099] The plant monitoring device 20 according to the above embodiment may be composed of a single computer 90 , or the structure of the plant monitoring device 20 may be distributed among a plurality of computers 90 , and the plurality of computers 90 may cooperate with each other to perform the functions of the plant monitoring device 20 .

[0100] <Note>

[0101] The plant monitoring device, the plant monitoring method, and the program described in each embodiment can be understood, for example, as follows.

[0102] (1) According to the first embodiment, a plant monitoring device (20) comprises: a sensor value acquisition unit (201) for acquiring a bundle of detection values ​​of each sensor value among a plurality of sensor values ​​related to the plant (1); a distance calculation unit (203) for calculating the Mahalanobis distance of the bundle of detection values ​​acquired by the sensor value acquisition unit (201) based on a unit space formed by gathering the bundle of detection values ​​of each sensor value among the plurality of sensor values; a plant abnormality determination unit (204) for determining whether the operating state of the plant (1) is normal or abnormal based on whether the Mahalanobis distance is within a predetermined threshold value; and a high value abnormality / low value abnormality determination unit (207). , in the case where it is determined that the operating state of the complete set of equipment is abnormal, it is determined whether at least one sensor value in the beam of detection values ​​that is inferred to be the cause is a high-value abnormality or a low-value abnormality, the high-value abnormality being an abnormality caused by a high detection value, and the low-value abnormality being an abnormality caused by a low detection value; an abnormality cause inference unit (209), for each sensor value in the at least one sensor value, infers the abnormality cause based on whether it is a low-value abnormality or a high-value abnormality and a fault location inference database, the fault location inference database stores the relationship between multiple abnormal causes that may occur in the complete set of equipment and the multiple sensor values ​​of each trend; and an output unit (210), outputs the inferred abnormal cause.

[0103] Thus, the plant monitoring device can estimate the cause of the abnormality by classifying each sensor value as either high-value abnormality or low-value abnormality. Therefore, the plant monitoring device can eliminate events with low probability of occurrence from the estimated results of the abnormality cause.

[0104] "Acquire" means to obtain a new value. Examples of "acquire" include receiving a value, accepting input of a value, reading a value from a storage medium, and calculating another value based on a value.

[0105] "Determining" means using the first value to determine a second value from a range of possible values. For example, "determining" includes calculating the second value based on the first value, retrieving the second value corresponding to the first value from a fault location estimation database, searching for the second value using the first value as a search request, and selecting the second value from multiple candidates based on the first value.

[0106] (2) According to the second embodiment, the complete equipment monitoring device (20) involved in the first embodiment can be equipped with an expected SN ratio calculation unit (205), which calculates the expected SN ratio of the multiple sensor values ​​based on the beam of the detection values, and the fault location inference database can store the amount of information indicating the increase / decrease in the possibility of the occurrence of the abnormal cause in a manner associated with the abnormal cause and the sensor value. The abnormal cause inference unit (209) can calculate, for each of the multiple sensor values, a value obtained by multiplying the amount of information associated with the low-value abnormality / high-value abnormality determined for the sensor value in the fault location inference database and the expected SN ratio related to the sensor value, and infer the abnormal cause based on the sum of the obtained values.

[0107] This increases the probability of abnormal causes with a high sensor value containing a large amount of information, while decreasing the probability of abnormal causes with a low amount of information. Consequently, the plant monitoring device can eliminate events with a low probability of occurrence from the estimated abnormal cause.

[0108] (3) According to the third method, in the complete equipment monitoring device (20) involved in the first or second method, in the fault location inference database, the positive amount of information can be associated with the abnormal cause and the abnormality that is more likely to occur when the abnormal cause occurs among the low-value abnormality and the high-value abnormality, and the negative amount of information can be associated with the abnormal cause and the abnormality that is more likely not to occur when the abnormal cause occurs among the low-value abnormality and the high-value abnormality.

[0109] As a result, the plant monitoring device can actively reduce the probability that a highly likely abnormality cause will not occur. Therefore, the plant monitoring device can eliminate events with a low probability of occurrence from the results of the inference of the abnormality cause.

[0110] (4) According to the fourth embodiment, in the complete equipment monitoring device involved in the third embodiment, in the fault location inference database, among low-value anomalies and high-value anomalies, the absolute value of the amount of information associated with anomalies that are more likely not to occur when the abnormal cause occurs may be greater than the absolute value of the amount of information associated with anomalies that are more likely to occur when the abnormal cause occurs.

[0111] (5) According to the fifth aspect, a program is used to cause a computer to execute the following steps: a step of acquiring a bundle of detection values ​​of each of a plurality of sensor values ​​associated with a plant; a step of calculating a Mahalanobis distance of the bundle of detection values ​​acquired by the acquisition unit based on a unit space formed by clustering the bundle of detection values ​​of each of the plurality of sensor values; a step of determining whether the operating state of the plant is normal or abnormal based on whether the Mahalanobis distance is within a predetermined threshold value; a step of determining whether at least one sensor value in the bundle of detection values ​​that is presumed to be a cause is a high-value abnormality or a low-value abnormality when it is determined that the operating state of the plant is abnormal, the high-value abnormality being an abnormality caused by a high detection value, and the low-value abnormality being an abnormality caused by a low detection value; a step of presuming a cause of the abnormality based on whether the cause is a low-value abnormality or a high-value abnormality and a fault location presumption database for each of the at least one sensor value, the fault location presumption database storing relationships between a plurality of abnormal causes that may occur in the plant and the plurality of sensor values; and a step of outputting the presumed cause of the abnormality.

[0112] Industrial applicability

[0113] The plant monitoring device can more reliably estimate the true cause of a failure by setting information on an event with a low probability of being caused by an abnormality in a sensor to a negative value.

[0114] Explanation of symbols

[0115] 1-complete equipment, 20-complete equipment monitoring device, 201-sensor value acquisition unit, 202-unit space storage unit, 203-MD distance calculation unit, 204-complete equipment abnormality determination unit, 205-Wangda SN ratio calculation unit, 206-abnormal sensor extraction unit, 207-high value abnormality / low value abnormality determination unit, 208-fault location inference database, 209-abnormality cause inference unit, 210-abnormality cause output display unit.

Claims

1. A complete plant monitoring device comprising: a sensor value acquisition unit that acquires a bundle of detection values ​​of each of a plurality of sensor values ​​related to the plant; a distance calculation unit that calculates a Mahalanobis distance of the acquired bundle of detection values ​​with reference to a unit space formed by clustering a bundle of detection values ​​of each of the plurality of sensor values; a plant abnormality determination unit for determining whether the operating state of the plant is normal or abnormal based on whether the Mahalanobis distance is within a predetermined threshold; a high-value abnormality / low-value abnormality determination unit that, when determining that the operating state of the plant is abnormal, determines whether at least one sensor value in the bundle of detection values ​​that is presumed to be a cause is a high-value abnormality or a low-value abnormality, wherein the high-value abnormality is an abnormality occurring due to a high detection value, and the low-value abnormality is an abnormality occurring due to a low detection value; an abnormality cause inferring unit that, for each of the at least one sensor value, infers the abnormality cause based on whether the value is a low-value abnormality or a high-value abnormality and a fault location inferring database, wherein the fault location inferring database stores relationships between a plurality of abnormality causes that may occur in the equipment and the plurality of sensor values; and an output unit for outputting the inferred cause of the abnormality, The abnormality cause inference unit calculates, for each of the multiple sensor values, a value obtained by multiplying the amount of information associated with the low-value abnormality / high-value abnormality determined for the sensor value in the fault location inference database and the expected SN ratio related to the sensor value, and infers the abnormality cause based on the sum of the calculated values.

2. The plant monitoring device according to claim 1, comprising an SN ratio calculation unit, The SN ratio calculation unit calculates the SN ratio of the plurality of sensor values ​​based on the bundle of detection values, The failure location estimation database stores information indicating an increase or decrease in the probability of occurrence of the abnormality cause in association with the abnormality cause and the sensor value.

3. The plant monitoring device according to claim 1 or 2, wherein: In the fault location inference database, when a high-value anomaly / low-value anomaly occurs, the positive amount of information is associated with the anomaly whose cause is more likely to occur than usual, and the negative amount of information is associated with the anomaly whose cause is less likely to occur than usual.

4. A complete set of equipment monitoring method, comprising: The step of obtaining a bundle of detection values ​​for each of a plurality of sensor values ​​associated with the set of equipment; a step of determining the Mahalanobis distance of the acquired bundle of detection values ​​with reference to a unit space formed by clustering a bundle of detection values ​​of each of the plurality of sensor values; a step of determining whether the operating state of the complete set of equipment is normal or abnormal according to whether the Mahalanobis distance is within a prescribed threshold; If it is determined that the operating state of the plant is abnormal, determining whether at least one sensor value in the bundle of detection values ​​that is presumed to be a cause is a high-value abnormality or a low-value abnormality, wherein the high-value abnormality is an abnormality occurring due to a high detection value, and the low-value abnormality is an abnormality occurring due to a low detection value; a step of inferring a cause of the abnormality for each of the at least one sensor value based on whether the value is low or high and a fault location inference database, wherein the fault location inference database stores relationships between a plurality of possible abnormal causes in the equipment and the plurality of sensor values; and The step of outputting the inferred cause of the abnormality, In the step of inferring the cause of the abnormality, for each of the multiple sensor values, a value is obtained by multiplying the amount of information associated with the low-value abnormality / high-value abnormality determined for the sensor value in the fault location inference database and the expected SN ratio related to the sensor value, and the cause of the abnormality is inferred based on the sum of the obtained values.

5. A computer-readable recording medium having a program recorded thereon, the program causing a computer to execute the following steps: The step of obtaining a bundle of detection values ​​for each of a plurality of sensor values ​​associated with the set of equipment; a step of determining the Mahalanobis distance of the acquired bundle of detection values ​​with reference to a unit space formed by clustering a bundle of detection values ​​of each of the plurality of sensor values; a step of determining whether the operating state of the complete set of equipment is normal or abnormal according to whether the Mahalanobis distance is within a prescribed threshold; If it is determined that the operating state of the plant is abnormal, determining whether at least one sensor value in the bundle of detection values ​​that is presumed to be a cause is a high-value abnormality or a low-value abnormality, wherein the high-value abnormality is an abnormality occurring due to a high detection value, and the low-value abnormality is an abnormality occurring due to a low detection value; a step of inferring a cause of the abnormality for each of the at least one sensor value based on whether the value is low or high and a fault location inference database, wherein the fault location inference database stores relationships between a plurality of possible abnormal causes in the equipment and the plurality of sensor values; and The step of outputting the inferred cause of the abnormality, In the step of inferring the cause of the abnormality, for each of the multiple sensor values, a value is obtained by multiplying the amount of information associated with the low-value abnormality / high-value abnormality determined for the sensor value in the fault location inference database and the expected SN ratio related to the sensor value, and the cause of the abnormality is inferred based on the sum of the obtained values.

Images