Monitoring method, method for calculating signal-to-noise ratio gain, monitoring device, and program

Abstract

Provided is a method for calculating an SN ratio gain, in which the magnitude level of the SN ratio gain based on an orthogonal array size for each unit space in a multi-MT method is corrected. This method for calculating an SN ratio gain involves creating a plurality of unit spaces by using data detected by a plurality of sensors provided to a plant, and evaluating the state of the plant using the unit spaces by a multi-MT method, and includes: a step for allocating, for each of the unit spaces, one of the sensors used for creating the unit space to a 2-level orthogonal array, and generating orthogonal arrays for the respective unit spaces; a step for calculating, for each of the orthogonal arrays for the respective unit spaces, an SN ratio for each line of the orthogonal array; and a step for calculating, for each of the sensors in the orthogonal arrays for the respective unit spaces, an SN ratio gain by calculating a difference between a total value of SN ratios based on a larger-is-better response in a first level and a total value of SN ratios based on a larger-is-better response in a second level.

Description

Technical Field

[0001] The present disclosure relates to a method for calculating the signal-to-noise gain in the Mahalanobis-Taguchi method, a method for monitoring a plant or the like using the signal-to-noise gain calculated by the calculation method, a monitoring device, and a program. The present disclosure claims priority based on Japanese Patent Application No. 2022-012655 filed in Japan on January 31, 2022, and incorporates the content herein by reference.

Background Art

[0002] Patent Document 1 discloses a monitoring device that acquires detection data representing the state of a plant detected by a plurality of sensors provided in the plant to be monitored, calculates the Mahalanobis distance based on the unit space in the MT method (Mahalanobis-Taguchi method), and determines that the operating state of the plant is abnormal if the Mahalanobis distance is greater than or equal to a threshold value. When this monitoring device determines that the operating state of the plant is abnormal, it calculates the signal-to-noise gain using an orthogonal array for each sensor, and identifies the sensor with a large signal-to-noise gain value as an item related to the cause of the abnormality. As disclosed in Patent Document 2 (paragraphs 0105 to 0107, etc.), generally, the signal-to-noise gain of a two-level orthogonal array is calculated by the difference between the average value of the signal-to-noise ratio of the first level (using that item) and the average value of the signal-to-noise ratio of the second level (not using that item) for each item. Further, Patent Document 1 divides the blocks at the time of plant startup and rated load operation, creates unit spaces using the detection data detected by different sensor groups for each block, and discloses a multi-method of the MT method (also called the multi-MT method, multi-level MT method, split-combination method, etc.) for performing monitoring based on the Mahalanobis distance from each unit space.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Patent Document 2

Summary of the Invention

[0004] The signal-to-noise ratio (SNR) gain calculated using conventional methods is susceptible to the size of the orthogonal array; for example, a larger orthogonal array tends to result in a lower SNR gain. When using the multi-MT method, a threshold for the SNR gain common to all blocks is sometimes set for operation. In this case, if the orthogonal array sizes of each block differ, the threshold will be set to match the block with the highest SNR gain (the block with the smallest orthogonal array size). However, setting the threshold in this way tends to result in false detections for blocks with small orthogonal array sizes and detection delays or missed detections for blocks with large orthogonal array sizes.

[0005] This disclosure provides a monitoring method, a method for calculating signal-to-noise ratio gain, a monitoring device, and a program that can solve the above-mentioned problems. [Means for solving the problem]

[0006] The monitoring method of this disclosure includes the steps of: acquiring unit space creation data for creating a plurality of unit spaces predetermined according to the operating mode and / or the object to be monitored, detected by a plurality of sensors installed in the plant; creating a plurality of unit spaces based on the unit space creation data detected by the sensors necessary for creating each of the unit spaces; acquiring evaluation data, which is a collection of data for evaluating the state of the plant, detected by the plurality of sensors; determining the Mahalanobis distance of the evaluation data based on at least a portion of the plurality of unit spaces; determining the state of the plant based on the Mahalanobis distance and a predetermined threshold; and, if an abnormality is determined in the step of determining the state of the plant, estimating the cause of the abnormality. The steps for estimating the cause of the abnormality include: creating an orthogonal array for each unit space used to determine the state, assigning each of the sensors used to create the unit space to a two-level orthogonal array, with the first level being the use of the sensor and the second level being the non-use of the sensor; calculating the maximum signal-to-noise ratio (SNR) for each row of the orthogonal array for each unit space; calculating the SNR gain for each sensor in each of the orthogonal arrays for each unit space by calculating the difference between the sum of the maximum SNRs for the first level and the sum of the maximum SNRs for the second level; and identifying the sensor whose calculated SNR gain exceeds the threshold for the SNR gain as the sensor related to the cause of the abnormality, based on the SNR gain calculated for each sensor and a predetermined threshold for the SNR gain.

[0007] The method for calculating the signal-to-noise ratio gain of the present disclosure includes the steps of: creating a plurality of unit spaces based on data detected by a plurality of sensors installed in a plant, and evaluating the state of the plant by the multi-MT method based on at least a portion of the plurality of unit spaces, and for each of the unit spaces, creating an orthogonal array for each unit space by assigning each of the sensors used to create the unit space to a two-level orthogonal array, with the first level being the use of the sensor and the second level being the non-use of the sensor; calculating the maximum signal-to-noise ratio for each row of the orthogonal array for each of the unit spaces; and calculating the signal-to-noise ratio gain for each sensor in each of the orthogonal arrays for each unit space by calculating the difference between the sum of the maximum signal-to-noise ratios at the first level and the sum of the maximum signal-to-noise ratios at the second level.

[0008] The monitoring device of this disclosure includes means for acquiring unit space creation data for creating a plurality of unit spaces predetermined according to the operating mode and / or the object to be monitored, detected by a plurality of sensors installed in the plant; means for creating a plurality of unit spaces based on the unit space creation data detected by the sensors necessary for creating each of the unit spaces; means for acquiring evaluation data, which is a collection of data for evaluating the state of the plant, detected by the plurality of sensors; means for determining the Mahalanobis distance of the evaluation data based on at least a portion of the plurality of unit spaces; means for determining the state of the plant based on the Mahalanobis distance and a predetermined threshold; and means for estimating the cause of the abnormality if an abnormality is determined in the step of determining the state of the plant. The means for estimating the cause of the abnormality includes, for each unit space used to determine the state, assigning each of the sensors used to create the unit space to a two-level orthogonal array, with using the sensor being a first level and not using the sensor being a second level, thereby creating an orthogonal array for each unit space; calculating the desired signal-to-noise ratio for each row of the orthogonal array for each unit space; calculating the signal-to-noise ratio gain for each sensor in each orthogonal array for each unit space by calculating the difference between the sum of the desired signal-to-noise ratios at the first level and the sum of the desired signal-to-noise ratios at the second level; and identifying the sensor whose calculated signal-to-noise ratio gain exceeds the threshold for signal-to-noise ratio gain as the sensor related to the cause of the abnormality, based on the signal-to-noise ratio gain calculated for each sensor and a predetermined threshold for signal-to-noise ratio gain.

[0009] The program of this disclosure includes the steps of: obtaining unit space creation data for creating a plurality of unit spaces predetermined according to the operating mode and / or the object being monitored, detected by a plurality of sensors installed in the plant; creating a plurality of unit spaces based on the unit space creation data detected by the sensors necessary for creating each of the unit spaces; obtaining evaluation data which is a collection of data for evaluating the state of the plant, detected by the plurality of sensors; determining the Mahalanobis distance of the evaluation data based on at least a portion of the plurality of unit spaces; determining the state of the plant based on the Mahalanobis distance and a predetermined threshold; and, if an abnormality is determined in the step of determining the state of the plant, estimating the cause of the abnormality. The step of estimating the cause of the abnormality includes the steps of: creating an orthogonal array for each unit space used to determine the state, assigning each of the sensors used to create the unit space to a two-level orthogonal array, with the first level being the use of the sensor and the second level being the non-use of the sensor; calculating the maximum signal-to-noise ratio (SNR) for each row of the orthogonal array for each unit space; calculating the SNR gain for each sensor in each of the orthogonal arrays for each unit space by calculating the difference between the sum of the maximum SNRs for the first level and the sum of the maximum SNRs for the second level; and identifying the sensor whose calculated SNR gain exceeds the threshold for the SNR gain, based on the SNR gain calculated for each sensor and a predetermined threshold for the SNR gain, as the sensor related to the cause of the abnormality. [Effects of the Invention]

[0010] According to the monitoring method, signal-to-noise ratio gain calculation method, monitoring device, and program described above, it is possible to calculate a signal-to-noise ratio gain that is less affected by the size of the orthogonal array. According to the monitoring method, monitoring device, and program described above, it is possible to estimate the cause of anomalies early and with high accuracy. [Brief explanation of the drawing]

[0011] [Figure 1] This figure shows an example of a monitoring system according to the embodiment. [Figure 2] This figure shows an example of a two-level orthogonal array. [Figure 3] This figure illustrates the relationship between the magnitude of the signal-to-noise ratio gain and the size of the orthogonal array according to the embodiment. [Figure 4A] This figure shows the general relationship between signal-to-noise ratio gain and threshold in the multi-MT method. [Figure 4B] This figure shows the relationship between the corrected signal-to-noise ratio gain and the threshold according to the embodiment. [Figure 5A] This is a flowchart showing an example of plant monitoring processing according to the embodiment. [Figure 5B] This flowchart shows an example of the calculation process for the corrected signal-to-noise ratio gain according to the embodiment. [Figure 6] This figure shows an example of the hardware configuration of the monitoring device according to the embodiment. [Modes for carrying out the invention]

[0012] The monitoring method and signal-to-noise ratio gain calculation method of this disclosure will be described below with reference to Figures 1 to 6. In the following description, components having the same or similar functions will be denoted by the same reference numerals. Duplication of these components may be omitted.

[0013] (Configuration of the monitoring system) FIG. 1 is a diagram showing an example of a monitoring system according to an embodiment. The monitoring system 10 shown in FIG. 1 includes a plant 20 to be monitored and a monitoring device 30. The plant 20 has machines 21 such as gas turbines and generators, and sensors 1 to n for detecting the state of the machines 21 and the environment in which the machines 21 operate. The plant 20 and the monitoring device 30 are communicably connected by a network or the like. For example, the detection data detected by the sensors 1 to n is transmitted to the monitoring device 30 in real time, and the monitoring device 30 acquires these detection data. The monitoring device 30 has a data acquisition unit 31, a unit space creation unit 32, a state evaluation unit 33, a cause estimation unit 34, an output unit 35, and a storage unit 36.

[0014] The data acquisition unit 31 acquires the detection data detected by the sensors 1 to n. For example, when the facility 21 includes a gas turbine and a generator, the sensors 1 to n are sensors for detecting the temperature, pressure, vibration, rotational speed of each part of the gas turbine, the output of the generator, and the like. The data acquisition unit 31 acquires these detection data and stores them in the storage unit 36. The detection data acquired by the data acquisition unit 31 includes unit space creation data for creating a unit space in the MT (Mahalanobis-Taguchi) method and evaluation data for evaluating the operating state of the plant 20. The unit space creation data is, for example, detection data detected by the sensors 1 to n when the plant 20 was operating normally in the past. The evaluation data is detection data representing the current state of the plant 20 detected by the sensors 1 to n of the plant 20 during operation.

[0015] The unit space creation unit 32 creates a unit space for calculating the Mahalanobis distance using the unit space creation data acquired by the data acquisition unit 31. A unit space is a collection of data used as a criterion for determining whether the operating state of the plant 20 is normal or not. The method for creating a unit space is well known, so it is omitted from this specification. When monitoring the plant 20 using the multi-MT method, the unit space creation unit 32 creates a unit space for each operating mode and monitoring target of the plant 20, for example. The operating mode includes different operating forms of the plant 20, such as the magnitude of the operating load, starting, and stopping. For example, regarding the operating mode of the plant 20, the unit space creation unit 32 uses the unit space creation data detected by at least some of the sensors 1 to n when the plant 20 starts up normally to create a unit space for determining whether the operating state of the plant 20 during startup is normal or not. For example, the unit space creation unit 32 uses unit space creation data detected by at least some of sensors 1 to n when the plant 20 is operating normally at rated load to create a unit space (referred to as the rated load operation unit space) for determining whether the operating state of the plant 20 at rated load is normal. Similarly, the unit space creation unit 32 may create unit spaces for each operating load of the plant 20. Regarding the evaluation target, if the plant 20 includes a gas turbine, the unit space creation unit 32 uses unit space creation data measured by sensors 1 to n necessary for evaluating the monitored target for each monitored part or state quantity such as blade path temperature, bearings, disk cavity temperature, shaft vibration, and filters to create a unit space for each monitored target. The examples given here, such as startup, rated load operation, blade path temperature, bearings, disk cavity temperature, shaft vibration, and filters, are examples of blocks (groups) in the multi-MT method. The unit space for each block is created using unit space creation data detected by different types of sensor groups. The unit space creation unit 32 stores each created unit space in the storage unit 36.

[0016] The state evaluation unit 33 determines whether the state of the plant 20 is normal based on the unit space created by the unit space creation unit 32 and the evaluation data acquired by the data acquisition unit 31. For example, the state evaluation unit 33 extracts detection data detected by the same type of sensors as those used for creating the unit space from the evaluation data, and calculates the Mahalanobis distance from the unit space of the extracted detection data group (an aggregate of data). The Mahalanobis distance is a measure representing the magnitude of the difference between a reference sample represented as a unit space and a newly obtained sample (the extracted detection data group). Since the method for calculating the Mahalanobis distance is well-known, the description thereof is omitted in this specification. For example, when the sensors used for creating the unit space for the rated load operation of the plant 20 are sensors 1 to 11, the state evaluation unit 33 extracts the detection data group detected by sensors 1 to 11 from the evaluation data acquired by the data acquisition unit 31 when evaluating the state of the plant 20 during the rated load operation, and calculates the Mahalanobis distance between the extracted detection data group and the unit space for the rated load operation. Next, the state evaluation unit 33 determines whether an abnormality has occurred in the plant 20 during the rated load operation based on the calculated Mahalanobis distance. Specifically, when the Mahalanobis distance is less than or equal to a predetermined threshold value, the state evaluation unit 33 determines that the state of the plant 20 is normal, and when the Mahalanobis distance exceeds the threshold value, the state evaluation unit 33 determines that the state of the plant 20 is abnormal. The state evaluation unit 33 determines the state of the plant 20 using the unit spaces of a plurality of blocks necessary for monitoring. For example, during the rated load operation, the state evaluation unit 33 determines the state of the plant 20 based on the unit space for the rated load operation and the unit space for each monitoring target.

[0017] The factor estimation unit 34 estimates the cause of the abnormality when the state evaluation unit 33 determines that the operating state of the plant 20 is abnormal. The factor estimation unit 34 has an SN ratio gain calculation unit 341. The SN ratio gain calculation unit 341 calculates the SN ratio of the maximum-size characteristic based on the orthogonal array in the MT method and calculates the maximum SN ratio gain from the SN ratio of the maximum-size characteristic. At this time, the SN ratio gain calculation unit 341 calculates a maximum SN ratio gain that is less affected by the difference in size of the orthogonal arrays between blocks compared to the general maximum SN ratio gain. Hereinafter, the maximum SN ratio gain that is less affected by the difference in size of the orthogonal arrays between blocks will be referred to as the "corrected SN ratio gain". The SN ratio of the maximum-size characteristic may be referred to as the maximum SN ratio, and the general maximum SN ratio gain may be referred to as the SN ratio gain. The calculation method of the corrected SN ratio gain and the difference between the corrected SN ratio gain and the general SN ratio gain will be explained later using Figures 2 to 4B. The factor estimation unit 34 compares the corrected signal-to-noise ratio gain calculated by the signal-to-noise ratio gain calculation unit 341 with a predetermined alarm threshold. If the corrected signal-to-noise ratio gain exceeds the alarm threshold, it estimates the orthogonal array item (sensor) corresponding to that corrected signal-to-noise ratio gain as the cause of the abnormality. As disclosed in Patent Document 1, the signal-to-noise ratio gain calculated based on an orthogonal array has the property of being larger for items that are abnormal, and by checking the item with a large signal-to-noise ratio gain, the cause of the abnormality can be identified. The same applies to the "corrected signal-to-noise ratio gain" in this embodiment.

[0018] The output unit 35 outputs the judgment result of the condition evaluation unit 33 regarding the operating status of the plant 20, and the cause of the abnormality estimated by the cause estimation unit 34. Examples of output include display on a screen, output to an electronic file, transmission of data to an external source, printing on paper or a sheet, and audio output.

[0019] The storage unit 36 ​​stores computer programs and data for realizing the monitoring method of the plant 20 according to this embodiment, data acquired by the data acquisition unit 31, and unit spaces created by the unit space creation unit 32. The storage unit 36 ​​may be provided outside the monitoring device 30, and configured so that the monitoring device 30 can access the storage unit 36 ​​via a communication line.

[0020] (Method for calculating corrected signal-to-noise ratio gain) Next, with reference to Figure 2, the method for calculating the corrected signal-to-noise ratio gain according to this embodiment will be described. Figure 2 shows an example of a two-level orthogonal array. Each column of the orthogonal array illustrated in Figure 2 is called an item, and each item is assigned a factor to be analyzed. The possible values ​​that each item can take are called levels, and in a two-level system, each item can take either a value of "use that item" (level 1) or "do not use that item" (level 2). The values ​​1 and 2 in each cell of the orthogonal array mean that 1 means "use that item" and 2 means "do not use that item". In this embodiment, the factors are sensors, and each column is assigned one of sensors 1 to n. In the example in Figure 2, sensors 1 to 11 are assigned. Each row of the orthogonal array is a number (experiment number) assigned to each combination of factors. For example, the first row (No=1) of the orthogonal array in Figure 2 shows the combination when all sensors 1 to 11 are used, and the second row (No=2) shows the combination when only sensors 3, 7, 8, 9, and 11 are used. In the multi-MT method, orthogonal arrays are created for each block. The orthogonal array shown in Figure 2 is an example of an orthogonal array created for a block in which a unit space was created using the unit space creation data detected by sensors 1 to 11.

[0021] In the MT method, the signal-to-noise ratio (SNR) is calculated for each row. The desired SNR is calculated using the following formula (1). η = -10·log{(1 / D1 2 +1 / D2 2 +···+1 / D m 2 ) / m} ...(1) Here, D x 2(x=1~m) is the square of the Mahalanobis distance (MD), and m is the number of data points. For example, suppose detection data from sensors 1 to 11 can be acquired at a 1-minute cycle. If the optimal signal-to-noise ratio (SNR) is calculated using the three data points going back from the most recently acquired detection data, then m=3. If the optimal SNR is calculated using only the most recently acquired detection data, then m=1. The SNR gain calculation unit 341 calculates the optimal SNRs η1 to η12 for each row of the orthogonal array using the above equation (1).

[0022] Next, the signal-to-noise ratio gain calculation unit 341 calculates the corrected signal-to-noise ratio gain for each of the sensors 1 to 11 using the desired signal-to-noise ratios η1 to η12. The corrected signal-to-noise ratio gain is calculated by the difference between the sum of the desired signal-to-noise ratios of the first level and the sum of the desired signal-to-noise ratios of the second level, which are calculated for each row of the orthogonal array of the two-level system. For example, the corrected signal-to-noise ratio gain for sensor 1 is calculated by the following equation (2). Corrected signal-to-noise ratio gain of sensor 1 = (η1 + η3 + η5 + η6 + η7 + η11) - (η2 + η4 + η8 + η9 + η10 + η12) ... (2) Here, η1, η3, η5, η6, η7, and η11 are the maximum signal-to-noise ratios calculated for rows where sensor 1 is at level 1, and η2, η4, η8, η9, η10, and η12 are the maximum signal-to-noise ratios calculated for rows where sensor 1 is at level 2.

[0023] Similarly, the corrected signal-to-noise ratio gain of sensor 2 is calculated by (η1+η4+η6+η7+η8+η12)-(η2+η3+η5+η9+η10+η11), and the corrected signal-to-noise ratio gain of sensor 11 is calculated by (η1+η2+η4+η5+η6+η10)-(η3+η7+η8+η9+η11+η12). The same applies to the other sensors 3, etc.

[0024] For comparison, let's explain the general signal-to-noise ratio (SNR) gain. The general SNR gain is calculated by the difference between the average of the maximum SNR values ​​for the first level and the average of the maximum SNR values ​​for the second level, among the maximum SNR values ​​calculated for each row of a two-level orthogonal array. For example, the corrected SNR gain of sensor 1 is calculated by the following equation (2'). The signal-to-noise ratio gain of sensor 1 = ((η1+η3+η5+η6+η7+η11)÷6) - ((η2+η4+η8+η9+η10+η12)÷6)···(2')

[0025] As described above, in the general calculation of signal-to-noise ratio (SNR) gain, the average value of the desired SNR at the first and second levels is used, so in the example above, it is divided by "6". This value "6" is related to the size of the orthogonal array (for example, in the case of Figure 2, the size of the orthogonal array can be "12", the number of rows) (number of rows ÷ 2), and the size of the orthogonal array has a positive correlation with the number of factors, i.e., the number of sensors. Figure 3 shows the relationship between the standard deviation of the general SNR gain and the size of the orthogonal array. In the graph of Figure 3, the vertical axis y represents the standard deviation of the general SNR gain, and the horizontal axis x represents the size of the orthogonal array. Each point in the graph of Figure 3 is a plot of the relationship between the size of the orthogonal array and the standard deviation of the SNR gain, calculated on paper for a block targeting a given group of sensors, when the size of the orthogonal array is changed (the number of sensors is changed). As a result of this analysis, it was confirmed that the size of the orthogonal array and the standard deviation of the SNR gain are inversely proportional, as shown in the figure. In general, the standard deviation tends to increase as the mean value of the data increases. Combining this general trend with the analysis results in Figure 3, we can deduce that a larger orthogonal array size tends to result in a smaller signal-to-noise ratio (SNR) gain, while a smaller orthogonal array size tends to result in a larger SNR gain. Since the size of the orthogonal array is positively correlated with the number of sensors, this finding can be rephrased as follows: a larger number of sensors used to create the unit space of a block tends to result in a smaller SNR gain calculated for the sensors in that block, while a smaller number of sensors tends to result in a larger SNR gain calculated for the sensors in that block.

[0026] Consequently, in each block configured according to the operating mode and monitoring target of Plant 20 as exemplified above, the number of sensors differs from block to block. Therefore, the signal-to-noise ratio (SNR) gain calculated using a general method for blocks with a large number of sensors tends to be smaller than the general SNR gain calculated for blocks with a small number of sensors. Figure 4A shows examples of general SNR gains calculated for blocks with a large number of sensors and blocks with a small number of sensors. The vertical axis of Figure 4A shows the magnitude of the SNR gain calculated using a general method, and the horizontal axis shows the blocks. Gb1 to Gb6 are the SNR gains for each block. For example, SNR gain Gb1 represents the SNR gain calculated for a sensor in block b1. Comparing block b1 and block b6 among blocks b1 to b6, block b1 has more sensors than block b6. As shown in the figure, as expected from the above findings, the SNR gain Gb1 is smaller than the SNR gain Gb6. Here, we assume that the same alarm threshold Th0 is set for blocks b1 to b6 and that the cause of the anomaly is estimated. In other words, it is presumed that sensors whose signal-to-noise ratio gain exceeds the alarm threshold Th0 are related to the cause of the anomaly. For example, for the sensor in block b1, the difference L1 between the alarm threshold Th0 and the signal-to-noise ratio gain Gb1 is excessive, making detection delays or missed detections more likely. For the sensor in block b6, the difference L1 between the alarm threshold Th0 and the signal-to-noise ratio gain Gb6 is insufficient, making false detections more likely.

[0027] In contrast, the corrected signal-to-noise ratio (SNR) gain according to this embodiment, which applies a correction to reduce the influence of differences in orthogonal array sizes between blocks, may reduce the difference in magnitude of SNR gain between blocks based on orthogonal array size. For example, if the number of rows in the orthogonal array is used as the size of the orthogonal array, and the size (number of rows) of the orthogonal array of block b1 is V1 and the size of the orthogonal array of block b6 is V6, then V1 > V6. In the calculation of the corrected SNR gain described above, the corrected SNR gain is calculated by the difference in the sum of the desired SNR values, without dividing by a value proportional to the orthogonal array size (number of rows ÷ 2). As a result, the corrected SNR gain for block b1 is SNR gain Gb1 × (V1 ÷ 2), and the corrected SNR gain for block b6 is SNR gain Gb6 × (V6 ÷ 2). The relationship between the corrected SNR gain Gb1' for block b1 and the corrected SNR gain Gb6' for block b6 may be improved compared to the relationship illustrated in Figure 4A.

[0028] Figure 4B shows an example of how the difference in the magnitude of the signal-to-noise ratio gain between blocks, caused by the effect of orthogonal array size as explained in Figure 3, can be improved by calculating a corrected signal-to-noise ratio gain instead of the general signal-to-noise ratio gain. The corrected signal-to-noise ratio gains Gb1' to Gb6' calculated for blocks b1 to b6 exemplified in Figure 4B are all of similar value. In such a case, if the same alarm threshold Th1 is set for blocks b1 to b6 and the cause of the anomaly is estimated, the difference between the alarm threshold Th1 and the corrected signal-to-noise ratio gain will be almost constant for all blocks, and the occurrence of detection delays and false detections can be suppressed.

[0029] Thus, by calculating the signal-to-noise ratio gain (corrected signal-to-noise ratio gain) using the difference between the sum of the desired signal-to-noise ratios at the first level and the sum of the desired signal-to-noise ratios at the second level, without calculating the average value by dividing by a value corresponding to the size (number of sensors) of the orthogonal array, if the signal-to-noise ratio gain levels between blocks can be made to be roughly the same, then by setting a common alarm threshold across all blocks in the multi-MT method, the cause of any anomaly can be identified early and accurately for any block.

[0030] (Modified S / N ratio gain) The method for calculating the corrected signal-to-noise ratio gain is not limited to the method described above. For example, it may be calculated as follows. That is, the signal-to-noise ratio gain calculation unit 341 calculates the corrected signal-to-noise ratio gain by multiplying the average value of the first level's maximum signal-to-noise ratio, which is calculated for each row of the two-level orthogonal array, by the number of sensors used to create the unit space corresponding to that orthogonal array, and by multiplying the average value of the second level's maximum signal-to-noise ratio, which is multiplied by the number of sensors used to create the unit space corresponding to that orthogonal array. For example, with respect to the correspondence table in Figure 2, the corrected signal-to-noise ratio gain of sensor 1 is calculated by the following equation (2). The corrected signal-to-noise ratio gain for sensor 1 = ((η1+η3+η5+η6+η7+η11)÷6×11)-((η2+η4+η8+η9+η10+η12)÷6×11)···(2'') where η1, η3, η5, η6, η7, η11 are the desired signal-to-noise ratios calculated for rows where sensor 1 is at the first level, and η2, η4, η8, η9, η10, η12 are the desired signal-to-noise ratios calculated for rows where sensor 1 is at the second level. In this way, when there are 11 sensors, the average value of the desired signal-to-noise ratios for the first and second levels is multiplied by "11". When there are 4 sensors used in other blocks, the signal-to-noise ratio gain calculated for the sensors in that block is multiplied by "4". This may reduce the difference in signal-to-noise ratio gain between the two blocks. A reference value may be set for the number of sensors, and the average value of the desired signal-to-noise ratios for the first and second levels may be multiplied by a value normalized by this reference value. For example, if the reference value for the number of sensors is set to 20, then for blocks with 11 sensors, 11÷20 is multiplied by the average value of the desired signal-to-noise ratios for the first and second levels instead of 11, and for blocks with 4 sensors, 4÷20 is multiplied.

[0031] (A modified example of how to set alarm thresholds) The signal-to-noise ratio gain can be calculated using a general method, and alarm thresholds may be set for each block. More specifically, alarm thresholds may be set for each block so as to have a negative correlation with the orthogonal array size of the block or the number of sensors used to create the unit space of the block. For example, alarm thresholds may be set so as to be inversely proportional to the orthogonal array size or the number of sensors of each block.

[0032] (operation) Next, with reference to Figures 5A and 5B, the flow of the monitoring process of the plant 20 by the monitoring device 30 will be described. Figure 5A is a flowchart showing an example of the plant monitoring process according to this embodiment. Prior to monitoring plant 20, a unit space to be used in the multi-MT method is created. The data acquisition unit 31 acquires previously detected data for creating unit spaces (step S10). The data acquisition unit 31 acquires data for creating unit spaces detected by sensors 1 to n when plant 20 is operating normally, and stores the acquired data in the storage unit 36. Next, the unit space creation unit 32 reads the data for creating unit spaces from the storage unit 36 ​​and creates multiple unit spaces (step S11). The unit space creation unit 32 creates unit spaces for each operating load (unit space for rated load, unit space for 50% load, etc.) according to the operating mode of plant 20, for example, when the plant is operating at a target operating load, using data for creating unit spaces detected by sensors necessary for evaluating plant 20 at that operating load. For example, the unit space creation unit 32 creates a unit space for each monitored object in the plant 20 using unit space creation data detected by sensors necessary for evaluating that object (unit space for blade path temperature, unit space for bearings, etc.).

[0033] Once multiple unit spaces are created, the monitoring device 30 begins monitoring the state of the plant 20. The data acquisition unit 31 acquires evaluation data detected by sensors 1 to n from the operating plant 20 (step S12). Next, the state evaluation unit 33 selects the unit spaces necessary for determining the state of the plant 20 and calculates the Mahalanobis distance (MD) between the evaluation data acquired in step S12 and each selected unit space (step S13). For example, if the plant 20 is currently operating at rated load, and it is stipulated that a unit space for rated load operation and unit spaces for each of the multiple monitoring targets are used for monitoring the plant 20 during rated load operation, and the unit space for rated load operation is created using unit space creation data detected by sensors 1 to 11, and among the unit spaces for each of the multiple monitoring targets, the unit space for blade path temperature is created using unit space creation data detected by sensors 12 to 40, then the state evaluation unit 33 calculates the Mahalanobis distance between the collection of evaluation data detected by sensors 1 to 11 from the evaluation data acquired by the data acquisition unit 31 and the unit space for rated load operation. Furthermore, the state evaluation unit 33 calculates the Mahalanobis distance between the collection of evaluation data detected by sensors 12-40 from the evaluation data acquired by the data acquisition unit 31 and the unit space for blade path temperature. The state evaluation unit 33 similarly calculates the Mahalanobis distance for the unit spaces of other monitored objects.

[0034] Next, the state evaluation unit 33 compares the Mahalanobis distance for each unit space calculated in step S13 with a predetermined threshold. If the Mahalanobis distance is less than or equal to the threshold (step S14; Yes), it determines that the state of plant 20 is normal (step S15). If the Mahalanobis distance exceeds the threshold (step S14; No), it determines that the state of plant 20 is abnormal (step S16).

[0035] If the state of plant 20 is determined to be abnormal, the signal-to-noise ratio gain calculation unit 341 calculates the corrected signal-to-noise ratio gain (step S17). The signal-to-noise ratio gain calculation unit 341 calculates the corrected signal-to-noise ratio gain for each unit space and each sensor using the method described with reference to Figure 2. The flow of this process will be explained next with reference to Figure 5B.

[0036] Next, the factor estimation unit 34 estimates the cause of the anomaly (step S18). For example, the factor estimation unit 34 compares a predetermined alarm threshold set in common for all unit spaces with the corrected signal-to-noise ratio gain for each unit space and sensor calculated in step S17, and extracts corrected signal-to-noise ratio gains that exceed the alarm threshold. If there are no corrected signal-to-noise ratio gains that exceed the alarm threshold, a predetermined number of corrected signal-to-noise ratio gains with values ​​close to the alarm threshold may be extracted in order. The factor estimation unit 34 estimates that the detection data detected by the sensor corresponding to the extracted corrected signal-to-noise ratio gain indicates the cause of the anomaly, and identifies the sensor as a sensor related to the cause of the anomaly.

[0037] Next, the output unit 35 outputs the determination result to a display device or the like (step S19). For example, if the plant status is determined to be normal, the output unit 35 outputs that the plant 20 is normal. If the plant status is determined to be abnormal, the output unit 35 outputs that the plant 20 is abnormal, as well as the sensors identified as being related to the cause of the abnormality and the detection data detected by those sensors.

[0038] Next, the calculation process for the corrected signal-to-noise ratio gain (step S17) will be explained with reference to Figure 5B. Figure 5B is a flowchart showing an example of the corrected signal-to-noise ratio gain calculation process according to the embodiment. First, the signal-to-noise ratio gain calculation unit 341 creates an orthogonal array for each unit space (each block) (step S20). The signal-to-noise ratio gain calculation unit 341 assigns each of the sensors used to create the unit space from among sensors 1 to n to each column of a two-level orthogonal array for each unit space used to determine the state of the plant 20, and creates an orthogonal array for each unit space. Next, the signal-to-noise ratio gain calculation unit 341 calculates the maximum signal-to-noise ratio for each row of the orthogonal array created for each unit space using the above formula (1) (step S21). Next, the signal-to-noise ratio gain calculation unit 341 calculates the corrected signal-to-noise ratio gain (step S22). For example, the SN ratio gain calculation unit 341 calculates the corrected SN ratio gain for each unit space and sensor by calculating the difference between the sum of the first level's maximum SN ratios in the orthogonal array of the unit space and the sum of the second level's maximum SN ratios in the unit space for each unit space and sensor. Alternatively, the SN ratio gain calculation unit 341 may calculate the corrected SN ratio gain for each unit space and sensor by calculating the difference between a value obtained by multiplying the average value of the first level's maximum SN ratio by a value corresponding to the number of sensors used to create the unit space (for example, the number of sensors or the number of sensors divided by a reference value) and a value obtained by multiplying the average value of the second level's maximum SN ratio by a value corresponding to the number of sensors used to create the unit space. This suppresses variations in the magnitude of the SN ratio gain due to differences in the number of sensors in each unit space and aims to make it more uniform.

[0039] (effect) As explained above, when estimating the cause of an anomaly using the signal-to-noise ratio gain based on an orthogonal array, variations in the magnitude of the signal-to-noise ratio gain occur for each of the multiple unit spaces due to differences in the number of sensors used to create the unit space. In contrast, the monitoring device 30 of this embodiment calculates the corrected signal-to-noise ratio gain by performing a correction that reduces the variation among multiple unit spaces. As a result, even if a common threshold (alarm threshold) is set for multiple unit spaces, the cause of the anomaly can be estimated accurately and early.

[0040] In the flowchart of Figure 5A, the signal-to-noise ratio gain calculation unit 341 calculates the corrected signal-to-noise ratio gain. However, a threshold value inversely proportional to the number of sensors used to create each unit space may be set for each unit space, and the signal-to-noise ratio gain calculation unit 341 may calculate a general signal-to-noise ratio gain. Even with this configuration, it is possible to suppress the occurrence of false detections and detection delays due to variations in signal-to-noise ratio gain across multiple unit spaces.

[0041] Figure 6 shows an example of the hardware configuration of the monitoring device according to the embodiment. The computer 900 includes a CPU 901, main memory 902, auxiliary memory 903, input / output interface 904, and communication interface 905. The monitoring device 30 is implemented in the computer 900. The functions described above are stored in auxiliary storage device 903 in the form of a program. The CPU 901 reads the program from auxiliary storage device 903, loads it into main memory 902, and executes the above processing according to the program. The CPU 901 allocates storage space in main memory 902 according to the program. The CPU 901 also allocates storage space in auxiliary storage device 903 to store the data being processed according to the program.

[0042] A program to implement all or part of the functions of the monitoring device 30 may be recorded on a computer-readable recording medium, and the program recorded on this recording medium may be loaded into a computer system and executed to perform processing by each functional unit. Here, "computer system" includes hardware such as the OS and peripheral devices. If a WWW system is used, "computer system" also includes the homepage provisioning environment (or display environment). "Computer-readable recording medium" refers to portable media such as CDs, DVDs, USBs, and storage devices such as hard disks built into the computer system. If this program is distributed to the computer 900 via a communication line, the computer 900 that receives the program may load it into the main memory 902 and execute the above processing. The above program may be for implementing some of the functions described above, and may also be for implementing the above functions in combination with a program already recorded in the computer system.

[0043] As described above, several embodiments relating to this disclosure have been explained, but all of these embodiments are presented as examples and are not intended to limit the scope of the invention. These embodiments can be carried out in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their variations are included in the scope and spirit of the invention, as well as in the claims and their equivalents.

[0044] <Note> The monitoring method, signal-to-noise ratio gain calculation method, monitoring device, and program described in each embodiment can be understood, for example, as follows.

[0045] (1) The monitoring method according to the first embodiment includes the steps of: (1) acquiring unit space creation data for creating a plurality of unit spaces predetermined according to the operating mode and / or the object to be monitored, detected by a plurality of sensors 1 to n provided in the plant 20 (S10); creating a plurality of unit spaces based on the unit space creation data detected by the sensors necessary for creating each of the unit spaces (S11); acquiring evaluation data, which is a collection of data for evaluating the state of the plant, detected by the plurality of sensors (S12); determining the Mahalanobis distance of the evaluation data based on at least a portion of the plurality of unit spaces (S13); determining the state of the plant based on the Mahalanobis distance and a predetermined threshold (S14 to S16); and, if an abnormality is determined in the step of determining the state of the plant, estimating the cause of the abnormality. The steps include (S17~S18), and the step of estimating the cause of the abnormality includes: creating an orthogonal array for each unit space used to determine the state, assigning each of the sensors used to create the unit space to a two-level orthogonal array with the first level being the use of the sensor and the second level being the non-use of the sensor; calculating the maximum signal-to-noise ratio for each row of the orthogonal array for each unit space; calculating the signal-to-noise ratio gain for each sensor in each of the orthogonal arrays for each unit space by calculating the difference between the sum of the maximum signal-to-noise ratios at the first level and the sum of the maximum signal-to-noise ratios at the second level; and identifying the sensor whose calculated signal-to-noise ratio gain exceeds the threshold for signal-to-noise ratio gain (alarm threshold), based on the signal-to-noise ratio gain calculated for each sensor and a predetermined threshold for signal-to-noise ratio gain (alarm threshold), as the sensor related to the cause of the abnormality. This allows for earlier estimation of the cause of an anomaly in a monitoring method that uses the multi-MT method to determine whether a plant is normal or abnormal by comparing the Mahalanobis distance and a threshold, and if an anomaly is determined, estimates the cause of the anomaly using the signal-to-noise ratio gain.

[0046] (2) The monitoring method according to the second embodiment is the monitoring method of (1), wherein in the step of calculating the signal-to-noise ratio gain, instead of calculating the difference between the sum of the desired signal-to-noise ratios of the first level and the sum of the desired signal-to-noise ratios of the second level, the signal-to-noise ratio gain is calculated by calculating the difference between a value obtained by multiplying the average value of the desired signal-to-noise ratios of the first level by a value corresponding to the number of sensors used to create the unit space and a value obtained by multiplying the average value of the desired signal-to-noise ratios of the second level by a value corresponding to the number of sensors used to create the unit space. This makes it possible to obtain the same effects as the monitoring method according to the first embodiment described above.

[0047] (3) The monitoring method according to the third embodiment is the monitoring method of (1), wherein in the step of calculating the signal-to-noise ratio gain, instead of calculating the difference between the sum of the signal-to-noise ratios of the first level and the sum of the signal-to-noise ratios of the second level, the signal-to-noise ratio gain is calculated by calculating the difference between the average value of the desired signal-to-noise ratios of the first level and the average value of the desired signal-to-noise ratios of the second level, and in the step of identifying the factors, for each unit space, the factors causing the abnormality are identified based on a threshold for the signal-to-noise ratio gain that is determined to be inversely proportional to the number of sensors used to create the unit space. This makes it possible to obtain the same effects as the monitoring method according to the first embodiment described above.

[0048] (4) A method for calculating the signal-to-noise ratio gain according to the fourth embodiment includes the steps of: creating a plurality of unit spaces based on data detected by a plurality of sensors 1 to n provided in the plant 20, and when evaluating the state of the plant by the multi-MT method based on at least a part of the plurality of unit spaces, creating an orthogonal array for each unit space by assigning each of the sensors used to create the unit space to a two-level orthogonal array, with the first level being the use of the sensor and the second level being the non-use of the sensor (S20); calculating the maximum signal-to-noise ratio for each row of the orthogonal array for each unit space (S21); and calculating the signal-to-noise ratio gain for each sensor in each of the orthogonal arrays for each unit space by calculating the difference between the sum of the maximum signal-to-noise ratios at the first level and the sum of the maximum signal-to-noise ratios at the second level (S22). This allows for the calculation of the signal-to-noise ratio (SNR) gain in the multi-MT method, correcting for the level of difference in SNR gain due to the orthogonal array size for each unit space.

[0049] (5) A method for calculating the signal-to-noise ratio gain according to the fifth embodiment is the method for calculating the signal-to-noise ratio gain according to (4), wherein in the step of calculating the signal-to-noise ratio gain, instead of calculating the difference between the sum of the desired signal-to-noise ratios of the first level and the sum of the desired signal-to-noise ratios of the second level, the signal-to-noise ratio gain is calculated by calculating the difference between a value obtained by multiplying the average value of the desired signal-to-noise ratios of the first level by a value corresponding to the number of sensors used to create the unit space and a value obtained by multiplying the average value of the desired signal-to-noise ratios of the second level by a value corresponding to the number of sensors used to create the unit space. This makes it possible to obtain the same effect as the method for calculating the signal-to-noise ratio gain according to the fourth embodiment described above.

[0050] (6) The monitoring device 30 according to the sixth embodiment includes means (data acquisition unit 31) for acquiring unit space creation data for creating a plurality of unit spaces predetermined according to the operating mode and / or the object to be monitored, which are detected by a plurality of sensors installed in the plant; means (unit space creation unit 32) for creating a plurality of unit spaces based on the unit space creation data detected by the sensors necessary for creating each of the unit spaces; means (data acquisition unit 31) for acquiring evaluation data which is a collection of data for evaluating the state of the plant, which are detected by the plurality of sensors; means (state evaluation unit 33) for determining the Mahalanobis distance of the evaluation data based on at least a portion of the plurality of unit spaces; means (state evaluation unit 33) for determining the state of the plant based on the Mahalanobis distance and a predetermined threshold; and if an abnormality is determined in the step of determining the state of the plant, The system includes means for estimating the cause of the abnormality (cause estimation unit 34), and the means for estimating the cause of the abnormality (cause estimation unit 34, SN ratio gain calculation unit 341) assigns each of the sensors used to create the unit space to a two-level orthogonal array, with using the sensor as the first level and not using the sensor as the second level, to each of the unit spaces used to determine the state, to create an orthogonal array for each unit space, calculates the maximum SN ratio for each row of the orthogonal array for each unit space, calculates the SN ratio gain for each sensor in each of the orthogonal arrays for each unit space by calculating the difference between the sum of the maximum SN ratios at the first level and the sum of the maximum SN ratios at the second level, and identifies the sensor whose calculated SN ratio gain exceeds the threshold for SN ratio gain as the sensor related to the cause of the abnormality, based on the SN ratio gain calculated for each sensor and a predetermined threshold for SN ratio gain.

[0051] (7) A program according to the seventh aspect includes the steps of: acquiring unit space creation data for creating a plurality of unit spaces predetermined according to the operating mode and / or the object being monitored, which is detected by a plurality of sensors installed in the plant; creating a plurality of unit spaces based on the unit space creation data detected by the sensors necessary for creating each of the unit spaces; acquiring evaluation data which is a collection of data for evaluating the state of the plant, which is detected by the plurality of sensors; determining the Mahalanobis distance of the evaluation data based on at least a portion of the plurality of unit spaces; determining the state of the plant based on the Mahalanobis distance and a predetermined threshold; and, if an abnormality is determined in the step of determining the state of the plant, estimating the cause of the abnormality. The step of estimating the cause of the abnormality includes the steps of: creating an orthogonal array for each unit space used to determine the state, assigning each of the sensors used to create the unit space to a two-level orthogonal array with the first level being the use of the sensor and the second level being the non-use of the sensor; calculating the maximum signal-to-noise ratio for each row of the orthogonal array for each unit space; calculating the signal-to-noise ratio gain for each sensor in each of the orthogonal arrays for each unit space by calculating the difference between the sum of the maximum signal-to-noise ratios at the first level and the sum of the maximum signal-to-noise ratios at the second level; and identifying the sensor whose calculated signal-to-noise ratio exceeds the threshold for signal-to-noise ratio gain is related to the cause of the abnormality, based on the signal-to-noise ratio gain calculated for each sensor and a predetermined threshold for signal-to-noise ratio gain. [Industrial applicability]

[0052] According to the monitoring method, signal-to-noise ratio gain calculation method, monitoring device, and program described above, it is possible to calculate a signal-to-noise ratio gain that is less affected by the size of the orthogonal array. According to the monitoring method, monitoring device, and program described above, it is possible to estimate the cause of anomalies early and with high accuracy. [Explanation of symbols]

[0053] 10. Surveillance system 20...plant 21...machine 1~n···Sensor 30...Monitoring device 31. Data Acquisition Unit 32. Unit Space Creation Section 33. Condition Evaluation Department 34. Factor Estimation Unit 341...SN ratio gain calculation section Output section of 35... 36...Storage section 900... Computer 901···CPU 902...Main memory 903...Auxiliary storage device 904... Input / Output Interface 905...Communication Interface

Claims

1. A step of acquiring unit space creation data for creating a predetermined number of unit spaces according to the operating mode and / or the object being monitored, detected by multiple sensors installed in the plant, A step of creating a plurality of unit spaces based on the unit space creation data detected by the sensor necessary for creating each of the unit spaces, A step of acquiring evaluation data, which is a collection of data for evaluating the state of the plant detected by multiple sensors, A step of determining the Mahalanobis distance of the evaluation data based on at least a portion of the multiple unit spaces, The steps include determining the state of the plant based on the Mahalanobis distance and a predetermined threshold, If an abnormality is determined in the step of determining the state of the plant, the steps include estimating the cause of the abnormality, Includes, The step of estimating the cause of the aforementioned anomaly is: For each unit space used to determine the state, assign each of the sensors used to create the unit space to a two-level orthogonal array, with the first level indicating the use of the sensor and the second level indicating the non-use of the sensor, thereby creating an orthogonal array for each unit space. For each of the orthogonal arrays in the unit space, the step of calculating the best signal-to-noise ratio for each row of the orthogonal array, The steps include calculating the signal-to-noise ratio gain for each sensor in each of the orthogonal arrays for each unit space by calculating the difference between the sum of the desired signal-to-noise ratios at the first level and the sum of the desired signal-to-noise ratios at the second level, The process includes the step of identifying a sensor whose calculated signal-to-noise ratio gain exceeds the threshold for signal-to-noise ratio gain, based on the signal-to-noise ratio gain calculated for each of the sensors and a predetermined threshold for the signal-to-noise ratio gain, as a sensor related to the cause of the abnormality. Monitoring method.

2. In the step of calculating the signal-to-noise ratio gain, instead of calculating the difference between the sum of the desired signal-to-noise ratios at the first level and the sum of the desired signal-to-noise ratios at the second level, The signal-to-noise ratio gain is calculated by multiplying the average value of the desired signal-to-noise ratio at the first level by a value corresponding to the number of sensors used to create the unit space, and by calculating the difference between the average value of the desired signal-to-noise ratio at the second level and a value corresponding to the number of sensors used to create the unit space. The monitoring method according to claim 1.

3. In the step of calculating the signal-to-noise ratio gain, instead of calculating the difference between the sum of the desired signal-to-noise ratios at the first level and the sum of the desired signal-to-noise ratios at the second level, The signal-to-noise ratio gain is calculated by calculating the difference between the average value of the desired signal-to-noise ratio at the first level and the average value of the desired signal-to-noise ratio at the second level. In the identifying step, for each unit space, the sensor is identified based on the signal-to-noise ratio gain threshold determined to be inversely proportional to the number of sensors used to create the unit space. The monitoring method according to claim 1.

4. When creating multiple unit spaces based on data detected by multiple sensors installed in the plant, and evaluating the state of the plant using the multi-MT method based on at least a portion of these unit spaces, For each of the aforementioned unit spaces, the steps include assigning each of the sensors used to create the unit space to a two-level orthogonal array, with the first level being the use of the sensor and the second level being the non-use of the sensor, thereby creating an orthogonal array for each of the aforementioned unit spaces, For each of the orthogonal arrays in the unit space, the step of calculating the best signal-to-noise ratio for each row of the orthogonal array, The steps include calculating the signal-to-noise ratio gain for each sensor in each of the orthogonal arrays for each unit space by calculating the difference between the sum of the desired signal-to-noise ratios at the first level and the sum of the desired signal-to-noise ratios at the second level, A method for calculating signal-to-noise ratio gain, including [the specified factor].

5. In the step of calculating the signal-to-noise ratio gain, instead of calculating the difference between the sum of the desired signal-to-noise ratios at the first level and the sum of the desired signal-to-noise ratios at the second level, The signal-to-noise ratio gain is calculated by multiplying the average value of the desired signal-to-noise ratio at the first level by a value corresponding to the number of sensors used to create the unit space, and by calculating the difference between the average value of the desired signal-to-noise ratio at the second level and a value corresponding to the number of sensors used to create the unit space. The method for calculating the signal-to-noise ratio gain according to claim 4.

6. Means for acquiring unit space creation data for creating a plurality of unit spaces predetermined according to the operating mode and / or the object being monitored, detected by a plurality of sensors installed in the plant, Means for creating a plurality of unit spaces based on the unit space creation data detected by the sensor necessary for creating each of the unit spaces, means for acquiring evaluation data, which is a collection of data for evaluating the state of the plant detected by multiple sensors, A means for determining the Mahalanobis distance of the evaluation data based on at least a portion of the plurality of unit spaces, A means for determining the state of the plant based on the Mahalanobis distance and a predetermined threshold, If the means for determining the state of the plant determines that an abnormality exists, the means for estimating the cause of the abnormality, It has, The means for estimating the cause of the aforementioned abnormality is, For each unit space used to determine the state, each of the sensors used to create the unit space is assigned to a two-level orthogonal array, with the first level indicating the use of the sensor and the second level indicating the non-use of the sensor, and an orthogonal array is created for each unit space. For each of the orthogonal arrays in the unit space, the maximum signal-to-noise ratio of each row of the orthogonal array is calculated. For each of the orthogonal arrays in each of the unit spaces, the signal-to-noise ratio gain is calculated by calculating the difference between the sum of the desired signal-to-noise ratios at the first level and the sum of the desired signal-to-noise ratios at the second level. Based on the signal-to-noise ratio gain calculated for each of the aforementioned sensors and a predetermined threshold for the signal-to-noise ratio gain, the sensors for which a signal-to-noise ratio gain exceeding the threshold for the signal-to-noise ratio gain is calculated are identified as sensors related to the cause of the abnormality. monitoring equipment.

7. On the computer, A step of acquiring unit space creation data for creating a predetermined number of unit spaces according to the operating mode and / or the object being monitored, detected by multiple sensors installed in the plant, A step of creating a plurality of unit spaces based on the unit space creation data detected by the sensor necessary for creating each of the unit spaces, A step of acquiring evaluation data, which is a collection of data for evaluating the state of the plant detected by multiple sensors, A step of determining the Mahalanobis distance of the evaluation data based on at least a portion of the multiple unit spaces, The steps include determining the state of the plant based on the Mahalanobis distance and a predetermined threshold, If an abnormality is determined in the step of determining the state of the plant, the steps include estimating the cause of the abnormality, Includes, The step of estimating the cause of the aforementioned anomaly is: For each unit space used to determine the state, assign each of the sensors used to create the unit space to a two-level orthogonal array, with the first level indicating the use of the sensor and the second level indicating the non-use of the sensor, thereby creating an orthogonal array for each unit space. For each of the orthogonal arrays in the unit space, the step of calculating the best signal-to-noise ratio for each row of the orthogonal array, The steps include calculating the signal-to-noise ratio gain for each sensor in each of the orthogonal arrays for each unit space by calculating the difference between the sum of the desired signal-to-noise ratios at the first level and the sum of the desired signal-to-noise ratios at the second level, A program that performs a process including the step of identifying a sensor whose calculated signal-to-noise ratio gain exceeds the threshold for signal-to-noise ratio gain, based on the signal-to-noise ratio gain calculated for each of the sensors and a predetermined threshold for the signal-to-noise ratio gain, as a sensor related to the cause of the abnormality.

Images