Data processing method, data diagnosis method, and data processing program

Abstract

The present invention relates to a data processing method for processing data for calculating a Mahalanobis distance. In this method, an objective function for linearizing a combination of items included in unit space data is calculated, and coordinate conversion parameters for coordinate conversion of the unit space data for each item for minimizing the objective function are calculated. Then, coordinate conversion is performed on the unit space data for each item using the calculated coordinate conversion parameters. And standardizing the unit space data after coordinate transformation according to each project.

Description

Technical Field

[0001] The invention relates to a data processing method, a data diagnosis method and a data processing program.

[0002] This application claims priority based on Japanese Patent Application No. 2022-166013 filed in Japan on October 17, 2022, and the contents are incorporated herein by reference. Background Art

[0003] As a method for determining normality / abnormality by taking into account the correlation between items (variables), the MT (Mahalanobis-Taguchi) method is known. In the MT method, for the signal space of the set to be determined, the Mahalanobis distance that quantitatively shows the degree of deviation from the unit space of the normal set is calculated as an index. The increase in the Mahalanobis distance means that the signal space deviates from the unit space, and it can be regarded as a criterion for determining normality / abnormality in the determination object that is difficult to quantify.

[0004] The set processed by the MT method is based on the premise of having a normal distribution, but in general, the set includes items with different units or items with non-normal distribution. In the MT method, in order to process a set containing such various items, the unit space or signal space is standardized (for example, refer to Patent Document 1). For example, the item X′ obtained by standardizing the item X is obtained by using its average value Xavg and standard deviation σ, and is obtained by the following formula.

[0005] X′=(X-Xavg) / σ (1)

[0006] The standardized item X' has a mean of "zero" and a variance of "1".

[0007] Previous technical literature

[0008] Patent Literature

[0009] Patent Document 1: Japanese Patent Application Publication No. 2012-252556 Summary of the invention

[0010] Technical issues to be solved by the invention

[0011] The MT method is a linear analysis method based on the premise that the items to be processed are normally distributed. Therefore, when a highly linear variable is used as the object, it can determine normality / abnormality with good reliability, but if it includes highly nonlinear items or items with non-normal distribution, its reliability may decrease.

[0012] At least one embodiment of the present invention is completed in view of the above situation, and its purpose is to provide a data processing method, a data diagnosis method and a data processing program, which can perform highly reliable abnormality judgment even when processing a collection of items including highly nonlinear items or items with non-normal distribution.

[0013] Means for solving technical problems

[0014] In order to solve the above-mentioned problem, a data processing method according to at least one embodiment of the present invention is a data processing method for processing data for calculating Mahalanobis distance, which comprises:

[0015] a step of calculating an objective function for linearizing a combination of items included in the unit space data;

[0016] a step of calculating coordinate transformation parameters for performing coordinate transformation on the unit space data according to each of the items in order to minimize the objective function;

[0017] A step of performing coordinate transformation on the unit space data according to each of the items using the coordinate transformation parameters; and

[0018] A step of normalizing the unit space data after the coordinate transformation for each of the items.

[0019] In order to solve the above-mentioned problems, a data diagnosis method according to at least one embodiment of the present invention includes:

[0020] A data processing method according to at least one embodiment of the present invention;

[0021] A step of performing coordinate transformation on signal space data using the coordinate transformation parameters;

[0022] a step of calculating the Mahalanobis distance as a degree of deviation of the signal space data after coordinate transformation relative to the unit space data after coordinate transformation; and

[0023] A step of diagnosing the soundness of the signal space data based on the Mahalanobis distance.

[0024] In order to solve the above-mentioned problem, a data processing program according to at least one embodiment of the present invention is a data processing program for processing data for calculating Mahalanobis distance, wherein:

[0025] The following processes can be performed using a computer device:

[0026] a step of calculating an objective function for linearizing a combination of items included in the unit space data;

[0027] a step of calculating coordinate transformation parameters for performing coordinate transformation on the unit space data according to each of the items in order to minimize the objective function;

[0028] A step of performing coordinate transformation on the unit space data according to each of the items using the coordinate transformation parameters; and

[0029] A step of normalizing the unit space data after the coordinate transformation for each of the items.

[0030] Effects of the Invention

[0031] According to at least one embodiment of the present invention, a data processing method, a data diagnosis method and a data processing program can be provided, which can perform highly reliable abnormality judgment even when processing a collection of items including highly nonlinear items or items with non-normal distribution. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 This is an overall configuration diagram of a plant monitoring device according to one embodiment.

[0033] Figure 2 yes Figure 1 An example of signal space data stored in a signal space file.

[0034] Figure 3 yes Figure 1 An example of unit space data stored in the unit space file.

[0035] Figure 4 This is an example of a unit space corresponding to unit space data.

[0036] Figure 5 Yes means through Figure 1 A flow chart of a complete set of equipment monitoring method implemented by a complete set of equipment monitoring device.

[0037] Figure 6 This is a flowchart showing a data processing method according to one embodiment.

[0038] Figure 7 Yes means Figure 6 Flow chart of the method for calculating the objective function in step S201.

[0039] Figure 8 It is a diagram showing the combination of evaluation items used to calculate the objective function.

[0040] Fig. 9 Yes means Figure 6Flow chart of another method for calculating the objective function in step S201.

[0041] Fig.10 It means that Figure 4 A diagram of a unit space corresponding to a result of data processing performed on the unit space data shown in FIG. DETAILED DESCRIPTION

[0042] Hereinafter, several embodiments of the present invention will be described with reference to the drawings. However, the configurations described as the embodiments or shown in the drawings are not intended to limit the scope of the present invention thereto, but are merely illustrative examples.

[0043] Figure 1 1 is an overall structural diagram of a plant monitoring device 100 according to an embodiment. The plant monitoring device 100 is a device for monitoring the operating state of a gas turbine power generation plant 1. The gas turbine power generation plant 1 includes a gas turbine 2 and a generator 6 that generates electricity by driving the gas turbine 2. The gas turbine 2 includes a compressor 3 that generates compressed air, a combustor 4 that mixes fuel and compressed air and burns them to generate combustion gas, and a turbine 5 that is driven to rotate by the combustion gas. The rotor of the turbine 5 is connected to the generator 6 via the compressor 3, and the generator 6 generates electricity by the rotation of the rotor.

[0044] The plant monitoring device 100 obtains the state quantity of each of the multiple evaluation items of the gas turbine power generation plant 1, and determines whether the operating state of the gas turbine power generation plant 1 is normal based on these state quantities. The plant monitoring device 100 uses the Mahalanobis-Taguchi method (hereinafter, appropriately referred to as the "MT method") to monitor the operating state of the gas turbine power generation plant 1. As evaluation items of the gas turbine power generation plant 1, there are, for example, gas turbine output, cavity temperature at multiple locations between the turbine rotor and the stationary part, blade passage temperature at multiple locations in the circumferential direction at the gas outlet of the turbine, displacement at multiple locations in the circumferential direction of the turbine rotor, and openings of various valves provided in the gas turbine. In order to detect these state quantities, the gas turbine power generation plant 1 is provided with various state quantity detection mechanisms such as sensors.

[0045] The plant monitoring device 100 is composed of a computer device, which includes: a CPU 10 for executing various calculation processes; a main storage device 20 such as a RAM that serves as a work area of ​​the CPU 10; an auxiliary storage device 30 such as a hard disk drive device that stores various data and programs; an input / output interface 40 for various state quantity detection mechanisms of the gas turbine power generation plant 1 or input / output devices such as a keyboard, mouse or touch panel, display, etc. that are not shown; and a recording / reproducing device 42 for recording or reproducing data on a disk-type storage medium such as a CD or DVD.

[0046] Various programs including a plant monitoring program 34, a data processing program 33, and an OS program for making the computer function as the plant monitoring device 100 are stored in advance in the auxiliary storage device 30. Various programs including the data processing program 33 and the plant monitoring program 34 are imported from a disk-type storage medium into the auxiliary storage device 30 via the storage / reproduction device 42.

[0047] Furthermore, these programs may be imported into the auxiliary storage device 30 from an external device via a portable memory such as a flash memory or a communication device (not shown).

[0048] Furthermore, during the execution of the data processing program 33 and the plant monitoring program 34, the following files are also set in the auxiliary storage device 30. Specifically, the auxiliary storage device 30 is provided with the following files: a signal space file 31 storing data (signal space data) of the state quantity of each of the plurality of evaluation items of the gas turbine power generation plant 1; and a unit space file 32 storing data (unit space data) of the unit space that serves as a reference when determining the operating state of the plant.

[0049] Figure 2 yes Figure 1 An example of signal space data 31a stored in the signal space file 31. The signal space file 31 is as follows: Figure 2 As shown in the figure, during the execution of the data processing program 33 and the plant monitoring program 34, the signal space data 31a of the set of state quantities X, Y, Z of each of the multiple evaluation items of the gas turbine power generation plant 1 is stored in a time series according to the acquisition time T1, T2,... of each state quantity.

[0050] Figure 3 yes Figure 1 An example of unit space data 32a stored in the unit space file 32. The unit space file 32 is as follows: Figure 3 As shown in FIG. 1 , during the execution of the data processing program 33 and the plant monitoring program 34, unit space data 32a of a set of state quantities X, Y, and Z for each of a plurality of evaluation items are stored in correspondence with the signal space data 31a.

[0051] The CPU 10 functionally comprises: a signal space data acquisition unit 11 for acquiring the signal space data 31a stored in the signal space file 31; a unit space data acquisition unit 12 for acquiring the unit space data 32a stored in the unit space file 32; a data processing unit 13 for performing data processing on the signal space data 31a and the unit space data 32a; a Mahalanobis distance calculation unit 14 for calculating the Mahalanobis distance using the data processed by the data processing unit 13; and a complete set of equipment state determination unit 15 for diagnosing the soundness of the operating state of the gas turbine power generation complete set 1 based on whether the Mahalanobis distance calculated by the Mahalanobis distance calculation unit 14 is within a prescribed threshold value.

[0052] Among the functional components of the CPU 10 described above, the data processing unit 13 functions by the CPU 10 executing the data processing program 33 stored in the auxiliary storage device 30. In addition, the signal space data acquisition unit 11, the unit space data acquisition unit 12, the Mahalanobis distance calculation unit 14, and the plant state determination unit 15 function by the CPU 10 executing the plant monitoring program 34 stored in the auxiliary storage device 30.

[0053] Next, the monitoring operation of the plant monitoring device 100 of this embodiment will be described. The plant monitoring of the plant monitoring device 100 utilizes the MT method as described above. Figure 4 The basic contents of the plant monitoring method based on the MT method are explained. Figure 4 This is an example of the unit space S corresponding to the unit space data 32a.

[0054] Here, it is assumed that the output of the generator 6 of the gas turbine power generation plant 1 and the intake air temperature of the compressor 3 are respectively set as the state quantities X and Y of the evaluation items (in this example, for ease of understanding, the case where the evaluation items are two state quantities X and Y is described, but it is also possible to have more than three state quantities by including the state quantity Z, etc.). In the MT method, the unit space S corresponding to the unit space data 32a of the aggregate of the bundles that serve as the reference of these state quantities is used as a reference, and the Mahalanobis distance D is calculated as an evaluation index of whether the signal space data 31a representing the operating state is abnormal. The greater the degree of abnormality of the monitored object, the larger the value of the Mahalanobis distance D. Therefore, in the MT method, whether the operating state of the plant is abnormal is determined based on whether the Mahalanobis distance D is within a predetermined threshold value Dc.

[0055] in addition, Figure 4 In FIG. 1 , the solid line surrounding the unit space S indicates the position where the Mahalanobis distance D becomes the threshold value Dc.

[0056] However, if Figure 4 As shown by the asterisk in the figure, the unit space data 32a shows nonlinear behavior by bending. The aforementioned MT method is a linear analysis method based on the premise that the items to be processed are normally distributed. Therefore, in the case of highly linear evaluation items, normality / abnormality can be determined with good reliability, but if highly nonlinear items or items with non-normal distribution are included, its reliability may be reduced. Such a problem can be appropriately solved by the data processing described later.

[0057] Figure 5 Yes means through Figure 1 Flow chart of a complete plant monitoring method implemented by the complete plant monitoring device 100.

[0058] First, the signal space data acquisition unit 11 acquires the signal space data 31a stored in the signal space file 31 (step S100). And the unit space data acquisition unit 12 acquires the unit space data 32a stored in the unit space file 32 (step S101).

[0059] Subsequently, the data processing unit 13 performs data processing on the signal space data 31a and the unit space data 32a acquired in step S100 (step S102). Generally speaking, the MT method is a linear analysis method based on the premise that the evaluation items to be processed are normally distributed. Therefore, if highly nonlinear evaluation items or evaluation items with non-normal distribution are included, their reliability may be reduced. The details of the data processing performed in step S4 will be described later, but by performing data processing, the signal space data 31a and the unit space data 32a used to calculate the Mahalanobis distance D are subjected to coordinate transformation, thereby improving the linearity of the evaluation items or approaching normal distribution.

[0060] Then, the Mahalanobis distance calculation unit 14 calculates the Mahalanobis distance D using the signal space data 31a and the unit space data 32a subjected to data processing (step S103). Then, the plant state determination unit 15 determines the soundness of the plant operation state by comparing the Mahalanobis distance D calculated in step S103 with the threshold value Dc (step S104).

[0061] Afterwards, Figure 5 The specific contents of the data processing method implemented in step S102 are described below. Figure 63 is a flowchart showing a data processing method according to an embodiment of the present invention. The data processing method implements the functions of the data processing unit 13 by executing the data processing program 33. In this case, the data processing unit 13 functionally includes a first normalization unit 16, an objective function calculation unit 17, a coordinate transformation parameter calculation unit 18, a coordinate transformation operation unit 21, and a second normalization unit 22.

[0062] First, the first standardization unit 16 standardizes the unit space data 32a obtained in step S100 (step S200). The state quantities X, Y, and Z of the multiple evaluation items contained in the unit space data 32a generally have different units or average values. In the standardization implemented in step S200, in order to compare these state quantities X, Y, Z, etc. on an equal basis, they are transformed in a manner such that the average value becomes "zero" and the standard deviation becomes "1". For example, the state quantity X of one of the evaluation items contained in the unit space data 32a uses the average value Xavg and the standard deviation σ, and is standardized by the following formula (the same applies to other state quantities Y and Z).

[0063] X′=(X-Xavg) / σ (2)

[0064] Then, the objective function calculation unit 17 calculates an objective function fx for linearizing the combination of evaluation items included in the unit space data 32a (step S201). Then, the coordinate transformation parameter calculation unit 18 calculates coordinate transformation parameters for minimizing the objective function fx calculated in step S201 (step S202).

[0065] In step S202, the coordinate transformation parameters contained in any coordinate transformation formula can be used, but as one method, when using SHASH (hash value) transformation, the coordinate transformation parameters 6 and ε are used, and the state quantity X′ of the evaluation item is transformed by the following formula ("X" is the state quantity X′ after the coordinate transformation).

[0066] X″=sinh(δ·asinhX′-∈) (3-1)

[0067]

[0068] In another embodiment, Yeo-Johnson transformation can be used as coordinate transformation. In this case, the coordinate transformation parameter λ is used, and the state quantity X′ of the evaluation item is transformed by the following equation (“X″” is the state quantity X′ after coordinate transformation).

[0069]

[0070] In another embodiment, as the coordinate transformation, Boltzmann transformation, double Boltzmann transformation, or broken line transformation can be used. Each of the coordinate transformations exemplified above may be used alone or in combination of two or more.

[0071] Then, the coordinate transformation calculation unit 21 uses the coordinate transformation parameters calculated in step S202 to transform the unit space data 32a for each item (step S203). Then, the second normalization unit 22 normalizes the unit space data 32a transformed in step S203 for each item (step S204). The second normalization performed in step S204 is substantially the same as the first normalization performed in step S200.

[0072] Here, a specific calculation example of the objective function fx in step S201 will be described. Figure 7 Yes means Figure 6 Flow chart of the method for calculating the objective function fx in step S201, Figure 8 It is a diagram showing combinations of evaluation items for calculating the objective function fx.

[0073] First, the correlation coefficient R of each combination of the evaluation items included in the unit space data 32a acquired in step S100 is calculated (step S300). Figure 8 In the example of FIG. 1 , the correlation coefficient R for each combination of the state quantities X, Y, and Z of the three evaluation items is shown. Specifically, the correlation coefficient R corresponding to the combination of the state quantity X and the state quantity Y is shown. X-Y , the correlation coefficient R corresponding to the combination of state quantity X and state quantity Z X-Z and the correlation coefficient R corresponding to the combination of state quantity Y and state quantity Z Y-Z .

[0074] Then, the determination coefficient R is calculated based on the correlation coefficients R calculated in step S300. 2 (Step S301), in addition, according to the determination coefficient R 2 To calculate the self-information (step S302). Self-information is a concept in information theory, which represents a measure of the difficulty of an event (phenomenon) when it occurs. Self-information can also be regarded as a measure of how much information the event essentially has. In this embodiment, the self-information I is defined by the following formula (when the determination coefficient R 2 When it is close to "1" (when linearized), the self-information I is maximized).

[0075] I=-log 10 (1-R 2 ) (5)

[0076] Then, the target function fx is obtained by multiplying the sum of the self-information calculated in step S303 by "-1" (step S303).

[0077] and Fig. 9 Yes means Figure 6 Flow chart of another method for calculating the objective function fx in step S201.

[0078] First, the objective function calculation unit 17 calculates the skewness and kurtosis for each evaluation item included in the unit space data 32a acquired in step S100 (step S400). Specifically, the skewness and kurtosis are calculated by the following equations.

[0079]

[0080] Then, the objective function calculation unit 17 calculates the objective function fx by adding the values ​​obtained by square-raising the skewness and kurtosis calculated in step S400 and multiplying them by different weighting coefficients (step S401). Specifically, the objective function fx is calculated by the following equation.

[0081] fx = 0.8 × (kurtosis) 2 +0.2×(skewness) 2 (7)

[0082] By calculating the coordinate transformation parameters so as to minimize the objective function fx calculated from the skewness and kurtosis, it is possible to obtain coordinate transformation parameters for making the skewness and kurtosis of each evaluation item close to zero, that is, close to normal distribution.

[0083] As described above, according to the above-mentioned embodiments, by performing data processing on the signal space data 31a and the unit space data 32a used to calculate the Mahalanobis distance D, the signal space data 31a and the unit space data 32a including highly nonlinear evaluation items or evaluation items with non-normal distribution are subjected to coordinate transformation. As a result, the linearity of the signal space data 31a and the unit space data 32a can be improved, or the normal distribution can be approached. For example, Fig.10 It means that Figure 4 FIG. 3 is a diagram of a unit space S′ corresponding to the result of data processing of the unit space data 32a shown in FIG. 3 , but in FIG. Figure 4 The nonlinear behavior is shown by the bending as shown by the asterisk. Fig.10 As a result, by calculating the Mahalanobis distance D based on the signal space data 31a and the unit space data 32a after the coordinate transformation, it is possible to diagnose the health condition with high accuracy.

[0084] Furthermore, within the scope not departing from the spirit of the present invention, the constituent elements in the above-described embodiments can be appropriately replaced with well-known constituent elements, and the above-described embodiments can be appropriately combined.

[0085] The contents described in the above-mentioned embodiments can be understood, for example, as follows.

[0086] (1) A data processing method according to one embodiment is a data processing method for processing data for calculating a Mahalanobis distance, comprising:

[0087] a step of calculating an objective function for linearizing a combination of items included in the unit space data;

[0088] a step of calculating coordinate transformation parameters for performing coordinate transformation on the unit space data according to each of the items in order to minimize the objective function;

[0089] A step of performing coordinate transformation on the unit space data according to each of the items using the coordinate transformation parameters; and

[0090] A step of normalizing the unit space data after the coordinate transformation for each of the items.

[0091] According to the method of (1) above, an objective function for linearizing the combination of each item included in the unit space data is calculated, and coordinate transformation parameters for linearizing the combination of each item included in the unit space data are calculated in a manner that minimizes the objective function. By using the coordinate transformation parameters calculated in this way to transform the unit space data, the unit space data can be linearized even when the unit space data contains highly nonlinear items or items that are not non-normally distributed. In the MT method, by using the unit space data linearized in this way, a high-reliability soundness diagnosis can be performed based on a set of items containing highly nonlinear items.

[0092] (2) In another embodiment, in the embodiment of (1) above,

[0093] The process of calculating the objective function includes:

[0094] The step of calculating the correlation coefficient of each combination;

[0095] A step of calculating a determination coefficient based on the correlation coefficient;

[0096] A step of calculating the self-information amount based on the determination coefficient; and

[0097] A step of obtaining the objective function by multiplying the sum of the self-information by -1.

[0098] According to the above method (2), by multiplying the self-information of the determination coefficient calculated based on the correlation coefficient of each combination of the items contained in the unit space data by -1, the objective function for linearizing the combination of the items contained in the unit space data can be preferably obtained.

[0099] (3) In another embodiment, in the embodiment of (1) above,

[0100] The process of calculating the objective function includes:

[0101] The procedure for calculating the skewness and kurtosis for each of the items mentioned; and

[0102] A step of obtaining the objective function by adding values ​​obtained by square-raising the skewness and the kurtosis and values ​​obtained by multiplying the skewness and the kurtosis by different weighting coefficients.

[0103] According to the above method (3), by adding the values ​​obtained by square the skewness and kurtosis of each item contained in the unit space data and the values ​​obtained by multiplying them by different weighting coefficients, it is possible to preferably obtain an objective function for linearizing the combination of the items contained in the unit space data.

[0104] (4) In another embodiment, in any one of the above embodiments (1) to (3),

[0105] The coordinate transformation includes at least one of SHASH transformation, Yeo-Johnson transformation, Boltzmann transformation, double Boltzmann transformation or broken line transformation.

[0106] According to the above-mentioned aspect (4), by calculating the coordinate transformation parameters used in these transformation methods so as to minimize the objective function, it is possible to preferably linearize the combination of the items included in the unit space data.

[0107] In addition, the coordinate transformation may adopt any one of these transformation methods, may adopt a combination of at least two, or may adopt the same method multiple times.

[0108] (5) In another embodiment, in any one of the above embodiments (1) to (4), further comprising:

[0109] The process of standardizing the unit space data according to each of the items,

[0110] In the step of calculating the objective function, an objective function for linearizing a combination of the items included in the standardized unit space data is calculated.

[0111] According to the above aspect (5), the objective function is calculated using the pre-normalized unit space data, thereby effectively reducing the computational load associated with the objective function calculation.

[0112] (6) The data diagnostic methods involved in the first method include:

[0113] The data processing method involved in any of the above methods (1) to (3);

[0114] A step of performing coordinate transformation on signal space data using the coordinate transformation parameters;

[0115] a step of calculating the Mahalanobis distance as a degree of deviation of the signal space data after coordinate transformation relative to the unit space data after coordinate transformation; and

[0116] A step of diagnosing the soundness of the signal space data based on the Mahalanobis distance.

[0117] According to the method of (6) above, the coordinate transformation of the signal space data to be diagnosed is also performed using the coordinate transformation parameters calculated in a manner to linearize the unit space data. That is, the unit space data and the signal space data are respectively transformed using the common coordinate transformation parameters. By calculating the Mahalanobis distance based on the unit space and signal space data that have been coordinate transformed in this way, it is possible to perform a high-reliability soundness diagnosis on a set including highly nonlinear items or items with non-normal distribution.

[0118] (7) A data processing program according to one aspect is a data processing program for processing data for calculating a Mahalanobis distance, wherein:

[0119] The following processes can be performed using a computer device:

[0120] a step of calculating an objective function for linearizing a combination of items included in the unit space data;

[0121] a step of calculating coordinate transformation parameters for performing coordinate transformation on the unit space data according to each of the items in order to minimize the objective function;

[0122] A step of performing coordinate transformation on the unit space data according to each of the items using the coordinate transformation parameters; and

[0123] A step of normalizing the unit space data after the coordinate transformation for each of the items.

[0124] According to the method of (7) above, an objective function for linearizing the combination of each item included in the unit space data is calculated, and coordinate transformation parameters for linearizing the combination of each item included in the unit space data are calculated in a manner that minimizes the objective function. By using the coordinate transformation parameters calculated in this way to transform the unit space data, the unit space data can be linearized even when the unit space data contains highly nonlinear items or items that are not non-normally distributed. In the MT method, by using the unit space data linearized in this way, a high-reliability soundness diagnosis can be performed based on a set of items containing highly nonlinear items.

[0125] Explanation of symbols

[0126] 1-gas turbine power generation equipment, 2-gas turbine, 3-compressor, 4-combustor, 5-turbine, 6-generator, 10-CPU, 11-signal space data acquisition unit, 12-unit space data acquisition unit, 13-data processing unit, 14-Mahalanobis distance calculation unit, 15-equipment state determination unit, 16-first normalization unit, 17-objective function calculation unit, 18-coordinate transformation parameter calculation unit, 20-main storage device, 21-coordinate transformation operation unit, 22-second normalization unit, 30-auxiliary storage device, 31-signal space file, 31a-signal space data, 32-unit space file, 32a-unit space data, 33-data processing program, 34-equipment monitoring program, 40-input / output interface, 42-recording / reproducing device, 100-equipment monitoring device, D-Mahalanobis distance, Dc-threshold value.

Claims

1. A data processing method for processing data for calculating Mahalanobis distance, the data processing method comprising: a step of calculating an objective function for linearizing a combination of items included in the unit space data; a step of calculating coordinate transformation parameters for performing coordinate transformation on the unit space data according to each of the items in order to minimize the objective function; A step of performing coordinate transformation on the unit space data according to each of the items using the coordinate transformation parameters; and A step of normalizing the unit space data after the coordinate transformation for each of the items.

2. The data processing method according to claim 1, wherein: The process of calculating the objective function includes: The step of calculating the correlation coefficient of each combination; A step of calculating a determination coefficient based on the correlation coefficient; A step of calculating the self-information amount based on the determination coefficient; and A step of obtaining the objective function by multiplying the sum of the self-information by -1.

3. The data processing method according to claim 1, wherein: The process of calculating the objective function includes: The procedure for calculating the skewness and kurtosis for each of the items mentioned; and A step of obtaining the objective function by adding values ​​obtained by square-raising the skewness and the kurtosis and values ​​obtained by multiplying the skewness and the kurtosis by different weighting coefficients.

4. The data processing method according to claim 1 or 2, wherein: The coordinate transformation includes at least one of a hash value transformation, a Young-Johnson transformation, a Boltzmann transformation, a double Boltzmann transformation or a broken line transformation.

5. The data processing method according to claim 1 or 2, further comprising: The process of standardizing the unit space data according to each of the items, In the step of calculating the objective function, an objective function for linearizing a combination of the items included in the standardized unit space data is calculated.

6. A data diagnosis method, comprising: The data processing method according to claim 1 or 2; A step of performing coordinate transformation on signal space data using the coordinate transformation parameters; a step of calculating the Mahalanobis distance as a deviation of the signal space data after coordinate transformation relative to the unit space data after coordinate transformation; and A step of diagnosing the soundness of the signal space data based on the Mahalanobis distance.

7. A data processing program for processing data for calculating Mahalanobis distance, wherein: The following processes can be performed using a computer device: a step of calculating an objective function for linearizing a combination of items included in the unit space data; a step of calculating coordinate transformation parameters for performing coordinate transformation on the unit space data according to each of the items in order to minimize the objective function; A step of performing coordinate transformation on the unit space data according to each of the items using the coordinate transformation parameters; and A step of normalizing the unit space data after the coordinate transformation for each of the items.

Images