JPH1077948A - Monitoring device for hydraulic turbine generator - Google Patents

Abstract

PROBLEM TO BE SOLVED: To judge the normal/abnormal state of a hydraulic turbine generator objectively with accuracy by comparing measured output data with output data computed on the basis of a model that describes the relation between input data and output data. SOLUTION: Input data 102 to a hydraulic turbine generator and output data 103 from the hydraulic turbine generator are sent out to model constructing means 121 from first switching positions 112, 113 of an input-output data switching means 111. Model storage means 122 stores a model learned in the model constructing means 121. After storing the model, the input data 102 and output data 103 are sent out to model output computing means 131 from second switching positions 114, 115 of the input-output data switching means 111 so as to compute model output data 140. A residual computing means 132 computes the residual between the model output data 140 and actual measured output data 103, and index computing means 133 computes a numerical index showing the state of the hydraulic turbine generator from this residual to judge a normal/ abnormal state.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention learns a model in a normal state using data measured from a normally operating turbine generator, and uses the model output data and the output data measured from the turbine generator. The present invention relates to a turbine generator monitoring device having a function of accurately determining the normal / abnormal state of a turbine generator.

[0002]

2. Description of the Related Art FIG.
Vol. 6, No. 4, 1990 (published by the Joint Research Institute of Electric Power Industry,
FIGS. 9A and 9B are configuration diagrams showing portions related to the present invention, which are extracted from the conventional water turbine generator monitoring device shown in FIGS. Measurement data measured by
Reference numeral 601 denotes a determination unit that determines an abnormal state of the turbine generator from the measured data. FIG. 7 is a graph showing a time change and an alarm level of the temperature data measured by the water turbine generator. In the figure, reference numeral 701 denotes the measured temperature data;
02 is an alarm level determined by experience.

Next, the operation will be described. Here, the water turbine generator monitoring device shown in FIG. 6 automatically sets the measurement data 602 such as vibration data and temperature data measured in real time in the operating state of the water turbine generator for each measurement data 602. When the alarm level exceeds the alarm level, the judging means 601 judges that there is an abnormality and issues an alarm. An operation when the measurement data 602 is a temperature will be described.
As shown in FIG. 7, an alarm level 702 predetermined by experience is set in the judging means 601, and if the measured temperature data 701 exceeds this, the judging means 6
01 is determined to be abnormal. For other measurement data, an alarm level is set empirically for each measurement data,
If the measured data exceeds this alarm level, the judgment means 601
Is determined to be abnormal. As a result, an abnormal state can be detected at an early stage without human intervention.

In order to determine these alarm levels more accurately, it is necessary to simulate an actual abnormal state. However, such a state cannot be simulated by an actual turbine generator, The warning level is determined subjectively based on the experience of the abnormal state. Therefore, the warning level is usually set to a high level in order to prevent the warning from being issued too much, and it is difficult to find any signs in the initial state of the abnormality.

[0005]

Since the conventional turbine generator monitoring apparatus is constructed as described above, an index indicating the normal / abnormal state of the turbine generator cannot be obtained. Therefore, the determination means must determine the normal / abnormal state of the turbine generator based on the alarm level set subjectively based on the data of the past abnormal state, and determine the accurate state of the turbine generator. There was a problem that it could not be done.

Further, since the alarm level must be set after the accumulation of data in an abnormal state, the data must be accumulated for a considerable period of time, such as when there is no accumulated data or when the normal state changes. In addition, there is a problem that a monitoring device cannot be introduced promptly.

Further, an alarm level must be set for each element of each measurement data, and there has been a problem that a normal / abnormal state cannot be determined comprehensively from a plurality of types of data.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problems, and an object of the present invention is to provide a turbine generator monitoring device for accurately determining a normal / abnormal state of a turbine generator.

[0009]

According to a first aspect of the present invention, there is provided a water turbine generator monitoring apparatus, comprising: a model that describes a relationship between input data and output data measured by a water turbine generator; Model output calculation means for calculating the output data of the generator, residual calculation means for calculating the residual of the output data calculated by the model output calculation means and output data measured by the turbine generator, and residual calculation means An index calculating means for calculating an index indicating the normal / abnormal state of the turbine generator from the residual series calculated in the above, and a normal / abnormal state of the turbine generator is determined from the index calculated by the residual calculating means. It is provided with an abnormality determination means.

According to a second aspect of the present invention, there is provided a turbine generator monitoring device including a learning unit for learning a model describing a relationship between input data and output data measured by the turbine generator.

According to a third aspect of the present invention, there is provided a water turbine generator monitoring apparatus which receives input data and output data measured by the water turbine generator and selects data effective for monitoring the water turbine generator. Means.

A turbine generator monitoring apparatus according to a fourth aspect of the present invention uses a plurality of turbine generator bearing temperature fluctuation data as measurement data.

A turbine generator monitoring apparatus according to a fifth aspect of the present invention uses a plurality of turbine generator shaft vibration data as measurement data.

A turbine generator monitoring apparatus according to a sixth aspect of the present invention uses a plurality of turbine generator shaft vibration data and a plurality of turbine generator bearing temperature fluctuation data as measurement data.

[0015]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below. Embodiment 1 FIG. FIG. 1 is a block diagram showing a configuration of a water turbine generator monitoring device according to a first embodiment of the present invention. In the drawing, reference numeral 80 denotes learning means, 101 denotes a water turbine generator monitoring device, and 102 denotes an input to the water turbine generator. Data 103 is output data from the turbine generator. Reference numeral 111 denotes input / output data switching means, and 112 and 113 denote first switching positions, 1
Reference numerals 14 and 115 indicate second switching positions. 121 is a model construction means for learning a model representing the relationship between the input data 102 and output data 103 from the turbine generator when the input / output data switching means 111 is at the first switching position, and 122 is learned by the model construction means 121 Model storage means for storing a model. 131 is input / output data switching means 1
Model output calculating means for calculating output data of the turbine generator from the input data 102 and the output data 103 from the turbine generator using the model stored in the model storage means 122 when 11 is the second switching position , 140 are model output data calculated as output data of the turbine generator using the model by the model output calculation means 131, and 132 is a residual calculation means for calculating a residual between the output data 103 and the model output data 140. . 133 is a residual calculation means 132
Is an index calculating means for calculating an index indicating a normal / abnormal state of the water turbine generator from the time series value of the residual calculated in the above. 1
Reference numeral 34 denotes an abnormality determination unit that determines the normal / abnormal state of the turbine generator based on the index calculated by the index calculation unit 133.

Next, the operation will be described. Input data 102 to the turbine generator and output data 10 from the turbine generator
3 is a first switching position 11 of the input / output data switching means 111
2 and 113, are sent to the model construction means 121. The output data 103 can be represented by an autoregressive model represented by the equation (1).

[0017]

(Equation 1)

Here, t is time, u (t) is input data 102 at time t, y (t) is output data 103 at time t, e (t) is white noise, and na and nb are output data y This is the order of the model expressing (t). Here, u (t) and y (t) are families of random variables of a weak stationary process having a time parameter t, {u (t), − ， <t <
∞｝, ｛y (t), −∞ <t <∞｝. In the model construction means 121, the given input data 1
02. The coefficients a ′ _i and b ′ _j are determined so that the output data 103 satisfies the expression (1) as much as possible. This is called model learning in the model construction means 121, and the input data 10
2. The equation (2) for calculating y ′ (t) (model output data 140) from the time-series values of the output data 103 is called a model. Here, “′” is added to the learned coefficient or the calculated model output data. Also, this coefficient a '
_As a learning method for determining _i and b ′ _j, for example, the least square method can be used. The model storage unit 122 stores the model learned by the model construction unit 121. When the normal state of the turbine generator changes due to, for example, a change in the setting state, the model construction means 121 can learn a new model in the normal state each time. Therefore, a model that is always in the state of the turbine generator can be used in the present apparatus.

When the model is stored, the input data 102,
The output data 103 is sent from the second switching positions 114 and 115 of the data switching unit 111 to the model output calculation unit 131. The model output calculation means 131 calculates y ′ (t) (model output data 140) from the time series values of the input data 102 and the output data 103 using the model represented by the equation (2) stored in the model storage means 122. .

[0020]

(Equation 2)

On the other hand, the output data 103 is supplied from the second switching position 115 of the data switching
, And at the same time, y ′ (t) (model output data 140) calculated by the model output calculation means 131 is also sent to the residual calculation means 132. The residual calculating means 132 calculates the residual r (t) according to the equation (3).

R (t) = y (t) −y ′ (t) (3)

The residual r (t) calculated by the residual calculating means 132 is input to the index calculating means 133, and the residual r
An index for determining the state of the turbine generator is calculated from the time series value of (t). As an index for determining the state of the turbine generator, for example, there is a value of an autocorrelation function R (s) of a time-series value of a residual r (t) given by Expression (4).

R (s) = | E [r (t) r (t + s)] | (4)

Here, E is a mathematical expected value (set average), which is averaged at a constant time interval in the present embodiment, and s is called a lag. From the equations (1) and (2), when the relationship between the input and output data from the turbine generator is in a normal state, that is, the relationship between the input and output data of the turbine generator is expressed by a model represented by the equation (2). In this case, the time series value of the residual r (t) approaches the time series value of the white noise e (t), and the residual r
The value of the autocorrelation function R (s) of the time series value of (t) is lag s
Is substantially equal to 0 when is other than 0.

The autocorrelation function R (s) is a random variable normalized between 0 and 1, and the value of the autocorrelation function R (s) will not change even if the turbine is in a normal state. Along with the probability distribution. In the present embodiment, the value of the autocorrelation function R (s) calculated with the passage of time in the normal state of the water turbine generator has the highest probability of indicating 0, and decreases as the value approaches 1. Therefore, if we define an abnormal state as a state outside the normal state distribution,
It can be said that as the value of the autocorrelation function R (s) approaches 1, the probability of an abnormal state increases, and the state of the water turbine generator can be objectively grasped from the value of the autocorrelation function R (s). In the present embodiment, the autocorrelation function R (s) is used as an index for determining the state of the turbine generator, and the normal / abnormal state of the turbine generator is determined from the value.

As described above, the state of the turbine generator can be objectively determined from the value of the autocorrelation function R (s). Therefore, the abnormality determining means 134 may determine the state of the turbine generator from the value of the autocorrelation function R (s). For example, when the value of the autocorrelation function R (s) is equal to or larger than the value A, the turbine generator is determined to be in an abnormal state. In the present embodiment, only one index is calculated. However, when a plurality of indexes are calculated, if the value of any one index is equal to or more than the A value, the turbine generator is determined to be in an abnormal state. That is, the A value is 0 <A
It suffices if there is a relationship of <1. If the abnormality determining means 134 determines that the state is abnormal, for example, a buzzer or the like is sounded to issue an alarm. Although the single-input single-output system has been described in the first embodiment, it is easy to expand to a multiple-input single-output system, a single-input multiple-output system, and a multiple-input multiple-output system.

As described above, according to the first embodiment, the residual calculation means for calculating the residual between the model output data of the model learned by the model construction means and the actually measured output data of the turbine generator. And an index calculating means for calculating a numerical index indicating a state of the turbine generator from the residual calculated by the residual calculating means, and observing a change in the numerical index calculated by the index calculating means. Thus, it is possible to objectively determine a change in the state of the turbine generator, and to obtain an effect of accurately determining the normal / abnormal state of the turbine generator.

Further, since the state of the turbine generator is defined based on the deviation of the distribution of the normal state using the value of the numerical index, the turbine generator is unified irrespective of the type and number of measured data or the turbine generator. It is possible to judge the state. Therefore, unlike the related art, there is no need to change the alarm level for each type of data and for each turbine generator, and the monitoring device itself can be easily monitored and handled.

Further, in the conventional alarm device, the alarm level is set after the accumulation of the data in the abnormal state, so that when the normal state of the turbine generator changes, the data in the abnormal state needs to be accumulated again. However, since the present apparatus is provided with the model construction means, the alarm device can be quickly operated only by learning the model in the normal state again. Therefore, there is no need to determine the alarm level after waiting for the accumulation of data in an abnormal state as in the past.
The effect is obtained that this device can be quickly introduced.

Embodiment 2 FIG. 2 is a block diagram showing a water turbine generator monitoring device according to Embodiment 2 of the present invention. In the figure, 101 is a turbine generator monitoring device, 20
1, 202 and 203 are input / output data of the turbine generator, 21
1 is input / output data selection means, 212 is input data selected by the input / output data selection means 211, and 213 is output data selected by the input / output data selection means 211.

Next, the operation will be described. The input / output data selection unit 211 calculates a coherency value that approaches 1 as the correlation between the data on the frequency axis is stronger, based on equation (5), for all input / output data. Select combinations that exceed the threshold. Here, Suy used in equation (5) is used.
(Ω), Suu (ω) and Syy (ω) are u
(T) y (t + τ), u (t) u (t + τ), y (t)
This is the Fourier transform of the mathematical expected value of y (t + τ). For example, the mathematical expected value Suy of u (t) y (t + τ)
(Τ) is expressed by equation (6), and the mathematical expected value Suy
The Fourier transform Suy (ω) of (τ) is expressed by equation (7). t and τ are time, u (t) and y (t) are input data and output data, respectively, E is a mathematical expected value (set average)
Is shown.

[0033]

(Equation 3)

Suy (τ) = E [u (t) y (t + τ)] (6)

[0035]

(Equation 4)

For example, when there are n and m pieces of input data and output data, the coherency is calculated for all combinations of input and output data as shown in Table 1. Here, all combinations of m output data and n input data (mx
n). Next, input data indicating a coherency value equal to or greater than a certain value is selected for each output data. By selecting data indicating a high coherency value, a more accurate model can be learned. Next, in the same procedure as in the first embodiment, a model in a normal state is learned by the model construction means 121 using input data selected for each output data, and each residual Is calculated, and each index is calculated by the index calculation means 133. For example, there are m output data, each of which is 1
If more than one input data has been selected, m indices will be calculated.

[0037]

[Table 1]

The abnormality judging means 134 judges an abnormal state when any one of the calculated index values exceeds a set value. If the abnormality determining means 134 determines that the state is abnormal, a buzzer or the like sounds and an alarm is issued. When one index exceeds the set value, it can be understood that an abnormality has occurred between the input data and the output data used for calculating the index. Further, an abnormal state among a wide range of input / output data can be found by using a plurality of indices as compared with a case where only one index is used, so that a normal / abnormal state can be determined with higher accuracy.

As described above, according to the second embodiment, by selecting a combination of input / output data having a high coherency value, data effective for model construction can be selected from a plurality of data. Therefore, it is possible to learn a model for calculating model output data with higher accuracy than when a model is learned using single data or a plurality of unselected data. In addition, by using a plurality of indices, a wider range of abnormalities can be found than using a single index. Therefore, the normal / abnormal state can be comprehensively determined from a plurality of types of data, and the effect that the normal / abnormal state of the hydroelectric power plant can be determined more accurately can be obtained.

Further, an effect is obtained that an abnormal part is limited between input and output data of an index indicating an abnormal state.

Embodiment 3 FIG. 3 is a block diagram showing a water turbine generator monitoring device according to Embodiment 3 of the present invention. In the figure, 301, 302, and 303 are bearing temperature fluctuation data (turbine turbine bearing temperature fluctuation data) measured by the turbine generator, and 312 is the input / output data selector 21.
The input data 313 selected by 1 is the output data selected by the input / output data selection unit 211. The same or corresponding portions as those shown in FIGS. 1 and 2 are denoted by the same reference numerals, and redundant description will be omitted.

Next, the operation will be described. In the third embodiment, when a plurality of pieces of bearing temperature fluctuation data are observed, the input / output data selection unit 211 selects a combination of input / output data whose coherency value exceeds a certain threshold value. Here, the temperature fluctuation data is a minute fluctuation component obtained by removing a DC component of the temperature data. A model is learned from the data selected by the input / output data selection unit 211 in the same procedure as in the first embodiment, and a model output operation is performed.
Normal operation of turbine generator shaft vibration /
Judgment of the abnormal state is performed.

As described above, according to the third embodiment, an input / output model can be determined by selecting a set of input / output data of bearing temperature fluctuation data having a high coherency value. The effect of being able to construct a temperature abnormality detection system is obtained.

Embodiment 4 FIG. FIG. 4 is a block diagram showing a water turbine generator monitoring device according to Embodiment 4 of the present invention. In the figure, 401, 402, and 403 indicate shaft vibration data (turbine turbine shaft vibration data) measured by the turbine generator, 412 indicates input data selected by the input / output data selection unit 211, and 413 indicates input / output data selection unit 211. Is the output data selected in. The same or corresponding portions as those shown in FIGS. 1 and 2 are denoted by the same reference numerals, and redundant description will be omitted.

Next, the operation will be described. In the fourth embodiment, when a plurality of shaft vibration data are observed, the input / output data selection unit 211 selects a combination of input / output data whose coherency value exceeds a certain threshold. A model is learned from the data selected by the input / output data selection unit 211 in the same procedure as in the first embodiment, and model output calculation, residual calculation, and index calculation are performed to determine the normal / abnormal state of the shaft vibration of the turbine turbine. Is determined.

As described above, according to the fourth embodiment, since the input / output model can be determined by selecting a set of input / output data of the shaft vibration data having a high coherency value, a more reliable shaft vibration can be determined. An advantage is obtained that an abnormality detection system can be constructed.

Embodiment 5 FIG. 5 is a block diagram showing a water turbine generator monitoring device according to Embodiment 5 of the present invention. In the figure, reference numerals 501 and 502 denote shaft vibration data (turbine generator shaft vibration data) measured by the turbine generator, 50
3, 504 are bearing temperature fluctuation data measured by the turbine generator (turbine generator bearing temperature fluctuation data), 512 are input data selected by the input / output data selection means 211,
Reference numeral 13 denotes output data selected by the input / output data selection unit 211. The same or corresponding portions as those shown in FIGS. 1 and 2 are denoted by the same reference numerals, and redundant description will be omitted.

Next, the operation will be described. In the fifth embodiment, when a plurality of shaft vibration data and bearing temperature fluctuation data are observed, the input / output data selection means 211 selects a combination of input / output data whose coherency value exceeds a certain threshold value. Input / output data selection means 21
From the data selected in 1, the turbine generator monitoring device learns the model in the same procedure as in the first embodiment, performs model output calculation, residual calculation, and index calculation to determine the normal / abnormal state of the turbine generator shaft vibration. Is determined.

As described above, according to the fifth embodiment, not only one of the shaft vibration data and the bearing temperature fluctuation data, but also a set of input / output data having a high coherency value is selected from both data. Since the input / output model can be determined, there can be obtained an advantage that a more reliable turbine generator monitoring device can be constructed as compared with a case where only one of the shaft vibration data and the bearing temperature fluctuation data is used.

[0050]

According to the first aspect of the present invention, the output data of the turbine generator is calculated from the measured data by the model describing the relationship between the input data and the output data measured by the turbine generator monitoring device. Model output calculating means, a residual calculating means for calculating a residual between the output data calculated by the model output calculating means and the output data measured by the turbine generator, and a residual series calculated by the residual calculating means. From an index calculating means for calculating an index indicating the normal / abnormal state of the turbine generator, and an abnormality determining means for determining the normal / abnormal state of the turbine generator from the index calculated by the residual calculating means. With this configuration, a numerical index indicating that the probability of a normal state increases as the value approaches 0 and that the probability of an abnormal state increases as the value approaches 1 can be calculated by the index calculating means. From water wheel By objectively determine the state of the electric machine there is an effect of determining the state of normality / abnormality of accurately water turbine generator.

According to the second aspect of the present invention, the water turbine generator monitoring device is configured to include learning means for learning a model that describes a relationship between input data and output data measured by the water turbine generator. When a new turbine generator is installed or when the normal state of the turbine generator changes, it is possible to quickly learn the model in the normal state. Therefore, always use the model that was in the state of the turbine generator. Thus, when a turbine generator is newly installed, even when the normal state of the turbine generator changes, it is possible to quickly operate the turbine generator monitoring device.

According to the third aspect of the present invention, the input / output data for selecting the data effective for monitoring the turbine generator by receiving the input data and the output data measured by the turbine generator in the turbine generator monitoring device. Since it is configured to include the selection means, it is possible to select a combination of input and output data having a high coherency value, and it is possible to select data effective for model construction from a plurality of data, so that a more accurate model can be selected. There is an effect that a model for calculating output data can be learned.

According to the fourth aspect of the present invention, since a plurality of turbine generator bearing temperature fluctuation data are used as the measurement data, data effective for model construction is converted into a plurality of turbine generator bearing temperature fluctuation data. You can choose from
Since a model for calculating more accurate model output data can be learned, there is an effect that a highly reliable bearing temperature abnormality detection system can be obtained.

According to the fifth aspect of the present invention, since a plurality of turbine generator shaft vibration data are used as measurement data, data effective for model construction can be selected from the plurality of turbine generator shaft vibration data. Since a model for calculating more accurate model output data can be learned, there is an effect that a highly reliable shaft vibration abnormality detection system can be obtained.

According to the sixth aspect of the present invention, since a plurality of turbine generator shaft vibration data and a plurality of turbine generator bearing temperature fluctuation data are used as measurement data,
Data that is effective for model construction can be selected from multiple turbine generator shaft vibration data and multiple turbine generator bearing temperature fluctuation data, and either the turbine turbine shaft vibration data or the turbine generator bearing temperature fluctuation data can be used. Since a model for calculating model output data with higher accuracy can be learned than when used, the normal / abnormal state of the turbine generator can be more accurately determined.

[Brief description of the drawings]

FIG. 1 is a block diagram of a turbine generator monitoring device according to a first embodiment of the present invention.

FIG. 2 is a block diagram of a water turbine generator monitoring device according to a second embodiment of the present invention.

FIG. 3 is a block diagram of a water turbine generator monitoring device according to a third embodiment of the present invention.

FIG. 4 is a block diagram of a water turbine generator monitoring device according to a fourth embodiment of the present invention.

FIG. 5 is a block diagram of a water turbine generator monitoring device according to a fifth embodiment of the present invention.

FIG. 6 is a block diagram of a conventional water turbine generator monitoring device.

FIG. 7 is a diagram showing a relationship between an alarm level in a determination unit of a conventional water turbine generator monitoring device and temperature data.

[Explanation of symbols]

80 learning means, 101 turbine generator monitoring device, 131
Model output calculation means, 132 Residual calculation means, 133
Index calculation means, 134 abnormality determination means, 211 input / output data selection means, 301, 302, 303, 503,
504 bearing temperature fluctuation data (turbine turbine generator temperature fluctuation data), 401, 402, 403, 501, 50
2-axis vibration data (turbine generator shaft vibration data).

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Takenori Yakura 2-3-2 Marunouchi, Chiyoda-ku, Tokyo Within Mitsubishi Electric Corporation (72) Inventor Masayuki Uchi 2-3-2 Marunouchi, Chiyoda-ku, Tokyo Inside Mitsui Electric Co., Ltd. (72) Kenzo Sano 2-3-2 Marunouchi, Chiyoda-ku, Tokyo Mitsui Electric Co., Ltd. (72) Inventor Shunji Komon 3-2-2 Nakanoshima, Kita-ku, Osaka-shi, Osaka No. Kansai Electric Power Co., Inc. (72) Inventor Seiji Honda 3-3-22 Nakanoshima, Kita-ku, Osaka-shi, Osaka

Claims (6)

[Claims] 1. A model output calculating means for calculating output data of the turbine generator from data measured by the turbine generator according to a model describing a relationship between input data and output data measured by the turbine generator. A residual calculating means for calculating a residual between the output data calculated by the model output calculating means and the output data measured by the turbine generator; and a turbine generator based on the residual series calculated by the residual calculating means. A turbine generator monitoring device, comprising: index calculating means for calculating an index indicating a normal / abnormal state of the turbine; and abnormality determining means for determining a normal / abnormal state of the turbine generator from the index. 2. The turbine generator monitoring device according to claim 1, further comprising learning means for learning a model describing a relationship between input data and output data measured by the turbine generator. 3. An input / output data selection means for receiving input data and output data measured by the turbine generator and selecting data effective for monitoring the turbine generator. Or the turbine generator monitoring device according to claim 2. 4. The turbine generator monitoring system according to claim 3, wherein a plurality of turbine generator bearing temperature fluctuation data are used as the measurement data. 5. A turbine generator monitoring system according to claim 3, wherein a plurality of turbine turbine shaft vibration data are used as the measurement data. 6. A turbine generator monitoring apparatus according to claim 3, wherein a plurality of turbine turbine shaft vibration data and a plurality of turbine turbine bearing temperature fluctuation data are used as the measurement data.