JP3465022B2 - Abnormality diagnosis device, abnormality diagnosis method, and program storage medium - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an abnormality diagnosing device, an abnormality diagnosing method, and a program storage medium, which detect an abnormality of a measuring device by using a wavelet transform result of a measuring signal output from the measuring device.

[0002]

2. Description of the Related Art Various measuring instruments such as a liquid level gauge and a flow meter are installed in a process site such as a petrochemical plant. Since these measuring devices not only measure the states of various plants, but also control the target plant by obtaining the manipulated variable from the measured data, they have an extremely important role. By the way, when these measuring devices become abnormal, it is obvious that the stability and safety of the plant cannot be maintained if the abnormal measurement data is directly used for controlling the plant. Further, the abnormality of the measuring device includes a temporary change in performance due to accumulation of dirt whose function is restored by cleaning, and a change in performance due to actual deterioration of the device, and it is difficult to distinguish them.
Furthermore, before the measuring device becomes abnormal, there are signs that it will become abnormal, for example, the ground state of the signal output deviates from a predetermined value, but it is possible to take prompt measures before it becomes abnormal. is necessary.

Therefore, in general, the measuring instrument is monitored by an operator or an observer, and when it is judged to be abnormal, repair or replacement is carried out to cope with the occurrence of abnormality in the measuring instrument. However, each measuring device has a different service life, and even the same device has individual differences, and in addition, the number of installed devices is very large. For this reason, it is not desirable from a variety of viewpoints, such as an increase in manufacturing cost, to always take a monitoring and maintenance system for those measuring devices that do not know when an abnormality will occur or will break down. Therefore, until now, attempts have been made to analyze the respective signal components by performing Fourier transform on the process signals output from those measuring devices to detect failure or deterioration. However, the data for each frequency component obtained by the Fourier transform has no information on time, so the data for each frequency component is inverse Fourier transformed to reproduce and use the information on time. .

[0004]

However, it is not easy to distinguish and extract, from the information, a temporary change in the measurement result due to dirt and a change in the measurement result due to deterioration of the device itself. It was Further, in the method based on the Fourier transform or the like, the inverse Fourier transform is required as described above, or all frequency bands are inspected, so it takes time to verify one device. If you try to process these things with a computer, for example, in order to support inspection of all equipment,
Those with a high degree of processing power have to be used, such computers are expensive and the process monitoring is very expensive.

The present invention has been made in order to solve the above problems, and uses the signal of the measurement result output from the measuring device to accurately measure the state of the measuring device at a low cost. The purpose is to be able to understand well.

[0006]

SUMMARY OF THE INVENTION An abnormality diagnosing apparatus of the present invention is a time-frequency signal obtained by wavelet-transforming a measurement signal, which is a time-series signal output from a measuring instrument, for a predetermined time with a predetermined number of expansions. select a wavelet transform section for converting the wavelet transform signal indicating the relationship, at least two or more Webure <br/> Tsu preparative conversion signal of each deployment times the separator at predetermined time intervals,
A distributed processing unit that calculates the dispersion of the wavelet-transformed signal in each of the divided sections, and a selection
For each of the generated wavelet transform signals, an abnormality determination unit is provided for determining an abnormality when the variation in the interval of variance becomes similar to a preset abnormality pattern. With the above configuration, a vibration state different from the steady vibration state, which cannot be captured as it is by the time-series signal output from the measuring instrument, is determined to be abnormal.

Further, according to the abnormality diagnosis method of the present invention, a time-series signal, which is a time-series signal, output from a measuring device for a predetermined time is wavelet-transformed with a preset number of expansions to establish a relationship between time and frequency. a first step of converting the wavelet transform signal indicating the number of times each deployment
Chi at least two to select the wavelet transform signal separated at predetermined time intervals, a second step of obtaining the dispersion of the wavelet transform signal in that interval in each of which delimited section, respectively, which are selected U
For each wavelet-converted signal, a third step of determining an abnormality when the variation of the dispersion interval becomes similar to a preset abnormality pattern is configured. With the above configuration, a vibration state different from the steady vibration state, which cannot be captured as it is by the time-series signal output from the measuring instrument, is determined to be abnormal.

Further, the program storage medium of the present invention wavelet-transforms a measurement signal, which is a time-series signal for a predetermined time, which is a time-series signal, with a preset number of expansions, and shows the relationship between time and frequency. a first step of converting the wavelet transform signal indicating each exhibition
Selects at least two or more wavelet transform <br/> No. signal of open times separated at predetermined time intervals, the delimited section second obtaining the variance of the wavelet transform signal in the section in each Step and selected
For each wavelet-transformed signal, a program composed of a third step of judging an abnormality when the variation in the interval of variance becomes similar to a preset abnormal pattern is stored. With the above configuration, a vibration state different from the steady vibration state, which cannot be captured as it is by the time-series signal output from the measuring instrument, is determined to be abnormal.

[0009]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings. In a petrochemical plant, for example, a floating liquid level gauge is used to grasp the storage amount in a tank that stores crude oil before refining. In a certain tank, the intermediate product flows in at a predetermined flow rate and flows out to the next process at a predetermined flow rate. The inflow amount and the outflow amount are controlled so that the liquid level of the intermediate product in the tank becomes a predetermined value. Its simple structure is shown in Fig. 1.
As shown in (a), the liquid level gauge 100 includes a bearing unit 10
1, a torque tube 102 is fixed, and a lever 103 is fixed to the tip of the torque tube 102 so as to be rotatable around the torque tube 102 as an axis.
A tubular float 104 is fixed to the tip. Further, a rod 105 is fixed to a tip end portion of the torque tube 102 to which the lever 103 is attached, and one end of the rod 105 penetrates the inside of the torque tube 102 and the bearing portion 101.
Protruding to the other side.

Therefore, the rotation of the lever 103 about the torque tube 102 is detected as the rotation of the rod 105 at one end of the rod 105. For example, if the strain gauge 106 is attached to the tip of the rod 105, the rotation of the lever 103 about the torque tube 102 becomes
It is detected as a change in the electrical output of the strain gauge 106. That is, the vertical movement due to the change in the liquid level of the float 104 is
It is detected as a change in the electrical output of the strain gauge 106. Since the float 104 side and the strain gauge 106 side can be completely separated via the bearing portion 101, the torque tube 102, the lever 103, and the float 1 are separated.
If a material having a corrosion resistance such as stainless steel is used for 04, the liquid level gauge 100 can be used for detecting the liquid level even if the liquid is a corrosive liquid.

Then, as shown in FIG. 1B, the liquid level of the intermediate product 111 in the tank 110 is measured by the liquid level gauge 100, and the measured liquid level is within a predetermined range in the control unit 120. The amount of inflow to the tank 110 and the amount of outflow from the tank 110 are controlled so as to enter. For example, the control unit 120 opens the inflow valve 113 of the inflow pipe 112 through which the intermediate product 111 flows into the tank 110 and opens the outflow valve 115 of the outflow pipe 114 through which the intermediate product 111 flows out of the tank 110. Level and 100
It is controlled by the signal from. Then, in this embodiment, the measurement signal output from the liquid level gauge 100 is
The abnormality is diagnosed by the abnormality diagnosis device 130 to detect the abnormality.

In the abnormality diagnosing device 130, first, the input measurement signal, which is a time-series signal, is wavelet-transformed by the wavelet transformer 131. Next, in the dispersion processing unit 132, a part of the wavelet transform signal obtained by the wavelet transform is divided at intervals of 24 hours for each number of expansions to show the degree of variation of the wavelet transform signal in each section. The variance value is calculated as the thing (dispersion), and this is designated as variance V. The variance value may be obtained by the following formula 1 when n variables are x1, x2, ..., Xn.

[Equation 1]

Instead of taking the variance value in each section, the variance may be taken by taking the difference between the maximum value and the minimum value. Then, the abnormality determination unit 133 determines and detects the abnormality of the measurement signal based on the section change such as the abnormality pattern indicating the obtained abnormal state of the variance V. In addition, the input unit 134 sets the number of expansions in the wavelet transform unit 131, or sets the determination reference in the abnormality determination unit 133, for example. In addition, on the display unit 135,
The change in the dispersion V obtained by the dispersion processing unit 132 is displayed, and the determination result by the abnormality determination unit 133 is displayed.

Here, first, the wavelet transform unit 131
The wavelet transform in FIG. The wavelet transform unit 131 first performs wavelet transform on the input measurement signal, which is a time-series signal, by using a preset wavelet function that is also preset based on preset preset scale parameters. Wavelet transform signal W expanded in the time-frequency domain represented as a result of the wavelet transform
Generate (a, b). At the time of this generation, the definition equation of the wavelet transform and its condition are expressed by the following equation 2.

[0015]

[Equation 2]

By the way, in the wavelet transform unit 131, the wavelet transform signal W (a, b) is actually used.
_Is calculated as a two-dimensional array W _ij of size I × J as shown in the following Expression 3, and this is output as a first distribution W _ij which is a wavelet transformed signal.

[0017]

[Equation 3]

Note that I is the number of scale parameters a (the number of expansions) in the wavelet transform, and J is the number of shift parameters b (equal to the number of time-series signal points). Also, in this embodiment, the basic wavelet function Ψ
The Meyer function shown in FIG. 2 was used as (t). Note that Ψ ^* represents the complex conjugate of Ψ. Therefore, in this embodiment, the wavelet transform signal expanded in the time-frequency domain represented by a real number can be obtained.

Hereinafter, taking the measurement signal actually obtained from the liquid level gauge 100 as an example, the abnormality diagnosing device 13 of this embodiment will be described.
The operation of 0 will be described. First, as shown in FIG. 3A, it is assumed that the measurement results for about 30 days in a month of the year have been obtained. In the measurement results thus obtained, the level greatly changes to −5% on the 10th day in FIG. 3A, for example. This is due to the change of the set value.
The signal of the measurement result output from 0 fluctuates greatly. As described above, it is not easy to detect an abnormality from the liquid level gauge 100 from the signal output state, while there is a large fluctuation in the output signal that is not abnormal.

Therefore, in this embodiment, as described above, the wavelet transform unit 131 performs the wavelet transform with the preset number of expansions i, and the wavelet transform signal W _ij is divided every 24 hours, respectively. The variance V _n (n = 1, 2, 3 ... 30) is obtained by the variance processing unit 132 for the section. here,
When the measurement signal is the result of measures the liquid level in 5-minute intervals, result in measured 288 times in 24 hours, W _i1 ~
One variance V _n is _obtained for 288 data of W _i288 . Then, in the abnormality determination unit 133,
When the change in the variance V _{n for} the one day becomes similar to a preset abnormality pattern, it is determined that an abnormality has occurred in the liquid level gauge 100.

Here, the number of expansions i described above is set in advance by making a diagnosis as follows.
That is, for example, from the measurement signal obtained from the liquid level gauge 100 as shown in FIG. 3A, the wavelet transform unit 131 is used to perform wavelet transform up to a predetermined development number I, and the wavelet transform at each development number i is performed. The signal W _ij is _divided, for example, every 24 hours, and the dispersion processing unit 132 obtains the degree of variation for each section. That is, first, the wavelet transform is performed with the number of expansion times 1, the wavelet transformed signal W _1j is divided every 24 hours, and the variance processing unit 132 obtains the variance V _n for each interval.
As shown in (b), the variance V _n is changed every 24 hours, that is, every day.

Similarly, the wavelet transform is performed with the number of expansion times 2, the wavelet transformed signal W _2j is divided every 24 hours, and the variance processing unit 132 obtains the variance V _n for each interval, as shown in FIG. The variance V _n is changed every day. In addition, the wavelet transform is performed with the number of expansions 3 (i = 3), the wavelet transformed signal W _3j is divided every 24 hours, and the variance processing unit 132 performs the variance V _n for each interval.
Then, as shown in FIG. 3D, the variance V _n for each day is _calculated.
To change. Further, the wavelet transform is performed with the number of expansions 4 (i = 4), and the wavelet transformed signal W _4j is 24
The distribution processing unit 132 obtains the dispersion V _n for each time segment, and the dispersion V _n is changed every day as shown in FIG.

On the other hand, changes in the state of the liquid level gauge are successively observed during the days when the measurement results are obtained. Here, for example, the liquid level gauge is supposed to be cleaned at the beginning of the month and during the month. Hereinafter, the relationship between the dirt of the float and the measurement signal will be considered. Tank 1 shown in FIG. 1 (b)
In Fig. 10, the intermediate product 111 is constantly flowing in and out, and there are various sources of vibration in the process. For this reason, the liquid surface of the intermediate product 111 always has large and small undulations. Therefore, since the float 104 is constantly vibrating, the vibration also appears in the measurement signal output from the liquid level gauge 100. Therefore, when measuring the liquid level,
I try to consider those vibrations.

However, if dirt is attached to the float 104,
Since the buoyancy of the float 104 changes, its vibration state also changes, which adversely affects the measurement results. Therefore, in general, the dirt on the float 104 is regularly cleaned to remove the dirt. For example, in the case described above, the float is cleaned at the beginning of the month and during the month.
However, I do not know if this cleaning interval is proper. Then, when the diagnostic results shown in FIGS. 3B to 3E are compared with the change in the state of the liquid level gauge, the daily variance V _{n of} W _1j shown in FIG. 3B is obtained. In the change of, there are large changes in the part that hits the beginning of the month and the part that hits during the month. And this repeated dispersion V
The change in _n increases more rapidly than in the initial state where the variance V _n is small, and the increased value continues to be almost constant thereafter.

Therefore, as a result of this diagnosis, the number of expansions set in advance in the above-mentioned wavelet transform unit 131 for abnormality determination is set to 1. That is, a wavelet-transformed signal having a development count of 1 is selected. Further, the threshold value set in the abnormality determination unit 133 is set to 0.05 as a variation in the degree of variation.
Is set. These settings are performed by the input unit 134. By setting in this way, the above-described abnormality diagnosis device 130 determines an abnormality by the following. First, the measurement signal obtained by the wavelet transform unit 131 is wavelet transformed with the number of expansions 1, and the wavelet transformed signal W _1j is divided by the dispersion processing unit 132 every day to obtain the degree of variation V _n for each section. And
In the abnormality determination unit 133, when the variation of the variance V _n exceeds 0.05, for example, the float 103
It is determined that abnormality has occurred in the liquid level gauge 100 due to the attachment of dirt to the liquid level gauge 100. The determination result of this abnormality is displayed on the display unit 135.

Further, for example, by diagnosing a measurement signal of a longer number of days and setting the number of times of deployment and a threshold value corresponding to other abnormalities, it is possible to detect abnormalities of two or more different causes. . For example, the diagnosis is performed in the same manner as described above using the measurement results for about 60 days in a certain month shown in FIG. 4A. First, the wavelet transform is performed with the number of expansion times 1, the wavelet transformed signal W _1j is divided every 24 hours, and the variance processing unit 132 obtains the variance V _n for each interval, and as shown in FIG. , Ie 1
6 of daily variance V _n (n = 1, 2, 3, ... 60)
Change to 0 days.

Similarly, the wavelet transform is performed with the number of expansions of 2, the wavelet transformed signal W _2j is divided every 24 hours, and the variance processing unit 132 obtains the variance V _n for each interval, as shown in FIG. Variance of daily V _n of 6
Change to 0 days. Further, the wavelet transform is performed with the number of expansions 3 (i = 3), and the wavelet transformed signal W _3j
Is divided every 24 hours, and the distributed processing unit 132 is divided for each section.
Then, the variance V _n is obtained, and as shown in FIG. 4D, the variance V _n is changed every day for 60 days. The number of deployments is 4
Wavelet transform is performed at (i = 4), the wavelet transform signal W _4j is divided every 24 hours, and the variance V _n is obtained by the variance processing unit 132 for each interval, and as shown in FIG. Of the variance V _n of 60 days.

On the other hand, changes in the state of the liquid level gauge are successively observed within the 60 days when the measurement results are obtained. As a result,
For example, during the 60 days when the measurement results shown in FIGS. 4B to 4E are obtained, the torque tube 102 of the liquid level gauge 100 having the structure shown in FIG. 1 is deteriorated by the inspection on the 45th day. It turned out that When the torque tube 102 deteriorates, the force that pushes down the float 104 via the lever 103 decreases, and the liquid level height is detected higher than it actually is. However, this is
It is hardly seen in the measurement results shown in (a).

Then, comparing the diagnostic results shown in FIGS. 4 (b) to 4 (e) with the change in the state of the liquid level gauge 100, in FIG. 4 (d), after about 33 days, the dispersion V There is a change in _n . Therefore, by comparing the actual state of the liquid level gauge 100 with the results of FIG. 4 (b) to FIG. 4 (e), the wavelet transform is performed with the expansion number of 3 and the wavelet transform signal W _3j is divided every 24 hours. It is diagnosed that the deterioration of the torque tube 102 can be detected in the change of the variance V _n obtained for each section.

From the above, using the input unit 134,
The number of expansions set in advance in the wavelet transformation unit 131 for abnormality determination is set to 3, and 1 is set to the threshold value of variation in variation set in the abnormality determination unit 133. By setting in this way,
In the above-described abnormality diagnosis device 130, an abnormality due to deterioration of the torque tube 102 is determined by the following. First, the measurement signal obtained by the wavelet transform unit 131 is wavelet-transformed with the expansion count of 3, and the wavelet-transformed signal W _3j is divided by the dispersion processing unit 132 every day to obtain the degree of variation for each segment. Then, in the abnormality determination unit 133, when there is a change in the variance V _n that exceeds 1, for example, the torque tube 102.
It is determined that the liquid level gauge 100 is abnormal due to deterioration. Then, the determination result of this abnormality is displayed on the display unit 135.

As described above, according to the abnormality diagnosing device of this embodiment, the abnormality of the liquid level gauge due to the dirt attached to the float is measured by using the measurement signal output from the liquid level gauge. The abnormality of the liquid level gauge due to the deterioration of the torque tube can be detected separately. In addition,
In the above embodiment, the wavelet transform is performed up to the expansion number 4 for diagnosis, but the number of expansions is not limited to this, and the expansion may be performed five times or more. Further, the variance V _n is calculated for the wavelet transform result every 24 hours, but the present invention is not limited to this, and for example, it may be calculated for the wavelet transform result every 12 hours.

Further, in the above embodiment, the liquid level gauge using the float was described as an example, but the present invention is not limited to this, and it goes without saying that it can be used for abnormality detection of various other measuring instruments. Absent. That is, first, the measurement signal obtained from the measuring device is wavelet-transformed by the wavelet transform unit, and the variance V is obtained for each interval by dividing the measurement signal at predetermined intervals in each expansion number.
Data for _n time series changes. Then, by comparing these data with the actual state of the measuring device (such as an abnormal state that occurred) during the period when the measurement signal was obtained,
Diagnose at which deployment frequency the state change that occurred in the measuring device appears.

Then, by using the diagnostic result, the number of expansions of the wavelet transform unit, the threshold value of the abnormality judging unit, and the like are set, and the abnormality of the measuring instrument which is not exposed in the measurement signal is detected by this operation. It becomes possible to detect with the form abnormality diagnostic device. Further, not only the threshold value but also an abnormality pattern indicating an abnormal state that has actually occurred may be set in the abnormality determination unit, and determination may be made by comparison with the abnormality pattern. For example, in the above description, for example, when the variance increases above the threshold value, it is determined to be abnormal, but the state when the variance decreases below a predetermined value may be diagnosed. For example, when the vertical movement of the float is hindered for some reason, the vibration component that normally occurs may be reduced or eliminated. Regarding such an abnormality, it may be determined to be abnormal when the variance is smaller than a predetermined value.

By the way, the abnormality diagnosing device of the above-described embodiment is provided with the wavelet transform section, the distributed processing section, and the abnormality determining section.
May be performed in. That is, as shown in FIG. 5, the input time-series signal may be converted into a digital signal by the analog / digital converter 501, and the converted signal may be processed by the CPU 502.

In the CPU 502, first, in the first step, a time-series signal, which is a measurement signal for a predetermined time, is wavelet-transformed based on a wavelet function to obtain a wavelet signal W _ij showing a relationship between time and frequency. Next, in a second step, the wavelet signal W _ij is distributed at predetermined intervals, for example, 1 ≦ j ≦ 288, 289 ≦ j ≦ 5.
76, ..., And the variance V _n (n = 1, 2, 3, ...) Is obtained for each section by dividing it every 24 hours. Next, in the third step, when the change in the variance V _n becomes similar to the predetermined abnormal pattern, it is determined that an abnormality corresponding to the number of expansions of the first wavelet transform has occurred. Then, for example, in the fourth step, the fact that an abnormality has occurred is displayed on the monitor 506.

The CPU 502 is the bus 502.
The series of operations described above is performed by the program expanded in the memory 503 connected to a. In addition, the memory 503
The program expanded in is stored in the external storage device 504. In addition, the wavelet function used for the wavelet transform is also stored in the external storage device 50.
Just prepare for 4. Further, for example, the setting of parameters such as the number of expansions in the wavelet transform and the setting of the threshold value in the abnormality determination may be input using the keyboard 505 as an input unit. Then, the diagnosis result and the like are displayed on the monitor 506 as a display unit.

[0037]

As described above, according to the present invention, a time-series signal, which is a time-series signal output from a measuring instrument, is wavelet-transformed with a preset number of expansions to obtain time and frequency. a wavelet transform section for converting the wavelet transform signal indicating the relationship, the exhibition
Selects at least two or more wavelet transform <br/> No. signal of open times separated at predetermined time intervals, the distributed processing unit for determining the variance of the wavelet transform signal in that interval in each of which delimited section And each selected
Each wavelet transform signal is provided with an abnormality determination unit that determines an abnormality when the variation in the variance section becomes similar to a preset abnormality pattern.

Further, in the abnormality diagnosing method of the present invention, the time-series signal output from the measuring device is wavelet-transformed with a predetermined number of expansions to obtain a relationship between time and frequency. a first step of converting the wavelet transform signal indicating the number of times each deployment
Chi at least two to select the wavelet transform signal separated at predetermined time intervals, a second step of obtaining the dispersion of the wavelet transform signal in that interval in each of which delimited section, respectively, which are selected U
For each wavelet-converted signal, a third step of determining an abnormality when the variation of the dispersion interval becomes similar to a preset abnormality pattern is configured.

Further, the program storage medium of the present invention wavelet-transforms a measurement signal, which is a time-series signal output from a measuring instrument, for a predetermined period of time at a predetermined number of expansions, and shows the relationship between time and frequency. a first step of converting the wavelet transform signal indicating each exhibition
Selects at least two or more wavelet transform <br/> No. signal of open times separated at predetermined time intervals, the delimited section second obtaining the variance of the wavelet transform signal in the section in each Step and selected
For each wavelet-transformed signal, a program composed of a third step of judging an abnormality when the variation in the interval of variance becomes similar to a preset abnormal pattern is stored.

With the above configuration, a vibration state different from the steady vibration state, which cannot be captured as it is by the time series signal output from the measuring instrument, is determined to be abnormal. As a result, according to the present invention, the state of the measuring device can be grasped by the process of the wavelet transform and the process of obtaining the variance using the signal of the measurement result output from the measuring device. Therefore, according to the present invention, the amount of calculation can be greatly reduced as compared with the case where Fourier transform or the like is used, and the state of the measuring device can be grasped with high accuracy.

[Brief description of drawings]

FIG. 1 is a perspective view (a) showing a configuration of a liquid level gauge and a configuration diagram (b) showing a configuration of an abnormality diagnosis device.

FIG. 2 is an explanatory diagram showing a Mayer function used in a wavelet transform.

FIG. 3 is an explanatory diagram (a) showing a measurement result for 30 days by a liquid level gauge and an explanatory diagram showing an output example of the abnormality diagnosis device in the embodiment.

FIG. 4 is an explanatory diagram (a) showing a measurement result for 60 days by a liquid level gauge and an explanatory diagram showing an output example of the abnormality diagnosis device in the embodiment.

FIG. 5 is a configuration diagram showing another embodiment of the present invention.

[Explanation of symbols]

100 ... Liquid level gauge, 101 ... Bearing part, 102 ... Torque tube, 103 ... Lever, 104 ... Float, 105 ... Rod, 106 ... Strain gauge, 110 ... Tank, 111 ...
Intermediate product, 112 ... Inflow pipe, 113 ... Inflow valve,
114 ... Outflow pipe, 115 ... Outflow valve, 120 ... Control unit, 130 ... Abnormality diagnosis device, 131 ... Wavelet transform unit, 132 ... Distributed processing unit, 133 ... Abnormality determination unit, 1
34 ... Input part, 135 ... Display part.

─────────────────────────────────────────────────── ─── Continued Front Page (72) Yoshiyuki Shirakawa 2-10-1 Toranomon, Minato-ku, Tokyo Within Japan Energy Co., Ltd. (72) Inventor Hideki Sasaoka 2-12-19 Shibuya, Shibuya-ku, Tokyo Stock Company Yamatakenai (72) Inventor Hirohiko Kato 2-12-19 Shibuya, Shibuya-ku, Tokyo Yamatakeuchi (72) Inventor Junji Nishimura 1-1, Shinurashima-cho, Kanagawa-ku, Kanagawa Prefecture Yamatake Industrial Systems Co., Ltd. (56) Reference JP-A-8-219955 (JP, A) JP-A-9-73443 (JP, A) JP-A-9-139676 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) G01D 21/00 G05B 23/02 G06F 9/44 G06F 11/22 G06F 17/17 G01H G01M

Claims (3)

(57) [Claims] 1. A wavelet for wavelet-transforming a measurement signal of a predetermined time, which is a time-series signal output from a measuring instrument, with a preset number of expansions to a wavelet-transformed signal showing the relationship between time and frequency. a conversion unit, each select at least two or more of said wavelet transform signal of each deployment times separated at predetermined time intervals, determining the variance of the wavelet transform signal in that interval in each of which delimited section A dispersion processing unit, and an abnormality determination unit that determines an abnormality when the section change of the dispersion is similar to an abnormality pattern indicating an abnormal state set in advance for each of the selected wavelet transform signals. An abnormality diagnosis device comprising: 2. A method of wavelet-transforming a measurement signal of a predetermined time, which is a time-series signal output from a measuring instrument, with a preset number of expansions to a wavelet-transformed signal showing a relationship between time and frequency. and first step, a dispersion of at least two of said each selected wavelet transform signal separated for each predetermined time interval, the wavelet transform signal in that interval in each of which delimited section of the deployment times A second step of obtaining, and a third step of judging an abnormality for each of the selected wavelet-transformed signals when the variation in the interval of the dispersion is similar to a preset abnormal pattern indicating an abnormal state An abnormality diagnosis method comprising: 3. A method of wavelet-transforming a measurement signal for a predetermined time, which is a time-series signal output from a measuring instrument, with a preset number of expansions to a wavelet-transformed signal indicating a relationship between time and frequency. and first step, a dispersion of at least two of said each selected wavelet transform signal separated for each predetermined time interval, the wavelet transform signal in that interval in each of which delimited section of the deployment times A second step of obtaining, and a third step of judging an abnormality for each of the selected wavelet-transformed signals when the variation in the interval of the dispersion is similar to a preset abnormal pattern indicating an abnormal state And a program storage medium that stores a program configured by the steps of.