JP7415850B2 - Structural abnormality diagnosis device and structural abnormality diagnosis method - Google Patents

Description

The present invention relates to a technique for diagnosing structural system abnormalities using vibration waveform data of rotating machines.

Non-Patent Document 1 describes the calculation of structural feature parameters for identifying structural system abnormalities from vibration waveform data measured in multiple positions and multiple directions, and sequential multivariate analysis (principal component analysis) using the structural feature parameters. A diagnostic method for identifying abnormal conditions in a structural system is disclosed.

Seki Shōji, 5 others, "Rotating machine structural system abnormality diagnosis method using multi-position/multi-directional vibration signal fusion and sequential multivariate analysis", Journal of the Japan Society of Equipment Management, Vol. 29, No. 2

Non-Patent Document 1 has the following four problems (1) to (4).

(1) The type of structural abnormality is identified, but it has not been evaluated quantitatively.In Non-Patent Document 1, principal component analysis is performed using structural feature parameters, and the distribution of the state to be determined and the distribution at the time of diagnosis are at a certain distance. The status is determined based on whether it is within or above. Therefore, it is not quantitatively evaluated how abnormal the condition is, and it is not possible to determine whether the condition of the equipment requires an emergency or whether it is in a condition that can withstand operation a little more.

(2) Since states are sequentially determined, only one type of state can be identified. In Non-Patent Document 1, abnormality is determined by sequential multivariate analysis. The sequential determination is, for example, the processes (I) to (III).
(I) Distinguish between normal and other states.
(II) Distinguish between unbalanced and other states, excluding normal.
(III) Distinguish between misalignment and other conditions, excluding normality and imbalance.

With this method, once the state is determined in the first stage, the remaining states are not determined, so if there is vibration waveform data that can be determined as normal or misaligned, in the above example, it will only be determined as normal. .

(3) Rotational frequency deviation is not taken into account Many structural feature parameters use rotational frequency components or their higher-order components, and it is difficult to accurately determine the rotational frequency when performing multi-band pass filter processing. becomes important.

When a structural abnormality occurs, the rotational speed may slow down depending on the state of the abnormality, and the position of the rotational frequency may shift. As shown in FIG. 9, when the rotational frequency f _r in the normal state deviates from the rotational frequency f' _r in the abnormal state, the harmonics 2f _r , 3f _r , 4f _r , 5f _r in the normal state and the harmonic 2f in the abnormal state ' _r , 3f' _r , 4f' _r , and 5f' _r are many times larger, so when applying a multi-band pass filter, there are cases where high-order harmonic components of the rotational frequency cannot be left behind by filtering. be.

(4) Some of the structural characteristic parameters can be calculated for each measurement axis. There is no discussion on whether parameters for the five directions (direction and horizontal direction) should be used.

Diagnosis is performed using structural characteristic parameters in Non-Patent Document 1, and these structural characteristic parameters include the “horizontal and vertical amplitude ratio” and the “left and right (load side/anti-load side) vertical There are parameters that are obtained using information in a plurality of axis directions, such as "amplitude ratio", and parameters that are obtained for each axis, such as "rotational frequency component ratio" and "difference in skewness". Regarding the latter parameter, Non-Patent Document 1 does not mention which axial parameter should be used.

In addition, in some structural system abnormalities, it is not known in which direction the vibration will increase, so if it is set in one direction, it may be determined that the abnormal state is normal even though it is abnormal, or it may be determined that the vibration is in a different type of abnormality. It is determined that there is.

In view of the above, it is a problem for structural system abnormality diagnostic devices to quantitatively calculate and diagnose the degree of abnormality for each type of abnormality.

The present invention has been devised in view of the above-mentioned conventional problems, and one aspect of the present invention is a structural abnormality diagnosis device for diagnosing structural abnormalities in rotating machines, which include vertical, horizontal, and axial directions. a waveform data dividing unit that divides the learning vibration waveform data and the diagnostic vibration waveform data; and extracting a rotational frequency component of the divided learning vibration waveform data and the diagnostic vibration waveform data and a harmonic of the rotational frequency component. and a structural feature parameter that calculates a plurality of structural feature parameter reference data and a plurality of diagnostic structural feature parameters based on the filtered learning vibration waveform data and the diagnostic vibration waveform data. a calculation unit; and a structural feature parameter abnormal value that calculates abnormal values of the plurality of diagnostic structural feature parameters based on the average and variance of the structural feature parameter reference data and the average and variance of the diagnostic structural feature parameters. a calculation unit; and a structural abnormality abnormal value calculation unit that selects and averages some of the abnormal values of the plurality of diagnostic structural feature parameters to calculate an abnormal value for each abnormal state. Features.

Further, as one aspect thereof, the rotation frequency having the highest frequency component within a range of several Hz before and after the normal state rotation frequency of the divided learning vibration waveform data and the diagnostic vibration waveform data is determined as the corrected rotation frequency. It is characterized in that it includes a frequency correction section, and the multiband-pass filter processing section extracts the corrected rotational frequency and harmonics of the corrected rotational frequency.

Further, as one aspect thereof, abnormal values in the vertical direction, horizontal direction, and axial direction of the diagnostic structural feature parameter for which abnormal values in the vertical direction, horizontal direction, and axial direction are calculated are averaged and output as one abnormal value. It is characterized by being equipped with a triaxial abnormal value averaging processing section.

Further, as one aspect thereof, the structural feature parameter calculation unit calculates the structural feature parameter reference data and the diagnostic structural feature parameters such as a rotational frequency component ratio, a second harmonic rate of the rotational frequency, and a three-dimensional harmonic of the rotational frequency. Harmonic rate, higher harmonic rate of rotation frequency, high and low frequency component ratio, difference in skewness of vibration waveform data after filtering, difference in kurtosis of vibration waveform data after filtering, vertical and horizontal directions The amplitude ratio of rotational frequency components in the vertical direction and the axial direction, and the amplitude ratio of the rotational frequency components in the horizontal direction and the axial direction are calculated.

According to the present invention, in the structural system abnormality diagnosing device, it becomes possible to quantitatively calculate and diagnose the degree of abnormality for each type of abnormality.

FIG. 3 is a diagram showing analysis processing of the structural abnormality diagnosis device in the first embodiment. The figure which shows an example of the division method of vibration waveform data. The figure which shows the other example of the division method of vibration waveform data. FIG. 3 is a schematic diagram showing processing contents of a multi-band-pass filter processing section. FIG. 7 is a diagram illustrating analysis processing of the structural abnormality diagnosis device in Embodiment 2. FIG. 3 is a schematic diagram showing processing contents of a rotation frequency correction section. FIG. 7 is a diagram showing analysis processing of the structural abnormality diagnosis device in Embodiment 3; FIG. 3 is a schematic diagram showing abnormal value averaging processing of structural feature parameters obtained along three axes. FIG. 3 is a diagram showing rotational frequencies and higher harmonics during normal and abnormal times.

Embodiments 1 to 3 of the structural abnormality diagnosing apparatus and structural abnormality diagnosing method according to the present invention will be described in detail below with reference to FIGS. 1 to 8.

[Embodiment 1]
FIG. 1 shows the analysis process of the structural abnormality diagnosis device in the first embodiment. In the first embodiment, the device to be diagnosed is a rotating machine, and the abnormality to be diagnosed is a structural abnormality (misalignment, imbalance, etc.). In the first embodiment, a vibration sensor is attached to a rotating machine, and characteristic structural feature parameters specific to a structural abnormality of the rotating machine are calculated from vibration waveform data obtained from the vibration sensor.

Embodiment 1 shows a method that uses these structural feature parameters to quantify (digitize) and diagnose the degree of structural system abnormality for each type of abnormality, and solves the problems of Non-Patent Document 1 (1). ) to (2) are solved.

The method of the first embodiment has two functions: a learning function and a diagnostic function.

In the learning function, reference data (average and variance) for each structural feature parameter at the time of learning (normal time) is generated by performing the flow of the learning function shown in FIG.

Similarly, the diagnostic function calculates the average and variance of each structural feature parameter during diagnosis, and uses the average and variance of each structural feature parameter obtained by the learning function and diagnostic function to find abnormal values.

In addition, when calculating structural feature parameters, the spectrum of vibration waveform data is obtained and the structural feature parameters have a specific frequency component. However, if the input vibration waveform data is for one measurement, the structure using the spectrum result is Since the mean and variance of the feature parameters cannot be calculated, a waveform data dividing section is provided for both learning and diagnosis to enable calculation.

Hereinafter, the function and operation of each function of the structural system abnormality diagnosis device in the first embodiment will be explained.

[Learning function]
[Learning vibration waveform data 1]
In the first embodiment, the learning vibration waveform data 1 requires vibration waveform data in three directions: vertical, horizontal, and axial directions, and one measurement or multiple vibration waveform data are considered for each direction. It will be done. At this time, if the learning vibration waveform data 1 is vibration waveform data in a normal state in which no structural abnormality has occurred, the diagnostic accuracy will be improved.

[Waveform data division section 2]
The waveform data dividing section 2 divides the learning vibration waveform data 1. The division here may be, for example, equally divided as shown in FIG. For example, one piece of learning vibration waveform data 1 is divided into eight parts to obtain eight pieces of learning vibration waveform data 1 with N pieces of data.

Furthermore, as shown in FIG. 3, a method of obtaining a plurality of waveform data by dividing (in this case, "cutting out") the waveform data while superimposing them while retaining the number N of data can be considered. For example, from one learning vibration waveform data 1, eight learning vibration waveform data 1 having a data number N are obtained.

[Multi-bandpass filter processing unit 3]
The multi-band-pass filter processing unit 3 performs multi-band-pass filter processing on the learning vibration waveform data 1, as in Non-Patent Document 1. FIG. 4 shows the processing of the multi-band-pass filter processing section 3. The multi-band-pass filter processing unit 3 extracts only the rotational frequency component and its harmonics used in the structural feature parameters, and sets the spectrum values of other frequencies to zero.

The spectrum waveform of the learning vibration waveform data 1 (rotation frequency is f _r ) shown in FIG. 4(a) is shown in FIG. 4(b). As shown in FIG. 4(b), waveforms can be seen in addition to the rotational frequency _fr , the second harmonic _2fr , the third harmonic _3fr , the fourth harmonic _4fr , and the fifth harmonic _5fr . As shown in FIG. 4(c), the spectrum waveform filtered by the multi-band-pass filter processing unit 3 has a rotational frequency f _r , a second harmonic _2fr , a third harmonic _3fr , a fourth harmonic 4fr _, Only a few Hz around the fifth harmonic 5f _r is extracted, and the rest are removed. The learning vibration waveform data 1 filtered by the multi-band-pass filter processing unit 3 has a waveform as shown in FIG. 4(d).

[Structural feature parameter calculation unit 4]
In the first embodiment, of the 12 types of structural characteristic parameters used in Non-Patent Document 1, 8 types (p ₁ to p ₈ shown below) are used as they are, and the relationship between the vibration waveform data in the three axial directions is 10 types of structural feature parameters are defined, including two types of parameters (p ₉ and p ₁₀ shown below) that take into account.

Let the ten types of structural feature parameters be p ₁ to p ₁₀ and let F _n (f) be the frequency component at frequency f of the learning vibration waveform data 1. Further, the rotation frequency is defined as f _r and the order of the harmonic component is defined as i.

[Rotation frequency component ratio p ₁ ]

[Second harmonic rate p ₂ of rotation frequency]

[Third harmonic rate p ₃ of rotation frequency]

[Higher harmonic rate of rotation frequency (4th to 10th harmonic) p ₄ ]

[High and low frequency component ratio p ₅ ]

[Distortion degree p ₆ of vibration waveform data after filter processing]

*The skewness of the learning vibration waveform data 1 after filter processing is S _n .

[Kurtosis p ₇ of vibration waveform data after filter processing]

*The kurtosis of the learning vibration waveform data 1 after filter processing is K _n .

Furthermore, the frequency component in the vertical direction at the frequency f of the learning vibration waveform data 1 is F _nv (f), the frequency component in the horizontal direction is F _nh (f), and the frequency component in the axial direction is F _na (f).

[Amplitude ratio p ₈ of vertical and horizontal rotational frequency components]

[Amplitude ratio p ₉ of vertical and axial rotational frequency components]

[Amplitude ratio p ₁₀ of horizontal and axial rotational frequency components]

The above structural feature parameters p ₁ to p ₁₀ are calculated for each of the plurality of divided learning vibration waveform data 1, and the number of learning structural feature parameter reference data 5 is equal to the number of learning vibration waveform data 1 for each structural parameter. are obtained respectively.

[Diagnostic function]
[Diagnostic vibration waveform data 6]
Similar to the learning vibration waveform data 1, the diagnostic vibration waveform data 6 requires vibration waveform data in three directions: vertical, horizontal, and axial directions, and one or more vibration waveforms for each direction are required. Data can be considered.

At this time, assuming that diagnostic vibration waveform data 6 is obtained periodically from a permanently installed vibration sensor, diagnosis is performed using one measurement of diagnostic vibration waveform data 6, and the quantified abnormal state values are analyzed in a time series. Status trends can be monitored by arranging them.

[Waveform data division section 7]
This is similar to the waveform data dividing section 2.

[Multi-bandpass filter processing unit 8]
This is similar to the multi-band pass filter processing section 3.

[Structural feature parameter calculation unit 9]
This is similar to the structural feature parameter calculation unit 4. As many diagnostic structural characteristic parameters 10 as the diagnostic vibration waveform data 6 are obtained for each structural characteristic parameter. Hereinafter, the 10 types of structural characteristic parameters will be referred to as p ₁ to p ₁₀ , and the frequency component at the frequency f of the diagnostic vibration waveform data 6 will be referred to as F _d (f).

[Rotation frequency component ratio p ₁ ]

[Second harmonic rate p ₂ of rotation frequency]

[Third harmonic rate p ₃ of rotation frequency]

[Higher harmonic rate of rotation frequency (4th to 10th harmonic) p ₄ ]

[High and low frequency component ratio p ₅ ]

[Distortion degree p ₆ of vibration waveform data after filter processing]

*The skewness of the diagnostic vibration waveform data 6 after filter processing is defined as S _d .

[Kurtosis p ₇ of vibration waveform data after filter processing]

*The kurtosis of the diagnostic vibration waveform data 6 after filter processing is K _d .

Further, the frequency component in the vertical direction at the frequency f of the diagnostic vibration waveform data 6 is F _dv (f), the frequency component in the horizontal direction is F _dh (f), and the frequency component in the axial direction is F _da (f).

[Amplitude ratio p ₈ of vertical and horizontal rotational frequency components]

[Amplitude ratio p ₉ of vertical and axial rotational frequency components]

[Amplitude ratio p ₁₀ of horizontal and axial rotational frequency components]

The above structural feature parameters p ₁ to p ₁₀ are calculated for each of the plurality of divided diagnostic vibration waveform data 6, and as many diagnostic structural feature parameters 10 as the number of diagnostic vibration waveform data 6 are obtained for each structural parameter. .

[Structural feature parameter abnormal value calculation unit 11]
Based on the mean and variance of each structural feature parameter reference data 5 obtained by the learning function and the mean and variance of each diagnostic structural feature parameter 10 obtained by the diagnostic function, abnormal values for each diagnostic structural feature parameter are determined. Perform the following calculation.

Y is the abnormal value of a certain diagnostic structural feature parameter 10, μ ₀ is the average value of the certain structural feature parameter standard data 5 obtained by the learning function, σ ₀ is the variance value, and is a certain diagnostic structural feature obtained by the diagnostic function. Let μ be the average value of parameter 10, and let σ be the variance value. Furthermore, a determination coefficient K is defined, which is set according to the installation status and operating status of the equipment.

As a result, the abnormal value Y of the diagnostic structural characteristic parameter 10 of the diagnostic vibration waveform data 6 is calculated.

[Structural abnormality abnormal value calculation unit 12]
Some of the abnormal values Y of the plurality of diagnostic structural feature parameters 10 are selected and averaged to obtain an abnormal value for each abnormal state. The optimal parameters for selecting the structural feature parameters 10 for diagnosis for calculating the abnormal value of this abnormal state vary depending on the installation status and size of the rotating machine to be diagnosed. Table 1 shows an example of abnormal value results when a certain device is unbalanced. Trend monitoring becomes possible by creating a time-series graph of the abnormal values for each abnormality.

As described above, according to the first embodiment, by obtaining an abnormal value for each abnormal state without performing sequential processing, it is possible to prevent each abnormal state from being overlooked.

Additionally, since the degree of abnormality for each abnormal state has been quantified, trends can be monitored by monitoring in chronological order.

Furthermore, by dividing the waveform, diagnosis can be made with one measurement worth of vibration waveform data, so the condition can be diagnosed with a small amount of data.

Furthermore, most structural abnormalities are events that appear periodically in vibration waveform data, so even if the waveform is divided, it will appear in each divided waveform. Therefore, when the variance of each structural feature parameter is calculated, the value of the variance becomes small, and a high abnormal value tends to appear in the abnormal value calculation formula. In other words, it becomes easier to detect structural abnormalities.

[Embodiment 2]
As mentioned in Problem (3) of Non-Patent Document 1, many of the structural feature parameters use rotational frequency components or higher-order components thereof, and it is difficult to set the rotational frequency correctly when performing multi-band pass filter processing. What is required becomes important. In the second embodiment, a function to obtain the rotation frequency before the multi-band pass filter processing is added to solve the problem (3) of Non-Patent Document 1.

FIG. 5 shows the analysis process of the structural abnormality diagnosis device in the second embodiment. Rotational frequency correction units 13 and 14 are added before the multiband-pass filter processing units 3 and 8 for both the learning function and the diagnosis function.

Only the functions of the added rotational frequency correction units 13 and 14 will be explained. A schematic diagram of the process is shown in FIG. The rotational frequency correction units 13 and 14 input the learning vibration waveform data 1, the diagnostic vibration waveform data 6, and the rotation frequency f _r (or rotational speed) of the device in its normal state, and correct the input vibration waveform data. Find the rotational frequency f' _r .

As shown in the schematic diagram of FIG. 6, the process calculates a spectrum waveform from the input learning vibration waveform data 1 and diagnostic vibration waveform data 6, and calculates the frequency components in the range of several Hz before and after the rotation frequency f _r in the normal state. Find the highest frequency. This frequency is defined as the corrected rotational frequency _f'r . The multi-band-pass filter processing units 3 and 8 perform filter processing to extract the corrected rotational frequency _f'r and harmonics of _the corrected rotational frequency _f'r based on the corrected rotational frequency f'r. .

As described above, according to the second embodiment, by correctly determining the corrected rotational frequency _f'r before the multi-band-pass filter processing units 3 and 8, the corrected rotational frequency f'r and the corrected rotational frequency _f'r Filtering is performed so that harmonics of the rotational frequency f' _r remain, making it possible to obtain structural characteristic parameters with higher accuracy.

[Embodiment 3]
As stated in problem (4) of Non-Patent Document 1, the abnormal values of the structural characteristic parameters are obtained in each of the three axes, and it is not known in which axis the abnormal vibration will appear depending on the abnormal state. It is not possible to use abnormal values of structural feature parameters only in one axis direction, such as only in the horizontal direction. In the third embodiment, problem (4) of Non-Patent Document 1 is solved by taking the average of each abnormal value in order not to lose information on abnormal values of structural feature parameters in each axis direction.

FIG. 7 shows the analysis process of the structural abnormality diagnosis apparatus in the third embodiment. A triaxial abnormal value averaging processing section 15 is added after the structural feature parameter abnormal value calculation section 11 in the diagnostic function. Further, the flow of processing of the triaxial abnormal value averaging processing section 15 is shown in FIG.

The triaxial abnormal value average processing section 15 receives the triaxial abnormal values of a particular diagnostic structural feature parameter 10 calculated by the structural feature parameter calculation section 9. The specific diagnostic structural feature parameters 10 herein refer to seven types of structural feature parameters p ₁ to p ₇ of the structural feature parameters in the first embodiment. That is, it is a structural feature parameter for diagnosis that requires abnormal values in the vertical, horizontal, and axial directions.

As shown in FIG. 8, vibration waveform data in the h (horizontal) direction, v (vertical) direction, and a (axis) direction is obtained from a rotating machine by a vibration sensor. From the vibration waveform data in these three directions, structural feature parameters p1h, p2h, ... in the h (horizontal) direction, structural feature parameters p1v, p2v, ... in the v (vertical) direction, structural feature parameters p1a in the a (axial) direction, respectively. , p2a,... are obtained. For the diagnostic structural characteristic parameters 10 from p ₁ to p ₇ , an abnormal value Y is obtained for each axis direction (horizontal direction, vertical direction, axial direction). The three-axis direction abnormal value averaging processing section 15 averages the abnormal values Y in each axis direction and outputs the averaged value as one abnormal value.

By taking the average of the abnormal values Y of each diagnostic structural feature parameter 10 obtained from the diagnostic vibration waveform data 6 in three axial directions, it is possible to perform a diagnosis that takes into account vibration information in each direction. Therefore, the abnormality can be detected no matter which vibration direction of the three axes the abnormality occurs in.

Although only the specific examples described in the present invention have been described in detail above, it is obvious to those skilled in the art that various modifications and modifications can be made within the scope of the technical idea of the present invention. Naturally, such variations and modifications fall within the scope of the claims.

In Embodiments 1 to 3, the waveform data dividing units 2 and 7, the multiband-pass filter processing units 3 and 8, the structural feature parameter calculation units 4 and 9, and the rotational frequency correction units 13 and 14 are given different symbols, respectively. However, they do not need to be provided as separate parts.

1... Vibration waveform data for learning 2, 7... Waveform data division section 3, 8... Multi-band pass filter processing section 4, 9... Structural feature parameter calculation section 5... Structural feature parameter reference data 6... Vibration waveform data for diagnosis 10... Structural feature parameters for diagnosis 11... Structural feature parameter abnormal value calculation section 12... Structural system abnormality abnormal value calculation section 13, 14... Rotation frequency correction section 15... 3-axis direction abnormal value averaging processing section

Claims (5)

A structural abnormality diagnosis device for diagnosing structural abnormalities in rotating machinery,
a waveform data dividing unit that divides learning vibration waveform data and diagnostic vibration waveform data in the vertical direction, horizontal direction, and axial direction;
a multi-band pass filter processing unit that extracts a rotational frequency component of the divided learning vibration waveform data and the diagnostic vibration waveform data and a harmonic of the rotational frequency component;
a structural feature parameter calculation unit that calculates a plurality of structural feature parameter reference data and a plurality of diagnostic structural feature parameters based on the filtered learning vibration waveform data and the diagnostic vibration waveform data;
a structural feature parameter abnormal value calculation unit that calculates abnormal values of the plurality of diagnostic structural feature parameters based on the average and variance of the structural feature parameter reference data and the average and variance of the diagnostic structural feature parameters;
a structural abnormality abnormal value calculation unit that selects and averages some of the abnormal values of the plurality of diagnostic structural feature parameters to calculate an abnormal value for each abnormal state;
A structural abnormality diagnosis device characterized by comprising: comprising a rotational frequency correction unit that calculates, as a corrected rotational frequency, a frequency with the highest frequency component within a range of several Hz before and after the normal state rotational frequency of the divided learning vibration waveform data and the diagnostic vibration waveform data;
2. The structural abnormality diagnosis device according to claim 1, wherein the multi-band-pass filter processing unit extracts the corrected rotational frequency and harmonics of the corrected rotational frequency. Abnormal values in the vertical, horizontal, and axial directions are averaged in the vertical, horizontal, and axial directions of the structural feature parameter for diagnosis and output as one abnormal value. 3. The structural system abnormality diagnosing device according to claim 1, further comprising a section. The structural feature parameter calculation unit calculates, as the structural feature parameter reference data and the diagnostic structural feature parameters, a rotational frequency component ratio, a second harmonic rate of the rotational frequency, a third harmonic rate of the rotational frequency, and a rotational frequency. High-order harmonic rate, high and low frequency component ratio, difference in skewness of vibration waveform data after filter processing, difference in kurtosis of vibration waveform data after filter processing, amplitude ratio of vertical and horizontal rotational frequency components , the structural system abnormality diagnosis device according to any one of claims 1 to 3, characterized in that the amplitude ratio of rotation frequency components in the vertical direction and the axial direction is calculated, and the amplitude ratio of the rotation frequency components in the horizontal direction and the axial direction is calculated. . A structural abnormality diagnosis method for diagnosing structural abnormalities in rotating machinery, the method comprising:
The waveform data division section divides the learning vibration waveform data in the vertical, horizontal, and axial directions,
a multi-band pass filter processing unit extracts a rotational frequency component of the divided learning vibration waveform data and a harmonic of the rotational frequency component;
a structural feature parameter calculation unit calculates a plurality of structural feature parameter reference data based on the filtered learning vibration waveform data;
The waveform data division section divides the diagnostic vibration waveform data in the vertical, horizontal, and axial directions,
a multi-band pass filter processing unit extracts a rotational frequency component of the divided diagnostic vibration waveform data and a harmonic of the rotational frequency component;
a structural feature parameter calculation unit calculates a plurality of diagnostic structural feature parameters based on the filtered diagnostic vibration waveform data;
a structural feature parameter abnormal value calculation unit calculates abnormal values of the plurality of diagnostic structural feature parameters based on the average and variance of the structural feature parameter reference data and the average and variance of the diagnostic structural feature parameters;
a structural abnormality abnormal value calculation unit selects some of the abnormal values of the plurality of diagnostic structural feature parameters and averages them to calculate an abnormal value for each abnormal state;
A method for diagnosing structural system abnormalities.

Images