JP2021063769A - Abnormality diagnosis device, abnormality diagnosis method, and abnormality diagnosis program - Google Patents

Abstract

To diagnose abnormality in a self-circulating drive device more reliably than in the conventional art.SOLUTION: An abnormality diagnosis device includes: a data acquisition unit 131 that acquires first vibration characteristic data detected when a self-circulating drive device is normal and second vibration characteristic data detected at the time of diagnosis; a parameter calculation unit 132 that processes each of the first vibration characteristic data and the second vibration characteristic data with a statistics information filter, and calculates feature parameters including at least two or more feature parameters for each of the first vibration characteristic data and the second vibration characteristic data after processing; a principal component analysis unit 134 that extracts two principal components by a principal component analysis method for normalized feature parameters; and a first abnormality diagnosis unit 135 that diagnoses abnormality based on the Euclidean distance between a first point and a second point corresponding to the two principal components extracted from the feature parameters related to each of the first vibration characteristic data and the second vibration characteristic data.SELECTED DRAWING: Figure 2

Description

The present invention relates to an abnormality diagnosis device, an abnormality diagnosis method, and an abnormality diagnosis program. In particular, the present invention relates to an abnormality diagnosing device, an abnormality diagnosing method, and an abnormality diagnosing program for diagnosing an abnormality in a self-circulating drive device including a sleeve portion and a bearing portion.

In the drive device, when the abnormality diagnosis is performed based on the vibration characteristic data output from the sensor installed in the drive device, the S (signal) / N (noise) ratio of the obtained signal is low, and the abnormality is detected accurately. It has been pointed out that there is a problem that it cannot be done.
Therefore, the conventional abnormality diagnosis is performed exclusively by the five senses of the inspector, and a high degree of skill is required to ensure its accuracy.

On the other hand, the following method is known as a means for removing noise from a signal having a low S / N ratio.
For example, there are a bearing diagnosis method using a high-pass filter as described in Patent Document 1 and a method of extracting an abnormal signal based on a power spectrum ratio as described in Patent Document 2. However, the former method has a problem that it is difficult to determine the cutoff frequency and it is not possible to select and extract the abnormal signal, and it is difficult to extract the abnormal signal sufficiently accurately. Further, the latter method has a problem that it is not suitable for extracting an abnormal signal of equipment accompanied by shocking vibration, and it is difficult to extract an abnormal signal sufficiently accurately.

After that, the present inventors select and extract abnormal signals with respect to a drive device including a bearing that rotates at a low speed of 200 rpm or less, such as a transport device such as a bucket elevator and a drive unit such as a rotary kiln and a mining device. In addition to the processing by the statistical information filter, the data in the frequency domain obtained by the Fourier transform of the diagnostic signal is analyzed by a new feature parameter, and the data in the time domain is also analyzed by a specific feature parameter. It has been found that the above object can be achieved by evaluating the result by principal component analysis (Patent Document 3).

Japanese Unexamined Patent Publication No. 2001-033353 Japanese Patent No. 3920715 Japanese Unexamined Patent Publication No. 2017-194371

As described above, the present inventors have found a diagnostic technique for a drive device including a bearing that rotates at a low speed, but on the other hand, even in a self-circulation type drive device represented by a canned pump motor, due to the structure of the device. , Abnormal signals are difficult to detect, and diagnostic techniques that can reliably diagnose abnormalities have been required.

An object of the present invention is to provide an abnormality diagnosis device, an abnormality diagnosis method, and an abnormality diagnosis program capable of more reliably diagnosing an abnormality of a self-circulating drive device as compared with the prior art.

ｂ）第２特徴パラメータ

ｃ）第３特徴パラメータ

ｄ）第４特徴パラメータ

ｅ）第５特徴パラメータ

ｆ）第６特徴パラメータ

ｇ）第７特徴パラメータ

ただし、ここで

The present invention is an abnormality diagnosis device for diagnosing an abnormality in a self-circulation type drive device including a sleeve portion and a bearing portion, wherein a vibration sensor installed in the self-circulation type drive device and the self-circulation type drive device are used. A data acquisition unit that acquires the first vibration characteristic data detected by the vibration sensor at the time of normal operation and the second vibration characteristic data detected by the vibration sensor at the time of diagnosis of the self-circulating drive device, and the above. Each of the first vibration characteristic data and the second vibration characteristic data is processed by the statistical information filter, and the first vibration characteristic data and the second vibration characteristic data after the processing are described below. A parameter calculation unit that calculates a feature parameter including at least two or more feature parameters from Equations 1 to 7, a normalization processing unit that normalizes the feature parameter calculated by the parameter calculation unit, and the normalization unit. Regarding feature parameters, a principal component analysis unit that extracts two principal components by the principal component analysis method, a first point corresponding to the two principal components extracted from the feature parameters related to the first vibration characteristic data, and a first point. A first abnormality diagnosis unit that diagnoses the abnormality based on the Euclidean distance between the second point corresponding to the two principal components extracted from the characteristic parameter related to the second vibration characteristic data. Regarding the abnormality diagnostic device to be provided.
a) First feature parameter

b) Second feature parameter

c) Third feature parameter

d) Fourth feature parameter

e) Fifth feature parameter

f) Sixth feature parameter

g) 7th feature parameter

However, here

Further, in the present invention, the first abnormality diagnosis unit has the first point and the first point corresponding to the main component extracted from a plurality of feature parameters including at least the third feature parameter or the sixth feature parameter. The abnormality may be diagnosed using the point 2.

Further, in the present invention, the first abnormality diagnosis unit corresponds to the first principal component extracted from a plurality of feature parameters including all feature parameters of the first feature parameter to the seventh feature parameter. The abnormality may be diagnosed using the point and the second point.

Further, the present invention includes a classification unit that classifies the third vibration characteristic data detected by the vibration sensor according to the type of abnormality during learning of the self-circulating drive device, and the classification unit for each type of abnormality. By performing machine learning by the canonical discrimination analysis method using the teacher data in which the feature parameter calculated from the third vibration characteristic data and the type of abnormality are paired, the second vibration characteristic data can be obtained. Two canonical variables are extracted from the corresponding feature parameters by a learning unit that constructs a learning model that outputs the type of abnormality and a canonical discrimination analysis method for the feature parameters corresponding to the second vibration characteristic data. The canonical discrimination analysis unit, the third point corresponding to the two canonical variable scores extracted from the characteristic parameter related to the first vibration characteristic data, and the characteristic parameter related to the second vibration characteristic data. For the Euclidean distance between the four points corresponding to the two canonical variable scores extracted from, the first vibration characteristic using the principal component coefficient or the canonical variable coefficient obtained by the learning model. By multiplying the data and the characteristic parameters obtained from the second vibration characteristic data, the principal component score or canonical variable score is obtained, and the state identification is performed in the plane space to which the principal component score or canonical variable score corresponds. By doing so, a second abnormality diagnosis unit for diagnosing the abnormality may be further provided.

Further, in the present invention, the second abnormality diagnosis unit corresponds to the canonical variable score extracted from at least the first feature parameter or three or more feature parameters including the third feature parameter. And the fourth point may be used to diagnose the abnormality.

Further, in the present invention, the second abnormality diagnosis unit corresponds to the canonical variable score extracted from at least three or more feature parameters including the first feature parameter and the third feature parameter. And the fourth point may be used to diagnose the abnormality of the sleeve portion.

Further, in the present invention, the second abnormality diagnosis unit corresponds to the canonical variable score extracted from at least three or more feature parameters including the first feature parameter and the third feature parameter. And the fourth point may be used to diagnose the abnormality of the bearing portion.

Further, in the present invention, the second abnormality diagnosis unit includes at least three or more sets including the first feature parameter and the third feature parameter, or the second feature parameter and the third feature parameter set. Even if the abnormality related to the weight imbalance in the self-circulating drive device is diagnosed by using the third point and the fourth point corresponding to the canonical variable score extracted from the characteristic parameters of the above. Good.

ｂ）第２特徴パラメータ

ｃ）第３特徴パラメータ

ｄ）第４特徴パラメータ

ｅ）第５特徴パラメータ

ｆ）第６特徴パラメータ

ｇ）第７特徴パラメータ

ただし、ここで

Further, the present invention is an abnormality diagnosis method for diagnosing an abnormality in a self-circulation type drive device including a sleeve portion and a bearing portion, and is installed in the self-circulation type drive device when the self-circulation type drive device is normal. A data acquisition step of acquiring the first vibration characteristic data detected by the vibration sensor and the second vibration characteristic data detected by the vibration sensor at the time of diagnosing the self-circulating drive device, and the first Each of the vibration characteristic data and the second vibration characteristic data is processed by the statistical information filter, and the first vibration characteristic data and the second vibration characteristic data after the processing are each processed by the following equations 1 to 1. A parameter calculation step for calculating a feature parameter including at least two or more feature parameters in Equation 7, a normalization processing unit for normalizing the calculated feature parameter, and a principal component analysis method for the normalized feature parameter. A principal component analysis step for extracting two principal components, a first point corresponding to the two principal components extracted from the characteristic parameters related to the first vibration characteristic data, and the second vibration characteristic data. The present invention relates to an abnormality diagnosis method having a first abnormality diagnosis step of diagnosing the abnormality based on the Euclidean distance between the second points corresponding to the two principal components extracted from the feature parameters according to the above.
a) First feature parameter

b) Second feature parameter

c) Third feature parameter

d) Fourth feature parameter

e) Fifth feature parameter

f) Sixth feature parameter

g) 7th feature parameter

However, here

ｂ）第２特徴パラメータ

ｃ）第３特徴パラメータ

ｄ）第４特徴パラメータ

ｅ）第５特徴パラメータ

ｆ）第６特徴パラメータ

ｇ）第７特徴パラメータ

ただし、ここで

Further, the present invention is an abnormality diagnosis program for causing a computer to execute an abnormality diagnosis method for diagnosing an abnormality in a self-circulation type drive device including a sleeve portion and a bearing portion, and when the self-circulation type drive device is normal. , The first vibration characteristic data detected by the vibration sensor installed in the self-circulation type drive device, and the second vibration characteristic data detected by the vibration sensor at the time of diagnosis of the self-circulation type drive device are acquired. Data acquisition step, the first vibration characteristic data and the second vibration characteristic data are each processed by a statistical information filter, and the first vibration characteristic data and the second vibration characteristic after the processing are processed. For each of the data, a parameter calculation step for calculating a feature parameter including at least two or more feature parameters from the following equations 1 to 7, a normalization processing unit for normalizing the calculated feature parameter, and the normalization. The principal component analysis step of extracting two principal components by the principal component analysis method for the converted feature parameters, and the first principal component corresponding to the two principal components extracted from the feature parameters related to the first vibration characteristic data. A first abnormality diagnosis step of diagnosing the abnormality based on the Euclidean distance between the point and the second point corresponding to the two principal components extracted from the characteristic parameter related to the second vibration characteristic data. And, regarding the abnormality diagnosis program to make the computer execute.
a) First feature parameter

b) Second feature parameter

c) Third feature parameter

d) Fourth feature parameter

e) Fifth feature parameter

f) Sixth feature parameter

g) 7th feature parameter

However, here

According to the present invention, it is possible to more reliably diagnose an abnormality in a self-circulating drive device as compared with the prior art.

It is an overall block diagram of the abnormality diagnosis system which concerns on embodiment of this invention. It is a functional block diagram of the abnormality diagnosis apparatus which concerns on embodiment of this invention. It is a figure which shows the Euclidean distance calculated at the time of abnormality diagnosis in embodiment of this invention. It is a flowchart which shows the operation of the abnormality diagnosis apparatus which concerns on embodiment of this invention. It is a flowchart which shows the operation of the abnormality diagnosis apparatus which concerns on embodiment of this invention. This is a configuration example of a can motor pump for diagnosing an abnormality by the abnormality diagnosing device according to the embodiment of the present invention. It is a figure which shows the example of the installation place of the vibration sensor in the can motor pump. It is a graph which shows the result of the principal component analysis at the time of diagnosing the abnormality of the can motor pump by the abnormality diagnosis apparatus which concerns on embodiment of this invention. It is a graph which shows the result of the principal component analysis at the time of diagnosing the abnormality of the can motor pump by the abnormality diagnosis apparatus which concerns on embodiment of this invention. It is a graph which shows the result of the principal component analysis at the time of diagnosing the abnormality of the can motor pump by the abnormality diagnosis apparatus which concerns on embodiment of this invention. It is a graph which shows the result of the principal component analysis at the time of diagnosing the abnormality of the can motor pump by the abnormality diagnosis apparatus which concerns on embodiment of this invention. It is a graph which shows the result of the principal component analysis at the time of diagnosing the abnormality of the can motor pump by the abnormality diagnosis apparatus which concerns on embodiment of this invention. It is a graph which shows the result of the principal component analysis at the time of diagnosing the abnormality of the can motor pump by the abnormality diagnosis apparatus which concerns on embodiment of this invention. It is a functional block diagram of the abnormality diagnosis apparatus which concerns on embodiment of this invention. It is a figure which shows the Euclidean distance calculated at the time of abnormality diagnosis in embodiment of this invention. It is a flowchart which shows the operation of the abnormality diagnosis apparatus which concerns on embodiment of this invention. It is a flowchart which shows the operation of the abnormality diagnosis apparatus which concerns on embodiment of this invention. It is a flowchart which shows the operation of the abnormality diagnosis apparatus which concerns on embodiment of this invention. It is a graph which shows the result of the principal component analysis at the time of diagnosing the abnormality of the can motor pump by the abnormality diagnosis apparatus which concerns on embodiment of this invention. It is a graph which shows the result of the principal component analysis at the time of diagnosing the abnormality of the can motor pump by the abnormality diagnosis apparatus which concerns on embodiment of this invention. It is a graph which shows the result of the principal component analysis at the time of diagnosing the abnormality of the can motor pump by the abnormality diagnosis apparatus which concerns on embodiment of this invention. It is a graph which shows the result of the principal component analysis at the time of diagnosing the abnormality of the can motor pump by the abnormality diagnosis apparatus which concerns on embodiment of this invention. It is a graph which shows the result of the principal component analysis at the time of diagnosing the abnormality of the can motor pump by the abnormality diagnosis apparatus which concerns on embodiment of this invention. It is a graph which shows the result of the principal component analysis at the time of diagnosing the abnormality of the can motor pump by the abnormality diagnosis apparatus which concerns on embodiment of this invention.

[1 First Embodiment]
Hereinafter, the abnormality diagnosis system 1 and the abnormality diagnosis device 10 according to the first embodiment of the present invention will be described with reference to FIGS. 1 to 7F. FIG. 1 is an overall configuration diagram of the abnormality diagnosis system 1 according to the first embodiment. FIG. 2 is a functional block diagram of the abnormality diagnosis device 10. FIG. 3 is a diagram showing the Euclidean distance calculated at the time of abnormality diagnosis. 4A and 4B are flowcharts showing the operation of the abnormality diagnosis device 10. FIG. 5 is a configuration example of a can motor pump that diagnoses an abnormality by the abnormality diagnosing device 10. FIG. 6 is a diagram showing an example of a location where a vibration sensor is installed in the can motor pump of FIG. 7A to 7F are graphs showing the results of principal component analysis when the abnormality of the can motor pump of FIG. 5 is diagnosed by the abnormality diagnosis device 10.

[1.1 Configuration of Embodiment]
As shown in FIG. 1, the abnormality diagnosis system 1 according to the first embodiment includes an abnormality diagnosis device 10 and a drive device 20.

The abnormality diagnosis device 10 is a device for diagnosing an abnormality of the drive device 20. In particular, the abnormality diagnosis device 10 diagnoses the abnormality of the drive device 20 based on the vibration characteristic data acquired from the drive device 20.

The drive device 20 is, for example, a device including a motor driven by electricity. In particular, in the drive device 20, the motor coil is sealed in a can-like shape and integrated with the pump, and the liquid handled by the pump is circulated to the bearing portion of the motor without leaking to the outside, thereby lubricating the bearing portion. It is a self-circulating drive device such as a canned motor pump, which is also used as a bearing.

As shown in FIG. 2, the abnormality diagnosis device 10 includes a vibration sensor 11, a control unit 13, and a storage unit 15.

The vibration sensor 11 is a sensor that measures the vibration of the drive device 20 that is the object of diagnosing the abnormality by the abnormality diagnosis device 10. The vibration sensor 11 is configured to be able to output vibration information such as the measured amplitude and frequency of the vibration to the control unit 13 described later. The vibration sensor 11 is basically installed on the surface of the housing of the self-circulating drive device and measures the vibration on the surface, but is not limited thereto.

The control unit 13 is a processor that controls the abnormality diagnosis device 10 as a whole. The control unit 13 reads out the system program and the application program stored in the ROM (not shown) via the bus, and according to the system program and the application program, the data acquisition unit 131 and the parameter calculation unit 132 included in the control unit 13. , The functions of the normalization processing unit 133, the main component analysis unit 134, and the first abnormality diagnosis unit 135 are realized.

The data acquisition unit 131 acquires the first vibration characteristic data detected by the vibration sensor 11 installed in the drive device 20 when the drive device 20 is normal. Further, the data acquisition unit 131 acquires the second vibration characteristic data detected by the vibration sensor 11 installed in the drive device 20 at the time of diagnosis of the drive device 20.

The parameter calculation unit 132 divides each of the first vibration characteristic data and the second vibration characteristic data into n stages (n is a natural number) having the length of each stage as a multiplier of 2, and then filters the statistical information. Use to filter. Specifically, the parameter calculation unit 132 may extract only the data in the high frequency region from the vibration characteristic data by using, for example, a high-pass filter, with any frequency between 1500 Hz and 2000 Hz as the cutoff frequency.

ｂ）第２特徴パラメータ

ｃ）第３特徴パラメータ

ｄ）第４特徴パラメータ

ｅ）第５特徴パラメータ

ｆ）第６特徴パラメータ

ｇ）第７特徴パラメータ

ただし、ここで

である。 Further, the parameter calculation unit 132 calculates seven feature parameters by the following equations 1 to 7 using the filtered vibration characteristic data.
a) First feature parameter

b) Second feature parameter

c) Third feature parameter

d) Fourth feature parameter

e) Fifth feature parameter

f) Sixth feature parameter

g) 7th feature parameter

However, here

Is.

The first feature parameter is the average feature frequency. The second feature parameter is the frequency with which the time average is closed per unit time. The third characteristic parameter is the stability index of the waveform. The fourth characteristic parameter is the volatility. The fifth feature parameter is skewness. The sixth feature parameter is kurtosis. The seventh feature parameter is the root mean square ratio.

The normalization processing unit 133 normalizes the feature parameters calculated by the parameter calculation unit 132.

When the vibration of the drive device 20 is measured by the vibration sensor 11, the amplitude change of the measurement signal due to the voltage fluctuation of the amplifier, the amplification multiple, or the like causes variations in the values of the respective feature parameters. In order to eliminate these effects, the normalization processing unit 133 normalizes the spectrum by the following equation 10 and normalizes the feature parameters by the equation 11.

The principal component analysis unit 134 uses the vibration characteristic data at the time of normal and diagnosis to obtain the value of the first principal component and the first principal component by the principal component analysis method for the feature parameters normalized by the normalization processing unit 133. 2 Calculate the value of the principal component.

The first abnormality diagnosis unit 135 calculates the coordinates of the first point corresponding to the two principal components extracted from the feature parameters related to the first vibration characteristic data which is the normal vibration characteristic data.

More specifically, the first abnormality diagnosis unit 135 sets the value of the first principal component extracted from the feature parameter related to the vibration characteristic data at the normal time (hereinafter, this may be referred to as "reference data"). Coordinate values with the X coordinate and the value of the second principal component as the Y coordinate are calculated for each feature parameter. Further, the first abnormality diagnosis unit 135 sets the center of gravity of the plurality of coordinate values corresponding to these feature parameters as the “center of reference data” (N _x , N _y ).

In addition, the first abnormality diagnosis unit 135 calculates the coordinates of the second point corresponding to the two principal components extracted from the feature parameters related to the second vibration characteristic data, which is the vibration characteristic data at the time of diagnosis.

More specifically, the first abnormality diagnosis unit 135 sets the value of the first principal component extracted from the feature parameter related to the vibration characteristic data at the time of diagnosis (hereinafter, this may be referred to as "diagnosis data"). Coordinate values with the X coordinate and the value of the second principal component as the Y coordinate are calculated for each feature parameter. Further, the first abnormality diagnosis unit 135 sets the center of gravity of a plurality of coordinate values corresponding to these feature parameters as the “center of diagnostic data” (D _x , D _y ).

Then, the first abnormality diagnosis unit 135 sets (u _Nx , u _Ny ) and (σ _Nx , σ _Ny ) as the average value and standard deviation of the first principal component and the second principal component, respectively, as follows. When the equation 12 holds, it is diagnosed that the drive device 20 is in an abnormal state. The right side of the equation 12 indicates the distance from the normal state.

For example, as shown in FIG. 3, when the center of the reference data (N _x , N _y ) is the origin, the center of the diagnostic data is the center as the origin and one of the vertices is (x, y) = (u _Nx). When it is outside the rectangle of + 3σ _Nx , u _Ny + 3σ _Ny ), the first abnormality diagnosis unit 135 diagnoses that the drive device 20 is in an abnormal state.

The storage unit 15 has reference data and diagnostic data acquired by the data acquisition unit 131, first feature parameters to seventh feature parameters calculated by the parameter calculation unit 132, and reference data calculated by the principal component analysis unit 134. And the first principal component, the second principal component, etc. related to the diagnostic data are stored.

[1.2 Operation of the embodiment]
Hereinafter, the operation of the abnormality diagnosis device 10 will be described with reference to FIGS. 4A and 4B.

In step S11, the data acquisition unit 131 acquires reference data by the vibration sensor 11.

In step S12, the abnormality diagnosis device 10 calculates two main components from the feature parameters calculated using the reference data. As shown in FIG. 4B, step S12 includes steps S12a to S12e.

In step S12a, the parameter calculation unit 132 performs a Fourier transform on the reference data.

In step S12b, the parameter calculation unit 132 filters the reference data after the Fourier transform using the statistical information filter.

In step S12c, the parameter calculation unit 132 calculates the first feature parameter to the seventh feature parameter using the reference data after the filter processing.

In step S12d, the normalization processing unit 133 performs normalization processing on the first feature parameter to the seventh feature parameter.

In step S12e, the principal component analysis unit 134 performs principal component analysis for each of the first feature parameter to the seventh feature parameter after the normalization process.

In step S21, the data acquisition unit 131 acquires diagnostic data by the vibration sensor 11.

In step S22, the abnormality diagnosis device 10 calculates two main components from the feature parameters related to the diagnostic data. As shown in FIG. 4B, step S22 includes steps S22a to S22e. Further, step S22a is basically the same as step S12a, step S22b is basically the same as step S12b, step S22c is basically the same as step S12c, and step S22d is basically the same as step S12d. It is the same, and step S22e is basically the same as step S12e.

In step S31, the first abnormality diagnosis unit 135 calculates the center of the reference data and the center of the diagnostic data, and calculates the Euclidean distance between these two centers.

In step S32, the first abnormality diagnosis unit 135 diagnoses the abnormality based on the above formula 12.

[1.3 Demonstration test]
The applicant of the present invention conducted a verification test under the following conditions in order to verify the effectiveness of the abnormality diagnosis method by the abnormality diagnosis device 10 according to the first embodiment.

[1.3.1 Test method]
FIG. 5 shows a configuration of a self-circulating drive device 20 in which an abnormality is intentionally generated in the verification test and the abnormality is diagnosed by the abnormality diagnosis device 10.

The drive device 20 is a can motor pump, which is a pump that sucks the handling liquid from the suction port 22 installed in the pump casing 21 and discharges it to the discharge port 24 by the rotation of the impeller 23.

Further, the shaft 26 of the rotor 25 is installed coaxially with the central shaft of the suction port 22 and the central shaft of the impeller 23, and the rotor 25 rotates about the shaft 26. Further, the rotor 25 is surrounded by the stator 27, and the rotor 25 and the stator 27 are brought into a canned state by the frame 28. The rotor 25 is rotated by an electromagnetic force generated by a current flowing through the rotor 25 and a magnetic force generated between the stators 27. The shaft 26 is supported by the slide bearing 29, and the handling liquid flows in the direction of the rotor 25 via the contact portion between the shaft 26 and the slide bearing 29, so that the handling liquid flows between the shaft 26 and the slide bearing 29. Functions as a lubricant.

FIG. 6 shows the installation location of the vibration sensor 11 on the can motor pump. In this demonstration test, each of the two vibration sensors 11 is installed on the surface of the pump casing 21 or the frame 28 in the vicinity of the front and rear slide bearings 29, respectively. This installation location is just an example, and is not limited to this.

Further, as the assumed condition of the diagnostic state, the front-rear wear is assumed by using the slide bearing 29 having the same inner diameter (23.7 mm or 23.3 mm) for the front and the rear, and the front and the rear are different. Assuming front-rear wear by using slide bearings 29 with inner diameters (23.6 mm-23.9 mm, 23.9 mm-23.6 mm), using adjustment seats (0.5 mm or 1.5 mm) We assumed wear and contact on the bearing by shifting the position of the bearing, and assumed imbalance by installing a weight (10 g or 20 g). The weight is installed outside the impeller.

Furthermore, for each of the front and rear, the data of the sensor direction in the axial direction (axial direction), the vertical direction, and the horizontal direction were measured. That is, 6 channels of data were measured.

The sampling frequency is 100 kHz, the sampling time is 30 seconds, and the rotation speed of the motor is 3600 rpm. Further, each measurement signal of the diagnostic state and the reference state was divided into eight stages, the number of data of each stage is 16384 (2 ^14). The cutoff frequency is 2000 Hz.

[13.1 Test Results]
In each channel, the score in each state was plotted with the first principal component calculated from the feature parameters calculated using the reference data and the diagnostic data as the X coordinate and the second principal component as the Y coordinate. FIG. 7A is a front-axial graph. FIG. 7B is a front-vertical graph. FIG. 7C is a front-horizontal graph. FIG. 7D is a rear-axial graph. FIG. 7E is a rear vertical graph. FIG. 7F is a rear-horizontal graph. In each graph, the range of (x, y) = (u _Nx ± 3σ _Nx , u _Ny ± 3σ _Ny ) is shown.

Table 1 below shows the determination results based on the Euclidean distance in each channel.

As can be seen from Table 1, it was demonstrated that the diagnostic results by the abnormality diagnostic device 10 were correct in all channels. It was also shown that the diagnostic sensitivity was good, especially in the front-vertical channel.

In addition, the order of sensitivity of the feature parameters was calculated using the measurement data of the front-vertical channel. More specifically, the square root of the first principal component coefficient and the second principal component coefficient α _ij (i = 1, 2 j = 1, 2) _{obtained by the principal component analysis S j} = √ (α _ij ² + α _ij ²⁾ ) Was calculated, and the sensitivity of each feature parameter was evaluated.

a) Sleeve: 23.3-23.3
The square root _Sj values of the first and second principal component coefficients α _ij are shown in Table 2 below.

As shown in Table 2, the order of sensitivity of the feature parameters in the sleeve: 23.3-23.3 state was 3, 7, 6, 4, 5, 1, 2.

b) Sleeve: 23.6-23.9
The square root _Sj values of the first and second principal component coefficients α _ij are shown in Table 3 below.

As shown in Table 3, the order of sensitivity of the feature parameters in the sleeve: 23.6-23.9 state was 6, 4, 1, 2, 7, 3, 5.

c) Sleeve: 23.7-23.7
The square root _Sj values of the first and second principal component coefficients α _ij are shown in Table 4 below.

As shown in Table 4, the order of sensitivity of the feature parameters in the sleeve: 23.7-23.7 state was 6, 4, 7, 2, 1, 5, 3.

d) Sleeve: 23.9-23.6
The square root _Sj values of the first and second principal component coefficients α _ij are shown in Table 5 below.

As shown in Table 5, the order of sensitivity of the feature parameters in the sleeve: 23.9-23.6 state was 3, 6, 7, 4, 1, 5, 2.

e) Adjustment seat: 0.5
The square root _Sj values of the first and second principal component coefficients α _ij are shown in Table 6 below.

As shown in Table 6, the order of sensitivity of the feature parameters in the adjustment seat: 0.5 state was 3, 2, 7, 6, 1, 5, 4.

f) Adjustment seat: 1.5
The square root _Sj values of the first and second principal component coefficients α _ij are shown in Table 7 below.

As shown in Table 7, the order of sensitivity of the feature parameters in the adjustment seat: 1.5 state was 3, 2, 7, 1, 5, 6, 4.

g) Unbalance: 10g
The square root _Sj values of the first and second principal component coefficients α _ij are shown in Table 8 below.

As shown in Table 8, the order of sensitivity of the characteristic parameters in the unbalanced: 10 g state was 3, 7, 6, 4, 5, 1, 2.

h) Unbalance: 20g
The square root _Sj values of the first and second principal component coefficients α _ij are shown in Table 9 below.

As shown in Table 9, the order of sensitivity of the characteristic parameters in the unbalanced: 20 g state was 3, 2, 1, 7, 4, 5, and 6.

Table 10 below shows the order of sensitivity of the feature parameters in each single abnormal state.

From Table 10, when diagnosing the abnormality of the drive device 20 by performing the principal component analysis of the feature parameters in the abnormality diagnosis device 10, the third feature parameter or the sixth feature parameter is extracted, and the parameter having the highest sensitivity is extracted. It was shown that it was done.

ｂ）第２特徴パラメータ

ｃ）第３特徴パラメータ

ｄ）第４特徴パラメータ

ｅ）第５特徴パラメータ

ｆ）第６特徴パラメータ

ｇ）第７特徴パラメータ

ただし、ここで

[1.4 Effect]
The abnormality diagnosis device 10 according to the present embodiment includes a vibration sensor 11 installed in the self-circulating drive device 20, a first vibration characteristic data detected by the vibration sensor 11 when the drive device 20 is normal, and a drive. At the time of diagnosis of the device 20, the data acquisition unit 131 that acquires the second vibration characteristic data detected by the vibration sensor 11 and each of the first vibration characteristic data and the second vibration characteristic data are processed by the statistical information filter. Then, with respect to each of the first vibration characteristic data and the second vibration characteristic data after the processing, the parameter calculation unit 132 for calculating the characteristic parameter including at least two or more characteristic parameters from the following equations 1 to 7. , The normalization processing unit 133 that normalizes the feature parameters calculated by the parameter calculation unit 132, the principal component analysis unit 134 that extracts two principal components by the principal component analysis method for the normalized feature parameters, and the first. The first point corresponding to the two principal components extracted from the characteristic parameters related to the vibration characteristic data of, and the second point corresponding to the two principal components extracted from the characteristic parameters related to the second vibration characteristic data. A first abnormality diagnosis unit 135 for diagnosing an abnormality based on the Euclidean distance between the two.
a) First feature parameter

b) Second feature parameter

c) Third feature parameter

d) Fourth feature parameter

e) Fifth feature parameter

f) Sixth feature parameter

g) 7th feature parameter

However, here

This makes it possible to more reliably diagnose an abnormality in the self-circulating drive device as compared with the prior art. In particular, it is possible to detect an abnormality caused by wear of the bearing portion of the self-circulating drive device and imbalance of the impeller.

Further, the first abnormality diagnosis unit 135 diagnoses an abnormality using the first point and the second point corresponding to the main components extracted from a plurality of feature parameters including at least the third feature parameter or the sixth feature parameter. You may.

This makes it possible to diagnose an abnormality in the self-circulating drive device with a smaller amount of calculation.

Further, the first abnormality diagnosis unit 135 uses the first point and the second point corresponding to the main components extracted from a plurality of feature parameters including all the feature parameters of the first feature parameter to the seventh feature parameter. You may diagnose the abnormality.

This makes it possible to more accurately diagnose an abnormality in the self-circulating drive device.

[2 Second Embodiment]
Hereinafter, the abnormality diagnostic apparatus 10A according to the second embodiment of the present invention will be described with reference to FIGS. 8 to 11F. FIG. 8 is a functional block diagram of the abnormality diagnosis device 10A. FIG. 9 is a diagram showing the Euclidean distance calculated at the time of abnormality diagnosis. 10A to 10C are flowcharts showing the operation of the abnormality diagnosis device 10A. 11A to 11F are graphs showing the results of canonical discriminant analysis when the abnormality of the can motor pump of FIG. 4 is diagnosed by the abnormality diagnosis device 10A.

[2.1 Configuration of Embodiment]
The abnormality diagnosis system 1A according to the second embodiment includes the abnormality diagnosis device 10A shown in FIG. 8 instead of the abnormality diagnosis device 10 provided in the abnormality diagnosis system 1 according to the first embodiment. Hereinafter, the configuration of the abnormality diagnosis device 10A will be described with reference to FIG. 8, but basically, the same reference numerals are used for the same components as the abnormality diagnosis device 10, and the description of their functions is omitted. Then, a component different from the abnormality diagnosis device 10 will be described.

The abnormality diagnosis device 10A includes a control unit 13A instead of the control unit 13. Further, the control unit 13A includes a classification unit 136, a learning unit 137, a canonical discriminant analysis unit 138, and a second abnormality diagnosis unit 139, in addition to the components included in the control unit 13.

The classification unit 136 classifies the vibration characteristic data in the drive device 20 according to the type of abnormality.

The learning unit 137 performs machine learning by the canonical discriminant analysis method using the teacher data in which the feature parameters calculated from the second vibration characteristic data in the drive device 20 and the types of abnormalities are paired. A learning model that outputs the type of abnormality is constructed from the feature parameters corresponding to the vibration characteristic data of 2.

More specifically, in the first embodiment, when the first abnormality diagnosis unit 135 diagnoses that an abnormality has occurred in the drive device 20, for example, where the abnormality has occurred by disassembling the drive device 20. To identify. The learning unit 137 performs machine learning by the canonical discriminant analysis method using the teacher data that is a set of the feature parameter calculated from the second vibration characteristic data when the abnormality occurs and the type of the abnormality. By inputting the feature parameters calculated from the diagnostic data, a learning model that outputs the type of abnormality is constructed.
Further, the learning unit 137 stores the constructed learning model in the storage unit 15.

As an example, the learning unit 137 is a support vector machine (hereinafter referred to as SVM), which is one of the identification methods using supervised learning (learning in which correct answer data and non-correct answer data are given as teacher data). It can be realized by using (also called).

The canonical discriminant analysis unit 138 uses the vibration characteristic data (reference data) at the time of normal analysis and the vibration characteristic data (diagnosis data) at the time of diagnosis by the canonical discriminant analysis method using the learning model constructed by the learning unit 137. Then, the value of the first canonical variable and the value of the second canonical variable are calculated.

The second abnormality diagnosis unit 139 calculates the coordinates of the first point corresponding to the two canonical variables calculated from the feature parameters related to the first vibration characteristic data which is the normal vibration characteristic data.

More specifically, the second abnormality diagnosis unit 139 sets the value of the first canonical variable calculated from the feature parameters related to the vibration characteristic data (reference data) in the normal state as the X coordinate and the value of the second canonical variable. The coordinate value to be the Y coordinate is calculated for each feature parameter. Further, the second abnormality diagnosis unit 139 sets the center of gravity of the plurality of coordinate values corresponding to the plurality of feature parameters as the “center of the reference data” (N _x , N _y ).

Further, the second abnormality diagnosis unit 139 calculates the coordinates of the second point corresponding to the two canonical variables calculated from the feature parameters related to the second vibration characteristic data which is the vibration characteristic data at the time of diagnosis.

More specifically, the second abnormality diagnosis unit 139 sets the value of the first canonical variable extracted from the feature parameter related to the vibration characteristic data (diagnosis data) at the time of diagnosis as the X coordinate, and the value of the second canonical variable. The coordinate value to be the Y coordinate is calculated for each feature parameter. Further, the second abnormality diagnosis unit 139 sets the center of gravity of the plurality of coordinate values corresponding to the plurality of feature parameters as the “center of diagnostic data” (D _x , D _y ).

Then, the second abnormality diagnosis unit 139 sets (u _Nx , u _Ny ) and (σ _Nx , σ _Ny ) as the mean value and standard deviation of the first canonical variable and the second canonical variable, respectively. In this case, if the above equation 12 holds, it is diagnosed that the drive device 20 is in an abnormal state.

For example, as shown in FIG. 9, when the center of the reference data (N _x , N _y ) is set as the origin, the center of the diagnostic data is centered on the origin and (x, y) = (u _Nx + 3σ _Nx , When it is outside the circle whose radius is up to u _Ny + 3σ _Ny ), the second abnormality diagnosis unit 139 diagnoses that the drive device 20 is in an abnormal state.

[2.2 Operation of the embodiment]
Hereinafter, the operation of the abnormality diagnosis device 10A will be described with reference to FIGS. 10A to 10C.

First, the operation of the abnormality diagnosis device 10A during machine learning will be described with reference to FIG. 10A.

In step S41, the data acquisition unit 131 acquires past diagnostic data (hereinafter, this is also referred to as “third vibration characteristic data”).

In step S42, the classification unit 136 classifies the past diagnostic data according to the type of abnormality.

In step S43, the parameter calculation unit 132 filters the past diagnostic data for each type of abnormality by using the statistical information filter.

In step S44, the parameter calculation unit 132 calculates the first feature parameter to the seventh feature parameter using the past diagnostic data after filtering for each type of abnormality.

In step S45, the normalization processing unit 133 performs normalization processing on the first feature parameter to the seventh feature parameter.

In step S46, the learning unit 137 constructs a learning model using the teacher data in which the normalized first feature parameters to the seventh feature parameters and the type of abnormality are paired.

Next, the operation of the abnormality diagnosis device 10A at the time of diagnosis will be described with reference to FIGS. 10B and 10C.

In step S51, the data acquisition unit 131 acquires reference data by the vibration sensor 11.

In step S52, the abnormality diagnosis device 10A calculates two canonical variables from the feature parameters related to the reference data. As shown in FIG. 10C, step S52 includes steps S52a to S52e.

In step S52a, the parameter calculation unit 132 filters the reference data by using the statistical information filter.

In step S52b, the parameter calculation unit 132 calculates the first feature parameter to the seventh feature parameter using the reference data after the filtering.

In step S52c, the normalization processing unit 133 performs normalization processing on the first feature parameter to the seventh feature parameter.

In step S52d, the canonical discriminant analysis unit 138 performs canonical discriminant analysis on the normalized feature parameters.

In step S61, the data acquisition unit 131 acquires new diagnostic data by the vibration sensor 11.

In step S62, the abnormality diagnosis device 10A calculates two canonical variables from the feature parameters related to the diagnostic data. As shown in FIG. 10C, step S62 includes steps S62a to S62d. Further, step S62a is basically the same as step S52a, step S62b is basically the same as step S52b, step S62c is basically the same as step S52c, and step S62d is basically the same as step S52d. It is the same.

In step S71, the second abnormality diagnosis unit 139 calculates the center of the reference data and the center of the diagnostic data, and calculates the Euclidean distance between these two centers.

In step S72, the second abnormality diagnosis unit 139 obtains the characteristic parameters obtained from the diagnostic data using the principal component coefficient or the canonical variable coefficient obtained by the learning model constructed in step S46 with respect to the calculated Euclidean distance. By multiplying, the principal component score or canonical variable score is obtained, and in the plane space (first and second principal components, or first and second canonical variables) to which the principal component score or canonical variable score corresponds. Abnormality is diagnosed by identifying the state.

[2.3 Demonstration test]
The applicant of the present invention conducted a verification test under the following conditions in order to verify the effectiveness of the abnormality diagnosis method by the abnormality diagnosis device 10A according to the second embodiment.

[2.3.1 Test method]
Similar to the test method described in [1.3.1 Test method] according to the first embodiment, the vibration sensor 11 is installed at the position shown in FIG. 6 for the can motor pump shown in FIG. Was done.

In addition, as the assumed conditions of the diagnostic state, the front-rear wear is assumed by using the slide bearing 29 having the same inner diameter (23.7 mm or 23.3 mm) for the front and the rear, and the front and the rear are different. Assuming front-rear wear by using slide bearings 29 with inner diameters (23.6 mm-23.9 mm, 23.9 mm-23.6 mm), using adjustment seats (0.5 mm or 1.5 mm) We assumed wear and contact on the bearing by shifting the position of the bearing, and assumed imbalance by installing a weight (10 g or 20 g).

In addition, axial, vertical and horizontal data were measured for each of the front and rear. That is, 6 channels of data were measured.

The sampling frequency is 100 kHz, the sampling time is 30 seconds, and the rotation speed of the motor is 3600 rpm. Further, each measurement signal of the diagnostic state and the reference state was divided into eight stages, the number of data of each stage is 16384 (2 ^14). The cutoff frequency is 2000 Hz.

[13.1 Test Results]
In each channel, the score in each state was plotted with the first canonical variable calculated from the feature parameters calculated using the reference data and the diagnostic data as the X coordinate and the second canonical variable as the Y coordinate. FIG. 11A is a front-axial graph. FIG. 11B is a front-vertical graph. FIG. 11C is a front-horizontal graph. FIG. 11D is a rear-axial graph. FIG. 11E is a rear vertical graph. FIG. 11F is a rear-horizontal graph. In each graph, the range of (x, y) = (u _Nx ± 3σ _Nx , u _Ny ± 3σ _Ny ) is shown.

Table 11 below shows the determination results based on the Euclidean distance in each channel.

As can be seen from Table 11, it was demonstrated that the diagnostic results by the abnormality diagnostic device 10A were correct in all channels except the front-axial. It was also shown that the diagnostic sensitivity was good, especially in the front-vertical channel.

In addition, the order of sensitivity of the feature parameters was calculated using the measurement data of the front-vertical channel. More specifically, the square root of the coefficient of the first canonical variable and the coefficient of the second canonical variable α _ij (i = 1, 2 j = 1, 2) _{obtained by the canonical discriminant analysis S j} = √ (α) _ij ² + α _ij ² ) was calculated, and the sensitivity of each feature parameter was evaluated.

a) Sleeve: 23.3-23.3
The square root _Sj values of the first and second canonical variable coefficients α _ij are shown in Table 12 below.

As shown in Table 12, the order of sensitivity of the feature parameters in the sleeve: 23.3-23.3 state was 1, 3, 2, 4, 5, 7, 6.

b) Sleeve: 23.6-23.9
The square root _Sj values of the first and second canonical variable coefficients α _ij are shown in Table 13 below.

As shown in Table 13, the order of sensitivity of the feature parameters in the sleeve: 23.6-23.9 state was 1, 3, 2, 4, 7, 5, 6.

c) Sleeve: 23.7-23.7
The square root _Sj values of the first and second canonical variable coefficients α _ij are shown in Table 14 below.

As shown in Table 14, the order of sensitivity of the feature parameters in the sleeve: 23.7-23.7 state was 1, 3, 2, 4, 7, 5, and 6.

d) Sleeve: 23.9-23.6
The square root _Sj values of the first and second canonical variable coefficients α _ij are shown in Table 15 below.

As shown in Table 15, the order of sensitivity of the feature parameters in the sleeve: 23.9-23.6 state was 1, 3, 2, 4, 7, 5, 6.

e) Adjustment seat: 0.5
The square root _Sj values of the first and second canonical variable coefficients α _ij are shown in Table 16 below.

As shown in Table 16, the order of sensitivity of the feature parameters in the adjustment seat: 0.5 state was 1, 3, 4, 7, 2, 5, and 6.

f) Adjustment seat: 1.5
The square root _Sj values of the first and second canonical variable coefficients α _ij are shown in Table 17 below.

As shown in Table 17, the order of sensitivity of the feature parameters in the adjustment seat: 1.5 state was 1, 3, 2, 4, 7, 5, and 6.

g) Unbalance: 10g
The square root _Sj values of the first and second principal component coefficients α _ij are shown in Table 18 below.

As shown in Table 18, the order of sensitivity of the characteristic parameters in the unbalanced: 10 g state was 1, 3, 4, 7, 2, 5, and 6.

h) Unbalance: 20g
The square root _Sj values of the first and second principal component coefficients α _ij are shown in Table 19 below.

As shown in Table 19, the order of sensitivity of the characteristic parameters in the unbalanced: 20 g state was 3, 2, 1, 5, 4, 6, and 7.

Table 20 below shows the order of sensitivity of the feature parameters in each single abnormal state.

From Table 20, when diagnosing an abnormality of the drive device 20 by canonical discriminant analysis of the feature parameters in the abnormality diagnosis device 10A, the first feature parameter or the third feature parameter is the parameter with the highest sensitivity. It was shown to be.

In particular, when diagnosing an abnormality in the sleeve portion or the bearing portion of the drive device 20, it has been shown that the canonical discriminant analysis of the set of the first feature parameter and the third feature parameter is particularly effective.

Further, when diagnosing an abnormality caused by an imbalance of weight in the drive device 20, in particular, the canonical level of the set of the first feature parameter and the third feature parameter, or the set of the second feature parameter and the third feature parameter. Discriminant analysis was shown to be effective.

[2.4 effect]
The abnormality diagnosis device 10A according to the present embodiment includes a classification unit 136 that classifies the third vibration characteristic data detected by the vibration sensor according to the type of abnormality during learning of the self-circulation type drive device, and each type of abnormality. In addition, the second vibration characteristic data is supported by performing machine learning by the canonical discrimination analysis method using the teacher data that is a set of the feature parameters calculated from the third vibration characteristic data and the type of abnormality. The learning unit 137 that builds a learning model that outputs the type of anomaly from the feature parameters to be used, and the canonical that extracts two canonical variables by the canonical discrimination analysis method for the feature parameters corresponding to the second vibration characteristic data. It was extracted from the discriminant analysis unit 138, the third point corresponding to the two canonical variable scores extracted from the feature parameters related to the first vibration characteristic data, and the feature parameters related to the second vibration characteristic data. For the Euclidean distance between the fourth point corresponding to the two canonical variable scores, the first vibration characteristic data and the second vibration are obtained using the principal component coefficient or the canonical variable coefficient obtained by the learning model. The principal component score or canonical variable score is obtained by multiplying the characteristic parameters obtained from the characteristic data, and the abnormality is diagnosed by performing state identification in the plane space corresponding to the principal component score or canonical variable score. A second abnormality diagnosis unit 139 is further provided.

As a result, the abnormality diagnosis device 10A can perform machine learning by the canonical discriminant analysis method, and as the amount of learning in the machine learning increases, it becomes possible to diagnose the abnormality more accurately.

In addition, the second abnormality diagnosis unit 139 has at least the third point corresponding to the canonical variable score extracted from the first feature parameter or three or more feature parameters including the third feature parameter, and the third point. The abnormality may be diagnosed using the fourth point.

Alternatively, the second abnormality diagnosis unit 139 has at least the third point corresponding to the canonical variable score extracted from the first feature parameter and three or more feature parameters including the third feature parameter, and the third point and the said. The abnormality of the sleeve portion may be diagnosed using the fourth point.

Alternatively, the second abnormality diagnosis unit 139 has at least the third point corresponding to the canonical variable score extracted from the first feature parameter and three or more feature parameters including the third feature parameter, and the third point and the said. The fourth point may be used to diagnose the abnormality of the bearing portion.

Alternatively, the second abnormality diagnosis unit 139 is based on at least three or more feature parameters including the first feature parameter and the third feature parameter set, or the second feature parameter and the third feature parameter set. The third point and the fourth point corresponding to the extracted canonical parameter scores may be used to diagnose an abnormality related to weight imbalance in the self-circulating drive device.

This makes it possible to diagnose an abnormality in the self-circulating drive device with a smaller amount of calculation.

Each device included in each of the above embodiments can be realized by hardware, software, or a combination thereof. In addition, a diagnostic method performed by the cooperation of the devices included in each of the above embodiments can also be realized by hardware, software, or a combination thereof. Here, what is realized by software means that it is realized by a computer reading and executing a program.

Programs can be stored and supplied to a computer using various types of non-transitory computer readable media. Non-transitory computer-readable media include various types of tangible storage media. Examples of non-temporary computer-readable media include magnetic recording media (eg, flexible disks, magnetic tapes, hard disk drives), magneto-optical recording media (eg, magneto-optical disks), CD-ROMs (Read Only Memory), CD- Includes R, CD-R / W, and semiconductor memory (eg, mask ROM, PROM (Programmable ROM), EPROM (Erasable PROM), flash ROM, RAM (random access memory)). The program may also be supplied to the computer by various types of transient computer readable media. Examples of temporary computer-readable media include electrical, optical, and electromagnetic waves. The temporary computer-readable medium can supply the program to the computer via a wired communication path such as an electric wire and an optical fiber, or a wireless communication path.

Further, although the above-described embodiment is a preferred embodiment of the present invention, the scope of the present invention is not limited to the above-described embodiment, and various modifications have been made without departing from the gist of the present invention. It can be implemented in the form. For example, it can be implemented in a modified form such as the following modification.

[3 Modification example]
For example, the abnormality diagnosis devices 10 and 10A may further include an alarm unit that issues an alarm when an abnormality is diagnosed by the first abnormality diagnosis unit 135 or the second abnormality diagnosis unit 139.

Further, the learning unit 137 of the abnormality diagnosis device 10A may execute machine learning that does not depend on the support vector machine, for example, machine learning that uses perceptron or logistic regression.

1, 1A Abnormality diagnosis system 10, 10A Abnormality diagnosis device 11 Vibration sensor 13 13A Control unit 15 Storage unit 20 Drive device 131 Data acquisition unit 132 Parameter calculation unit 133 Normalization processing unit 134 Principal component analysis unit 135 First abnormality diagnosis unit 136 Learning Department 137 Classification Department 138 Canonical Discriminant Analysis Department 139 Second Abnormality Diagnosis Department

Claims (10)

An abnormality diagnostic device for diagnosing an abnormality in a self-circulating drive device including a sleeve portion and a bearing portion.
The vibration sensor installed in the self-circulating drive device and
Acquires the first vibration characteristic data detected by the vibration sensor when the self-circulation type drive device is normal, and the second vibration characteristic data detected by the vibration sensor when the self-circulation type drive device is diagnosed. Data acquisition department and
Each of the first vibration characteristic data and the second vibration characteristic data is processed by the statistical information filter, and the first vibration characteristic data and the second vibration characteristic data after the processing are described below. A parameter calculation unit for calculating a feature parameter including at least two or more feature parameters among the formulas 1 to 7 of
A normalization processing unit that normalizes the feature parameters calculated by the parameter calculation unit, and
A principal component analysis unit that extracts two principal components by the principal component analysis method for the normalized feature parameters, and a principal component analysis unit.
The first point corresponding to the two main components extracted from the characteristic parameter related to the first vibration characteristic data and the two main components extracted from the characteristic parameter related to the second vibration characteristic data. A first abnormality diagnosis unit that diagnoses the abnormality based on the Euclidean distance between the corresponding second points, and
An abnormality diagnostic device equipped with.

a) First feature parameter

b) Second feature parameter

c) Third feature parameter

d) Fourth feature parameter

e) Fifth feature parameter

f) Sixth feature parameter

g) 7th feature parameter

However, here

The first abnormality diagnosis unit uses the first point and the second point corresponding to the main component extracted from at least the third feature parameter or a plurality of feature parameters including the sixth feature parameter. The abnormality diagnosis device according to claim 1, wherein the abnormality is diagnosed. The first abnormality diagnosis unit has the first point corresponding to the main component extracted from a plurality of feature parameters including all the feature parameters of the first feature parameter to the seventh feature parameter, and the second feature parameter. The abnormality diagnosis device according to claim 1 or 2, wherein the abnormality is diagnosed using points. A classification unit that classifies the third vibration characteristic data detected by the vibration sensor according to the type of the abnormality during learning of the self-circulation type drive device.
By performing machine learning by the canonical discriminant analysis method using the teacher data in which the characteristic parameter calculated from the third vibration characteristic data and the type of abnormality are paired for each type of abnormality. A learning unit that builds a learning model that outputs the type of abnormality from the feature parameters corresponding to the second vibration characteristic data, and a learning unit.
A canonical discriminant analysis unit that extracts two canonical variables by a canonical discriminant analysis method for the feature parameters corresponding to the second vibration characteristic data.
A third point corresponding to the two canonical variable scores extracted from the characteristic parameter related to the first vibration characteristic data and two positives extracted from the characteristic parameter related to the second vibration characteristic data. For the Euclidean distance to the fourth point corresponding to the quasi-variable score, the first vibration characteristic data and the second vibration are used by using the principal component coefficient or the canonical variable coefficient obtained by the learning model. By multiplying the characteristic parameters obtained from the characteristic data, the principal component score or the canonical variable score is obtained, and the state identification is performed in the plane space corresponding to the principal component score or the canonical variable score to obtain the abnormality. The second abnormality diagnosis department to diagnose and
The abnormality diagnostic apparatus according to any one of claims 1 to 3, further comprising. The second abnormality diagnosis unit has the third point and the fourth point corresponding to the canonical variable score extracted from at least the first feature parameter or three or more feature parameters including the third feature parameter. The abnormality diagnosis device according to claim 4, wherein the abnormality is diagnosed by using the points of the above. The second abnormality diagnosis unit has the third point and the fourth point corresponding to the canonical variable score extracted from at least three or more feature parameters including the first feature parameter and the third feature parameter. The abnormality diagnosis device according to claim 4 or 5, wherein the abnormality of the sleeve portion is diagnosed by using the points of the above. The second abnormality diagnosis unit has the third point and the fourth point corresponding to the canonical variable score extracted from at least three or more feature parameters including the first feature parameter and the third feature parameter. The abnormality diagnosis device according to any one of claims 4 to 6, wherein the abnormality of the bearing portion is diagnosed by using the points of the above. The second abnormality diagnosis unit is extracted from at least three or more feature parameters including the set of the first feature parameter and the third feature parameter, or the set of the second feature parameter and the third feature parameter. Any of claims 4 to 7, wherein the third point and the fourth point corresponding to the canonical parameter score are used to diagnose an abnormality related to weight imbalance in the self-circulating drive device. The abnormality diagnostic device according to item 1. It is an abnormality diagnosis method for diagnosing an abnormality in a self-circulating drive device including a sleeve portion and a bearing portion.
The first vibration characteristic data detected by the vibration sensor installed in the self-circulation type drive device when the self-circulation type drive device is normal, and the first vibration characteristic data detected by the vibration sensor at the time of diagnosis of the self-circulation type drive device. The data acquisition step to acquire the second vibration characteristic data,
Each of the first vibration characteristic data and the second vibration characteristic data is processed by the statistical information filter, and the first vibration characteristic data and the second vibration characteristic data after the processing are described below. A parameter calculation step for calculating a feature parameter including at least two or more feature parameters from the formulas 1 to 7 of
A normalization processing unit that normalizes the calculated feature parameters,
A principal component analysis step of extracting two principal components by a principal component analysis method for the normalized feature parameters, and a principal component analysis step.
The first point corresponding to the two main components extracted from the characteristic parameter related to the first vibration characteristic data and the two main components extracted from the characteristic parameter related to the second vibration characteristic data. A first anomaly diagnosis step of diagnosing the anomaly based on the Euclidean distance to the corresponding second point,
Abnormal diagnosis method having.
a) First feature parameter

b) Second feature parameter

c) Third feature parameter

d) Fourth feature parameter

e) Fifth feature parameter

f) Sixth feature parameter

g) 7th feature parameter

However, here

An abnormality diagnosis program for causing a computer to execute an abnormality diagnosis method for diagnosing an abnormality in a self-circulating drive device including a sleeve portion and a bearing portion.
The first vibration characteristic data detected by the vibration sensor installed in the self-circulation type drive device when the self-circulation type drive device is normal, and the first vibration characteristic data detected by the vibration sensor at the time of diagnosis of the self-circulation type drive device. The data acquisition step to acquire the second vibration characteristic data,
Each of the first vibration characteristic data and the second vibration characteristic data is processed by the statistical information filter, and the first vibration characteristic data and the second vibration characteristic data after the processing are described below. A parameter calculation step for calculating a feature parameter including at least two or more feature parameters from the formulas 1 to 7 of
A normalization processing step for normalizing the calculated feature parameters, and
A principal component analysis step of extracting two principal components by a principal component analysis method for the normalized feature parameters, and a principal component analysis step.
The first point corresponding to the two main components extracted from the characteristic parameter related to the first vibration characteristic data and the two main components extracted from the characteristic parameter related to the second vibration characteristic data. A first anomaly diagnosis step of diagnosing the anomaly based on the Euclidean distance to the corresponding second point,
An anomaly diagnostic program that lets your computer run.
a) First feature parameter

b) Second feature parameter

c) Third feature parameter

d) Fourth feature parameter

e) Fifth feature parameter

f) Sixth feature parameter

g) 7th feature parameter

However, here

Images