KR20160109159A - Method for monitoring machinery health based on fictitious frequency response function and system using the same - Google Patents

Abstract

The present invention relates to an abnormality diagnosis method of a rotary machine based on a fictitious frequency response function, and an abnormality diagnosis apparatus using the same. The abnormality diagnosis method of a rotary machine based on a fictitious frequency response function comprises: (i) a step of measuring vibration signals on two points; (ii) a step of calculating a standard deviation of the vibration signals in two points measured in step (i); (iii) a step of calculating a covariance from the vibration signals measured in the step (i); (iv) a step of calculating a coefficient from the value calculated in step (ii) and step (iii); and (v) a step of selecting the positions for sensors to be mounted in two points from the coefficient calculated in step (iv). As such, the present invention stably and accurately diagnoses abnormality of a rotary machine by installing the sensors to measure vibration signals in two points for the abnormality diagnosis of the rotary machine in optimal positions.

Description

TECHNICAL FIELD [0001] The present invention relates to a rotating machine abnormality diagnosis method using a virtual frequency response function, and a rotating machine abnormality diagnosis apparatus using the virtual frequency response function.

The present invention relates to a method of diagnosing an abnormality of a rotating machine based on a virtual frequency response function for diagnosing an abnormality in a rotating machine based on a virtual frequency response function, and a rotating machine abnormality diagnosis apparatus using the same.

Generally, mechanical diagnosis methods using vibration signals are performed by measuring vibration signals by attaching sensors (mainly acceleration sensors) adjacent to main parts such as bearings and gears and then measuring the average (RMS) or standard deviation of vibration signals, Kurtosis System, power spectrum, or envelope spectrum to determine whether the machine is abnormal.

Especially, frequency analysis method using power spectrum is widely used.

Conventional methods simultaneously measure signals of a plurality of channels using a plurality of sensors, but the method of analyzing each channel is basically the same as a single channel data analysis technique.

These methods can vary sensitively depending on the location where the sensor is attached, and in some cases due to environmental constraints, such as temperature, humidity, electrical and magnetic factors, and structural constraints, It is often the case that it can not be attached. In this case, the reliability of the diagnosis result may be greatly deteriorated.

In most rotating machines, anomalous signals due to various defects are transmitted to other parts through the bearing part, and defects of the bearings among many parts occur most frequently.

Accordingly, an acceleration sensor (A1) is attached to a region adjacent to the bearing housing or the bearing housing to measure the vibration signal, and various types of analysis methods such as average (RMS) or standard deviation are used to identify the presence or absence of the defect.

The system is a transfer function corresponding to the path between the excitation point and the vibration signal of the response point. Since the excitation component corresponding to the input is generally impossible to measure, only the output component is measured, .

Therefore, it is necessary to attach the measurement sensor to a location close to the point where it is possible to sense the mechanical abnormality early, and if it is attached to a remote place due to an environmental or structural problem, the abnormality signal included in the output component is very small, Respectively.

The following Patent Document 1 discloses a method for diagnosing a mechanical abnormality based on a virtual frequency response function and a diagnostic system using the same.

The machine abnormality diagnosis system based on the virtual frequency response function disclosed in Patent Document 1 mounts a plurality of vibration sensors at arbitrary different positions (S1, S2) of the object to measure vibration signals.

Further, the virtual fault response function-based fault diagnosis method disclosed in Patent Document 1

I) measuring a first vibration signal;

Ii) measuring a second vibration signal different from the first vibration signal;

Iii) calculating a virtual frequency response function (virtual FRF) of the first vibration signal and the second vibration signal;

Iv) extracting a group delay from the calculated virtual FRF; And

And v) analyzing the virtual FRF and the group delay,

The step i)

Measuring a steady state signal at a plurality of different measurement points including a first measurement point and a second measurement point; And

And calculating a first cross spectrum from data obtained from the plurality of different measurement points.

The method of diagnosing a mechanical anomaly based on the virtual frequency response function of the patent document 1 and the diagnostic system using the same can independently detect the spectrum of the magnitude and the phase so that the minute defects of the object can be detected early.

In addition, by using a group delay to represent a phase change with respect to a virtual frequency response function in the case where there is no defect and in the case of a defect, a substantial shape of the complex data can be analyzed, It can indicate whether it is happening or not.

Korean Registered Patent No. 10-1406778 (issued on June 17, 2010)

However, according to the conventional art including the patent document 1 and the method for diagnosing the abnormality of the rotating machine, the device for diagnosing the abnormality of the rotating machine can not provide a solution to the optimum installation position of the sensors for measuring the vibration signal at two points.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems of the prior art and it is an object of the present invention to provide a method of diagnosing an abnormality of a rotating machine more stably and accurately And a rotating machine abnormality diagnosis apparatus using the virtual frequency response function based on the virtual frequency response function.

In order to achieve the above object, the present invention provides a method for diagnosing a rotating machine malfunction based on a virtual frequency response function

I) measuring vibration signals at two points;

Ii) calculating a standard deviation of the vibration signals of the two points measured in the step i);

Iii) calculating a covariance from the vibration signals at the two points measured in the step i);

Iv) calculating a correlation coefficient from the values calculated in the step ii) and the step iii);

And selecting a position of a sensor to be mounted at two points from the correlation coefficient calculated in the step iv).

The method for diagnosing a rotating machine malfunction based on the virtual frequency response function according to the present invention

(Vi) mounting the sensor at two points selected in the step (v);

Measuring the first vibration signal from the two low points mounted in step vi);

Measuring a second vibration signal different from the first vibration signal;

Calculating a virtual frequency response function of the first vibration signal and the second vibration signal;

X) extracting a group delay from the calculated virtual FRF; And

Xi) analyzing the virtual FRF and the group delay,

Measuring the steady state signal at a plurality of different measurement points including a first measurement point and a second measurement point; And

And calculating a first cross spectrum from the data obtained from the plurality of different measurement points.

The fault diagnosis method based on the virtual frequency response function according to the present invention

The above step (a)

Measuring a signal of a fault condition at a plurality of different measurement points including a first measurement point and a second measurement point; And

And calculating a second cross spectrum from the data obtained from the plurality of different measurement points.

The method for diagnosing an anomaly based on a virtual frequency response function according to the present invention is characterized in that the virtual FRF is calculated from a first mutual spectrum for the first vibration signal and a second mutual spectrum for the second vibration signal.

In the method for diagnosing an anomaly of a machine based on the virtual frequency response function according to the present invention, the method for calculating the correlation coefficient in the above (iv) is characterized by being calculated by the following equation (1).

ρ _xy = Cov ( _{x, y} ) / ρ _x. ρ _y (1)

(Where ρ _x .ρ _y is the standard deviation of the two-point signal, and Cov ( _{x, y} ) is the covariance of the two signals)

The method for diagnosing a malfunctioning machine based on the virtual frequency response function according to the present invention is characterized in that the correlation coefficient as a reference for selecting the sensor position in the step (v) is in a range of 0.05 to 0.25.

In the method for diagnosing a rotating machine abnormality based on the virtual frequency response function according to the present invention, the average of the signals detected in the calculation range of the standard deviation is 50 to 60 seconds in the step ii) And calculates a correlation coefficient by an average of 50 to 100 unit blocks.

The apparatus for diagnosing an abnormality of a rotating machine using a virtual frequency response function based on a virtual frequency response function according to the present invention is characterized in that the rotating machine has two compressors and one motor, The abnormality diagnosis of the rotating machine is performed based on the sensor attached to the single bearing (compressor 1) and the signal measured from the sensor attached to the vicinity of the two-stage bearing (compressor 2) of the first compressor.

The apparatus for diagnosing an abnormality of a rotating machine using a virtual frequency response function based on a virtual frequency response function according to the present invention is characterized in that the rotating machine has two compressors and one motor, The abnormality diagnosis of the rotating machine is performed based on the signals measured from the sensor (compressor 3) attached near the single bearing and the sensor (compressor 4) attached near the four-stage bearing of the second compressor.

The apparatus for diagnosing an abnormality of a rotating machine using a virtual frequency response function based on a virtual frequency response function according to the present invention is characterized in that the rotating machine has two compressors and one motor and two sensor mounting positions are located near the motor front bearing The abnormality diagnosis of the rotating machine is performed based on the two signals measured from the attached sensor (motor 1) and the sensor (motor 3) mounted on the side case of the motor.

An apparatus for diagnosing an abnormality of a rotating machine based on a virtual frequency response function based on a virtual frequency response function according to the present invention is characterized in that the rotating machine has two compressors and one motor and two sensor mounting positions are located near the motor rear bearing And the abnormality diagnosis of the rotating machine is performed based on the two signals measured from the sensor (motor 2) mounted and the sensor (motor 3) mounted on the side case of the motor.

The apparatus for diagnosing an abnormality of a rotating machine using a virtual frequency response function based on a virtual frequency response function according to the present invention is characterized in that the rotating machine has two compressors and one motor and two sensor mounting positions are located near the motor front bearing And the abnormality diagnosis of the rotating machine is performed based on the two signals measured from the sensor (motor 1) mounted and the sensor (motor 2) mounted near the motor rear bearing.

According to the present invention, there is provided a method for diagnosing a malfunction of a rotating machine using a virtual frequency response function and an apparatus for diagnosing a malfunction of a rotating machine using the same, wherein sensors for measuring vibration signals at two points are installed at optimal positions, It is possible to diagnose abnormalities.

1 is an experimental apparatus of a sensor for diagnosing a rotating machine malfunction,
FIGS. 2 to 4 are graphs showing power spectrum values measured by the respective sensors,
5 and 6 are graphs comparing the correlations between the sensors,
FIG. 7 is a top view of a rotary machine tested to apply the method of diagnosing a rotating machine abnormality based on the virtual frequency response function according to the present invention,
8 is a graph showing correlation coefficient trends calculated from signals detected at two points,
9 is a graph showing correlation coefficient trends according to combinations of different attachment positions of four sensors attached to a motor.

Hereinafter, a method of diagnosing a malfunction of a rotating machine based on a virtual frequency response function and a malfunction diagnostic apparatus of a rotating machine using the same will be described in detail as follows.

Generally, a plurality of vibration sensors should be mounted at different positions (S1, S2) in order to perform an error diagnosis of a rotating machine by a virtual frequency response function.

In the normal state, there is no problem because the mounting position of the vibration sensor does not affect much. However, in the abnormal state, the sensor includes other components such as the number of rotations of the motor, It can be difficult.

Therefore, it is very important to attach the mounted sensor in order to accurately diagnose the malfunction of the rotating machine based on the virtual frequency response function.

In the case of using the mechanical diagnosis method based on the virtual frequency response function, 'the correlation between the two signals must be very small' is used in order to find the optimum sensor attachment position according to the assumption between the signals detected from the sensors mounted at different positions The correlation coefficient is calculated by the following equation (1).

ρ _xy = Cov ( _{x, y} ) / ρ _x. ρ _y (1)

(Where ρ _x .ρ _y is the standard deviation of the two-point signal, and Cov ( _{x, y} ) is the covariance of the two signals)

The correlation coefficient is determined by the ratio of the covariance of the two signals to the product of the signal standard deviation of the two points.

The value of the correlation coefficient is preferably in the range of 0.05 to 0.25.

The value of the correlation coefficient includes both the steady state and the abnormal state. The meaning of including both cases is that if the correlation coefficient value in the steady state is 0.1, the correlation coefficient value in the abnormal state is also 0.1 , And the diagnostic method based on the virtual frequency response function is applied as the correlation coefficient values in the steady state / abnormal state are similar.

The method of obtaining the correlation coefficient is,

First, the vibration signal of two points is detected.

Second, the standard deviation of the detected vibration signals at two points is calculated.

In this case, the vibration signal of two points is continuously detected, but the detected signal is averaged in units of one block of data in a period of 50 to 60 seconds, and a standard deviation is calculated.

Third, calculate the covariance of the vibration signal at two points.

Fourth, the correlation coefficient is calculated from the second calculated standard deviation and the third calculated covariance.

Fifth, the correlation coefficient is calculated by averaging the values of the block units calculated in the fifth above by 50 to 100 block units.

The standard deviation and mean

The standard deviation is

The average value

to be.

The formula for obtaining the covariance

to be

1, an artificial large defect was made in the outer ring of the ball bearing, and then the vibration was measured at the following measuring points, and the results of the rotational machine abnormality diagnosis method using the power spectrum and the virtual response function were compared .

At this time, the rotor is rotating at 1800 rpm, and the frequency of the bearing outer ring defect is about 93 Hz.

The method of diagnosing the rotating machine abnormality based on the virtual frequency response function is applied to the signals measured by the two sensors installed in the experimental apparatus of FIG.

In the case of using the sensor mounting positions S ₁ and S _{2 in the first embodiment} and using the sensor mounting positions S ₁ and S _{3 in} FIG. 3, the sensor mounting positions S ₂ and S ₃ of FIG. 4 are used In all three cases, the bearing fault frequency component (93 Hz) shows a distinct peak.

Referring to FIGS. 2 to 4, the power spectral values measured by each sensor in the steady state show that the motor component is the main frequency component and the rotation frequency and other components are relatively small. However,

First, in the case of using the sensor mounting positions S ₁ and S ₂ , it is difficult to interpret the diagnosis results because it contains many other components in addition to the defective frequency component.

Second, although the results of using the sensor mounting positions S ₁ and S ₃ show somewhat improved results, frequency components near 300 Hz are not clearly recognized.

Third, in the case of using the sensor mounting positions S ₂ and S ₃ , only the defective frequency component is clearly shown, and the size is also very large, and the diagnostic result is best represented.

The correlation coefficient between the two signals detected at the two points based on the results of FIGS. 2 to 4 is analyzed. When the power of the defect component is small, a normal power spectrum can be used as a sensor directly attached to the bearing However, it has been verified that it is possible to grasp the defective frequency components by using the virtual FRF diagnosis method.

However, since the results can vary greatly depending on the position of the two sensors, it can be seen that the selection of the position of the sensor is very important.

The sensor location can be set based on the very small correlation between the two "assumptions" of the virtual FRF diagnostic method.

For example, in the case of this experiment, the correlation between the sensors is shown in FIG. 5, FIG. 6, and Table 1.

The sensor location can be set based on the small correlation between the two "assumptions" of the virtual FRF diagnostic method.

However, in case of too low (eg 0.01), the virtual FRF diagnostic method can not be used because the two signals are so independent that the correlation is very low.

First, the results of applying the virtual FRF diagnostic method are summarized as follows: S ₂ and S ₃ , where the correlation coefficient is 0.09, and S ₁ and S _3, And S ₀ and S ₃ with a correlation coefficient of 0.19.

Also, it can be seen that the correlation coefficient in the abnormal state (defect) should also be in the range and in the steady state without any large change.

From this result, the following conclusions can be obtained by comparing the results of the correlation coefficient analysis between the sensors.

First, the correlation between two signals is less than 0.1 in all cases, and the best correlation is obtained when there is no correlation with the state (normal / abnormal).

Second, it can be seen that the virtual FRF diagnosis method can be applied even when the correlation between two signals is relatively low (less than 0.2) and the correlation according to the state is relatively low.

Third, when the correlation between two signals is large, it is impossible to apply the virtual FRF diagnosis method.

Example 2 The rotating machine tested to apply the virtual frequency response function based on the virtual frequency response function of FIG. 7 is a 3-Multi Stage type air compressor, which is a rotating machine having two compressors and a motor.

Table 2 shows the sensor position and sensor name (arbitrarily assigned).

8 and Table 3 show correlation coefficients calculated from four sensors mounted on the compressor and sensors mounted on two different points as one pair. FIG. 8 shows a correlation coefficient calculated from signals detected at two points , And Table 3 shows the correlation coefficient average value of all data blocks.

The vertical axis in FIG. 8 represents the magnitude of the correlation coefficient, and the horizontal axis represents the average of the signals detected in 50 to 60 seconds in units of one data block, and shows the trend of the correlation coefficient trend in 50-block units.

Table 3 shows the correlation coefficients of the four sensors attached to the compressor according to different combinations of attachment positions.

According to Table 3, the correlation coefficient between the two signals measured from the sensor attached to the first stage bearing (compressor 1) of the first compressor and the sensor attached to the second stage bearing (compressor 2) of the first compressor is 0.178 The proposed method is suitable to apply the fault diagnosis method based on the virtual frequency response function.

Further, the correlation coefficient between the two signals measured from the sensor (compressor 3) attached near the three-stage bearing of the second compressor and the sensor (compressor 4) attached near the four-stage bearing of the second compressor is 0.021, Based on this study,

Therefore, when the rotational mechanical device is applied to the fault diagnosis method based on the virtual frequency response function, a suitable one pair of sensor attachment positions is located near the one-stage bearing of the first compressor and the two- And compressor 2), and another suitable pair of sensor attachment locations are near the three-stage bearing of the second compressor and near the four-stage bearing of the second compressor (compressor 3 and compressor 4).

FIG. 9 shows the correlation coefficient trends according to different combinations of attachment positions of the four sensors attached to the motor, and Table 4 shows the average value of the correlation coefficients of the entire data blocks. According to Table 4, a suitable one pair of sensor attachment positions is a correlation coefficient between the two signals measured from the sensor (motor 1) attached to the motor front bearing and the sensor (motor 3) mounted on the motor side case is 0.058 The proposed method is suitable to apply the fault diagnosis method based on the virtual frequency response function.

The sensor attachment position of the other pair is based on a virtual frequency response function with a correlation coefficient of 0.19 between the two signals measured from the sensor (motor 2) mounted near the motor rear bearing and the sensor mounted on the motor side case (motor 3) And it was found to be suitable to apply the diagnostic method of rotating machine failure.

The other pair of sensor attachment positions is a correlation coefficient between the two signals measured from the sensor (motor 1) mounted near the motor front bearing and the sensor mounted near the motor rear bearing (motor 2) is 0.10, And it is shown that it is suitable to apply the function based fault diagnosis method.

Therefore, when the rotational mechanical device is applied to the abnormality diagnosis method based on the virtual frequency response function, a suitable pair of sensor attaching positions are located near the motor front bearing, near the motor side case (motor 1 and motor 3) The position is near the motor rear bearing and near the motor side case (motor 2 and motor 3), and the other pair of sensors are located near the motor front bearing and near the motor rear bearing (motor 1 and motor 2)

Although the present invention has been described in detail with reference to the above embodiments, it is needless to say that the present invention is not limited to the above-described embodiments, and various modifications may be made without departing from the spirit of the present invention.

Claims (12)

A method for diagnosing a rotating machine malfunction based on a virtual frequency response function,
I) measuring vibration signals at two points;
Ii) calculating a standard deviation of the vibration signals of the two points measured in the step i);
Iii) calculating a covariance from the vibration signals at the two points measured in the step i);
Iv) calculating a correlation coefficient from the values calculated in the step ii) and the step iii);
And v) selecting a position of a sensor to be mounted at two points from the correlation coefficient calculated in step iv).
The method according to claim 1,
(Vi) mounting the sensor at two points selected in the step (v);
(D) measuring a first vibration signal from two points installed in step (vi);
Measuring a second vibration signal different from the first vibration signal;
Calculating a virtual frequency response function of the first vibration signal and the second vibration signal;
X) extracting a group delay from the calculated virtual FRF; And
Xi) analyzing the virtual FRF and the group delay,
Wherein the step (e) comprises: measuring a steady state signal at a plurality of different measurement points including a first measurement point and a second measurement point; And
And calculating a first cross spectrum from data obtained from the plurality of different measurement points. ≪ Desc / Clms Page number 20 >
3. The method of claim 2,
The step (i)
Measuring a signal of a fault condition at a plurality of different measurement points including a first measurement point and a second measurement point; And
And calculating a second cross spectral from the data obtained from the plurality of different measurement points. ≪ Desc / Clms Page number 20 >
3. The method of claim 2,
Wherein the virtual FRF is calculated from a first mutual spectrum for the first vibration signal and a second mutual spectrum for the second vibration signal.
The method according to claim 1,
Wherein the method for calculating the correlation coefficient in (iv) is calculated by the following equation (1).
ρ _xy = Cov ( _{x, y} ) / ρ _x. ρ _y (1)
(Where ρ _x .ρ _y is the standard deviation of the two-point signal, and Cov ( _{x, y} ) is the covariance of the two signals)
3. The method of claim 2,
Wherein the correlation coefficient as a reference for selecting the sensor position in the step (v) is in the range of 0.05 to 0.25.
The method according to claim 1,
The calculation range of the standard deviation in the step (ii) is one in which the average of the signals detected in 50 to 60 seconds is calculated in units of one data block, and the correlation coefficient is calculated as an average of 50 to 100 unit blocks Based on the virtual frequency response function.
8. An abnormality diagnosis apparatus for a rotating machine using a virtual machine response diagnostic function based on a virtual frequency response function, the apparatus comprising:
The rotating machine has two compressors and one motor,
The sensor mounting positions of the two points are based on signals measured from sensors attached to the first stage bearing (compressor 1) of the first compressor and sensors attached to the second stage of the first compressor (compressor 2) Wherein the abnormality diagnosis is performed based on the virtual frequency response function.
8. An abnormality diagnosis apparatus for a rotating machine using a virtual machine response diagnostic function based on a virtual frequency response function, the apparatus comprising:
The rotating machine has two compressors and one motor,
The sensor mounting positions of the two points are based on the signals measured from the sensor (compressor 3) attached near the three-stage bearing of the second compressor and the sensor (compressor 4) attached near the four-stage bearing of the second compressor Wherein the abnormality diagnosis is performed based on the virtual frequency response function.
8. An abnormality diagnosis apparatus for a rotating machine using a virtual machine response diagnostic function based on a virtual frequency response function, the apparatus comprising:
The rotating machine has two compressors and one motor,
The sensor mounting position of the two points is characterized in that the abnormality diagnosis of the rotating machine is performed based on the two signals measured from the sensor (motor 1) attached to the motor front bearing and the sensor (motor 3) mounted on the side case of the motor Based on the virtual frequency response function.
8. An abnormality diagnosis apparatus for a rotating machine using a virtual machine response diagnostic function based on a virtual frequency response function, the apparatus comprising:
The rotating machine has two compressors and one motor,
The sensor mounting position of the two points is characterized in that the abnormality diagnosis of the rotating machine is performed based on the two signals measured from the sensor (motor 2) mounted near the motor rear bearing and the sensor (motor 3) mounted on the side case of the motor Based on the virtual frequency response function.
8. An abnormality diagnosis apparatus for a rotating machine using a virtual machine response diagnostic function based on a virtual frequency response function, the apparatus comprising:
The rotating machine has two compressors and one motor,
The sensor mounting position of the two points is characterized by performing an error diagnosis of the rotating machine based on the two signals measured from the sensor (motor 1) mounted near the motor front bearing and the sensor (motor 2) mounted near the motor rear bearing Based on the virtual frequency response function.

Images