JP3373718B2 - Bearing damage detection method - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a bearing, and more particularly to a bearing flaw detection method for detecting flaws in a rolling bearing.

[0002]

2. Description of the Related Art In a bearing comprising an inner ring and an outer ring, at least one of which rotates, a rolling element rollingly contacting these two wheels and a retainer holding the rolling element, when the inner ring, the outer ring or the rolling element is damaged, Japanese Patent Publication No. 7-26942 discloses a device capable of detecting which has a scratch without disassembling.

This is because the inner ring is rotated with a radial load applied to the bearing, the vibration generated at this time is measured by a sensor, and the obtained vibration waveform is subjected to envelope processing and then compared with a reference value to obtain a reference value. A damage signal is output when it exceeds. The generation cycle of the damage signal is the rotation cycle of the inner ring (1 / fr) or the cycle at which one point on the inner ring passes the rolling element (1 / f1).
If it is in synchronism with the above, it is judged that the inner ring is damaged, and if the above-mentioned generation cycle is in synchronism with the cycle (1 / f2) in which the rolling element passes through the load zone of the outer ring under the radial load It is judged that the outer ring is damaged, and if the above-mentioned generation cycle falls within the allowable range of ± 1 / 2fc with respect to the retainer rotation cycle (1 / fc), it is judged that the rolling elements are damaged. is there.

[0004]

In this case, the period (1 / f1) at which one point of the inner ring passes the rolling element is 0.25 times the rotational period (1 / fr) of the rotating inner ring. When it is larger (fr / f1> 0.25), vibration occurs at least once in the load zone, which causes a problem that the vibration generation cycle due to scratches cannot be specified.
In a bearing that is accompanied by deformation of the inner and outer rings along with scratches,
Even if fr / f1 ≦ 0.25, there is a problem that the vibration generation cycle due to scratches cannot be specified because the vibration may never occur in the load zone.

The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a bearing flaw detection method capable of identifying a member having a flaw on any bearing. To provide.

[0006]

SUMMARY OF THE INVENTION In the present invention, however, the vibration waveform obtained by measuring the vibration generated in the bearing in which the inner ring or the outer ring is rotated by a sensor is subjected to envelope processing, frequency analysis, and rotation. A frequency f1 at which one point of the inner ring or the outer ring passes the rolling element, a frequency f2 at which one point of the fixed outer ring or the inner ring passes the rolling element, and a frequency fb at which one point of the rolling element contacts the inner ring and the outer ring, It is characterized in that the amplitude value at the frequency corresponding to the rotation frequency fr of the rotating inner ring or outer ring and the rotation frequency fc of the retainer is compared with a reference value. It is possible to identify a member having a scratch regardless of the value of the ratio between the frequency f1 and the frequency fr.

The frequency for extracting the amplitude value to be compared with the reference value is (n + 1) times (n is an integer of 1 or more) the rotational frequency fr of the rotating inner or outer ring, or (n + 1) of the rotational frequency fc of the retainer. ) Double the frequency,
You may make it add the frequency of n times the difference fr-fc of the rotation frequency fr of the rotating inner ring or outer ring, and the rotation frequency fc of the retainer. Further, as a frequency for extracting an amplitude value to be compared with the reference value, a beat component f1 ± n × f
r, f2 ± n × fr, fb ± n × fc, fb ± n × (f
You may make it add the frequency of r-fc).

The frequencies for extracting the amplitude value to be compared with the reference value are (n + 1) × f1, (n + 1) × f2, (n +
It is also preferable to add a frequency of 1) × fb. When comparing, the extracted amplitude values at multiple frequencies are integrated with an evaluation formula to compare with a reference value, or the extracted amplitude values at multiple frequencies are integrated with multiple evaluation formulas and used as a reference value. You may make it compare. An evaluation formula may be created for each scratch position and compared with the reference value, or an evaluation formula may be created for each type of abnormality and compared with the reference value. Furthermore, each element of the evaluation formula may be converted by a conversion formula and compared with the reference value. It is also preferable to use a learned neural network that inputs the extracted amplitude values at a plurality of frequencies and outputs the scratch position.

Further, according to the present invention, the vibration waveform obtained by measuring the vibration generated in the bearing rotating the inner ring or the outer ring with the sensor is compared with the reference value, and the time interval of the occurrence time of the vibration exceeding the reference value is determined. , A cycle in which one point of the rotating inner ring or outer ring passes the rolling element, an integral multiple cycle of the cycle 1 / f1, and a cycle in which one point of the fixed outer ring or inner ring passes the rolling element 1 / f2
Another feature is that it is compared with each cycle of an integral multiple cycle of 1 and the integer multiple cycle of 1 / fb in which one point of the rolling element contacts the inner ring and the outer ring.

In this case, it is preferable to use the maximum vibration amplitude value when the amplitude exceeds the reference value within a predetermined time after the amplitude exceeds the reference value. It is also preferable to compare the time interval pattern of the vibration occurrence time with the pattern set according to the deformed shape of the bearing. Furthermore, a learned neural network which inputs the time interval pattern of the vibration occurrence time and outputs the deformed shape of the bearing may be used.

When measuring the vibration, a radial load may be applied to the bearing. At this time, the outer ring or inner ring that is the fixed ring is rotated at a fixed angle and vibration is measured at each angle, or the outer ring or the inner ring that is the fixed ring is rotated at a different rotational speed from the inner ring or the outer ring that is rotating. Vibration measurement may be performed while rotating.

[0012]

DESCRIPTION OF THE PREFERRED EMBODIMENTS In FIG. 2, an outer ring 10 and an inner ring 1 are shown.
1 and a plurality of rolling elements 12 which are held by a retainer (not shown) and are in rolling contact with the outer ring 10 and the inner ring 11, the outer ring 10 is fixed and the inner ring 11 is fixed.
Is connected to the motor 2, and a vibration sensor 3 is attached to the outer ring 10. The vibration sensor 3 detects the vibration of the bearing 1 when the inner ring 11 is rotated by the motor 2.

As shown in FIG. 1, the obtained vibration waveform is first subjected to envelope processing and then frequency analysis by Fourier transform to extract the amplitudes of the following five frequencies. That is, the frequency f1 at which one point of the rotating inner ring 11 passes through the rolling element 12, and the fixed outer ring 10
Frequency f2 at which one point passes through the rolling element 12 and the rolling element 1
Frequency fb at which one point of 2 contacts inner ring 11 and outer ring 10
Then, the amplitude at the rotation frequency fr of the rotating inner ring and the rotation frequency fc of the retainer (revolution frequency of the rolling element 12) are extracted. Although these frequencies change depending on the number of rotations of the inner ring 11 by the motor 2, it is preferable to actually measure the number of rotations with an encoder or the like.

Then, the amplitude value of each of these frequencies is compared with a reference value respectively, and if all are less than the reference value, it is judged that there is no abnormality in the bearing, and the contact frequency f1, f2, fb of the former three rotates. When the amplitude value at the frequency f1 at which one point on the inner ring 11 passes the rolling element 12 exceeds the reference value, it is determined that the inner ring 11 is damaged, and one point on the fixed outer ring 10 is the rolling element 1.
When the amplitude value at the frequency f2 that passes 2 exceeds the reference value, it is determined that the outer ring 10 is damaged, and the amplitude value at the frequency fb at which one point of the rolling element 12 contacts the inner ring 11 and the outer ring 10 is the reference value. If it exceeds the value, it is determined that the rolling element 12 has a scratch.

Of the remaining two rotation frequencies fr and fc, when the amplitude value of the rotating inner ring 11 at the rotation frequency fr exceeds the reference value, the outer ring 10 is damaged and the inner ring 11 is deformed at the same time. When the amplitude value of the retainer at the rotation frequency fc exceeds the reference value, it is determined that the rolling element 12 is damaged and the inner ring 11 is deformed at the same time.

In this case, the rotation frequency fr and the contact frequency f
The scratches can be detected and specified regardless of the value of the ratio of the bearing 1 to 1 and the deformation of the fixed outer ring 10 and the rotating inner ring 11 depends on the contact frequencies f1, f2, fb, etc. Since the generated shock waveform appears in the form of amplitude-modulated with the rotation frequencies fr and fc, it is possible to detect a defect accompanied by deformation abnormality by extracting the amplitude value of the rotation frequencies fr and fc. Here, the outer ring 10 is fixed,
The inner ring 1 has been described as rotating the inner ring 11.
1 is fixed, the outer ring 10 is rotated, and the outer ring 1
By setting the rotational frequency of 0 to fr, the flaw detection of the bearing 1 can be performed in the same manner as above.

Another example is shown in FIGS. Here, the inner ring 11 of the bearing 1 is fixed by a fixed shaft 40, and a belt 43 is bridged between a pulley 41 fixed to the outer periphery of the outer ring 10 and a pulley 42 attached to the output shaft of the motor 2 to rotate the outer ring 10. I am able to let you do it. The vibration sensor 3 is attached to a mount 44 having the fixed shaft 40.

Even in this case, the damage of the bearing 1 is detected in the same manner as in the above case, but the load F in the radial direction is applied to the bearing 1 by the belt 43. At this time, the radial gap in the bearing 1 is an area A1 where the load is large and small due to the radial load F (right half of the bearing 1 in FIG. 4) and an area A2 where the load is small and large ( The left half of the bearing 1 in FIG. 4) occurs. And even if there are scratches on the bearing 1,
In the area A2 where the radial gap is large, almost no impact waveform is seen due to weak or no contact between the scratch and other parts, and conversely, the impact waveform having a relatively large amplitude is seen in the area A1 where the radial gap is small. It will be.

Therefore, the bearing 1 in which the rotating outer ring 10 is damaged
Then, as shown in FIG. 5, the amplitude is modulated by the rotation frequency fr of the outer ring 10, and the bearing 1 having a flaw in the fixed inner ring 11 has an angle of the flaw position with respect to the direction of the radial load F as shown in FIG. θ (θ = 0 ° in Fig. 6 (a), θ in Fig. 6 (b)
= 90 ° or 270 °, (c) has different amplitudes depending on θ = 180 °), and in the bearing 1 in which the rolling element 12 has a flaw, the vibration waveform has an amplitude at the rotation frequency fc of the retainer as shown in FIG. It will be modulated.

For this reason, the frequencies f1, f2, f
In comparing the amplitude values at b, fr, and fc with the reference value, if the amplitude value of the frequency f1 or the frequency fr exceeds the reference value, it is assumed that the outer ring 10 that is rotating has a scratch, and the amplitude of the frequency f2 is If the value exceeds the reference value, it may be determined that the fixed inner ring 11 is damaged, and if the amplitude value of the frequency fb or the frequency fc exceeds the reference value, the rolling element 12 may be damaged. For the further fixed inner ring 11, from the relationship between the flaw position and the direction of the radial load F,
Rotate the inner ring 11 by 360 ° at a constant angle to obtain each angle θ.
The vibration measurement is performed every time, and the amplitude value of the frequency f2 at the angle θ at which the maximum amplitude value is recorded is compared with the reference value. In order to efficiently perform this work, vibration measurement may be performed while rotating the inner ring 11 at a rotation speed different from the rotating outer ring 10. Then, for frequency analysis, the rotated inner ring 1
Inner ring 1 so that it can be considered that 1 is fixed
If the rotation frequency of 1 is fin (the rotation in the same direction as the outer ring 10 that is rotating is the positive direction), each frequency f1,
Instead of f2, fb, fr, and fc, f1 × (fr-fi
n) / fr, f2 × (fr-fin) / fr, fb ×
(Fr-fin) / fr, fr-fin, fc * (fr
-Fin) / fr The frequency amplitude is extracted. Even when the outer ring 10 is fixed and the inner ring 11 is rotated by the motor 2, it is possible to detect the scratch in the same manner.

If the outer ring 10 and the inner ring 11 are distorted or deformed, as shown in FIG. 8, the frequencies for extracting the amplitude are f1, f2, f
In addition to b, fr, and fc, the frequency (n + 1) × fr (n
Is preferably an integer of 1 or more). This frequency (n +
1) xfr vibration is generated by rotating outer ring 10 or inner ring 11
When the rotating outer ring 10 is distorted at three points as shown in FIG. 9, the vibration waveform due to the scratches on the fixed inner ring 11 is as shown in FIG.
As shown in 0, the vibration waveform is amplitude-modulated at a frequency of 3 × fr. Therefore, by extracting the amplitude of the frequency (n + 1) × fr and comparing it with the reference value, the bearing 1 in which the fixed inner ring 11 or outer ring 10 is damaged and the rotating outer ring 10 or inner ring 11 is deformed, This can be detected and the presence of scratches and deformations can be pointed out.

Similarly, frequencies (n + 1) × fc and n ×
If the amplitude of (fr-fc) is also extracted and compared with the reference value, the inner ring 1 in which the rolling element 12 is damaged and is fixed
1 or outer ring 10 is deformed Bearing 1 and rolling element 12 are damaged and outer ring 10 or inner ring 11 is rotating
It is possible to detect the scratch and the deformation of the bearing 1 having the deformation.

Further, as a frequency for extracting an amplitude value to be compared with the reference value, a beat waveform component f1 ± n × fr,
f2 ± n × fr, fb ± n × fc, fb ± n × (fr−
If the frequency fc) is applied, it is possible to detect scratches and deformations with higher accuracy. Further, when considering the case where there are a plurality of flaws, the frequencies for extracting the amplitude value to be compared with the reference value are (n + 1) × f1 and (n + 1) × f.
2, it is advisable to add a frequency of (n + 1) × fb. Even if there are a plurality of scratches, the contact frequency f1 is obtained if the rotating outer ring 10 is damaged, the contact frequency f2 is obtained if the fixed inner ring 11 is damaged, and the contact frequency f is obtained if the rolling element 12 is damaged.
Since the influence appears in b, the flaw can be detected. However, when two flaws are present at a distance of π radian with respect to each contact frequency, 2 × f1, 2 × f2, 2 are obtained.
Although the frequency of × fb appears, the frequencies f1 and f
2, it will not appear in fb.

Next, the amplitude values p1 and p at each of the above frequencies
In comparing with the reference value of 2 ... pn, the reference value is not set for each frequency and compared, but a plurality of amplitude values p
1, p2 ... pn integrated evaluation formula E (p1,
You may make it compare by the value of p2 ... pn). For example, considering a vector having amplitude values p1, p2, ... Pn of all frequencies as elements, the magnitude of the vector, that is, the square root value of the sum of squares of each element is compared as a reference value as an evaluation formula. . At this time, each element of the vector is converted by the conversion formula, and by comparing the size of the vector composed of the converted elements with the reference value as the evaluation formula, it is possible to perform standardization and weighting of each element, Further, it becomes possible to detect scratches and deformations with higher accuracy. The coefficient at this time is
For example, the frequency amplitude value of the bearing to be inspected is pa '= (pa-AVGpa) / STDpa pa': normalization based on the distribution (average value or standard deviation) of each frequency amplitude value obtained from the vibration of a plurality of reference bearings. Frequency amplitude value pa: Frequency amplitude value AVGpa: Average value of frequency amplitude value pa of the reference bearing STDpa: Standard deviation value of frequency amplitude value pa of the reference bearing may be standardized.

Amplitude values (p1,
p2 ... pi, pi + 1, pi + 2 ... pm, pm + 1, p
m + 2, ... pn) are integrated by a plurality of evaluation expressions E1 (p1, p2 ... pi) E2 (pi + 1, pi + 2 ... pm) E3 (pm + 1, pm + 2 ... pn), and the values of these evaluation expressions are compared with a reference value. Then, it is possible to obtain an evaluation result that cannot be obtained by a single evaluation formula.

When creating a plurality of evaluation formulas, the first
In the evaluation formula of, the fundamental frequency generated by the damage or deformation of the bearing, for example, f1, f2, fb, n × fr, n × fc,
The amplitude value of n × (fr−fc) is integrated, and in the second evaluation formula, the frequency of the amplitude modulation component generated by the scratch or deformation of the bearing, for example, f1 ± n × fr, f2 ± n × fr, fb ±
When the amplitude values of n × fc and fb ± n × (fr−fc) are integrated, bearings are roughly classified into good / defective / necessary check of the bearing in the first evaluation formula, and the second evaluation only for the bearing of check required. It is possible to adjust the abnormality detection accuracy such as performing a highly accurate inspection using a formula.

In addition to the above, in the first evaluation formula, the frequency generated by the scratch on the rotating outer ring (or inner ring), for example, f
1, n × fr, f1 ± n × fr are integrated, and in the second evaluation formula, a frequency generated by a scratch on the fixed inner ring (or outer ring), for example, f2, n × fr, f2 ± n
The amplitude value of × fr is integrated, and the frequency generated by the scratch of the rolling element by the third evaluation formula, that is, fb, n × fc, f
If the amplitude values of b ± n × (fr−fc) are integrated, it becomes possible to specify the scratch position with high accuracy. Furthermore, if an evaluation formula is created for each type of abnormality, that is, the first Frequencies f1, f2, fb generated by scratches on each part in the evaluation formula
Are integrated, and the frequencies n × fr, n × fc, and n × (fr−f generated by the deformation of each part in the second evaluation formula are integrated.
By integrating the amplitude values of c), the cause of the abnormality can be specified with high accuracy.

Here, the reference value may be set by a statistical method such as three times the standard deviation value. In this case, the optimum setting of the reference level is repeatedly empirically repeated. It is necessary to do so, and it takes a lot of time and effort. For this purpose, when an amplitude value of each frequency as shown in FIG. 11 is input, and a flaw position is output, a neural net that has learned about the presence or absence of a flaw and the identification and deformation of the flaw position is used to easily set the optimum reference value. It can be performed.

As shown in FIG. 12, after the vibration waveform from the vibration sensor 3 is envelope-processed, it is compared with the reference value without frequency analysis to find the time when the reference value is exceeded, and the time interval between the exceeded times is calculated. The scratch position can also be specified by comparing (t1, t2, t3 in FIG. 13) with the following three cycles. That is, if one point of the rotating inner ring passes through the rolling element and is equal to an integer multiple cycle n / f1 of the cycle 1 / f1, the rotating inner ring is damaged, and one point of the fixed outer ring is rolling element. If the time interval matches the cycle n / f2, which is an integer multiple of the cycle 1 / f2, the outer ring is damaged and one point of the rolling element contacts the inner ring and the outer ring.
If the time interval matches the integer multiple cycle of fb, n / fb, it is determined that the rolling element has scratches. If there is no corresponding cycle, it is unsteady vibration, so it is determined that foreign matter is not mixed into the bearing. To do.

In the case of making such a judgment, it is possible to detect a defect in which the inner ring or the outer ring is deformed, as well as the scratches on each part of the bearing. If the inner ring or outer ring is deformed,
Since the shock vibration waveform due to the scratches causes a beat, the shock waveform having a constant amplitude does not appear in a constant cycle and becomes an aperiodic waveform as shown in FIG. In this case, the contact frequencies f1, f2, fb of each part are less likely to be affected and the analysis becomes difficult, but an integral multiple of the contact cycle of each part of the bearing n / f1, n
This can be dealt with by making a comparison with / f2 and n / fb.

Here, it is the reference value exceeded time for determining the time interval, but in the case of the vibration waveform as shown in FIG.
Since only a preset reference value s0 is used, a plurality of reference value excess times are detected for a certain shock vibration waveform. Therefore, at the time T1 when the reference value is exceeded, the maximum amplitude value of the rising at that time is updated. If the reference value is set to s1 and the reference value is again exceeded within a certain time t, then the maximum amplitude value of the rising waveform of T2 is set to the new reference value s2 again and the reference value exceeded time is updated from T1 to T2. When the reference value s3 is not exceeded within the fixed time t after the last reference value excess time T3, the reference value excess time is set to T3 and the reference value is returned to the initial value s0. To do. This makes it possible to obtain a single reference value excess time for a given shock vibration waveform.

Further, when the cycle of the beat superimposed on the shock vibration waveform differs depending on the deformed shape, the time interval t
The patterns of 1, t2 and t3 are also different. Therefore, if the time interval pattern is compared with a preset pattern according to the deformed shape of the bearing, the deformed shape can be detected. Then, as shown in FIG. 15, if a neural network that is trained by the deformed shape and the time interval pattern is provided and the time interval pattern is input to the neural network, the deformed shape of the bearing is output. It becomes easy to set a pattern for each deformed shape.

[0033]

As described above, in the present invention, the vibration waveform obtained by measuring the vibration generated in the bearing by the sensor is subjected to the envelope processing and then the frequency analysis is performed, and one point of the rotating inner ring or outer ring is the rolling element. Frequency f1 that passes through, the frequency f2 that one point of the fixed outer ring or the inner ring passes through the rolling element, the frequency fb that one point of the rolling element contacts the inner ring and the outer ring, and the rotation frequency fr of the rotating inner ring or the outer ring. In order to compare the amplitude value at the frequency corresponding to the rotation frequency fc of the retainer with the reference value, it is possible to identify the member having a flaw regardless of the value of the ratio between the frequency f1 and the frequency fr. In addition, it is possible to detect a defect accompanied by deformation abnormality from the frequencies fr and fc.

When a frequency that is (n + 1) times the rotation frequency fr of the rotating inner or outer ring (n is an integer of 1 or more) is also used as the frequency for extracting the amplitude value to be compared with the reference value, one of the inner ring and the outer ring is used. When there is a scratch and the other has a deformation, the detection of the scratch and the deformation can be performed. As the frequency for extracting the amplitude value to be compared with the reference value, the frequency is (n + 1) times the rotation frequency fc of the retainer, and n times the difference fr-fc between the rotation frequency fr of the rotating inner ring or outer ring and the rotation frequency fc of the retainer. When the frequency of is also used, it is possible to detect the scratch and the deformation when there is a scratch on the rolling element and the inner ring or the outer ring is deformed.

The frequencies for extracting the amplitude value to be compared with the reference value are f1 ± n × fr and f2 ± n × which are beat components.
When frequencies of fr, fb ± n × fc, and fb ± n × (fr-fc) are also used, scratches and deformations can be detected more accurately, and further, as a frequency for extracting an amplitude value to be compared with a reference value. , (N + 1) × f1, (n + 1) × f2,
When the frequency of (n + 1) × fb is also used, the scratch position can be specified even when there are a plurality of scratches.

In comparison, if the amplitude values at a plurality of extracted frequencies are integrated by an evaluation formula and compared with a reference value, the accuracy of flaw detection can be improved, and the extracted frequencies at a plurality of frequencies can be compared. If the amplitude value is integrated by a plurality of evaluation expressions and compared with the reference value, the accuracy can be further improved. In particular, when creating a plurality of evaluation expressions, when creating an evaluation expression for each scratch position or creating an evaluation expression for each type of abnormality, it is possible to more accurately determine the cause of the abnormality.

Furthermore, if each element of the evaluation formula is converted by the conversion formula and compared with the reference value, the standardization and weighting of each element can be performed, so that more accurate detection of scratches and deformations is possible. It will be possible. When using a learned neural network that inputs the extracted amplitude values at a plurality of frequencies and outputs the scratch position, it becomes easy to set the reference value. The vibration waveform is compared with a reference value, and the time interval of the occurrence of vibration that exceeds the reference value is fixed as an integral multiple cycle of cycle 1 / f1 at which one point of the rotating inner ring or outer ring passes through the rolling element and the fixed outer ring. Or when comparing each cycle of an integer multiple cycle of cycle 1 / f2 where one point of the inner ring passes the rolling element and an integer multiple cycle of cycle 1 / fb where one point of the rolling element contacts the inner ring and the outer ring Also,
Regardless of the value of the ratio between the frequency f1 and the frequency fr, it is possible to identify a member having a scratch and also detect a defect accompanied by deformation abnormality.

At this time, since the maximum vibration amplitude value when the amplitude exceeds the reference value within a predetermined time after the amplitude exceeds the reference value is used, the time interval can be accurately captured. It is possible to detect flaws with high accuracy and identify the deformed shape when comparing the time interval pattern of the vibration occurrence time with the pattern set according to the deformed shape of the bearing, and the time interval pattern of the vibration occurrence time. If a learned neural network that inputs the input and outputs the deformed shape of the bearing is used, it becomes easy to set the pattern according to the deformed shape.

The radial load may be applied to the bearing when measuring the vibration. The influence of the scratch on the vibration becomes more prominent, and the scratch can be easily detected. At this time, by rotating the outer ring or the inner ring which is the fixed wheel at a constant angle and measuring the vibration at each angle,
It is possible to detect scratches on the outer ring or inner ring as a fixed ring with high accuracy, and if the vibration is measured while rotating the outer ring or inner ring as the fixed ring at a different rotation speed than the rotating inner ring or outer ring, it is efficient. Good scratch detection can be performed.

[Brief description of drawings]

FIG. 1 is a flowchart of an example of an embodiment of the present invention.

2A and 2B show a vibration measurement structure of the above, wherein FIG. 2A is a front view and FIG. 2B is a side view.

FIG. 3 is a perspective view showing another vibration measurement structure of the above.

FIG. 4 is a schematic diagram of the above.

FIG. 5 is a time chart of an example of the above-mentioned vibration waveform.

FIG. 6 shows another example of the above vibration waveform, in which (a) is θ.
= 0 °, θ = 90 ° or 270 ° in (b), θ = 1 in (c)
It is a time chart at the time of 80 degrees.

FIG. 7 is a time chart of another example of the vibration waveform of the above.

FIG. 8 is a flowchart of another example.

FIG. 9 is a schematic diagram of the above.

FIG. 10 is a time chart of an example of the above-mentioned vibration waveform.

FIG. 11 is an explanatory diagram of an example of a neural network.

FIG. 12 is a flowchart of another example.

FIG. 13 is a time chart of an example of the above-mentioned vibration waveform.

FIG. 14 is an enlarged time chart of the vibration waveform of the above.

FIG. 15 is an explanatory diagram of another example of a neural network.

[Explanation of symbols]

1 bearing 2 motor 3 Vibration sensor 10 outer ring 11 inner ring 12 rolling elements

─────────────────────────────────────────────────── ─── Continuation of the front page (56) Reference JP 62-270820 (JP, A) JP 6-186136 (JP, A) JP 1-152335 (JP, A) JP 55- 54429 (JP, A) JP-A-7-318457 (JP, A) JP-A-5-157664 (JP, A) JP-A-6-58849 (JP, A) JP-A-6-213771 (JP, A) Japanese Patent Publication 7-26942 (JP, B2) (58) Fields investigated (Int.Cl. ⁷ , DB name) G01M 13/04 G01H 17/00 G01M 1/14 G01M 19/00 G01N 29/14

Claims (18)

(57) [Claims] 1. A flaw detection method for a bearing comprising an inner ring and an outer ring and a rolling element rolling contacting at least one of the rotating two wheels and a retainer holding the rolling element, wherein the inner ring or the outer ring is rotated. The vibration waveform obtained by measuring the generated vibration with the sensor is subjected to the frequency analysis after the envelope processing, and the frequency f1 at which one point of the rotating inner ring or the outer ring passes through the rolling element and the one point of the fixed outer ring or the inner ring are An amplitude value at a frequency corresponding to the frequency f2 passing through the rolling element, the frequency fb at which one point of the rolling element contacts the inner ring and the outer ring, the rotation frequency fr of the rotating inner ring or outer ring, and the rotation frequency fc of the retainer. A method for detecting flaws in a bearing, characterized in that 2. A frequency that is (n + 1) times (n is an integer of 1 or more) the rotational frequency fr of the rotating inner or outer ring is added as a frequency for extracting the amplitude value to be compared with the reference value. Item 1. A bearing flaw detection method according to Item 1. 3. As a frequency for extracting an amplitude value to be compared with a reference value, a frequency that is (n + 1) times the rotational frequency fc of the retainer (n is an integer of 1 or more) and a rotational frequency fr of the rotating inner ring or outer ring. The bearing flaw detection method according to claim 1, wherein a frequency n times a difference fr-fc of the rotation frequency fc of the retainer is added. 4. A beat component f1 ± n × fr, f2 ± n as a frequency for extracting an amplitude value to be compared with a reference value.
4. The method for detecting flaws in a bearing according to claim 2, wherein frequencies of xfr, fb ± n × fc, and fb ± n × (fr−fc) (n is an integer of 1 or more) are added. 5. As frequencies for extracting an amplitude value to be compared with a reference value, (n + 1) × f1, (n + 1) × f2, (n
The bearing flaw detection method according to claim 1, wherein a frequency of (+1) × fb (n is an integer of 1 or more) is added. 6. The bearing flaw detection method according to claim 1, wherein the extracted amplitude values at a plurality of frequencies are integrated by an evaluation formula and compared with a reference value. 7. The bearing flaw detection according to claim 1, wherein the extracted amplitude values at a plurality of frequencies are integrated by a plurality of evaluation expressions and compared with a reference value. Method. 8. The method for detecting flaws in a bearing according to claim 6, wherein an evaluation formula is prepared for each flaw position and compared with a reference value. 9. The bearing flaw detection method according to claim 6, wherein an evaluation formula is created for each type of abnormality and compared with a reference value. 10. The method for detecting flaws in a bearing according to claim 6, wherein each element of the evaluation formula is converted by a conversion formula and compared with a reference value. 11. The method for detecting flaws in a bearing according to claim 1, wherein a learned neural net which inputs the amplitude values at a plurality of extracted frequencies and outputs the flaw position is used. 12. A method for detecting damage to a bearing, comprising an inner ring and an outer ring, and rolling elements rollingly contacting these wheels and a retainer holding the rolling element, wherein a vibration generated in a bearing rotating the inner ring or the outer ring is detected by a sensor. The vibration waveform obtained by measurement is compared with a reference value, and the time interval of the occurrence time of vibration exceeding the reference value is an integer multiple of 1 / f1 which is the period when one point of the rotating inner ring or outer ring passes through the rolling element. And an integer multiple cycle of the cycle 1 / f2 in which one point of the fixed outer ring or inner ring passes the rolling element, and cycle 1 in which one point of the rolling element contacts the inner ring and the outer ring
A method for detecting flaws in a bearing, characterized by comparing each cycle with an integer multiple cycle of / fb. 13. The method for detecting flaws in a bearing according to claim 12, wherein a maximum vibration amplitude value when the amplitude exceeds the reference value within a predetermined time after the amplitude exceeds the reference value is used as the reference value. 14. The method for detecting flaws in a bearing according to claim 12, wherein a time interval pattern of vibration occurrence times is compared with a pattern set according to the deformed shape of the bearing. 15. The method for detecting flaws in a bearing according to claim 12, wherein a learned neural net that inputs a time interval pattern of vibration occurrence times and outputs a deformed shape of the bearing is used. 16. The method for detecting flaws in a bearing according to claim 1, wherein the vibration is measured while a radial load is applied to the bearing. 17. The method for detecting flaws in a bearing according to claim 16, wherein the outer ring or the inner ring serving as the fixed ring is rotated at a constant angle and vibration measurement is performed at each angle. 18. An outer ring or an inner ring serving as a fixed ring,
17. The method for detecting flaws in a bearing according to claim 16, wherein the vibration measurement is performed while rotating the inner ring or the outer ring at a different rotation speed than the rotating inner ring or outer ring.