JPH0730787B2 - Failure detection method for rolling bearings - Google Patents

Description

TECHNICAL FIELD The present invention relates to a method of detecting a cause of failure of a rolling bearing by a fuzzy inference method.

[Prior Art] Rolling bearings having rolling elements (balls or rollers) between an outer ring and an inner ring, such as ball bearings and needle bearings, will rotate if any of the contact surfaces of the outer ring, the inner ring and the rolling elements is damaged. Vibration occurs inside. This vibration has a different natural frequency depending on whether the scratch exists on the outer ring, the inner ring or the rolling element. This vibration is generated, for example, when the rolling element has a scratch, and the scratch contacts the rolling surface of the outer ring or the inner ring during rotation. If there is a scratch on the rolling surface of the outer or inner ring,
Vibration is caused by a plurality of rolling elements sequentially coming into contact with the scratch during rotation. Therefore, the frequencies of these vibrations change according to the rotational speed of the rolling bearing. The rotation speed of the rolling bearing is represented by the rotation speed of the inner ring with respect to the outer ring. Generally, when a rolling bearing is used, either the inner ring or the outer ring is fixed and the other is rotated.

The theoretical value of each frequency when the outer ring, the inner ring, or the rolling element has a flaw is represented by the following well-known formulas (1), (2), and (3). That is, the frequency fout when the outer ring is scratched is The frequency fin when there is a scratch on the inner ring is The frequency fball when there is a scratch on the rolling element is Represented by

Where fo: Number of revolutions of rolling bearing D: Pitch diameter of rolling bearing d: Diameter of ball when rolling element is ball Z: Number of balls α: Contact angle.

Vibration during rotation of the rolling bearing can be detected by attaching a vibration sensor to the rolling bearing and converting the vibration into an electric signal. The detected signal is frequency-analyzed by a frequency analyzer to obtain a frequency spectrum distribution represented by a plurality of frequencies and signal strength data of each frequency.

A plurality of frequencies obtained by the frequency analysis are compared with a reference spectrum distribution, which is a set of frequencies calculated from the equations (1), (2), and (3), and there is a match or approximation. Investigate whether or not. If there is a match or an approximation as a result, the signal strength at that frequency is compared with a judgment reference level that has been empirically determined in advance for judging a failure. If the frequency exceeds the determination reference level, it is determined that a failure having a cause corresponding to the frequency has occurred.

During the actual rotation of the rolling bearing, slips occur between the outer ring and the rolling elements and between the inner ring and the rolling elements, which causes minute vibrations. Since the detection output of the vibration sensor also includes the minute vibration, the actually measured frequency spectrum distribution does not match the reference spectrum distribution obtained from the equations (1), (2) and (3).

Therefore, a reference frequency band having a predetermined width is provided above and below the theoretical value of the frequency obtained by the above calculation. The reference frequency band is represented by the product of the resolution Δf of the frequency analyzer used and the empirically determined constant ± n ± n · Δf.

The failure location of the rolling bearing is determined by comparing the measured spectral distribution with the predetermined reference frequency band thus determined.

[Problems to be Solved by the Invention] The constant ± n is determined based on experience as the most appropriate value. However, the vibration frequency is represented by ± n · Δf depending on the condition of scratches on the rolling bearing. In some cases, it may not fall within the reference frequency band. In the judgment by the conventional technology, if the measured frequency does not fall within the reference frequency band, it is not judged as abnormal, so if the measured frequency deviates from the reference frequency band due to an unpredictable cause, there is a scratch. There is a possibility that it may be determined that there is no failure despite the fact that there is a failure such as.

Further, when the measured vibration level is compared with the determination reference level, the measured value of the vibration level changes depending on the distance between the vibration generation point and the measurement point (sensor position), and the longer the measured value, the smaller the measured value. Also, the higher the frequency, the greater the attenuation during propagation, resulting in a smaller measured value. Therefore, if the presence or absence of a failure is determined by making a determination at a fixed determination reference level, the failure may not be detected correctly.

[Means for Solving the Problems] A method for detecting abnormality of a rolling bearing according to the present invention includes a step of detecting vibration during rotation of a rolling bearing and outputting a detection signal, a step of A / D converting the detection signal, and the step of A step of performing a frequency spectrum analysis of the detection signal based on the D / D-converted detection signal; a step of selecting a predetermined number of frequencies in order of increasing spectral intensity from a plurality of frequencies subjected to the frequency spectrum analysis; Regarding the membership function, the horizontal axis is frequency and the vertical axis is grade, where the outer ring of the rolling bearing has scratches, the inner ring has scratches, or the rolling element has scratches Setting in memory a first membership function having three triangles with grade 1 points at each abscissa value of the frequency as vertices;
Determining a grade of each of the predetermined number of frequencies subjected to the frequency spectrum analysis based on a first membership function, with respect to the second membership function, the horizontal axis represents the spectrum intensity, and the vertical axis represents In the coordinates that represent the grade, pass through the origin of the coordinates, and grade 1
Setting in memory a second membership function represented by a curve that gradually increases toward, a grade for each spectral intensity of the predetermined number of frequencies based on the second membership function, A grade for each of the predetermined number of frequencies determined based on the first membership function, and a grade for each spectral intensity of the predetermined number of frequencies determined based on the second membership function. The step of obtaining the product of each of the predetermined number of frequencies and the inner ring grade, the outer ring grade, and the rolling element grade, and the algebraic sum of the product values for each of the inner ring, the outer ring, and the rolling element. , A step of determining a possibility that an abnormality has occurred in the inner ring, the outer ring and the rolling element.

[Operation] The frequency of vibration of the rolling bearing is obtained from the first membership function for each grade of the plurality of frequencies subjected to the frequency spectrum analysis, and each grade of the spectrum intensity of the plurality of frequencies is obtained from the second membership function. The probability that an abnormality has occurred is calculated by using the two grades and performing a predetermined calculation based on the graded flow.

[Embodiment] FIG. 1 shows an experimental vibration data collecting apparatus in an embodiment of the method for detecting abnormality of a rolling bearing according to the present invention. In the figure, a shaft 1 is supported by a normal ball bearing 2 and a defective ball bearing 3 under test, one end of which is connected to a motor 5 via a coupling 4. A disk 6 as a load is fixed to the other end of the shaft 1. Mounting housing for trial ball bearing 3A 3A
Is provided with an acceleration sensor 7, and detects vibration of the ball bearing 3 under test when the shaft 1 is rotated by the motor 5. The detected signal is input to the abnormality detection device 17. In the abnormality detecting device 17, the input signal is envelope-detected, A / D-converted, converted into a digital signal, and input to the computer system. In the computer system, the frequency spectrum analysis of the input signal is performed by a well-known spectrum analyzer function of predetermined software.

FIG. 2 is a plan view of the test ball bearing 3. The specifications of this ball bearing 3 are: shaft diameter (Φ1) 25 mm, pitch circle diameter (D) 36.0 mm, ball diameter (d) 6.35 mm, number of balls (Z) 10, contact angle (α) 0 It is ゜. In addition, a hole 9 having a diameter of about 3 mm is provided in the outer ring 8 of the ball bearing 3 for test to intentionally bring it into an abnormal state.

FIG. 3 shows a spectrum obtained by detecting the vibration of the ball bearing 3 under test with the acceleration sensor 7 when the shaft 1 is rotating at 640 RPM and performing spectrum analysis. In the figure, the frequency of vibration having the highest acceleration (G) (1 G = 980 cm / sec ² ) is 43.5 HZ, and its acceleration is 0.0795 (G). Table 1 shows the top 10 frequencies in the analyzed spectrum and the magnitude of their acceleration (G).

On the other hand, the theoretical value of the frequency of vibration when the outer ring 8, the inner ring 10 or the ball 11 of the ball bearing 3 under test has a flaw is given by the formulas (1), (2) and (33) given in the above-mentioned "Prior Art". ) Is required. Table 2 shows the fundamental frequency and the harmonics up to the 5th order.

Based on the three fundamental waves (44.0HZ, 62.9HZ, 58.1HZ) in Table 2, the fuzzy set for expressing the frequency distribution in Table 1 is shown in Fig. 4 as a membership function with three triangles. Is defined and is set in the memory in the abnormality detection device 17. This membership function is based on each fundamental frequency (44.0H
Z, 62.9HZ, 58.1HZ) with a grade 1 peak,
Three triangles connecting the apex and the point of the grade zero of another fundamental wave frequency that becomes the fundamental wave frequency of the apex, and the triangles at both ends are represented by isosceles triangles.

Table 3 shows the result of obtaining the grade value of the set of frequencies shown in Table 1 by the membership function shown in FIG.

Next, FIG. 5 shows a fuzzy set for each frequency component spectrum intensity (acceleration: G) in the spectrum of Table 1. The membership function in the figure is generally used in fuzzy diagnosis and is represented by equation (3).

This membership function has a spectral intensity (G) of 0.02
At that time, the grade becomes 0.5. This membership function is also set in the memory inside the abnormality detection device 17.

Table 4 shows the grades obtained by applying the value of the spectrum intensity (acceleration: G) at each frequency of the spectrum in Table 1 to the membership function in FIG.

Next, the product of each grade in Table 3 (inner ring grade, outer ring grade, ball grade) and the grade in Table 4 was obtained for each number, and the resulting grade is shown in Table 5.
The grades shown in Table 5 are grades that include both the frequency grade and the acceleration grade at each frequency.

In Table 5, the algebraic sums of the grades of the inner races numbered 1 to 10 are obtained by the algebraic sum calculation formula shown in the equation (4) to obtain the grade indicating the possibility of the inner race being damaged.

a ₁ + a ₂ + ... a _n = 1- (1-a ₁ ) (1-a ₂ ) ... (1-a _n ) ...
(4) In Table 5, the value is 0.000.

The above formula (4) shows an example of the calculation method, and the calculation formula is not limited to this.

Similarly, by obtaining the algebraic sum of the outer ring grade and the ball grade, the grades indicating the possibility of scratches on the outer ring and the possibility of scratches on the ball are obtained. The values in Table 5 are 0.919 and 0.0, respectively.
81. The term "likelihood" as used herein is substantially synonymous with "probability."

FIG. 6 is a flow chart of abnormality diagnosis expressed using the grades showing these possibilities. In this flowchart, the possibility of yes (displayed in Y) or no (displayed in N) is indicated by a numerical value indicating the grade in the vicinity thereof at each step indicating judgment. Fifth
From the table, the grade of the inner ring is zero, and the grades of the outer ring and the ball are 0.919 and 0.081, respectively. In the flowchart and the following description, the values of 0.919 and 0.081 are rounded off to the nearest 0.92 and 0.08, respectively.

In step (1) of the flowchart, the possibility that the inner ring is damaged is determined. The possibility of scratches on the inner ring is 0.00, so the grade of yes (hereinafter abbreviated as Y) is 0.0.
The grade of 0 and No (hereinafter abbreviated as N) is 1.00.

In step (2), the possibility that the outer ring is damaged is determined. The likelihood of scratches on the outer ring is 0.92, so the Y grade is 0.92 and the N grade is 0.08. Further, in step (3), the possibility of ball damage is judged. Since the possibility of ball damage is 0.08, the Y grade is 0.08 and the N grade is 0.92. As a result, the grade in which the inner ring, the outer ring and the ball are all scratched is 008 as shown in box (a) of FIG.

Similarly, grade 6 with both inner and outer rings is scratched
It becomes 0.92 as shown in the box (b) of the figure. Similarly, grades with both inner ring and ball scratches are 0.08 shown in box (c) in Fig. 6, grades with only inner ring scratches are 0.92 (box (d)), both outer ring and balls have scratches. Grade is 0.08 (box (e)), grade with only outer ring is 0.92 (box (f)), grade with only ball is 0.08 (box (G)) It becomes 0.92 (box (h)). “Others” also includes the case where no abnormality is found, for example.

Next, in each of the boxes (a) to (h), a set of grades showing the possibility of each of the inner ring scratch, the outer ring scratch, and the ball scratch was obtained based on the flowchart, and they are shown in Table 6.

For example, the grade set of box (a) is step (1)
Y grade, step (2) Y grade and step (3) Y grade. Similarly, the set of grades in box (b) is step (1).
Y grade, step (2) Y grade, step (3) N grade. The set of grades in box (c) is obtained by the Y grade in step (1), the N grade in step (2), and the Y grade in step (3). In this way, grade sets are obtained for the other boxes (a) to (h). Next, the minimum grade is selected from the set of grades by the minimum method. The results are shown in the bottom row of Table 6. From these results, the following four diagnostic results are obtained.

1. There is a 0.92 chance that the outer ring is scratched.

2. There is a probability of 0.08 that the outer ring and balls are damaged.

3. There is a 0.08 chance of scratching the ball.

4. There are other causes with a probability of 0.08.

FIG. 7 shows a flowchart of processing for performing the above processing. The envelope of the output signal detected by the vibration sensor 7 is detected by the envelope detector (step 11) (step 12), and unnecessary high frequency components are removed by the LPF (low-pass filter) (step 13). Next, it is sampled by the sample and hold circuit (step 14). The sampled signal is converted into a digital signal by the A / D converter (step 15) and input to the computer system (step 16).

In the computer system, the spectrum analysis, the calculation of the membership function, and the output of the diagnostic result are performed by a predetermined software (step 16).

[Effect of the Invention] According to the present invention, when an abnormality occurs in the rolling bearing, it is possible to determine whether the abnormality occurs in the outer ring, the inner ring or the rolling element which is a component of the rolling bearing. Is represented by a numerical value. Therefore, especially if there are abnormalities in multiple components (eg outer ring and rolling elements)
In the case of occurrence in the above, the degree of these abnormalities can be individually known by the probability.

[Brief description of drawings]

FIG. 1 is a side view of a vibration data collecting apparatus used in an embodiment of the present invention, FIG. 2 is a plan view of a ball bearing, FIG. 3 is a view showing a spectrum distribution of vibration of a ball bearing for trial use, and FIG. Is a diagram of the membership function for the spectral distribution of vibration of the co-test ball bearing,
FIG. 5 is a membership function relating to the spectral intensity of vibration of the ball bearing, FIG. 6 is a flow chart for abnormality diagnosis, and FIG. 7 is an abnormality detection device for carrying out this embodiment.

Claims (1)

[Claims] 1. A step of detecting vibration of a rolling bearing during rotation and outputting a detection signal, a step of A / D converting the detection signal, and a frequency spectrum of the detection signal based on the A / D converted detection signal. A step of performing analysis, a step of selecting a predetermined number of frequencies in order of increasing spectral intensity from a plurality of frequencies subjected to frequency spectrum analysis, with respect to the first membership function, the horizontal axis represents frequency and the vertical axis represents 0 At the coordinates representing the grade 1 value, the lateral vibration of the theoretical fundamental frequency for each of the three cases where the outer ring of the rolling bearing has a scratch, the inner ring has a scratch, or the rolling element has a scratch Setting in memory a first membership function having three triangles each of which has a vertex whose grade value is 1 in the coordinate values; Determining a grade of each of the predetermined number of frequencies subjected to spectrum analysis of the frequency based on the membership function, with respect to the second membership function, a horizontal axis represents spectrum intensity and a vertical axis represents grade. In the memory, setting a second membership function represented by a curve that gradually increases toward the grade 1 through the origin of the coordinates, based on the second membership function, Determining a grade for each spectrum intensity, each grade of the predetermined number of frequencies obtained based on the first membership function, and the predetermined number obtained based on the second membership function The product of the frequency and the spectral intensity for each spectral intensity is calculated by By determining the frequency and inner ring grade, outer ring grade, rolling element grade, and algebraic sum of the product values for each inner ring, outer ring, and rolling element, the inner ring, outer ring, and rolling element are calculated. A rolling bearing failure detection method including a step of determining a possibility that an abnormality has occurred in a moving body.