WO2007101432A2

WO2007101432A2 - Method for the analysis of bearing damage

Info

Publication number: WO2007101432A2
Application number: PCT/DE2007/000416
Authority: WO
Inventors: Rupert Stitzinger
Original assignee: Schaeffler Kg
Priority date: 2006-03-09
Filing date: 2007-03-08
Publication date: 2007-09-13
Also published as: DE102006010847A1; WO2007101432A3

Abstract

The invention relates to a method for analyzing bearing damage with the aid of an envelope demodulation (HKD) signal. A threshold value (SW) that is characteristic of a bearing damage is determined in one step of said method while at least one first calculated value (SF) is formed in a second step by taking into account the characteristic threshold value (SW), said calculated value making it possible to draw conclusions about a damaged bearing.

Description

Schaeffler KG Industriestraße 1 - 3, 91074 Herzogenaurach

Name of the invention

Method for investigating bearing damage

description

Field of the invention

The present invention relates to a method for examining bearing damage and in particular bearing damage in rolling bearings or linear guides.

The inventive method is applicable to a variety of rolling bearing types such as ball bearings, cylindrical roller bearings, tapered roller bearings or spherical roller bearings. The rolling bearings to be examined can be used, for example, in electric motors, railway wheel sets, gearboxes, paper machine test stands and the like.

Rolling bearings can cause various damage as a result of wear. An example of such damage are so-called "pittings", ie material breakages of the inner or outer ring or the rolling elements, as well as surface damage, since such pittings or damage as a result of further use quickly lead to rapid deterioration of the rolling bearing There is a need to investigate such damage using appropriate measurement techniques The prior art discloses various methods for investigating conditions of a rolling bearing. Such a diagnostic method is, for example, the so-called envelope analysis with subsequent Fourier transformation. In the process, a so-called envelope demodulation signal (HKD signal) is evaluated to evaluate the bearing condition. The recording of such signals is possible, for example, with piezoelectric sensor devices which can be screwed, glued or held by magnetic holders to the bearing housing.

With the help of the envelope analysis and subsequent Fourier transformer, for example, periodic repeats of bumps, as generated by pittings in rolling bearings, can be detected and monitored, thus inferring damage. However, with the method known from the prior art, no recognition of a wide variety of damage factors is possible. Such envelope amplitude modulation signals have hitherto been evaluated by an observer, especially at speeds that are not exactly known, and this can be concluded with the aid of certain experience of the existence of specific damage.

The present invention is therefore an object of the invention to provide a method that allows the detection of several different damage and their size in bearings.

This is achieved according to the invention by a method according to claim 1. Advantageous developments are the subject of the dependent claims.

In the method according to the invention for examining bearing damage with the aid of an envelope demodulation signal, a characteristic threshold value is determined from the envelope amplitude modulation signal in one method step, this threshold value depending on the existence of bearing damage and in a further method step at least one first calculation value taking into account the characteristic value. formed terist threshold value, this first calculation value allows conclusions on existing bearing damage. The threshold is a value dependent on the type of damage. According to the nature and extent of exceeding or falling short of the value, the damage can be deduced.

Preferably, a second calculation value is formed taking into account the characteristic threshold value, and particularly preferably a third calculation value independent of the threshold value. Based on these three values, a more detailed classification into different types of damage and sizes can be made.

Preferably, the first calculation value is a value determined from the ratio of an effective value above the threshold value to a total effective value. Instead of RMS values, mean values could also be evaluated. This first calculation value is also referred to below as the threshold factor. The second calculation value, which is also referred to below as the sum pulse width, is a percentage total pulse width based on a predetermined measurement interval, for example a measurement interval of 5 seconds. In this case, the pulses considered are formed from the signal components above the threshold or the threshold value and their widths are measured and added up.

The third calculation value, also referred to below as CF, is a ratio of a moving peak value and an overall effective value, whereby the DC component is not taken into account here. However, instead of an overall rms value, CF may be formed taking into account a total mean value.

Preferably, a result value is determined from a mathematical relationship between the calculation values, which for the presence Of bearing damage and the type of bearing damage is characteristic. This result value is also referred to below as the damage factor. The mathematical relationship is preferably the weighted product of the abovementioned calculation values. Extensive research has shown that the result value allows a very good classification into different damage categories. Since significant calculation values increase with higher damage, their product also increases.

In a further preferred representation, it is also possible for the three calculation values to be plotted in a three-dimensional coordinate system. Within this diagram, a clear separation between intact rolling bearings, rolling bearings with inner ring damage, rolling bearings with massive outer ring damage and roller bearings with surface damage is possible. Also, such a plot in a three-dimensional coordinate system, a separation between light and area damage possible. To evaluate the individual calculation values, it is advantageous to use a program in a tree structure. In particular, it is advantageous to use fuzzy logic for the evaluation, since the transitions between the individual types of damage are blurred in their effects on the calculation values and, on the other hand, different types of damage can occur simultaneously.

Instead of the above-mentioned product as a mathematical relationship, however, the individual calculation values within a mathematical relationship can be weighted differently.

The threshold value is preferably determined with the aid of an RMS value of the HKD signal, this RMS value being determined from that portion of the HKD signal which is greater than an overall RMS value of the HKD signal. This means that first a total RMS value or overall mean value of the HKD signal within a predetermined Time window is determined. In this case, the DC component of the HKD signal is also taken into account. For calculating or determining the effective value, only those parts of the HKD signal which are greater than the above-mentioned total RMS value are considered. From these shares, the effective value is calculated. The calculation is preferably carried out by time averaging.

The threshold value is preferred according to the mathematical relationship

SW = (RMSp _0S -αx corr) x V (1)

where α is a weighting factor, V is a gain factor, and Korr is a correction factor. More specifically, V is a variable gain factor and may be 18, for example, for a particular measurement made. Since the amount within the bracket of Equation (1) shown above can mathematically also assume values smaller than 0, but this is not technically meaningful, the minimum value for the threshold value is set to 0.

Preferably, the correction factor is determined from its sum of RMS values that lie within predetermined amplitude bands of the HKD signal, these amplitude bands being determined as a function of the RMS value and the overall RMS value. More specifically, two positive voltages B ⁺ and B ₊ and two equal but negative voltages B ^' and B. are generated from the total RMS value and the RMS value. To generate this voltage are the relationships

B ⁺ = ax (RMSDC + RMSpos), B = B ⁺ (2) and,

used. The values for a and b were determined empirically and values of a between 1 and 5 and between 1 and 4 were found to be particularly favorable within the scope of the measurements carried out. The value RMS _D c denotes the total RMS value: and the value RMS _p0S indicates the RMS value. Subsequently, those correction effective values Korr are found which lie within the amplitude bands which are formed on the one hand by B ⁺ and B ₊ and on the other hand by B. and B ^' , that is to say the effective values of the voltage which lie within this voltage range.

Based on the equations (1) - (3), the threshold is determined.

The determined threshold depends, as mentioned above, on the type of damage of the rolling bearing.

The present invention is further directed to a computer program for carrying out a method of the type described above. Preferably, this computer program causes an HKD signal, which can also be formed by this computer program, to be read in a predefined time window and, based on this signal, first of all to calculate a threshold value and then, using the threshold value, to calculate the individual calculation values shown above.

Further advantages and variants of these methods of the invention will become apparent from the accompanying drawings. Hereby show:

FIG. 1 shows an HKD time signal for an intact bearing; FIG.

2 shows an HKD time signal for a bearing with surface damage;

3 is a HKD timing chart for a bearing with outer ring fittings; and

4 shows an evaluation of the different bearings with the aid of a three-dimensional coordinate system.

FIG. 1 shows an example of an HKD time signal for an intact rolling bearing. In this case, the horizontal line 3 denotes the above threshold. The horizontal line 5 refers to the positive voltage B ⁺ , the reference numeral 7 to the corresponding negative voltage B ^" (which coincides in magnitude with the voltage B ⁺ ), the reference numeral 4 to the second positive voltage B ₊ and the reference numeral 6 to the second negative voltage B- which coincides in magnitude with B ₊ , reference numeral 2 refers to the HKD time signal taken in the temporal window shown in Fig. 1, that is, time t is plotted on the X coordinate and on the Y-coordinate voltage values are plotted, which here range from -500 to +500 mV, the reference numeral 8 designating the zero axis.

As mentioned above, it is possible for the skilled person to infer at least roughly from the time signal in the manner of any damage that may be present. The present invention thus provides a method to this inference by machine d. H. without allowing the user to monitor the signal. The threshold value 3, which, as mentioned above, is needed to determine the two individual quantities SF and IS, must be set automatically, ie without knowledge of the respective HKD signal characteristic. Also should not be necessary intervention by the user.

At this threshold 3, two main requirements have to be met. On the one hand, the threshold for intact bearings must be so high that at most the (rare) highest signal peaks exceed this threshold. In the case of the illustration shown in FIG. 1, this is the case only for the voltage peak 9. In this way, small calculation values for the Threshold factor SF and the sum pulse width IS achieved. Conversely, in the case of bearings with area damage, the threshold should be as low as possible to reach maximum values of SF and IS.

These contradictory requirements can be solved by generating a correction voltage that is small in intact bearings and large in heavily damaged bearings.

FIG. 2 shows an HKD time signal for a bearing with a planar damage. In this case, the threshold value 3 is much lower than in the representation shown in FIG. It can be seen that in this case, a large number of voltage peaks exceeds the threshold value.

To calculate the threshold value, the effective value RMSDC of the HKD signal is calculated in a first method step, wherein the DC component HKDDC is also taken into account. Subsequently, that part of the HKDDC signal is considered which is higher than the effective value RMSDC and from which in turn its effective value RMS _{pos is} calculated.

From these two effective values RMSDC and RMS _pos , two positive voltages B ⁺ and B ₊ and two equal but negative voltages B ^' and B are calculated according to the above equations (2) and (3).

In a last method step, the sum of the RMS values Korr of those parts of the HKD signal without the DC component (HKD _ac ) are calculated, which lie within the amplitude bands which are formed by B ⁺ , B ₊ on the one hand and B-, B ^" on the other hand In FIGS. 1-3 these are the areas denoted by 11 and 12.

Finally, the threshold is determined according to the relationship (1). Furthermore, as mentioned above, a clamp diode is used to prevent the threshold voltage from taking values less than 0 volts. It should be noted that the described procedure is a primarily empirical procedure, that is to say that in the course of the development, the abovementioned values were determined in a large number of experimental steps in order to split the representations into the greatest possible degree to achieve different damage classes. However, other quantities derived from the HKD signal, such as the HKD DC component, the RMS negative of the HKD ₃₀ signal, or the crest factor, could be used to generate the threshold. Finally, combinations of these values could be used. Thus, it will be apparent to those skilled in the art that various methods are suitable for determining the above threshold. However, all of these methods are equal in that each threshold is calculated from characteristics of the HKD time signal.

In the present approach, the symmetry of the HKD ₃₀ signal with respect to the zero axis 4 is utilized. This symmetry depends on the condition of the bearing. While the HKD ₃₀ signal of an intact bearing is almost zero-symmetric, this signal of a damaged bearing lies predominantly in the positive amplitude range. In the case of an intact bearing, the rms values of the signal components in the positive and negative range almost cancel each other out so that only a small correction margin is generated. For damaged bearings, the effective value of the positive wall portion of the negative wall portion far exceeds. This allows a high correction voltage to be generated, which can significantly reduce and sometimes even exceed the above-mentioned RMS _pos signal component. This could even lead to a negative threshold voltage. However, as mentioned above, this is prevented by clamping the threshold voltage to zero.

FIG. 3 shows an HKD time signal for a roller bearing, which has outer ring fittings. It can be seen that here, at periodic intervals, the threshold value 3, which is just above the noise, is exceeded. The regions 9a marked in gray are those regions of the HKD time signal which exceed the voltage value B ₊ . As mentioned above, these values, or the RMS value from these values, are used to calculate the value Korr, which in turn is used to calculate the threshold.

The time signal shown in FIG. 3 thus leads to a comparatively high value for Korr and thus to a very low threshold value. Due to the strong asymmetry of the time signal, in particular in the time signal shown in FIG. 3, the effective value RMS _{pos is also} very high.

FIG. 4 shows a three-dimensional coordinate system on whose axes the values for the threshold factor SF, the CF and the sum pulse width IS have been plotted. One recognizes a total of four groups of rolling bearings, each represented by different hatchings. The reference I designates intact bearings. It can be seen that here the damage factor SF, which results from a product of the values IS, CF and SF, lies between 0 and 3. In the group of rolling bearings marked Il, it can be seen that their value is significantly higher for both SF, CF, and IS than for the group marked I. This group identifies the rolling bearings with slight damage. The damage factor SF is between 4 and 10.

The reference numeral III denotes bearings with flat damage. The damage factor here lies between 10 and 30. Reference numeral IV denotes roller bearings with heavy inner ring damage, the damage factor here being between 1 and 8. The reference symbol V denotes bearings with heavy outer ring damage, whereby the damage factor lies between 40 and up to 200 here. Thus, by the application shown in FIG. 4 in a 3D diagram with three spatial axes, a clear separation between intact rolling bearings, rolling bearings with inner ring damage, rolling bearings with massive outer ring damage, rolling bearings with surface damage is possible. A separation between rolling bearings with light and rolling bearings with surface damage is possible. The above-mentioned damage factor is at least suitable as a crude means of classification. Thus intact rolling bearings show damage factors between 0 and 2, rolling bearings with slight damage damage factors between 40 and 10, rolling bearings with massive inner ring damage or inner ring cracks damage factors between 1 and 8, rolling bearings with massive outer ring fittings damage factors between 40 and 200 and rolling bearings with area damage damage factors between 10 and 30.

Nevertheless, a differentiation of the individual damage using the individual sizes makes sense. This can be achieved particularly advantageously by a program with a tree structure, whereby, for example, so-called fuzzy logic is suitable since the transitions between the individual types of damage are fluid and rolling bearings with a plurality of different types of damage can occur.

A significant advantage of the method described here is that it is essentially independent of rotational speed and can be used both at very low rotational speeds and at high rotational speeds. Advantageously, in order to improve the selectivity between very light and area heavy damage, a separation monitoring can be performed, or it is the absolute RMS levels used to better distinguish these types of damage can. Furthermore, the method is independent of the exact transfer function of the machine because it does not use absolute values. This is especially important if you are not constantly monitoring because you do not have to look at changing absolute sizes.

All disclosed in the application documents features are claimed as essential to the invention, provided they are new individually or in combination over the prior art.

LIST OF REFERENCE NUMBERS

2 HKD - time signal

3 threshold

4 second positive voltage B ₊

5 voltage B ⁺

6 second negative voltage B.

7 negative voltage B ^"

8 zero axis

9 voltage peak

9a Range of the HKD signal

11, 12 area within the amplitude bands

SF threshold factor

IS sum pulse width

SW threshold

I - V rolling bearing groups

Claims

Schaeffler KG Industriestraße 1-3, 91074 HerzogenaurachPatent claims

1. A method for examining bearing damage by means of an envelope detection (HKD) - signal, wherein in one step of the HKD signal for a characteristic threshold (SW) is determined, this threshold (SW) depends on the presence of bearing damage and in a further method step, at least one first calculation value (SF) is formed taking into account the characteristic threshold value (SW), whereby this calculation value makes it possible to draw conclusions about existing bearing damage.

2. The method according to claim 1, characterized in that a second calculation value (IS) is formed taking into account the characteristic threshold value (SW).

3. The method according to at least one of the preceding claims, characterized in that a third of the threshold value (SW) independent calculation value (CF) is formed.

4. The method according to claim 3, wherein a result value is determined from a mathematical relationship between the calculation values (SF, IS, CF), which is used for the

Presence and the type of bearing damage is characteristic.

5. The method according to claim 4, characterized in that the mathematical relationship is the product of the calculation values (SF, IS, CF).

6. The method according to at least one of the preceding claims, characterized in that the first calculation value (SF) is a threshold factor, which from the

Ratio of the RMS value above the threshold and the total RMS value.

7. The method according to at least one of the preceding claims, characterized in that the second calculation value (IS) is a sum pulse component of pulse components above the threshold value (SW).

8. The method according to at least one of the preceding claims, characterized in that the third calculation value (CF) is formed from the ratio between a sliding peak value and a total RMS value.

9. Method according to claim 1, characterized in that the threshold value (SW) is determined with the aid of an effective value (RMS _pos ) of the HKD signal, wherein the effective value (RMS _pos ) is determined from that portion of the HKD signal which is greater than a total rms value (RMSDC) of the HKD signal.

10. The method of claim 6, wherein the threshold value (SW) is determined according to the mathematical relationship

where α is a weighting factor, V is a gain factor, and Korr is a correction factor.

11.A method according to at least one of the preceding claims. 9

-10, characterized in that the minimum value for the threshold is 0.

12. The method according to claim 11, characterized in that the correction factor is determined from a sum of RMS values which lie within predetermined amplitude bands of the HKD signal, these amplitude bands being determined as a function of the RMSp ₀ S and RMSDC rms values.

13. Computer program for carrying out a method according to at least one of the preceding claims.