The present application is a 35 U.S.C. §§ 371 national phase conversion of PCT/DE2004/000095, filed 23 Jan. 2004, which claims priority of German Application No. 103 03 877.9, filed 31 Jan. 2003. The PCT International Application was published in the German language.
The invention relates to a method for detection of structure-borne noise events in a roller bearing.
DE 199 43 689 A1 discloses a method and an apparatus for monitoring and/or diagnosis of moving machines and/or machine parts. The aim of such measurement and evaluation methods is to identify damage to a machine or to a machine part such as a roller bearing as early as possible in order to make it possible to replace this item, when appropriate, before its total failure.
In this known method, oscillations and structure-borne sound waves which are produced by the machines or machine parts are recorded by means of an acceleration sensor as a broadband time signal, and are passed to an evaluation device. The measurement signal is digitized in this evaluation device, and a relatively large range of such digital values are stored. This data is subjected to frequency analysis by means of a fast Fourier transformation in a subsequent calculation process. The process of carrying out a fast Fourier transformation such as this is highly complex and requires a comparatively large amount of computation capacity and memory space.
Furthermore, this document discloses the broadband time signal being subjected to discrete filtering using different analysis methods.
During this process, frequency components of the signal components caused by the damage are filtered out from the measurement signal, and an amplitude profile of this component is evaluated. Changes in the frequencies can extend the evaluation to adjacent signal components in such a way that the local amplitude maxima can be determined. The computation complexity for carrying out discrete filtering is admittedly less than that of such a conventional Fourier transformation, but it is nevertheless comparatively high since the cut-off frequencies of the filter must be dynamically matched to different rotation speeds of the shaft/bearing system.
Finally, DE 101 36 438 A1 discloses a sensor arrangement in which strain gauges are arranged in a circumferential groove in a stationary roller bearing ring, where they are connected to signal-electronic modules, by means of which a signal evaluation process can be carried out directly in the circumferential groove in the roller bearing ring. Since this circumferential groove is comparatively small, the signal-electronic modules which are arranged there may have only a quite modest storage and computation capacity, so that the evaluation methods described above are not carried out in situ and, so to speak “on-line”.
Against this background, the object of the invention is to provide a method for detecting structure-borne sound waves and structure-borne sound or noise events for a measurement apparatus which, by way of example, is arranged in a roller bearing. By this method, the required analysis steps can be carried out directly at the point at which the measurement signals are obtained, without a large amount of computation capacity and without requiring a large memory space. A further aim is for the measurement signal from suitable sensors on the roller bearing to be converted such that it is possible to make an unambiguous statement about the actual presence of bearing damage in a better manner than in the past. A final aim is to be able to carry out this method sufficiently quickly that suddenly occurring bearing damage can be detected preferably “online”, that is to say without any time delay.
These objects are achieved by the features in the main claim, while advantageous refinements and developments of the invention can be found in the dependent claims.
Accordingly, the method provides that, in order to detect structure-borne sound or noise events in a roller bearing, the measurement signal from a pressure-sensitive or strain-sensitive sensor which is arranged on a roller bearing is first of all supplied to a frequency filter in order to filter out undesired signal components. A first variance value is then calculated from the digital values of the frequency filter output signal, and at least one second variance value is determined from subsequent digital values of the frequency filter output signal. The weighted arithmetic mean of these variance values is then calculated from these at least two variance values (new variance, old variance) with the aid of a recursive calculation. If the weighted arithmetic mean value is above a preselected variance threshold value, this is assessed as a structure-borne sound event, which has been caused by mechanical damage to the roller bearing.
In this context, the expression “recursive calculation” means that result values from a first variance calculation are included in the calculation of a second variance value. This allows consistent estimation of the structure-borne sound amplitude on the basis of only two sample values without any need to provide additional storage capacity in an evaluation device.
The weighted arithmetic variance mean value is determined with the aid of the following basic recursive equation:
Ē{X2}(k+1)=Ē{X2}(k)+Cx2[X2(k+1)−Ē{X2}(k)] (Equation 1)
where Ē{.} represents the expected value of the weighted arithmetic mean value, k is a running variable, x is a digital sample value of the frequency filter output signal, and Cx is an adaptation constant.
With regard to the adaptation constant C, a value is provided which is less than unity and is greater than zero, and which can be calculated from the equation for the so-called adaptation rate:
t={1/Cx2−½}*T (Equation 2).
In this expression, t indicates how quickly the true variance can be determined with a tolerable error rate.
The basic recursive equation (equation 1) can be used to form a simplified recursive equation
nE=aE+aK*(nA−aE) (Equation 3)
in which nE represents a new result value, equal to Ē{x2}(k+1), aE represents an old result value, equal to Ē{x2}(k), aK represents the adaptation constant, equal to Cx, and nA represents a new sample value, equal to x2(k+1). The “new result value” is in this case the value to be calculated at any given time, while the “old result value” denotes the value calculated one sample clock period early.
This simplified recursive equation can be expressed for the evaluation process as an evaluation equation:
nV=aV+aK*(nA−aV) (Equation 4)
where nV represents the new variance, equal to nE, aV represents the old variance, equal to aE, aK represents the adaptation constant, and nA represents the new sample value, equal to nA.
A new variance calculated in this way is assessed as an estimated value for the true variance of the measurement signal from which the presence of structure-borne sound events in the roller bearing is deduced. In this case, a high new variance value represents major damage in the roller bearing.
As is evident from the above statements, the method according to the invention can be used to detect the presence of structure-borne sound waves and structure-borne sound events from a measurement apparatus that is arranged in a roller bearing, by means of an evaluation device which does not have a large memory space and has quite a small computation capacity. Evaluation devices such as these may be microcomputers which are arranged directly at the point at which the measurement signal is obtained, for example in the circumferential groove in a roller bearing, and which carry out the necessary analysis steps in order to detect bearing damage.
Furthermore, the method according to the invention allows the measurement signal from a suitable sensor, such as a strain gauge, on the roller bearing, to be converted such that an unambiguous statement about the actual presence of bearing damage is possible in a better manner than with previously known methods. Finally, the method makes it possible to identify structure-borne sound waves using only a small number of measurement values sufficiently quickly that suddenly occurring bearing damage can be detected preferably “on-line”, that is to say without any time delay.
The method according to the invention, its interaction with a measurement roller bearing and an evaluation device can best be explained on the basis of exemplary embodiments of the invention, which are illustrated in a drawing, in which:
Measurement resistors R1, R2, R3, R4 are thus attached to the circumferential surface of the outer ring 1 in a measurement area 5. These resistors change their electrical resistance as a function of strain and by means of which the deformation of the outer ring 1 can thus be detected as the roller bodies 3 roll over each measurement resistor R1, R2, R3, R4. One important factor in this context, however, is that, in terms of the distances between them, the resistors R1, R2, R3, R4 must not be separately aligned.
As can be seen from
Unfortunately, in the case of measurement signals such as these, the signal level of the sinusoidal basic signal M is very much higher than that of the higher-frequency signal component. The measurement signal M thus has to be evaluated separately in order to separate the higher-frequency signal from the basic signal component and to make it possible to unambiguously distinguish it from a noise signal which is always present and which has an amplitude that is often of similar magnitude. For this purpose, the measurement signal M from the measurement apparatus 8 is first passed to a frequency filter, which removes the undesirable frequency components from the measurement signal M. A frequency filter such as this can be designed such that it filters the analog or else preferably the already digitized measurement signal using appropriately preselected prior settings, and emits digital signals.
Since this noise separation is very small, the detection sensitivity for bearing damage that is to be detected is initially also comparatively poor. The evaluation method according to the invention now overcomes this in that, inter alia, it creates a very much greater separation between the structure-borne sound signal and the noise signal. Furthermore, an unambiguous statement about the presence of bearing damage can be made with only a very small number of digital measurement values.
For this purpose, the invention provides that a first variance value is first calculated from digital values of the output signal from the frequency filter (the signal profile shown in
Since a large weighted arithmetic variance mean value means that the investigated measurement signal component differs to a major extent from the sinusoidal basic signal or the noise signal, its exceeding of a predetermined variance threshold value VSW can be assessed as a structure-borne sound event which has been caused by mechanical damage to the roller bearing.
Number | Date | Country | Kind |
---|---|---|---|
103 03 877 | Jan 2003 | DE | national |
Filing Document | Filing Date | Country | Kind | 371c Date |
---|---|---|---|---|
PCT/DE2004/000095 | 1/23/2004 | WO | 00 | 8/29/2005 |
Publishing Document | Publishing Date | Country | Kind |
---|---|---|---|
WO2004/068100 | 8/12/2004 | WO | A |
Number | Name | Date | Kind |
---|---|---|---|
5365787 | Hernandez et al. | Nov 1994 | A |
5602761 | Spoerre et al. | Feb 1997 | A |
5852793 | Board et al. | Dec 1998 | A |
6208944 | Franke et al. | Mar 2001 | B1 |
6567752 | Cusumano et al. | May 2003 | B2 |
6883373 | Dyer | Apr 2005 | B2 |
6889553 | Robinson et al. | May 2005 | B2 |
6999884 | Astley et al. | Feb 2006 | B2 |
7039557 | Mayer et al. | May 2006 | B2 |
7136794 | Bechhoefer | Nov 2006 | B1 |
20020013664 | Strackeljan et al. | Jan 2002 | A1 |
20060110086 | Morita et al. | May 2006 | A1 |
Number | Date | Country |
---|---|---|
199 02 326 | Aug 2000 | DE |
19943689 | Mar 2001 | DE |
10136438 | Mar 2002 | DE |
WO03023721 | Mar 2003 | DE |
0 889 316 | Jan 1999 | EP |
1 514 792 | Jun 1978 | GB |
363304127 | Dec 1988 | JP |
Number | Date | Country | |
---|---|---|---|
20060150737 A1 | Jul 2006 | US |