DE102009024981A1 - Damages determining and analyzing method for e.g. rolling bearing, involves dividing data signals into number of different timing signals, and evaluating amplitudes of bispectrum, where amplitudes exceed predetermined value - Google Patents

Abstract

The method involves recording data signals with a sampling frequency, where the data signals are emitted during unwinding of a machine element. The data signals are divided into a number of different timing signals. The timing signals of the data signals are transferred into a frequency domain by Fourier-transformation. A bispectrum is calculated, and amplitudes (7, 8) of the bispectrum are evaluated, where the amplitudes exceed a predetermined value. The data signals are selected from one of acceleration signals, speed signals or path signals. An independent claim is also included for a device for determining and analyzing damages at a machine element.

Description

The The invention relates to a method for the determination and analysis of Damage to rotating machine elements, in particular for monitoring and diagnosing circulating rolling and sliding bearings.

Around Malfunctions and unexpected costs due to premature and unscheduled failing machine elements to limit to a minimum, on the one hand, attempts by constructive Measures and preventive maintenance damage on moving components, in particular on bearings, in principle to avoid. On the other hand, a damage analysis allows the prevention of far-reaching damages.

The Knowledge about type, location and size of damage allows in connection with the function of the warehouse or Machine element a risk assessment. Especially when incipient signs of wear can during monitoring of the warehouse, the remaining time remaining to be estimated. The accuracy of a bearing is preferably already monitored during the manufacturing process.

numerous The diagnostic systems known from the prior art use these the components occurring vibrations to assess the operating condition various machine elements.

The Intensity and thus the amplitude height of a For example, caused by a rolling bearing vibration is in particular of load and speed as well as the surrounding Construction dependent. In addition, the types to distinguish the vibrations.

For the description of the vibrations assumed to be harmonious are basically three mechanical sizes to select. These are vibration travel, speed and acceleration.

at the rolling bearing diagnosis of the built-in roller bearing In operation, different ways of vibration analysis can be achieved recorded functions are evaluated. For example, these are Time function, amplitude spectrum, envelope, cepstrum and Autocorrelation function. The Fourier transformation takes place today For example, their use in the signal evaluation for detection advanced errors, whereas the envelope detection in the early detection of initial damage is used.

Previous Methods do not have sufficiently high detection reliability in case of errors in machine elements, if noise is present in the signal are. Then often the error signals are in the range of Noise, making detection difficult and initial regularly very small error signals almost is excluded.

The currently the main method of surveillance used of machine vibration and rolling bearing condition is this Hüllkurvendetektionsverfahren. This is a detection from Z. B. Rolling damage, gear damage (Toothing) possible. Typical, but not exclusive Fields of application are the monitoring of pumps, electrical Motors, fans, machine tools, compressors, gearboxes and spindles. In the process, vibration signals are generated at previously determined measuring points recorded with a sensor and from it z. B. the RMS values calculated by vibration velocity and vibration acceleration. From these characteristics can be machine and component state derived.

The so-called envelope detection or envelope demodulation is a signal processing with downstream spectral analysis. It uses the amplitude modulation of the carrier signal from, the z. B. in rolling bearings by shock excitations when rolling over bearing damage arises. By the signal processing becomes unimportant for the damage analysis Areas filtered and the modulation frequencies, the inferences on the bearing damage, from the carrier frequency separated and moved. In the edited spectrum can be the amplitudes of the individual rollover frequencies directly read and evaluate.

From the DE 10 2005 015 465 A1 a method and a measuring and analyzing device for categorizing rolling bearing damage is known. In this case, an analysis function is derived from the recorded signal of the so-called "jerk", which is evaluated statistically. The analysis function is the second time derivative of the velocity-proportional structure-borne sound signal or a higher derivative level in order to base the damage categorization on it. This corresponds in principle to the examination of properties in the extreme value range of the recorded signal.

Other algorithms for signal evaluation in various industries are known. More recently, high order spectra (HOS, for example) have been used for signal analysis. The practical use of such algorithms is relatively new, as they require extremely high computing power, which has only been achieved in recent years. The algorithms were first used in seismological investigations. Meanwhile, they come in various research areas of oceanography, Astro nomie, medical technology, image processing, laser technology, data transmission and armament, here among others in radar and sonar technology, for use.

The JP 58 193 425 A deals with the evaluation of vibrations on rotating parts, such as shafts or bearings. The detected vibrations are evaluated in a power spectrum analyzer and, if certain values are exceeded, in a bispectrum analyzer. The quantities of the bispectrum are input to a discriminator and displayed on a monitor. The bispectrum analysis is carried out at "suspicious" exceeding certain specifications.

The DE 199 33 924 A1 describes a microsystem for condition monitoring and diagnosis of machines, plants and assemblies. The focus of the monitoring lies in the area of rolling bearings, whereby a determination of the remaining operating time should be possible. Acceleration signals are recorded and evaluated via different signal paths and searched for characteristic values in the envelope spectrum, for characteristic frequencies in the spectrum, possibly for volatile components in the frequency domain and for features in the time signal. These features undergo a trend analysis, which leads to a classification of possible errors. Due to the necessary evaluation of different signal paths, the evaluation is extensive and complicated.

In the DE 694 23 006 T2 describes a system for damage analysis of shafts and turbine blades. From the recorded vibration signal of the rotating parts of a spectrum is created by discrete Fourier transform on. Resonance curves of turbine blade sets are monitored. The shapes of the resonance curves are analyzed according to statistical features such as skewness, variance and others. The resonance curves are also monitored for the formation of secondary peaks as shoulders of the main peaks. Changes in these characteristics may be indications of incipient damage to turbine parts. Noise of the signal is minimized by different filters.

From the JP 2006 153 761 A is a system for monitoring rotating parts, such as bearings known. An oscillation signal is recorded and from this a simplified higher-order cumulant is calculated. The result is compared with a simplified absolute indexed higher order cumulant, which is calculated from a noiseless signal.

Of the Invention is based on the object, a method and a device to create the rolling bearing diagnosis, the justifiable Computing effort a reliable early error detection on rotating machine elements such as sliding and rolling bearings allows, even if the signal to be analyzed a high noise component or the signal of smaller signal energy to be examined is, than the otherwise measurable signal components.

The The object is achieved by a method as well as by a device with the features of the respective independent Claim solved.

in the The scope of the method according to the invention is first by means of a sensor, preferably an accelerometer, recorded a data signal, which oscillates on a rotating Machine element, for example, on a built-in a plant Rolling bearing rep presented. Of course you can the signal acquisition also for quality control be used in the manufacturing process of the rolling bearing. In addition, the data signal can be sent to a route signal or be a speed signal.

The recording is done discretely with a sampling frequency f _x .

In In a preferred embodiment, the sampling frequency becomes selected from a range between 10 kHz and 500 kHz, often at 50 kHz.

The Data signal contains in addition to the important signal components many annoying signal components that are also from others Plant parts, as the rolling bearing can come from. This does not come from an error or a corruption of the Bearing originating signal components are general below referred to as noise.

from recorded data signal is made more accurate by means of a later mathematical algorithm calculated a bispectrum and graphically, for example. The bispectrum is one so called higher order spectrum and provides couplings between frequency components of a signal. On the basis of the bispectrum symmetrically distributed amplitudes and their position and height can draw conclusions about the Condition of a warehouse to be monitored.

Advantages are to be seen in particular in the fact that the method according to the invention enables a reliable error detection even with a comparatively high noise component of the signal (in relation to the useful signal). This helps to avoid consequential damage in the system, to minimize downtimes and to ensure a safe functional sequence in the system to ensure complex installations.

The Detection reliability of errors is many times higher and more reliable than the usual in the prior art Method.

A special importance comes to the invention Procedures in the monitoring of complex systems such as for example in the printing industry or in power plants. So it is conceivable, for example, with several inventive Measuring systems in a large system to perform a fault location. For example, this could include three measuring systems on the system be arranged to error location by evaluation of the Perform phase information and the signal transit times.

The basis for calculating the bispectrum B (f ₁ , f ₂ ) are the Fourier transforms of time signals x _i (t) recorded at different epochs.

The data signal is thus first divided into N epochs, so that N time signals x _i (t) are present as vector or matrix.

The Number N of epochs is preferably between 50 and 200, is but depending on the individual application.

From the time signals, the spectra X _i (f ₁ ) and X _i (f ₂ ) are calculated by Fourier transformation. The bispectrum B represents a spectral correlation and is calculated as follows:

N is the number of epochs into which the data signal is divided has been.

As a result of the calculation of the bispectrum, values are output that are arranged, for example, in a three-dimensional value space. The abscissas of this value range form factors f1 and f2. These are multiplied by the sampling frequency f _x and thus indicate the frequencies of the divided signals. The ordinate value indicates the respective degree of phase and frequency coupling between two frequencies.

The Representation of the bispectrum can be as 3D or contour plot, with Representation of the contour lines, done. A vertical cut of the 3D plot at a certain frequency allows one closer examination of the amplitude levels.

To recognize in the bispectra several characteristic positive and negative amplitudes representing the searched errors. The bispectrum indicates due to the applied calculation method several axes of symmetry. Therefore, it is for troubleshooting sufficient to consider only one quadrant of the contour plot, whereby the evaluation effort can be reduced. Of the considered quadrant can be one of the possible result quadrant be.

The Amplitudes in the bispectrum would ideally have to be in the distance the modulation frequency (ie the disturbance) of the carrier frequency are located. However, in systems to be monitored not only the bearing, but also the entire structure in the time signal detected. It is therefore in the evaluation of bispectrum of a Base frequency from the complete structure of the monitored Plant is dependent.

at the further investigation of the bispectrum is assumed that the base frequency and different natural frequencies of the monitored Systems are known. On this basis, in the amplitudes of the Bispektrums are searched for characteristic deviations, which indicate storage damage.

The Assignment of characteristic damage to the bearing is now by analyzing the positions of the amplitudes in the contour plot of the bispectrum possible. The amplitude distribution in the typical Damage of a bearing is preferably determined experimentally. There are different types of damage with different types of bearings prevail. Depending on the type of storage other frequencies of the Components excited. Within a storage type, the distributions remain the typical amplitudes are constant for certain errors, whereby a damage allocation and also a comparison of the amount of damage within a type are possible.

at It is generally sufficient to monitor installations to detect an error in time. there It does not matter, the exact place of damage within the Locate bearing, since the defective stock exchanged anyway must become. Deeper diagnoses are in the expanded state the camp also possible with conventional methods.

These Cognition makes it possible the dissolution of the Biszubktrums to choose so that at reasonable computing times a reliable error detection is possible.

The inventive method and the invention Device will be closer in the following with reference to figures explained.

It demonstrate:

1 : a typical test setup for recording structure-borne sound vibrations on moving machine parts;

2 : Graphs of a bispectrum in different views;

3 : two representations the bispectrum according to 2 as a detail view of a contour plot;

4 Three bispectives of bearings with different damages.

1 shows a basic structure for structure-borne noise measurement according to the prior art. The structure-borne noise can be recorded with various sensors such as displacement, acceleration or speed sensors. Such sensors are, for example, eddy current sensors, electrodynamic velocity sensors or acceleration sensors according to the piezoelectric principle.

The basic measuring chain is identical for the individual sensor types. One at a machine 01 attached sensor 02 converts the mechanical vibrations, which among other things of a rolling bearing 03 resulting in electrical data signals. These are in a signal processing 04 amplified, filtered and sent to an evaluation unit 06 forwarded. The evaluation unit 06 can be a PC with a measurement card, a test computer in production or a mobile evaluation unit.

Measuring chains for structure-borne noise measurement are known to the person skilled in the art and are described, for example, in US Pat VDI Guideline VDI 3832: "Structure-borne noise measurements for the condition assessment of rolling bearings in machines and plants" described.

The evaluation unit 06 For carrying out the method according to the invention, unlike the prior art, preferably comprises a signal processor with a processor optimized for the calculation of the bispectrum and for the mathematical operations required for this purpose.

2 shows a calculated from the recorded data signal bispectrum in various forms of representation. In this case, Fig. A) shows a 3D plot, Fig. B) shows a contour plot with contour lines, and Fig. C) shows a vertical section through the bispectrum corresponding to the section line AA in Fig. B).

The bispectrum shows symmetrically arranged positive amplitudes 07 and negative amplitudes 08 , The contour plot (Fig. B) is preferably shown in color to give negative amplitudes 07 and positive amplitudes 08 to distinguish better.

The abscissae of the bispectrum form the factors f1 and f2. These are multiplied by the sampling frequency f _x and thus indicate the frequencies of the divided signals. The ordinate value (magnitude) is the respective degree of phase and frequency coupling between two frequencies.

In Fig. B) it can be seen that the distribution of the amplitudes 07 . 08 is symmetrical in the bispectrum. With dashed lines symmetry axes are shown. For the determination of the positions or the frequencies of the amplitudes, for example, the first quadrant I of the contour plot is considered. The frequency of the amplitude in the area below the symmetry line in quadrant I is read at f ₁ , above the line of symmetry at f ₂ . The reading can also be done in the sectional view.

to Sectional view in Fig. C) it should be noted that the consideration a section for a statement about the storage condition is not enough, because not all amplitudes in one cut can be represented. To determine the heights the amplitudes, however, the sectional representations are very suitable Good.

In The present examples were calculated using the computer algebra program MATLAB carried out. The result of the MATLAB calculation of the bispectrum is in complex form. Graphically, however, either the real or the imaginary part are displayed.

In 3 is once again the bispectrum 2 shown. In the bispectrum in Fig. A) and b) are a base frequency 09 of 2,500 Hz, the natural frequency of the outer ring 11 at 5,600 Hz and a natural frequency of the inner ring 12 entered at 9,700 Hz. These frequencies are in good agreement with the highest amplitudes.

In Fig. B) of 3 The highest amplitudes in the first quadrant of the bispectrum were again marked below the line of symmetry (✶). These are positive amplitudes 13 . 14 . 16 . 17 at 1,500 Hz, 2500 Hz, 7,500 Hz and 10,000 Hz and negative amplitudes 18 . 19 . 21 and 22 readable at 5,000 Hz, 5,500 Hz, 6,000 and 7,000 Hz.

Based on 4 shows what is to be read from the position and the distribution of the amplitudes or to be determined by calculation. Fig. A) shows the bispectrum of a bearing with an outer ring damage, Fig. B) the bispectrum of a bearing of the same type with an inner ring damage and Fig. C) the bispectrum of a bearing with an inner and outer ring damage.

Characteristic for outer ring damage (see Fig. a) the small number of significant amplitudes that only near the base frequency at 3,000 Hz and 1,500 Hz are located.

One Inner ring damage (see Fig. B) has the largest Number of significant amplitudes. These are located in near the base frequency and the natural frequencies of the Outer and inner ring. In addition, amplitudes occur at 1,500 and 7,500 Hz.

Of the mixed damage (see Fig. c) also has significant amplitudes near the base frequency and the natural frequency of the Outer ring on, next to it are amplitudes at 1,500 Hz and 7,500 To recognize Hz. In this damage, however, no amplitudes occur near the natural frequency of the inner ring.

01

machine

02

sensor

03

roller bearing

04

signal conditioning

05

06

evaluation

07

positive amplitude

08

negative amplitude

09

base frequency

10

11

natural frequency of the outer ring

12

natural frequency of the inner ring

13

positive amplitude

14

positive amplitude

15

16

positive amplitude

17

positive amplitude

18

negative amplitude

19

negative amplitude

20

21

negative amplitude

22

negative amplitude

QUOTES INCLUDE IN THE DESCRIPTION

This list The documents listed by the applicant have been automated generated and is solely for better information recorded by the reader. The list is not part of the German Patent or utility model application. The DPMA takes over no liability for any errors or omissions.

Cited patent literature

• DE 102005015465 A1 [0011]
• - JP 58193425A [0013]
• - DE 19933924 A1 [0014]
• - DE 69423006 T2 [0015]
• - JP 2006153761 A [0016]

Cited non-patent literature

• - VDI Guideline VDI 3832: " Structure-borne noise measurements for the condition assessment of rolling bearings in machines and plants" [0048]

Claims (11)

A method for detecting and analyzing damage to a rotating machine element, comprising the steps of: - recording a data signal at a sampling frequency (f _x ) which is emitted when the machine element is unrolled; - decomposing the data signal into a number N of different time signals (x _i (t)); - Transferring the time signals (x _i (t)) of the data signal in the frequency domain by means of Fourier transform; - calculation of the bispectrum (B (f ₁ , f ₂ )) according to the following rule

where X _i (f ₁ ) and X _i (f ₂ ) are the Fourier transform of the N time signals (x _i (t)); - Evaluation of significant amplitudes ( 07 . 08 ) of the bispectrum (B (f ₁ , f ₂ )) exceeding a predetermined value. Method according to claim 1, characterized in that that the data signal is an acceleration signal, a speed signal or a path signal. Method according to claim 1 or 2, characterized that the sampling frequency at 10 kHz to 500 kHz, preferably at 50 kHz. Method according to one of the claims 1 to 4, characterized in that the number N of the time signals (x _i (t)) determines the resolution of the bispectrum (B (f ₁ , f ₂ )) and N is selectable from 50 to 200. Method according to one of claims 1 to 5, characterized in that only one of the possible quadrants of the bispectrum (B (f ₁ , f ₂ )) is used to evaluate the significant amplitudes, in particular the first quadrant. Method according to one of claims 1 to 3, characterized in that the real part of the complex bispectrum (B (f ₁ , f ₂ )) is used to evaluate the significant amplitudes. Method according to one of claims 1 to 6, characterized in that a base frequency and natural frequencies of the machine element are determined and with the significant amplitudes ( 07 . 08 ) of the bispectrum (B (f ₁ , f ₂ )). A method according to claim 7, characterized in that an error signal is generated when the height and location of the significant amplitudes ( 07 . 08 ) in the bispectrum (B (f ₁ , f ₂ )) exceed given values. Device for detecting and analyzing bearing damage with a sensor ( 02 ) for detecting a data signal (x _i (t)), - a signal conditioning and conversion unit ( 04 ), - a computing and evaluation unit ( 06 ), characterized in that an algorithm for calculating a bispectrum (B (f ₁ , f ₂ )) from the data signal (x _i (t)) is implemented in the computing and evaluation unit and significant amplitudes ( 07 . 08 ) of the bispectrum (B (f ₁ , f ₂ )) are compared in their position and height with typical amplitudes which occur in damaged bearings and are stored in a database. Device according to claim 9, characterized in that the sensor ( 02 ) is an accelerometer, speed sensor or transducer. Device according to claim 9 or 10, characterized that the computing and evaluation unit a method according to one of Claims 1 to 8 executes.