Diagnostic Techniques for the Vibration Analysis of Bearings E. Estupiñan Professor University of Tarapacá - eestupin@uta.
cl Casilla 6 – D, Arica - CHILE P. Saavedra Professor University of Concepción - email@example.com Vibrations Laboratory – Mechanical Engineering Department Casilla 160 – C, Concepción – CHILE Abstract The vibration signal emitted by a bearing can contain spectral components that are related to the geometry of the bearing, the number of rolling elements, the speed of rotation, the location of the defect and the type of applied load; which marks the difference with a journal bearing type. It is of great importance in the modern industry to detect defects in the bearings of critical machines in an incipient fault stage. In the present work we have used some of the modern techniques of vibration analysis included today in some commercial vibration analyzers. For the experimental study, good shape ball bearings and localized defect in the outer race ball bearings, were tested under different levels of fault severity and various load and speed conditions. Normal spectral analysis, demodulation, PeakVue and real zoom analysis were the techniques used for the analysis. Key Words – Bearings, Faults Diagnostic, Predictive Maintenance, Vibration Analysis, Demodulation, PeakVue, Spectral analysis.
Introduction The bearings act as a source of noise and vibration due to varying compliance as much as the presence of defects in them, which may be classified into distributed and localized defects . The fact that the load distribution on the bearings varies, as the rolling element set rotates round the rings, causes the bearings to behave themselves as a vibration generator. This behavior may arise as much as from a geometrically perfect bearing as one which possesses imperfections from manufacture, installation, lubrication or inadequate ambient running conditions or some other factor that may help to cause wear or fatigue . Localized defects The most common type is the crack of the races or rolling elements, mainly caused when a crack due to fatigue originated sub-superficially is propagated towards the surface until a metal piece is detached causing a small defect. The fault caused by superficial fatigue is accelerated when the bearing is overloaded or submitted to shock or impact loads during their functioning or installation and also with the increase of the rotational speed. Distributed defects Within these we find: the surface roughness, waviness, misaligned races and unequal rolling elements. They can be caused by manufacturing errors, inadequate installation or due to wear. The variation of the contact forces between the rolling elements and the races cause an increase in the vibration level. Because it is difficult to discern when the vibration is caused by either distributed or localized defects, it is highly recommended to make an analysis based on; one hand the frequency and, on the other the amplitude of the spectral components . Evolution of the fault due to crack The first symptom that appears when the fault arises, at an early stage, is a vibration with high frequency components (generally higher than 5 KHz) due to the generation of stress waves and
other kinds, which can excite natural frequencies of housing and races of the bearings or the measuring sensor. In this stage, there is no temperature increase and the cracks are not visible, it is not necessary to change the bearings. In a second stage the cracks start to be visible and the bearings produce audible sound and in some cases the temperature arises. In this fault stage spectral components related to fault frequency of bearings in the low and intermediate frequency range (less than 5 KHz) appear. These frequencies are commonly named as “characteristic defect frequencies” and they are designed as BPFO (ball pass frequency outer race), BPFI (ball pass frequency inner race), BSF (ball spin frequency) and FTF (fundamental train frequency), depending on the location of the defect; whether it is on the inner race, outer race or on one of the rolling elements [3, 4]. They are determined based upon the geometry and rotational speed of the bearing and can be calculated from simple mathematical expressions or using commercial software [3, 4]. It is necessary in this fault stage to program the change of the bearing. In a third stage, close to a catastrophical failure, the noise increases significantly an overheating may arise, the vibration in the high frequency range decreases, the spectral components of the low frequency range increases and it will be necessary to change the bearing immediately [5, 6]. Although this, the most usual way for it to show when a bearing develops a fault, each bearing could have different fault modes with a different progression rate, it depends of the load, rotational velocity and lubrication conditions mainly . Diagnosis techniques To detect a defect in an early fault stage, different techniques and instruments have been developed depending on the range of frequencies within which the vibration analysis will be carried out. Some techniques have the high range frequency analysis approach; others have the medium and low range approach. The techniques which analyze the high frequency zone are based on the excitation of the natural frequency of the sensor, bearing parts and housing structures due to when such a defect on one surface strikes its mating surface, a pulse of short duration is produced (see Fig. 1). When the bearing rotates with a constant rotational speed, these pulses are generated periodically and the frequency is the characteristic defect frequency.
Figure 1. High frequency waves produced due to a defect in the outer race. The main symptom that allows detecting incipient defects is the presence of components of high frequency. When the fault progresses, the impulsive excitation of type increases even up to a time until when the edges, corners or rims of the defects smooth down due to the wear and the impact levels diminishes and they can even get to disappear. Most of the techniques of high frequency analysis give only an overall value of the energy caught by the sensor, which is compared with reference values and related to the severity of the defect. Within these techniques, there are the shock pulse measurement (SPM), the detection of acoustic emissions (AE), emitted spectral energy (SEE), technique of detection of high frequency (HFD) and the ultrasound.
Other techniques of high frequency analysis related to the excitation of natural frequencies are the techniques of enveloping or demodulation and PeakVue or peak value analysis. Techniques that will be described briefly later in this work. In the high frequency analysis there are two difficulties mainly, the first one is the low amplitude of the high frequency bursts compared with the components of low frequency. And second it is the difficult to identify the characteristic defect frequencies of the bearings due to the noise and the bad spectral resolution. To facilitate the tasks of diagnosis, it is necessary to isolate the high frequency vibrations from others using a high pass or band pass filters around the natural frequency that is excited. Thus the vibration generated only by the impacts can be analyzed identifying the discrete frequencies and evaluating the severity of fault. In this work we have analyzed two of the techniques that can be used for this purpose, the demodulation and PeakVue analysis.
Figure 2a. Enveloped detector.
Figure 2b. Peak value detector.
Demodulation Basically this technique consists of using a high pass or band pass filter on the temporal signal, with the purpose of isolating resonant zones of high frequency, then to rectify the modulated signal, to eliminate de DC component and finally using an enveloped detector (Fig. 2a) and a low pass filter the modulator signal is obtained, how it is shown in Fig. 3.
Figure 3. Demodulation process. PeakVue This is a technique that captures the peak value of the stress waves that are produced (see Fig. 2b), and then via a spectral analysis the repetition frequency of the impacts is obtained. PeakVue detects the presence of the stress waves mainly due to metal to metal contact during an early
stage of the failure. Like the demodulation, this technique isolates the resonance zones by means of high pass or band pass digital filters, but it is differentiated from the demodulation technique in that in the final stage the enveloped detector is not used, rather using a high frequency sampling (100 kHz) catches the peak value for each interval of the normal sampling time , see Fig. 4.
Figure 4. Peak Value detection using a high sampling frequency. Real Zoom The real zoom analysis, allows by means of digital filters to acquire the signal with a narrow bandwidth around the resonant frequency, which must previously be identified. A spectrum with a good resolution is obtained this way, identifying the modulator frequencies that actually cause the excitation of the resonant zone in an easier manner. In Fig. 7, the obtained result of an analysis with real zoom around the resonant frequency of 60 Kcpm is shown. Experimental study In this work good shape ball bearings and localized defect in the outer race ball bearings, were tested under different levels of fault severity and various load and speed conditions. In these tests the normal spectral analysis, demodulation, PeakVue and real zoom analysis, were used. A testing station as the one shown on Fig. 5 (characteristics are shown on Table I), was used to carry out the testing. For the vibration analysis a CSI-2120 analyzer, high and low frequency accelerometers (0.1 mV/g and 0.5 mV/g) and analog signals digital recorder were used.
Figure 5. Testing station Table I. Characteristics of testing station.
A B C D E
Designation Outer bearing Inboard bearing Pulley Coupling Motor 1400 rpm
Description Ball bearings SKF Ref. 6306-2z. Ref. 6207-2z Flexible Induction AC (1.5hp) Controlled with a frequency converter.
The measured vibrations taken in the housing of the outboard bearing in vertical direction were compared, between faultless and faulty bearings. The testing was carried out for the speed
condition of 1000 rpm. In addition, 60 rpm and a 200-Kg over the pulley were applied. In the faulty bearings the defects was intentionally made in order to simulate a defect at an early stage. In Fig. 6 the spectra in acceleration units, obtained from a faultless and faulty bearing for the turning velocity of 1000 rpm is shown. It can be clearly observed in the faulty bearing spectrum a high energy vibration zone around the 60 Kcpm, frequency which corresponds to a natural frequency. When a real zoom is carried out around this frequency, sidebands can be distinguished in the spectrum with a separation between them to the BPFO frequency showing a fault in the outer race (see Fig. 7).
Figure 6. Acceleration spectrum (Running speed = 1000 rpm)
Figure 7. Real Zoom Analysis. (see the sidebands) Observing the acceleration waveform, the periodicity of the impact produced due to the passing of the rolling elements over the defect is identified as shown in Fig. 8.
Figure 8. Impacts identified in the waveform time. In Figs. 9 and 10, the obtained results from the demodulation and PeakVue analysis are shown. In the case of the faulty bearing the BPFO components and its harmonics are clearly identified, confirming the presence of a defect in the outer race.
Figure 9. Demodulation (Running speed = 1000 rpm)
Figure 10. PeakVue (Running speed = 1000 rpm).
When these techniques are applied to the low speed turning bearings (60 rpm), in the acceleration spectrum the excitation of the resonance zone in a faulty bearing is observed (see Fig. 11). Applying the PeakVue and the demodulation techniques, the presence of faults is identified but not clearly as in the former case (see Figs. 12 and 13).
Figure 11. Acceleration spectrum. (Running speed = 60 rpm)
Figure 12. Demodulation spectrum. (Running speed = 60 rpm)
Figure 13. PeakVue spectrum. (Running speed = 60 rpm)
Discussion From the results obtained in the different tests, we may conclude the following: - Independently of the running speed, in the fault bearing, an excitation over the high frequency zone is produced. However, a lower intensity excitation is produced and the lower natural frequencies are excited in the cases of low rotational velocity. - Using the real zoom and waveform time analysis, it is possible to identify the periodically produced impacts. In the low speed cases the results obtained using these techniques were not satisfactory due to the low amplitude of the impacts produced and they were also mixed with the low frequency spectral components. - The results delivered from the demodulation and PeakVue techniques were similar. Nevertheless, the amplitude obtained using PeakVue were higher than when using demodulation. In low velocity applications it is more advantageous to use these techniques rather than the traditional ones in which case the results obtained were not so good. It is important to keep in mind that this work was carried out under controlled noise and vibration conditions. In the industry, more difficulties can be found especially in the low rotation speed machines due to the various noise sources, vibrations coming from other machines and the speed and load variation during the measurements. Therefore the traditional spectral analysis must be complemented with the use of other techniques which are nowadays available in most
commercial vibration analyzers such us the synchronic time average, the tracking filter and order tracking analysis. Conclusions It is possible to detect in the bearings the presence of defects when these are at an early stage, using vibration analysis integrated with the demodulation and /or PeakVue analysis techniques which are centered on the high frequency analysis. The demodulation and PeakVue techniques are more useful than the normal spectral analysis in the early fault detection for the analysis of low velocity turning bearings. References  TANDON N., NAKRA B. C. (1992). Vibration and acoustic monitoring techniques for the detection of defects in rolling element bearings – a review. The Shock and Vibration Digest. Vol. 24 (3), 3-11.  AKTÜRK N., GOHAR R. (1998). Vibrations associated with ball bearings. ImechE, Conference Transactions, Multi-Body Dynamics. December 10-11 No.13, 43-64.  TANDON N., CHOUDHURY. (1999). A review of vibration and acoustic measurement methods for the detection of defects in rolling element bearings. Tribology International. Vol.32, 469 –480.  HARRIS T. (1966). Rolling bearing analysis. NewYork : John Wiley and Sons.  JONES R. M. (1996). Enveloping for bearing analysis. Sound and Vibration. February, 10-15.  SAAVEDRA P. (1998). Análisis de vibraciones de máquinas rotatorias (Nivel III). Cap 4. 1017. Universidad de Concepción - Chile.  CANADA R., PIETY K., ROBINSON J.C. (1998). New Methodology for bearing fault defect: PeakVue analysis. Application Paper, Computational System Incorporated. (CSI).