Fluid Film Bearing Diagnostics Using Envelope Spectra


Anton Azovtsev, Alexei Barkov, and Duncan Carter


Spectrum analysis of the envelope of the high frequency vibration excited by friction forces is becoming a common method for the diagnostics and condition prediction of rolling element bearings. Measurement of the vibrations excited by bearing friction forces present all the necessary information for diagnostics of bearing condition including installation problems and the quality of lubrication.

Although not in common use, similar demodulation techniques, with modification, are practical for the diagnostics of fluid film bearings. A number of reasons exist that can limit the use of such techniques. First is an ambiguous dependence between the features of friction forces and the defect severity due to the strong dependency between friction forces and the thickness of the lubrication layer. The thickness of lubrication varies with the type of bearing. The second reason is the high probability and unpredictability of shaft oscillation relative to the bearing. These oscillations, which depend on numerous parameters of the machine, its lubrication system, etc., modulate the friction forces and make the diagnostic process by the envelope spectra more complicated. Despite these problems, a number of relationships between the defects of journal bearings and the features of friction forces and the induced high frequency vibrations have been found which make diagnostics of journal bearings practical without the necessity of installing displacement sensors in the bearings. These relationships and comparisons with rolling element methods are discussed in the paper together with several case studies of journal bearing diagnostics in different types of machines.

I. The sources of random vibrations used for demodulation in rolling element and fluid film bearings.

Spectrum analysis of the envelope of signals is becoming widely used in a number of fields. If the content of the signal to be demodulated is limited to representative signals of the physical phenomena of interest, extraction of information containing numerous complex components having different features can be accomplished with very accurate results[1,2,3] but this process may not be adequately supported in most current vibration measurement equipment. The main application of the envelope detection method is the analysis of low frequency oscillations of higher frequency signals. In the case of rolling element bearing condition monitoring and diagnostics, it is used to extract the information about the modulation of rolling friction forces and the power of the random vibration extracted by these forces[4,5]. With modifications, similar demodulation techniques may be used for the condition monitoring and diagnostics of fluid film bearings.

The main sources of random vibration in rolling element bearings are rolling friction and shock pulses. The friction forces can be considered as a combination of great many micro-shock pulses randomly distributed in time that do not lead to a breakdown of lubrication layer. Shock pulses that appear as a result of good and defective friction surfaces’ interaction are much more rare, but their power may exceed the power of these micro-pulses by several orders and break the lubrication layer. The last feature is the reason for the fact that, in the high frequency domain above 10KHz to 20 KHz, components of shock pulses can prevail in the vibration signal.

In a fluid film bearing, the source of the random vibrations to be demodulated are pressure pulsations in the lubrication flow between stationary and rotating surfaces. There are two distinguishable types of such pulsations. The first type is pressure pulsation in the border region of the laminar flow. Pulsations of this kind occur when there are no significant vibration displacements of the shaft relative to the bearing shell. The second type is the appearance of short term turbulence in the lubrication flow due to the rapid changes in the lubrication layer and, consequently, in the flow speed in the region of turbulence. This source of pressure pulsation can be named, analogously to the rolling bearing situation, the hydrodynamic "shock" pulse.

We can conclude that the amplitude of the rolling element bearing emitted random vibrations, in cases of constant rotation speed and lubrication quality, depends on the friction coefficient and bearing load. There are three main reasons for possible dependency of the random vibration power on the friction surface’s rotation angle. The first reason, which is the most frequent, is the dependency of the friction coefficient ton the rotation angle due to non-uniform surface wear or appearance of cracks (spalls). The second is connected with improper bearing installation that may lead to the appearance of additional dynamic loads on the bearing dependent on the rotation angle of friction elements. The third reason has the same origin as the previous one. This is appearance of additional loads on the proper mounted bearing that may change their direction or value because of machine operation peculiarities of defects of other machine units. The most typical example is shaft wobbling due to improper coupling of the shafts of two or more machines.

In the fluid film bearing case, the main reason for random vibration power changes in time is shaft oscillations relative to the stationary bearing parts i.e. changes of lubrication layer thickness and the consequent oscillation of flow speed within the lubrication layer. A second reason can be the movement of the oil wedge on the stationary friction surface. By the analysis of the modulation of random vibration power, we can find out not only characteristics of shaft oscillations relative bearing shells, but also the oil wedge movement.

II. Detected Defects

As shown above, the analysis of a fluid film bearing’s high frequency vibration envelope spectrum can be used to detect two main peculiarities of its operation, the existence of vibration displacements between stationary and rotating surfaces and appearance of hydrodynamic shock pulses. Both of these are determined by the defects of either the bearing or of the shaft and other connected units of the machine.

Parameters of friction surface oscillation such as frequency and amplitude can be measured. Amplitude can be measured within a limited accuracy in the units relative to the lubrication thickness. One difference between demodulation techniques and displacement probes installed in the bearing is that the envelope spectrum can not be used to determine the path of shaft axis oscillation-orbits.

For hydrodynamic shock pulses, their frequency, amplitude, and form (duration) can be measured. The last parameter is closely connected with the number of harmonics of the pulse frequency in the envelope spectrum. Note that sometimes single hydrodynamic shock pulses randomly distributed in time can be observed in fluid film bearings.

As a result of several years field experience in the condition diagnostics of rotating machines with fluid film bearings using random vibration envelope spectra and auto spectra measured on bearings, diagnostic symptoms were found for the following group of defects:

The peculiarities of hydrodynamic shocks formation in the lubrication layer allowed a division of the last defect into two groups:

The first one is an indication of the problems in lubrication functionality and the second is an indication of a near breakdown situation.

The variety of defects identified by the limited number of symptoms in the envelope spectrum of fluid film bearing high frequency vibration is enhanced by the fact that these symptoms can be supplemented by the information derived by the analysis of bearing vibration auto spectra. The peculiarities of hydrodynamic shocks formation in the lubrication layer are such that their development does not always change the parameters of both the low and high frequency domains of the bearing vibration spectrum. That's why the algorithms of fluid film bearings diagnostics are based on the joint analysis of auto spectra and envelope spectra of the bearing vibration.

III. Examples of Diagnostics

Our experience with diagnostics of rotating machines having fluid film bearings in different industries in Russia showed that there are two basic groups of machines, each having their own peculiarities. The first group is horizontal machines with massive rotors. The second group is small, not very powerful machines with limited loading of friction surfaces. The examples presented below are typical for the machines mostly in first group.

Fig. 1a, 1b, 1c, 1d (below). Envelope and auto spectra of one of the supports of a turbogenerator.

Fig. 1a(above). Envelope spectra of a turbogenerator before the defect was detected.

Fig. 1b(above). Auto spectra of a turbogenerator before the defect was detected.

Fig. 1c(above). Envelope spectra of a turbogenerator after the detection of increased low frequency vibration.

Fig. 1d(above). Auto spectra of a turbogenerator after the detection of increased low frequency vibration.

Consider figure 1. Here you can see auto spectra and envelope spectra measured on one of the bearings of a 300 MW turbo-generator of a conventional power plant in the city of Cherepovets, Russia. The vibration was measured on the bearing of generator next to the turbine. The bearing has increased vibration levels on the lower frequencies. The diagnostics showed that the reason for is shaft wobbling because of the joint coupling defect.

In the autospectrum, the shaft wobbling is indicated by the increase of rotation frequency and, especially, by its higher harmonics and in the envelope spectrum, by the presence of series of rotation speed harmonics that occur due to the hydrodynamic shocks in the lubrication layer with the rotation frequency. Also, there is no significant increase in the high frequency vibration levels which can be indication of "dry" shocks in the bearing.

Fig. 2a, 2b, 2c, 2d (below). Envelope and auto spectra of two supports of a turbogenerator after an increase of the temperature of a bearing.

Figure 2a(above). Envelope spectra of good bearing.

Figure 2b(above). Auto spectra of good bearing.

Figure 2c(above). Envelope spectra of overheated bearing.

Figure 2d(above). Auto spectra of overheated bearing.

Figure 2 presents an example of different defect detection in a journal bearing. This is a case history from one of 500 MW turbo-generators from a power plant of St. Petersburg, Russia. The experts were invited by the maintenance staff when, after repair of the machine, one of the bearings became overheated, but the vibration levels were within specified levels for a good bearing. The first measurements of bearing high frequency envelope spectrum showed that the bearing shell was installed mis-aligned. The diagnostic symptom in this case is the second harmonic of rotation frequency that prevails over all other shaft rotating frequency harmonics. The reason for this is that, due to the natural movements, the shaft "contacted" the shell twice per revolution, though the lubrication layer was not broken. After bearing disassembly, it was observed that the shells were rather burned in just a few days of bearing operation and some dangerous structural changes developed in the bearing. An autospectrum showed no indications of the bearing defect.

Fig. 3a, 3b, 3c, 3d, 3e, and 3f(below). Envelope and auto spectra from periodic condition monitoring and diagnostics of a turbogenerator exciter bearing.

Fig. 3a(above) Envelope spectra, initial state (February, 1995).

Fig. 3b(above) Auto spectra, initial state (February, 1995).

Fig. 3c(above) Envelope spectra, medium wear of the bearing (August 1996), the symptoms of the initial stage of shell spalls are present.

Fig. 3d(above) Auto spectra, medium wear of the bearing (August 1996), the symptoms of the initial stage of shell spalls are present.

Fig. 3e(above) Envelope spectra, bearing before repair (December 1996), the symptoms of self-sustained rotor oscillations in the bearing and dry shocks exist.

Fig. 3f(above) Auto spectra, bearing before repair (December 1996), the symptoms of self-sustained rotor oscillations in the bearing and dry shocks exist.

The third example, illustrated in figure 3, is the detection of a much more dangerous defect that has similar diagnostic symptoms. This defect was found on a bearing of a turbo-generator exciter at one of power plants in Kiev, Ukraine. During condition monitoring measurements of bearing vibration, an increase of low - and high frequency components was detected in the vibration auto spectrum simultaneous with the increase of harmonic components in high frequency vibration envelope spectrum. In other words, all symptoms indicated the presence of "dry" shocks in the bearing. The bearing was disassembled and inspected. The shells had a number of spalls which caused the appearance of shock pulses. With the analysis of data obtained by periodic condition diagnostics, recommendations for bearing repair could be issued three months before the final diagnostics, just after appearance of symptoms of severe bearing wear - randomly distributed in time hydrodynamic shock pulses.

Figure 4(below).

Fig. 4. Envelope spectrum of a turbogenerator support in the mode of self-sustained oscillations.

Figure 4 shows as additional example from journal bearing diagnostic practice - detection of self-sustained shaft oscillations. Self-sustained shaft oscillations in journal bearings can occur in a single bearing and not necessarily in the most worn one. Major problems can be caused by self-sustained oscillations during rotor field balancing i.e. balancing in rotor's own supports. This situation is typical for when an expert is called and, after significant financial investments and labor efforts have been done in balancing, but appropriate results are not attained. Figure 4 shows how self-sustained oscillations were detected. Before that, balancing of the rotor in its own supports had been attempted for almost four days with nearly zero results. After bearing shell replacement in two out of eight lemon (elliptical) bearings of turbo-generator, the rotor was successfully balanced in four runs.

IV. Peculiarities of Diagnostics

The main peculiarity of fluid film bearing diagnostics using the spectra of the vibration envelope, as mentioned earlier in the examples from practice, is the use of additional diagnostic symptoms from the auto spectra of bearing vibration. The combination of diagnostic symptoms from auto spectra and envelope spectra of bearing vibration is necessary for highly reliable identification of defect type and severity.

The second peculiarity is connected with the diagnostics of the bearings of machines with more than one shaft line connected with gear, belt or other transmission devices. These transmissions may produce additional dynamic loads on the bearing that introduce their own peculiarities of journal oscillation relative stationary parts of the bearing. These peculiarities may cause significant changes in the process of pressure pulsations in the lubrication layer and fluid film bearing vibration.

Figure 5(below).

Fig. 5. Envelope spectrum of a journal bearing in a one stage gearbox with defects on both gears. Here Fz is a gear mesh frequency.

Figure 5 presents an envelope spectrum measured on the journal bearing from a gearbox with wobbling of both driver and driven shafts together with defects of the gearwheels of this transmission. For instance, the spectrum components on the sub-harmonics of the shafts' rotation speed are the diagnostic symptoms, not of the self-sustained shaft oscillations, but of the gear wear[6]. This peculiarity demonstrates the peculiarity of the hydrodynamic shock pulses in the lubrication layer on the combination frequencies when, due to existence of two or more oscillation processes, the overall vibration exceeds the level of shock pulses excitation.

As a result, the overall number of the defects identified as a defects of fluid film bearings decrease. Their symptoms occur in conjunction with other symptoms for the mechanical transmission defects. In particular, such defects as self-sustained oscillation of the shaft or shell misalignment are not considered as a bearing defects in mechanical transmissions, but are included in a defect group called defects of gear wheels, belts, etc. At the same time, these defects are not missed in the machine, but their identification requires significant complication of the diagnostic measurements and processing algorithms.

One more peculiarity in diagnostics is connected with the choice of levels for the defect detection by the envelope spectrum analysis. These levels are mostly dependent on the lubrication layer thickness. In turn, the lubrication layer thickness may differ greatly according to the bearing design. This results in the need of defect levels adaptation for the different designs of the fluid film bearings. A rather simple method was developed for this purpose. It binds the increase of the vibration components in the auto spectra and the levels used in the analysis of vibration envelope spectra. Using this method, a customer easily corrects the levels for defects on the adaptation stage of the automatic condition diagnostic systems to his equipment.

Some differences in the choice of measurement points location for high frequency vibration envelope measurements between rolling and fluid film bearings should also be mentioned. Diagnostics is done by the high frequency vibration excited by pulsations in lubrication layer which propagate through the stationary bearing parts. When the bearing shells are mounted in a special way that attenuates high frequency vibration while it propagates to the bearing shield, it is impossible to measure the signals of interest without special means that allow rigid connection between the accelerometer and the bearing shells. Sometimes, a change in shell design is required.

A final main peculiarity of fluid film bearings diagnostics should be mentioned. In Russia, this problem postponed the wide introduction of condition diagnostics systems for several years. This peculiarity is connected with the possibility of fully automating the diagnostics process with the aim of replacing a highly qualified expert with software for automatic condition diagnostics. Such an approach is widely used in Russia where there is a lack of qualified experts in vibration diagnostics. For the diagnostics of fluid film bearings, it is not sufficient to analyze a single envelope spectrum. In addition, an autospectrum of the bearing vibration should also be analyzed together with a number of envelope spectra. The increased complexity of the diagnostics requires much more computational power from the computers. After appearance of more powerful personal computers, it became possible in Russia to develop and widely use systems for automatic diagnostics of nearly all types of rotating machines units.


The analysis of results of development and practical application of condition diagnostics methods for fluid film bearings without installation of special sensors enables the following conclusions:

1. Envelope detection methods based on the spectrum analysis of high frequency random vibration envelope that are well known to be successful in the diagnostics of rolling element[7] bearings can be efficiently used for the fluid film bearings condition diagnostics as well.

2. Condition diagnostics of fluid film bearings in comparison to rolling element bearings should be done by the joint analysis of auto spectra of bearing vibration and envelope spectra of its high frequency components.

3. The probability of missing a dangerous situation using this type of diagnostics is very low.

4. Automatic condition diagnostic systems based on the joint analysis of auto spectra and envelope spectra require adaptation to the particular design of the bearings i.e. the thickness of lubrication layer. The purpose for this is to determine the levels for dangerous defects. This adaptation may be made by a customer.

5. These systems are capable of automating the process of condition diagnostics and safe operation time forecast. Tests of such systems in Russia demonstrated their high efficiency when used by customers who had no special training in diagnostics.


  1. Duncan L. Carter, U. S. Patent Number 5,477,730, "Rolling Element Bearing Condition Testing Method and Apparatus" issued December 26, 1995.
  2. Duncan L. Carter, "A New Method For Processing Rolling Element Bearing Signals", presented at the 20th annual meeting of the Vibration Institute, June, 1996.
  3. Azovtsev A. Yu., Barkov A. V., Carter D. L., "Improving the accuracy of Rolling Element Bearing Condition Assessment", Proceedings of the 20th Annual Meeting of the Vibration Institute, Saint Louis, Missouri, USA, 1996, pp. 27-30.
  4. Barkov A. V., Barkova N. A., Mitchell J. S., "Condition Assessment and Life Prediction of Rolling Element Bearings", Sound & Vibration, 1995, June pp.10-17, September, pp. 27-31.
  5. A. A. Alexandrov, A. V. Barkov, N. A. Barkova, V. A. Shafransky, Vibration and Vibrodiagnostics of Electrical Equipment in Ships, -Sudostroenie (Shipbuilding), Leningrad, 1986.
  6. A.V. Barkov, N. A. Barkova, "Diagnostics of Gearings and Geared Couplings Using Envelope Spectrum Methods", Proceedings of the 20th Annual Meeting of the Vibration Institute, Saint Louis, Missouri, USA, 1996, pp. 75-83
  7. A. Azovtsev, A. Barkov, "Automatic computer system for roller bearings diagnostics", Computers in Railways V , Proceedings of the COMPRAIL-96 conference, 21-23 August 1996, Berlin, Germany, volume 2, pp. 543-550.

Back to articles/notes page

Back to VAST/Inteltech Enterprises home page