Anton Azovtsev, Alexei Barkov, VibroAcoustical Systems and Technologies, Inc.


Diagnostics of rotating machinery using the envelope spectrum of high frequency vibration exited by friction forces is becoming widely used by many companies and experts. In recent years, automatic diagnostic systems based on this method have been developed for rolling element bearings. These systems can provide detailed diagnostics and condition prediction of a bearing by a single vibration measurement. Automatic diagnostic systems for simultaneous diagnostics of a gears and rolling element bearings in a gear transmissions have also been developed. But until recently, there was no diagnostic systems that used envelope spectra of random vibration for detection and identification of defects in fluid film bearings.

The main problems in the use of standard enveloping methods developed for the diagnostics of rolling element bearings in the diagnostics of fluid film bearings are discussed in this paper. Methods for solution of these problems are analyzed together with the practical results achieved in this field.


The method of rolling element bearings diagnostics by the spectrum of high frequency vibration envelope is based on the analysis of characteristics in the formation of friction forces in good and defective bearings as well as in the features of shock pulses that appear in the interaction of rolling surfaces with cavities, spalls, or cracks in the bearing elements [1].

The friction forces depend on the rolling friction coefficient and the load on rolling elements. In good bearings, the friction is uniform in time, i.e. it does not depend on the rotation angle of the rotating race or of the cage.

In rolling element bearings with defects of installation including misalignment of races and non-uniform radial tension, loads on the rolling elements increase and, more importantly, these loads become dependent on the rotation angle of the rotating race and the cage. As a result, the friction forces, together with the random vibration excited by them, become amplitude modulated.

In bearings with non-uniform wear of inner and outer races and rolling elements, the friction coefficient in turn depends on the rotation angle of rotating race and cage which results in similar amplitude modulation of the friction forces and the resultant high frequency vibration.

Finally, shock pulses in bearings with cavities and cracks on rolling surfaces and races produce vibration as well. On the resonant frequencies of rolling elements and races, this vibration is actually attenuated self-oscillations that should not be considered as random vibration. At other frequencies, shock pulses excite random, fast attenuated vibration that is also modulated in amplitude.

As a result, all defects of bearing installation, wear, and cavities can be detected by the spectrum analysis of the high frequency vibration envelope. Figure 1 gives examples of envelope spectra from a good rolling element bearing (a), a bearing with an installation defect (b), with wear of a rolling surface (c), and with a crack on a rolling surface(d).

Figure 1a (above). Envelope spectrum of a good rolling element bearing.

Figure 1b (above). Envelope spectrum of a rolling element bearing with misalignment of the outer race.

Figure 1c (above). Envelope spectrum of a rolling element bearing with a worn outer race.

Figure 1d (above). Envelope spectrum of a rolling element bearing with an outer race crack.

The apparent simplicity of defect detection and identification by the envelope spectrum of a rolling element bearing vibration can not be fully realized in practice. There are two main characteristics of the friction forces and the resultant random vibration formation that are the reasons for this.

The first reason is connected with the characteristics of the loads applied to the bearings in the real machines. In addition to the normal load, the rotating load from the shaft wobbling of an unbalanced rotor may be applied to the bearing. This additional load may also depend on the rotation angle of the shaft which significantly complicates the problem of defect identification. For example, a bearing may be exposed to shock loads due to defects of a gear transmission.

Gearing defects, for example, are detected by measuring the results of this shock load on the rolling element bearings [2]. Figure 2 presents envelope spectra of a rolling element bearing from a gearbox with one defective gear. Here, you can see envelope spectra measured on the bearing of the shaft with defective gearing and the adjacent shaft. The defects of gearing and cracks on the bearing surfaces can be distinguished by the repetition frequencies of the shock loads sequence.

Figure 2a (above). Envelope spectrum a rolling element bearing from a gearbox with one defective gear on the bearing of the shaft with the defective gear.

Figure 2b (above). Envelope spectrum a rolling element bearing from a gearbox with one defective gear on the bearing of the adjacent shaft.

The second reason for the complexity of fault detection and identification is connected with the necessity of detecting the envelope spectrum produced by only the random components of bearing vibration and excluding from consideration any of the harmonic components from either the bearing under diagnostics or from other machine units. Special methods for signal processing or careful choice of frequency band for enveloping should be used for this purpose [3].


High frequency random vibration is excited by friction forces in both rolling element and fluid film bearings. When defects of friction surfaces develop in these types of bearings, the friction forces and high frequency vibration acquire amplitude modulation and thus bearing defects can be detected by the analysis of envelope spectrum of this vibration. In the same situation, in the case of fluid film bearings, there are much more problems in the detection and identification of defects than in rolling element bearings. The first problem is connected with the limited diagnostic information that can be derived from the modulation frequencies. Rolling element bearings have at least 4 types of friction surfaces with different rotation speeds. These are outer and inner races, rolling elements, and the cage. We can list three fundamental bearing frequencies of rotation and all combinations of them. In case of fluid film bearings, there are only two friction surfaces and one fundamental frequency.

The second problem is due to the characteristics of pressure pulsation formation in the lubrication layer and the vibration excited by this pulsation. The pulsation power is determined by the velocity gradient in the lubrication layer, but it increases not only with rotation speed, but also with the decrease of lubrication layer thickness. The thickness of the lubrication layer, in turn, depends on the bearing design and on the relative position of the shaft axis and bearing shell axis. In a number of bearings, the shaft axis moves and, as a result, the high frequency vibration, may be modulated even in good bearings.

The third problem occurs in the detection of the defect severity. In the case of fluid film bearings, you should take into account not only the modulation index in the envelope spectrum, but the thickness of the lubrication layer as well. Reliable information about the thickness of the lubrication layer, as a rule, is not available for a user, thus the levels for defects have to be adapted for every machine type that has some special design characteristics of the bearings or of the machine.

Following many years of investigations and practical diagnostics of bearings in rotating machines using the envelope spectrum of high frequency vibration, we came to the conclusion that the above problems are not solvable in practice. A number of diagnostic symptoms were found which allow successful condition diagnostics of fluid film bearings [4]. These are based on the characteristics of shaft journal oscillations in the lubrication layer of the fluid film bearing. Consider three of these characteristics.

The first characteristic is the possible appearance of short pulses with increased oil pulsation during the movement of oil wedge on the bearing shell surface. The rotation of the oil wedge is an indication of, first of all, of shaft wobbling. In a good bearing, such a pulse can appear when the oil wedge passes joints of the shell sectors. In the case of a worn bearing, when the oil wedge passes non-uniform wear zones on the bearing shell, cracks, and so fourth, such short pulses can be considered as shock pulses in fluid film bearings analogous to the shock pulses in rolling element bearings. An envelope spectrum of the journal bearing high frequency vibration in such a case in presented on figure 3 (b).

The second characteristic is possible appearance of shaft vibration at frequencies different from the harmonics of rotation speed. Most often, this is a self-sustained oscillation of the shaft in bearings with loose clearances or defects in the oil supply system. In most practical cases, self-sustained oscillations of the shaft synchronize with one of the sub-harmonics of the rotation speed. Sometimes you can also observe pendulum shaft oscillation in very loose bearings which also, as a rule, synchronize with one of rotation speed sub-harmonics. An example of the envelope spectrum in this case is presented in figure 3 (c).

And the third characteristic is the appearance of ultra-low frequency random oscillations of the shaft relative to the bearing shell surface. This situation can be found in the bearings with non-uniformly worn shells. These oscillations are defined by the unstable shape of the oil wedge with small and random changes. This can be detected by changes in the shape of the envelope spectrum and background level increases on frequencies below the shaft rotation speed. Such increases of background level in the envelope spectrum should be considered as an effective symptom of the bearing shell wear. An example of such an envelope spectrum is presented on figure 3 (d).

The use of the above diagnostic symptoms allow the fluid film bearings diagnostics without needing a relative displacement transducer installation in the bearing unit. At the same time, these methods do not eliminate the problems stated above, namely, detection of the defect severity and distinguishing between defective bearings and good bearings where the oil wedge moves with the shaft rotation on the bearing shell due to the bearing design peculiarities or shaft faults.

Figure 3a (above). Envelope spectra of a good journal bearing.

Figure 3b (above). Envelope spectra of a worn journal bearing with "shock pulses".

Figure 3c (above). Envelope spectra of a journal bearing with self-sustained oscillations.

Figure 3d (above). Envelope spectra of a journal bearing with a worn shell.


The following problems should be considered when the diagnostic measurements are planned:

In the case of rolling element bearings, all the above problems can be successfully solved by the analysis of envelope spectrum of high frequency vibration measured on the bearing housing. In this case, no other measurements of vibration or other parameters are needed. The only concern is to comply with the periodicity of measurements. The intervals between measurements can be rather long, about a few months when the service life of the bearing is 2 to 3 years and when the machine is operated in its standard modes. Possible cases of machine operation that should be considered differently occur when there are shock loads on the machine and overheating. If the above conditions are met, it is enough to make 10 to 20 measurements of the high frequency vibration envelope spectra during the whole service life of a bearing to eliminate its un-predicted failure [1].

The levels for defects that are used for the estimation of the detected defect severity in rolling element bearings can be similar for all types of machines and bearings. They monotonously increase with the increase of rotation speed and bearing dimensions. For example, the levels for the bearing dimensions from some millimeters up to a few meters and rotation speeds from a dozen of revolutions per minute up to hundreds of Hertz differ by just 3 to 5 times. According to the importance of a particular machine, the customer can correct the alarm levels. According to our experience with customers in from different industries, such corrections typically do not exceed 2 to 3 times.

A number of different problems occur in the diagnostics of fluid film bearings using an envelope spectrum of high frequency vibration. These problems are solved by the added analysis of diagnostic information from the autospectra of bearing housing vibration.

The first problem is connected to the design characteristics of some machines where the shaft motion oscillates in the stationary bearing shell, even with no defects present. In this case, an envelope spectrum at this bearing will contain harmonic components proportional to the rotation frequency in spite of the absence of defect. The characteristics of these components can undergo spasmodic changes during machine operation and defects in the initial stage of their development may not be detected. In this case, the amplitude changes of these components should be analyzed by using both autospectra and envelope spectra in parallel. This significantly increases the probability of incipient defect detection. The vibration measurements, in turn, should be done more often, i.e. with time intervals of about 1 month or less.

The second problem is closely connected with the variety of fluid film bearings designs, each of which has its own lubrication layer thickness. For this reason, similar defects in the bearings with different lubrication layer thickness lead to different changes in the normalized thickness and thus, different modulation of random vibration. The introduction of corrections for the lubrication layer thickness in the defect levels calculation may not always provide good results. The best solution is to correct levels for defects according to the data derived from the analysis of autospectra, i.e. the increase of vibration components after the detection of severe defects in similar bearings of other machines.

For the reasons discussed above which are different from the rolling element bearing case, fluid film bearings are best diagnosed by the parallel analysis of autospectra and envelope spectra of the bearing shield vibration.


Comparative analysis of possibilities for rolling element and fluid film bearings diagnostics by the analysis of high frequency vibration envelope spectra allows the following conclusions:

  1. Vibration diagnostics is the most efficient method for detection and identification of incipient defects in rolling element and fluid film bearings. A accumulation experience acquired during many years diagnostics of rolling element bearings and during several years of broad investigations on similar type diagnostics of fluid film bearings makes practical similar diagnostic techniques for fluid film bearings.
  2. The typical interval between measurements for rolling element bearings in absence of detected defects is several months and for fluid film bearings is about 1 month.
  3. The levels for the detection of defects in rolling element bearings and estimation of their severity are nearly independent on the machine and bearing design. Only some small dependence of levels for defects on the rotation speed and bearing dimensions exists. For machines with fluid film bearings, the levels for defects also depend on the bearing design.


  1. A. Barkov, N. Barkova, J. Mitchell, "Condition Assessment and Life Prediction of Rolling Element Bearings", Sound & Vibration, 1995, June pp.10-17, September, pp.27-31.
  2. A. Barkov, N. 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
  3. A. Azovtsev, A. Barkov, D. Carter, "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. A. Azovtsev, A. Barkov, D. Carter, "Journal Bearing Diagnostics by the Envelope Spectra", Accepted for publication and presentation at the 21st Annual Meeting of the Vibration Institute. June 1997.

Back to Inteltech Enterprises, Inc. home page. Back to the articles/notes age.