AUTOMATIC DIAGNOSTICS AND CONDITION PREDICTION OF ROLLING ELEMENT BEARINGS USING ENVELOPING METHODS

by

Dr. Yury A. AZOVTSEV, Dr. Alexej V. BARKOV, Dr. Iosif A. YUDIN

Abstract

A method has been developed to accurately identify the presence, type and magnitude of defects within rolling element bearings from enveloped random vibration spectra. In an automatic mode, the Diagnostic Rolling Element Module (DREAM) can be used effectively by inexperienced personnel. In the hands of an expert, it becomes a powerful instrument for vibration analysis.

DREAM was developed, verified and refined with 20 years of data that included the detailed analysis of more than 100,00 bearings in different types of rotating machines. During the past 3 years, DREAM has been tested in more than 30 commercial facilities primarily within the former USSR. The module has proven highly efficient and is currently being used in more than 50 facilities.

1. METHOD OF DIAGNOSTICS

The diagnostic method is based on the relationship of friction forces in a rolling element bearing. Friction forces are affected (modified) by the presence, type and depth of defects on the friction surfaces. With defects present, periodic variations occur in the friction forces. These variations in force are observed externally as amplitude modulation of the force excited by the random vibration of the bearing housing.

Friction forces are modulated by the periodic changes in the friction coefficient and in the pressure forces. The former contains detectable characteristics of the wear of bearing elements including spalls and cracks. The latter is related to mounting defects of new bearings.

All typical properties of the friction forces for each type of defect are detected by analysis of the envelope spectrum of the bearing vibration in a 1/3 octave frequency band of about 10 kHz. This method was proposed in the beginning of the 80s and from there it was developed and tested in many industries.

At the present time, 11 main defects of the rolling element bearings which strongly influence their service life in normal operation conditions, are detected, identified and their depth estimated by this method. The twelfth type of defect - lubrication ageing and contamination - are detected by the increase up to 20 dB of 1/3 octave high frequency vibration level compared to the mean value. The 11 defect types lead to modulation of friction forces and vibration by certain groups of frequencies which are shown in Table I. As these groups of frequencies depend on the type of load applied to the bearing, rotating frequency, and the size of bearings, in the table are represented only typical groups of frequencies that usually appear in the envelope spectra of more than 50% of the faulty bearings.

According to the depth of the main defects they are divided into three levels:

If, during diagnostics, no medium or severe defects are detected, then the typical residual service life exceeds 20% of the mean time before failure (MTBF) of the bearing in the diagnosed machine. This defines the period of the long-term prediction and the periodicity of the diagnostic measurements of the bearing unit. When medium defects appear, the predicted residual service life decreases by not less then two times. For certain types of severe defects, recommendations will be generated to replace the bearing or correct the defect as soon as possible.

The severe defect levels for all types of defects except the defect of lubrication, are defined in terms of the excess DL in harmonic lines in the envelope spectrum of the high frequency vibration (Fig.1). These uniquely define the depth of modulation of friction forces and vibration. The magnitude of severe defect levels for different types of defects, bearings and rotation frequencies are different and usually are in the limits of 10dB<L<30dB . The particular values of levels in terms of the modulation depth of vibration are calculated automatically according to the bearing data.


Fig. 1. An example of the envelope spectrum of high frequency vibration of the bearing unit.

Only one level - severe - is defined for the slip of the race in its housing, and only two levels - severe and medium - are defined for a lubrication defect.

In most practical cases, a failing bearing will have several defects. In this case, components of all the defect types are added linearly in the envelope spectrum. But, when a group of developed defects appears, there appear new criteria. These criteria are defined by nonlinear summation of the components responsible for each certain defect type. This factor defines the eleventh defect called "Defect on several bearing friction surfaces" that `is included in the group of possible defects. Identification of this defect means that there are at least two defects intended to develop rapidly in the bearing. The main criteria of such a defect is the appearance of clearly expressed components with frequencies multiple to the sum of the main frequencies of each defect type in the envelope spectrum. It can be a series of components with frequencies k(BPFO+RPM) which are due to a significant wear or spalls on the outer and inner races, and a series of harmonics with the frequencies k(BPFO+FTF), k(BPFO+BSF), k(BPFO+BPFI)=kNbRPM and others. This series should not be mixed with the series of the main components modulated by the frequencies rpm, FTF and others.

Diagnostic criteria or symptoms, mentioned above, are the most frequently used by the automatic identification of the defect type. All in all about 500 criteria for different defect combinations and different loads on the bearings are used in the algorithm for the identification of the defect type. As mentioned, there are some diagnostic criteria that can be similar for different defect combinations and loads. In some particular situations that can appear in practice, this diagnostic criteria can diminish the possibility of the exact identification of a defect type.

For example, consider a simple situation that is rather typical for the diagnostics of bearings with rotating inner races in machines with flexible couplings or when several bearings are installed in one bearing housing. The main diagnostic criteria for the wear of the inner race is a series of harmonics of the rotating frequency kRPM. However, if there is a revolution of the rotor around the outer race, then this series kRPM is a symptom for the wear not of the inner, but for the outer, race. The revolution of the rotor around the outer race with a probability of 20% is due to the defects of flexible couplings, mounting defects of shafts between rotors or due to noncoincidence (misalignment) of the axes of different bearings, mounted in one support.

Table I.

The main frequencies of defective bearings' vibration modulation

No Type of defect The main frequencies of modulation Comments
1 Revolution around outer (fixed) race RPM
2 Radial tension of bearing 2*RPM
3 Misalignment of outer race 2*BPFO
4 Wear of outer race BPFO
5 Spalls (cracks) on outer race k*BPFO; k=1,2,3...
6 Wear of inner race k*RPM; k=1,2,3... the decay of amplitudes with increase of K
7 Spalls (cracks) on inner race k*BPFI; k=1,2,3...
8 Wear of cage and rolling elements FTF or (RPM-FTF)
9 Spalls on rolling elements k*BSF; k=1,2,3...
10

Defects of several bearing surfaces

BPFO+BPFI
or BPFO+RPM
or (RPM-FTF)(Nb+1)

no BPFO-RPM
no (RPM-FTF)(Nb-1)
11 Slip of race in the mounting seat k*RPM; k=1,2,3... there are no other frequencies of modulation
12 Defects of lubrication
the increase of overall vibration level

RPM - rotor frequency;

FTF - cage frequency;

BSF - rolling element rotation frequency;

BPFO - rolling element pass frequency outer race;

BPFI - rolling element pass frequency inner race;

Nb - number of rolling elements.

In doubtful cases, the DREAM program uses an algorithm which defines the defect type as one which has the most severe influence on the service life of a bearing. In the above situation, the wear of inner race will be identified because it develops more rapidly than the wear of outer race. This is the reason why the type of the defect after changing the bearing and visual diagnostic in some cases ( about 20 % ) can differ from one detected by the program. But the prediction of the bearing's condition will be optimal from the standpoint of correctly identifying the most dangerous defects. In automatic mode of operation the program uses 20 years statistical data of the development rate of all 12 defect types.

To make a decision on whether to change a bearing or correct a defect, the operator can take into account additional information according to the rules formulated below. These rules are the result of analyzing a lot of information about the models of defect development and diagnostic criteria.

According to the probability of exact identification ( not the detection!) of the defect type, all defects can be divided into four main groups:

  1. Spalls on the inner, outer races and on rolling elements are identified exactly. The probability that an identified defect type coincides with the existing defect type is more than 90%.
  2. Wear of the outer race, wear of rolling elements and lubrication defects. The probability of the exact identification exceeds 80%.
  3. The following defects are identified with a probability of more than 70%: a misalignment of the outer race and a defect of several friction surfaces.
  4. The revolution of the rotor around the outer race, radial tension of bearing, wear of the inner race and slip of the race are identified with a probability not less than 60%.

Attention All the above mentioned probabilities are those of the correct identification of the defect type, but not the probabilities of the defect detection, which are much higher. For example, experience indicates that the probability of missing a defect is not more than 1-5% for different types of machines. Experience also indicates that the probability of a bearing failure during the predicted period is also very low, less than 10% for the default severe defect levels and can be decreased down to 3-5% by reducing the severe defect levels. The probability of an error in the defect levels is also below 20% and all these probabilities can be lowered by reducing intervals between measurements.

All groups of defects and the probabilities of their correct identification are presented in Table II. For all main defect types there are presented all other defect types that, in some cases, can have similar diagnostic criteria with the main type. The results of diagnostics of a great number of bearings in different industries have shown that, if one of the main types of defects was detected in the automatic mode of diagnostics, then there is a probability that the bearing has a different type of defect with similar diagnostic criteria. The maximum values of such probabilities are presented in Table II.

It is possible to increase the reliability of the defect-type identification by considering additional information that was not included in the algorithms of the automatic identification. The following additional information is worth considering as the most valuable:

If after the first measurement of a newly mounted bearing, a defect, usual for a faulty bearing, was detected in automatic mode, there is a great probability that this defect is a complex combination of several mounting defects. The presence of such defects after replacing a bearing, will influence the identification of additional defects during future operation of the machine.

The results of other bearings diagnostics in the same machine should be considered in the following way. Compare all spectra of the bearings mounted on one shaft. If, in several bearings, there are
Table II.

Probabilities of the correct identification of a defect type and a possible error.

No Main type of defect detected by DREAM Probability Other possible defect types with the same diagnostic criteria Error
probability
1 spalls (cracks) on outer race >90% wear of outer race <5%
strain hardening <5%
defect of surface processing + strong radial tension. <5%
2 spalls (cracks) on inner race >90% wear of inner race <5%
strain hardening <5%
defect of surface processing + strong radial tension <5%
3 spalls on rolling elements >90% nonuniform wear of rolling elements <5%
defect of surface processing + strong radial tension. <10%
4 wear of outer race >80% spalls on outer race <10%
misalignment of outer race <5%
defect of surface processing + strong radial tension. <10%
5 wear of rolling elements and cage >80% nonuniform size of rolling elements + strong radial tension <15%
autooscillations of rotor in the bearings. <10%
6 lubrication defect >80% slip of race in the seat <5%
cage destruction <10%
near-failure condition of rolling surfaces <10%
7 misalignment of outer race >70% wear of outer race <15%
spalls (strain hardening) on the outer race <10%
defect of surface processing + strong radial tension. <10%
8 several friction surfaces defect >70% defect of surface processing + strong radial tension <10%
defects of flexible coupling <15%
other units of machine defect <10%
9 revolution around outer race >60% defects of flexible coupling <20%
strong radial tension <5%
wear of inner race <10%
other units of machine defect <10%
10 strong radial tension >60% defects of flexible coupling <20%
revolution <5%
wear of outer race <10%
other units of machine defect <10%
11 wear of inner race >60% defects of flexible coupling <10%
wear, spalls on outer race <20%
slip of race <5%
other units of machine defect <10%
12 slip of race >60% defects of flexible coupling <10%
wear of inner race <15%
wear, spalls on outer race <10%
other units of machine defect <10%

harmonics of frequencies Krpm with very close amplitudes then, usually, defects of the fourth group are less probable than the defects of flexible couplings, mechanical transmissions and other units.

If peculiarities of the machine construction are taken into account, then it is possible to determine factors that can influence the bearing and must be considered. Some peculiarities of bearings' diagnostics in machines with different units that can influence the results of diagnostics are discussed below.

If an operator has to make a valid diagnosis in a certain situation for which the automatic mode of operation of the program is not intended to be used, it is necessary to detect not only the type of the defect but also the depth of it.

The levels of the severe defects when there is a possibility (but not a certainty)  of the bearing's failure are determined in the depth of modulation for eleven defect types, concerning the surfaces condition and the quality of bearing's mounting, and in dB of the increase of the high vibration level for the defect of lubrication. The level for the defect of the lubrication is defined according to the experience in the diagnostics and the standards used in different countries. It is 20 dB, but can be corrected by the user according to the economic consequences of the bearing failure. The magnitude of the levels of severe defects are varying in the limits of 8-30%  depending on the type of the defects, the rotation rate and the diameter of the bearing. With a sufficient approximation they can be considered to be equal to 15% for all defect types for the machines with the rotation rate 1000-15000 rev/min. With a decrease of the rotation frequency, the defect level decreases too. The larger a bearing's diameter, the more the difference between the levels of the inner and outer races' defect levels. The levels for the outer race are higher because the outer race is traditionally fixed and the decay of this vibration during its transfer from the outer race to the measurement point is less then during vibration transfer from the inner race or the rolling elements. The user can correct the recommended defect levels according to the peculiarities of the machine construction and his own experience.

When the defect is detected by a group of components in the envelope spectrum, which characterize this type of defect, it is recommended to use the maximum depth of the modulation of the components which evidently belong to the identified defect.

2. PROGRAM OPERATION

The program is customized for several different instruments. The process of operation depends on the type of instruments, but there are common features.

The operation begins with configuration of the data base. All the equipment is divided into Machine Groups in such a way as it is convenient for the operators working with the program. After this the Measurement Points are configured. Here the process parameters of the diagnosed equipment and the geometric parameters of the bearings are included in the data base.

The signals defined by the Measurement Points are investigated next to choose the center frequency of the 1/3 octave band pass filters and the full scale frequency band of the analyzer.

When this has been completed the operator is ready to do the diagnostics and condition prediction of the bearings in the automatic mode of operation.

Diagnostics and condition predictioin are accomplished in three steps (Fig.2):

When the envelope spectrum is displayed on the screen, the background is marked and can be edited by the operator. All the frequency lines are identified, the defects are detected and identified, the date of next measurement is defined according to the condition of the bearing and the recommendations are given for maintenance if necessary. All these data are included in the History of the Bearing. (Figs.3.)

On the basis of all the measurements the program generates a list of bearings' condition where the bearings are listed with the results of the diagnostics, recommendations for their maintenance and the date of next measurement. (Fig.4)

Fig.2a. The spectrum with marked background (RMS of overal vibration)


Fig.2b. The lines are identified automaticaly


Fig.2c. The defects are detected and identified




Fig.3. List of Bearing's History

                                                    Feb 13,  1992 19:14
                B E A R I N G'S  H I S T O R Y
                                  Machine Group : EXAMPLES OF DEFECTS
                              Measurement Point : Rolling elements

  RECOMMENDATIONS :
                   Replace the Bearing
  The date of next measurement : 12 Feb 1992

Date 1 2 3 4 5 6 7 8 9 10 11 12 DV NID
July 01 ,1991 09:05
Sep 02 ,1991 09:03
Feb 12 ,1992 14:40

I
I



I
I



M
M


S




0
0
0
  1. REVOLUTION AROUND OUTER RACE
  2. NONUNIFORM RADIAL TENSION OF THE BEARING
  3. MISALIGNMENT OF OUTER RACE
  4. WEAR OF OUTER RACE
  5. CAVITIES ON OUTER RACE
  6. WEAR OF INNER RACE
  7. CAVITIES ON INNER RACE
  8. WEAR OF BALLS, ROLLERS AND CAGE
  9. CAVITIES, SPALLINGS ON ROLLING ELEMENTS
  10. DEFECTS ON SEVERAL BEARING SURFACES
  11. SLIP OF RACE
  12. DEFECTS OF LUBRICATION
    DV - DATA VARIATIONS
    NID - NOT IDENTIFIED DEFECT
    S - Severe defect
    I - Incipient defect
    M - Medium defect
    Y - Yes



Fig.4. List of Bearings' Condition

                                                    Okt 29,  1993 11:20
         B E A R I N G S'  C O N D I T I O N
                                Machine Group : KHM-NPP
                                 Included are : T*
Measurement Point ID RB RL CM CT Date of next Measurement 1 2 3 4 5 6 7 8 9 10 11 12 DV NID
TB30D01-1
TB30D04-1
TF33D01-4
TL25D02-2
Y

Y
Y



14 Apr 1993
02 Apr 1993
12 Mar 1993
22 Mar 1993
0
21
13
0
0
0
0
0
0
0
0
0
0
7
0
0
20
0
0
16
0
0
0
0
0
0
0
0
9
0
0
8
0
0
23
0
0
0
0
10
0
0
0
0
0
0
0
3


Y
0
0
0
0
     RB  - Replace the bearing
     RL  - Replace the lubrication
     CM  - Correct the misalignment of outer race
     CT  - Correct the nonuniform radial tension
       1. REVOLUTION AROUND OUTER RACE
       2. NONUNIFORM RADIAL TENSION OF THE BEARING
       3. MISALIGNMENT OF OUTER RACE
       4. WEAR OF OUTER RACE
       5. CAVITIES ON OUTER RACE
       6. WEAR OF INNER RACE
       7. CAVITIES ON INNER RACE
       8. WEAR OF BALLS, ROLLERS AND CAGE
       9. CAVITIES, SPALLINGS ON ROLLING ELEMENTS
      10. DEFECTS ON SEVERAL BEARING SURFACES
      11. SLIP OF RACE
      12. DEFECTS OF LUBRICATION
          DV - DATA VARIATIONS
         NID - NOT IDENTIFIED DEFECT
           S - Severe defect
           I - Incipient defect
           M - Medium defect
           Y - Yes


SUMMARY

The Dream program is being used successfully in many former USSR facilities including nuclear and steam electric power plants, paper and pulp mills, metallurgical, chemical, tube plants and many others to reduce maintenance expences and ensuring safety. It is also being used in western facilities in Canada, Italy, Austria and other countries.


Return to Inteltech Enterprises home page.

Return to articles/notes page.