A.V. Barkov, N.A. Barkova, A.Y. Azovtsev

Association VAST, Ltd.,
140, Stachek, St. Petersburg, 198207, Russia,
tel +7 (812) 1587513, fax +7 (812) 324 6547

Presented on the COMADEM 2003 international conference, Vaxjo, Sweden.

Copyright () 2003, A.V. Barkov, N.A. Barkova, A.Y. Azovtsev , all rights reserved


The experience shows that the maintenance of the machines according to their real condition is possible only when the dangerous defects are detected on their incipient stage of development. The vibration diagnostics solves this problem most simple and efficient for machines with a single mode of operation, such as fans or pumps. A long term machine condition prediction for such machines is about six months and the periodicity of measurements is 2-3 months. One operator can make a diagnosis and condition prediction for 20-30 machines of average power.

The diagnostics for transport machines is more complicated by the following reasons:

  • often it is impossible to diagnose the machine during its motion on specified mode of operation;
  • often there is no information about non-observance of the specified operation conditions between the measurements;
  • not typical presentation of defect symptoms in the machine vibration signals during diagnostics on special stands and modes of operation;
  • the compactness of most machines that does not enable to control the vibration in optimal points.

Nowadays the detailed diagnostics problem with the described peculiarities is solved for two typical cases. The first one is the diagnostics of the wheel-pairs and gearboxes of railroad locomotives, and the second one the diagnostics of aviation and marine gas turbine motors used in gas industry. The case histories of machine condition diagnostics in such cases are discussed.


For more than 10 years, "Association VAST" has designed and introduced systems for machinery preventive vibration diagnostics that allow condition based maintenance. While the main task of traditional methods is to define the reasons for machinery and equipment vibration increases registered by the condition monitoring systems, for condition based maintenance it is vitally important to detect all possible types of defects in the incipient stage of their development and provide reliable prediction of safe operation period for the machine. This will allow being well prepared for the needed repair work in a timely manner. Individual algorithms and/or means of different incipient defect types detection are used in preventive diagnostics. As incipient defects do not produce oscillating forces that can produce significant vibration increases, preventive diagnostics is mostly oriented to analysis of high frequency vibrations at measurement points located only on the diagnosed machine units.

Most often, preventive diagnostic systems are applied for a single mode machine's condition diagnostics and mostly for fan and pump equipment where the bearing shields are easily accessible for the installation of vibration sensors during measurements and the load on the machine units during measurements is the same as the load when the defects occurred and developed. Intermittent diagnostics of such machines are carried out for just 5 to 10 times between the repairs but the intervals between single measurements should not exceed 6 months. In this case, the probability for missing a dangerous defect usually is less than one percent.

The vibration diagnostics of machines in transport industries has its own peculiarities due to one of the following reasons or groups of reasons:

  • The presence of shock loads in some machines, for example, in reciprocating machines. The problem is that high shock loads totally cover the tiny vibration components used for the detection of the most incipient defects.
  • For multimode machines, typically it is impossible to conduct diagnostic measurements for all loads and modes of operation when the defects developed.;
  • Most of the optimal measurement locations are not accessible for diagnostic measurements in compact and high energy equipment.
  • A common situation is the absence of any data regarding extreme working conditions of the equipment between diagnostic measurements.

One of many groups of rotating equipment on the railway transport that requires adaptation of the vibration diagnostics systems are wheel-motor cartridges of locomotives and railway motorcars. The wheel-motor cartridge diagnostics should be done with the locomotive that lifted on special jigs for maintenance. Their own motors, powered from an external power supply, rotate the wheels. Such a mode of operation results in significant change of all static and dynamic loads on motor, gearbox and wheel pairs compared to the normal mode of operation. This brings up a high probability for changes of a number of defects that influence the machine and its components vibration characteristics.

The solution used by the Association VAST in the vibration diagnostics system, Vector-2000, for the railway transport is based on parallel and independent use of a number of the most effective methods for the defect detection by the analysis of high-frequency vibration, ultrasonic vibration and low frequency vibration of the bearing units. The ultrasonic vibration is most efficient in detection of lubrication problems, including the appearance of metal particles in lubrication as well as the detection of stationary rolling surfaces. At the same time, there are nearly no changes in ultrasonic vibration pattern following the development of defects on rotating friction surfaces. Parameters of the high frequency vibration envelope are much more sensitive to the defects of the rolling surfaces of the bearings as well as gears surfaces and gears interaction. At the same time, its sensitivity to the severe defects can be significantly lower than to the incipient ones. This may result in the increase of the probability of missed defects in cases where the machine undergoes extreme loads and working conditions between planned diagnostic measurements which can cause the defect development rate to be much higher that the typical. The low frequency vibration parameters that are used in most common condition monitoring systems and diagnostic systems produce significant diagnostic changes only when severe defects are present in the machine. These low frequency parameters are used to decrease the probability of missing dangerous defects of the motor-wheel cartridge of the locomotive.

Thus, condition diagnostic systems for railways use the following measurement data: intermittent measured autospectra of low- and medium frequency vibration (up to 1-3 kHz), envelope spectra of 5-10 kHz vibration and magnitudes of crest factor and RMS of ultrasonic vibration of 15-25 kHz. Nowadays, such systems of vibration diagnostics are being used in a number of locomotive depots in Russia and Belorussia. Diagnostics of each motor-wheel cartridge is done at rotating speeds of about 3-5 Hz. The interval between measurements and diagnostics for each unit is 1-3 months, depending on the results of the last diagnostics. Both diagnostics and condition prediction are done in automatic mode with no intervention of the operator. The operator's responsibility is to make a decision using the results of diagnostics to continue operation or to change lubrication or to make repair of the motor-wheel cartridge.

Figure 1 presents an example of data measured on the box bearing of the locomotive together with the results of automatic diagnostics and condition prediction in the form of the next measurement date. The safe operation period is defined as the period until the date of the next measurement.
Fig. 1 Envelope and auto-spectra measured on a box bearing of the locomotive together with the results of automatic condition diagnostics and recommendations on repair.

Aviation and sea gas turbine motors for transport form a different group of rotating equipment possessing their own specific peculiarities in vibration diagnostics. In these compact and high-speed rotating machines, there is no way to mount vibration sensors on most of bearing shields. As a result, it is not possible to use any high frequency vibration measurements for the diagnostics, neither envelope spectra nor ultrasonic data. The traditional methods of diagnostics by low frequency autospectra also provide no reliable results. The reason for this is that the most diagnostic information can be found in blade-pass and other high frequency vibration components and, besides that, in the autospectra measured at quite a distance from the vibration origin e.g. on the turbine suspension there are so many vibration components of different origin that it is really not possible to classify all of them.

Several years were spent in finding a solution for such problems. Finally, a means of vibration measurements capable of measuring narrow band spectra from a few Hz up to 25 kHz and automatic analysis modules were included in the diagnostic system. The spectra are measured on the turbine supports and the analysis modules are capable of identifying the vibration components produced by a group of up to three asynchronously rotating shafts with an unlimited number of working wheels on each of them. It was an especially difficult task to identify each vibration components produced by shafts with very close rotation speeds and especially when the speeds are known with very low accuracy.

Besides identification of several hundred vibration components in each measurement point, the problem of automatic formation of a diagnostic parameters standard was solved. The standard parameters for a non-defective machine are formed from the basis of spectra measurements on a group of similar machines among which, some defective machines may be present. These studies resulted in the introduction of an automatic diagnostic system for gas turbines with rolling element bearings. This system was tested on several dozens of aviation and sea based turbines used in Russian GAZPROM company in the gas pump stations on gas pipelines..

Figure 2 presents vibration autospectra measured on one of the supports of the GPU-10 gas pumping machine with a sea gas turbine engine together with the results of automatic diagnostics of the machine by the group of similar machines. The designed software is unique in the world and is used in the condition diagnostics systems of Association VAST based on the DREAM software. But the capabilities of the software are much wider; it can be used as part of the on-line condition monitoring system for gas turbines if the systems are capable of measurements of vibration spectra in the frequency band from 5 Hz to 25,000 Hz with a frequency resolution of 0.5-1 Hz.
Fig. 2 Autospectra measured on one of the supports of the GPU-10 gas pumping machine together with the results of automatic condition diagnostics and safe operation period prediction.

Naturally, the limitations on the use of high frequency vibration measurements of rolling element bearings decreases the possibility of incipient defect detection. At the same time, gas turbine engines that possess high power have compact dimensions such that even a small defect may significantly influence the vibration pattern of the machine as a whole. For this reason, it has become possible to use rather long intervals between intermittent diagnostic measurements for gas turbines as well. Thus for engines that have a service life of about 10000 hours, at first the diagnostic measurements should be done every 500 hours of operation. Continuously operating turbines have even longer service lives and can be diagnosed with intervals of about 1-2 months.


  • The problem of condition diagnostics of the machinery in transport industry is more complicated compared to the continuously working equipment in other industries.
  • The main problems in the design of the automatic diagnostic systems are not the ones connected to the choice of diagnostic symptoms and parameters but those connected to the choice of measurement points and operating modes of the machines during measurements.
  • In many cases, due to the extremely complicated composition of the vibration under investigation, the only effective systems are the ones that provide automatic analysis of the measurement data and automatic diagnostics and condition prediction for the diagnosed machinery.

Copyright () 2003, A.V. Barkov, N.A. Barkova, A.Y. Azovtsev

Back to articles and publications