Vibration diagnosis
In mechanical drive technology, vibration diagnosis is understood to mean vibration analysis processes and methods that are suitable for determining the state of damage to roller bearings, gear stages, shafts and other drive elements. Alternative terms are machine diagnosis , machine diagnosis and vibration diagnosis . The term condition monitoring has also been used frequently for a number of years, although this term can also include other processes. Vibration diagnosis is based on the simple connection that all mechanical processes in machines - including irregularities in components and even damage - result in force conversion processes that are passed on in the machine and ultimately reach the housing surface. They have a periodic character, i. H. they occur again and again at fixed time intervals.
Causes of vibration
The following phenomena typically lead to characteristic vibrations (see Dresig & Fidlin (2014)):
- Imbalance leads to sinusoidal oscillations with the rotational frequency.
- Alignment errors lead to sinusoidal oscillations with multiples of the rotational frequency.
- Loose parts that end up in contact, problems with the fit and damage to shafts lead to shock pulses that repeat themselves at the rotational frequency.
- Circumferential tooth damage (all tooth flanks of a gear have similar flank shape deviations, e.g. as a result of wear) lead to harmonic harmonics for meshing tooth vibration.
- Local tooth damage (less than all tooth flanks have a flank shape deviation, e.g. due to fatigue.) Lead to the modulation of the tooth meshing vibration with the rotational frequency of the damaged gear.
- Rolling bearing damage leads to shock pulse sequences, the repetition frequency of which corresponds to the kinematic frequency of the respective rolling bearing component.
- Electrical effects on electric motors lead to sinusoidal oscillations with the mains frequency or converter frequency and its multiples.
- Sinusoidal vibration excitations with the blade or blade passing frequency are generated on fans and pumps.
- Shaft precession (often referred to as oil whirl) can occur on plain bearings. This leads to sinusoidal oscillations with about 0.43-… 0.48 times the rotational frequency.
history
The fact that vibrations can give indications of specific mechanical causes must be assumed here as having been known for a long time. This relationship is described very clearly in a publication from 1964: “In many cases, bearing damage can be recognized by the fact that the running noise changes. In this case, one should try to describe the noise, for example using information such as: evenly or swelling, periodically or unevenly, humming, whistling, singing, knocking. If you notice a regularly recurring sequence of noises, try to describe the frequency with which it occurs. ... At low speeds it has often proven useful to tap a piece of paper with a pencil in the rhythm of the noise and then count the points after a certain number of seconds. "
Measuring devices for vibration-based condition monitoring of mechanical drives have been around since the 1960s at the latest. Initially, only parameters from offline measurements on a monthly or weekly basis were monitored. It was assumed that parameters are representative for the running condition of a drive or for any damage that occurs. Often the mere increase in a parameter in the course of trend monitoring was used as an indicator of an anomaly. During this time, the first corresponding regulation was created, the VDI guideline 2056, which has since been withdrawn and which has since been replaced by ISO 10816-3.
In the 1970s and 1980s, efforts were made to develop machine diagnostics into a practical product and to use it widely in industry. For this purpose, the measured signals were also examined for their contained frequencies and the signal shape. The correspondence of the frequency of a measured oscillation with an easily calculated damage frequency provided a detailed statement about the resulting irregularity. From today's point of view, the systems of this time were characterized by poor data quality and insufficient resolution in the visualization. This was due to the state of computing technology at the time and of course was not possible in any other way at the time.
In the 1990s, computing technology experienced rapid development. This also had an effect on the machine diagnosis. It was now possible to record and process long time data sets with high resolution. As a result, machine diagnostics became more sensitive, more precise and led to higher diagnostic reliability.
Also in the 1990s there were the first successful practical approaches in the field of remote diagnosis. Digital telephony and fast data networks opened up completely new possibilities - a development that even today does not seem to be over. Continuous online monitoring is now possible, provided that automated measurement interpretation is set up.
Parameter monitoring
The description of the machine status using parameters has been in use since the middle of the last century and has been partially successful. The VDI guideline 2056, which has since been withdrawn, is probably one of the best-known applications. There, the rms value of the vibration velocity is used to describe the influence of rotating machines on the foundation. In addition, the monitoring of the effective value of the vibration acceleration is interesting. There higher frequency phenomena are visualized more clearly. Effects caused by the onset of roller bearing irregularities, on the other hand, are represented in the peak value of the vibration acceleration. There are also a number of special diagnostic parameters, mostly for rolling bearing diagnostics, e.g. B. Crest factor, K (t), Kurtosis, SPM, Spike Energy, BCU, SEE.
The monitoring of parameters provides a quick overview, in principle allows an immediate reaction and does not require any knowledge of the kinematic relationships of the drive. However, the depth and reliability of the diagnosis are limited. The type and location of damage cannot usually be clearly determined, depending on the complexity of the drive.
Frequency selective monitoring
The frequency-selective monitoring of vibrations is much more informative. All mechanical exciters in drives, regardless of whether they can be traced back to normal operation or irregularities, have a certain frequency that is usually calculable and therefore known. It does happen that two different phenomena have the same frequency and thus cannot be differentiated, but it is rarely the case. And even then you can often make a rough limitation by comparing the vibration amplitudes at different measurement locations.
Kinematic frequencies are the rotational frequencies of the shafts that result from any known speed - usually this will be the drive speed - and the respective gear ratios of the gear stages. The gear mesh frequencies are still required for gear stages, and the rollover frequencies on the sun, the planet and the ring gear for planetary stages. For rolling bearings, the rollover frequencies on the inner ring, outer ring and rolling elements as well as the cage rotation frequency are calculated from the geometric rolling bearing data, as far as these are known, or these data can be obtained from the rolling bearing manufacturers.
Analysis tools are characteristic functions, the spectrum for analyzing high-energy sinusoidal vibrations and the envelope spectrum for analyzing shock pulse-shaped vibration events. As a result, unbalance, misalignments and irregularities in the gearing are found in the spectrum, while rolling bearing damage is beginning to appear, bearing seating problems and shaft damage are in the envelope curve spectrum.
Today one is able to analyze vibrations very finely. The comparison of the measured with the calculated kinematic frequencies reveals which drive element has which damage. The conclusion to the respective polluter based on the kinematic frequency is usually clear. The type and location of the damage can thus be precisely determined.
The frequency-selective monitoring requires the collection and processing of a certain amount of data and is therefore inevitably somewhat delayed. We're talking about 60 to 120 seconds. Knowledge of the kinematic relationships of the drive and - depending on the level of automation of the respective system - a certain amount of diagnostic knowledge are absolutely essential.
Order analysis
The frequency analysis is based on synchronously acquired data. This is sufficient for drives operated at constant speed. Since vibrations caused by kinematics are not generated synchronously, but synchronously with the speed, the frequency analysis leads to the distribution of spectral components over a more or less large range of spectral lines even with minimal speed ripple. This is a hindrance for the interpretation of spectra.
In the 1990s that was less of a hindrance. At that time, drives often ran at a fairly constant speed, and spectra were formed with relatively few lines, so that speed ripples caused smearing over only a few spectral lines. Back then, diagnosing slow-moving people was considered difficult anyway. Today spectra are formed with high resolution, i.e. with 32,768 lines or more. Even very small speed ripples mean that the energy of an oscillation is distributed over so many spectral lines that an unambiguous detection is practically impossible.
Order analysis provides a solution. Vibrations are not recorded synchronously, but synchronously with the angle of rotation. The spectra formed by this signal are called order spectra or envelope curve order spectra. The reference to a reference wave must be established, which then has the order one.
In fact, the method practiced is that vibrations are recorded synchronously and at the same time the course of the speed is saved over time. The vibration signal is resampled on the basis of the recorded speed curve. This approach is favored, for example, by the Alliance for Condition Monitoring Systems for Wind Turbines
The speed or angular momentum is recorded using inductive sensors, for example on the screw heads of a coupling, or using optical sensors and reflective marks. In order to record speed changes as precisely as possible, the speed should be measured on the fastest shaft of a drive.
As a result of this process, the order spectrum and the envelope order spectrum are created in addition to the spectrum and envelope curve spectrum. In this way, both time-synchronous and rotationally synchronous phenomena can be reliably diagnosed.
Automation of vibration diagnosis
It is generally not possible to set limit values for vibration amplitudes. There is no generally valid relationship between the height of a peak in the spectrum or envelope curve spectrum and a mechanical damage quantity. This only works in individual cases with drives for which there are reliable reference values, which are then usually based on sufficiently large statistical masses. On the other hand, the time required to maintain each individual system must be small, otherwise it is not worth using.
Various approaches to automation are common today for frequency-selective vibration diagnosis. The monitoring of spectra for amplitude limit values that are defined manually is very common. These then inevitably have a subjective character. Limit values are sometimes also generated automatically by learning algorithms on the basis of reference data, if corresponding patterns are available.
A very effective way is to use a modified significance analysis. Conspicuous spectral lines are extracted fully automatically. In a second step, it is then only checked for these conspicuous spectral lines whether their frequencies correspond to kinematic damage patterns. The significance analysis is known from mathematical statistics and was originally used for the evaluation of large statistical masses. This method is inherently largely independent of the load on the drive, so it works under full load as well as in partial load operation. The results are so reliable that manual analyzes are initially not necessary.
Individual evidence
- ^ R. Wirth: Machine diagnosis on industrial gears. Part II: Signal identification in practice. In: Drive technology. Volume 37, No. 11, 1998, pp. 77-81. (maschinendiagnose.de)
- ↑ Hans Dresig, Alexander Fidlin: Vibrations of mechanical drive systems - modeling, calculation, analysis, synthesis . 3. Edition. Springer Vieweg, 2014, ISBN 978-3-642-24116-1 , doi : 10.1007 / 978-3-642-24116-1 ( google.de [accessed on January 8, 2019]).
- ↑ Image source: GfM Gesellschaft für Maschinendiagnose mbH
- ↑ To identify the cause of damage in rolling bearings . In: FAG Kugelfischer (Ed.): Rolling bearing technology . No. 3 , 1964, pp. 19-21 .
- ↑ VDI guideline 2056 Assessment standards for mechanical vibrations of machines. VDI-Verlag, October 1964.
- ↑ DIN ISO 10816-3: Evaluation of machine vibrations by measuring non-rotating parts. Part 3: Industrial machines with rated outputs over 15 kW and rated speeds between 120 min -1 and 15000 min -1 for measurements on site. Beuth-Verlag, Berlin, December 1998.
- ^ A. Sturm, R. Förster, N. Hippmann, D. Kinsky: Rolling bearing diagnostics for machines and systems. Verlag Technik, Berlin 1985.
- ↑ J Kolerus: Condition monitoring of machines. expert publishing house, 1995.
- ↑ S. Billhardt: Envelope analysis for the detection of damage-related periodic excitations in the sound and vibration signal. Dissertation. TH Zittau, 1991.
- ↑ R. Wirth: Influences on the reliability of vibration diagnosis methods on rolling bearings. Dissertation. TH Zittau, 1994.
- ↑ G. Ellmer: Study diagnostic systems. Necessity and possibility of condition analysis and monitoring of units in drive technology. Research Association for Drive Technology , Frankfurt am Main 1993.
- ↑ G. Meltzer: New methods and devices for the diagnosis of gear drives. Research report 422/98. Technical University of Dresden, 1998.
- ↑ Image source: GfM Gesellschaft für Maschinendiagnose mbH
- ↑ K. Uchtmann, R. Wirth: Machine diagnosis on variable-speed drives using order analysis. In: Drive technology. Volume 38, No. 5, 1999, pp. 44-49. (maschinendiagnose.de)
- ↑ Th. Gellermann: Requirements for Condition Monitoring Systems for Wind Turbines. AZT investigation report No. 03.01.068. Munich 2003.
- ↑ R. Wirth: Condition Monitoring in the Context of Industry 4.0. In: B. Schlecht: DMK 2017, Dresdner Maschinenelemente Kolloquium, December 12 and 13, 2017. pp. 373–384. (maschinendiagnose.de)