Application note on Glitches:

"GLITCH"
Definition, Sources, and
Methods of Correcting

The term "GLITCH" is Bently Nevada Corporation's nomenclature

for describing all forms of vibration measurement error from an
observed shaft surface. This does not include other possible
noise and signal error sources, such as electrical line noise,
monitor problems, or problems associated with casing mounted
transducers. Casing mounted transducers, such as velocity coils
or accelerometers are also subject to signal error sources in the
form of cross axis vibrations, improper mountings, line noise as
well as several others particular to each type of transducer;
however, glitch deals with observed shaft surface anomalies
only. In order to clarify the sources of these vibration
measurement errors, sometimes referred to as runout, the
following discussion shall focus on the two categories of
"GLITCH", mechanical and electrical.

Sources of Glitch Mechanically Induced Runout:

Nonconcentric Surfaces/Bows
A shaft surface which has been improperly machined (egg shaped
or nonconcentric) will yield a sinusoidal displacement signal with a
frequency coincident with the rotational speed of the rotating
element. A second condition on a rotating element which will yield
the same apparent dynamic motion condition is manifest when the
 rotating element has been physically bent or bowed.

Sources of improperly machined surfaces can usually be traced to

a worn or defective set of bearings on the machine used for final
machining or grinding or a worn-out set of lathe centers on a lathe.
Bows in a rotating element are typically introduced due to
improper handling of the rotor during its manufacturing cycle. This
may be the result of a sudden or jarring load applied or due to
long-term storage of the element with improper supports for the
rotor. In the case of the latter, an improperly supported rotor may
introduce a permanent sag or bow due to gravity forces.

Bows introduced into a rotor from asymmetrical heating or

improper cooling are generally referred to as thermal bows, and
can usually be removed by placing the rotor on turning gear, and
or slowly starting and heat soaking the machine at low speeds.

Surface Irregularities or The presence of surface imperfections or irregularities will yield a

Imperfections runout condition as observed by the proximity transducer. The
surface imperfections discussed herein are in the form of
scratches, dents, burrs, etc.

In general, surface irregularities are created due to improper

handling of the rotor during the manufacturing cycle. Care should
be taken to protect the shaft surface to be used for dynamic
motion measurements. In essence, these surface areas should be
given the same protective measures used to protect a bearing
journal surface. Crane lifts should be made with cables attached to
shaft areas away from the probe measurement surfaces. Support
fixtures for storage of rotors should not introduce surface
scratches, dents, etc.

Occasionally, surface irregularities are introduced via a machine

cutting tool. If the tool is dull or the feed is too rapid, some tool
chatter may occur which can introduce small ripples in the shaft
surface.

Electrically Induced In general, proximity transducers will operate satisfactorily in the

Runout: presence of magnetic field, as long as the field is uniform or
symmetrical and not localized to a particular location on the rotor.
If a particular area or zone of the shaft surface is highly magnetic
Residual Magnetism and the remaining surface is nonmagnetic or at a much lower
value, an electrical runout condition will be manifest. This is due to
the resultant change in sensitivity on the shaft surface to the
 applied field from the proximity transducers.

Residual magnetism runout problems are seldom encountered.

However, various physical inspection techniques employed during
the manufacturing cycle, such as the use of magnetic chucks, can
introduce residual magnetism problems. The most common
inspection technique, where residual magnetism may be a
byproduct, is a Magnetic Particle inspection (Magnaflux®) to
check for cracks on castings, or weldments, or after other
manufacturing cycles. The magnetic field introduced to the rotor
for this inspection should be neutralized after the inspection
program is completed. This is done with the Magnaflux machine
and involves continuously reversing the polarity and passing a
current through the rotor at continuously decreasing amperes. If
done properly, this procedure should neutralize the magnetic
properties of the rotor. In some cases a proper polarity reversal is
not performed, and residual magnetism is produced.

Precipitation Hardening 17-4 pH steel nearly always presents an electrical runout problem.
Some form of material replacement (shrink a collar, overspray a
material) is normally required to eliminate glitch.

Other pH steels, such as 15-5 pH, seems less prone to glitch, but
any pH steel may cause difficulties.

Metallurgical Segregation The scale factor yielded from a proximity transducer is dependent
upon several variables. One variable involves the specific metal or
metallurgy it is to observe. Typically, steel alloys for shaft materials
contain a variety of alloying agents. In general, the final
metallurgical composition of these alloys is a homogenous
mixture. On some rotating elements microscopic segregation of
the steel alloys may occur. Since the proximity transducer
responds with different voltage outputs depending upon specific
metals, the lack of a homogenous metallurgical composition
around the circumference of a shaft may give rise to varying
electrical outputs.

Residual Stress During the normal manufacturing cycle of rotors, a variety of

Concentrations machining and surface treatment processes can introduce small
amounts of localized stress concentrations. Although these stress
areas do not adversely affect the mechanical properties of the
rotor, they may give rise to an electrical runout from the proximity
 transducers. Since one of the variables affecting voltage output
from the transducer is the resistivity of the observed shaft surface,
any deviation of the resistance around the circumference of the
shaft (due to varying stress levels) will produce a voltage change.

Measurement of Glitch The observation of mechanical runout at a rotor speed below

which dynamic vibration is eliminated (typically less than 10% of
the rotor operating speed) on an oscilloscope will yield a
sinusoidal dynamic waveform for non-concentric surfaces or shaft
bows. Surface irregularities or imperfections will appear on an
oscilloscope as sharp voltage spikes superimposed on the
dynamic wave form.

Nonconcentric surfaces or shaft bows may also be measured by a

dial indicator mounted in the probe area of the shaft which in turn
is mounted at the bearing journals, in vee blocks, or roller
bearings. The circumference of the probe area is marked off in 36
positions (10 degree intervals) with zero in line axially with the
thrust collar keyway. Dial gauge readings are recorded at each
indicated position. Two sets of readings are taken, approximately
half an inch apart axially, one on each side of the probe position
centerline. The two sets of readings are averaged to give a record
of mechanical runout. American Petroleum Institute (API)
Standard 670 recommends that "the combined total electrical and
mechanical runout does not exceed 25 percent of the maximum
allowed peak to peak vibration amplitude or 0.25 mil (6
micrometers), whichever is greater." The shaft surface finish
should be from 16 to 32 micro inches (0.4 to 0.8 micrometers) root
mean square, also per API 670.

The observation of a residual magnetism runout condition on an

oscilloscope can yield a sinusoidal motion indication. However, the
sine wave will be distorted and to some extent tending toward a
square wave. A final check for residual magnetism embraces the
use of a small handheld field strength indicator manufactured by
Magnaflux Corp. Holding this meter at the shaft surface and hand
turning the rotor will confirm the presence or absence of magnetic
fields of less than 2 gauss with variations less than 1 gauss.

An oscilloscope observation of metallurgical segregation will

typically indicate a somewhat sinusoidal waveform with high
voltage, high frequency spikes superimposed on the waveform.

Observation of residual stress concentrations on an oscilloscope

 will yield a sinusoidal waveform with high voltage, high frequency
spikes superimposed on the waveform.

It should be noted that the oscilloscope waveform in all the above

cases may also be very irregular, depending on the amount of
other shaft surface anomalies.

If desired, the electrical runout can be determined by subtraction

of mechanical runout from the total runout.

Glitch Reduction Various methods of reducing glitch are available and have been
successfully used by firms. It is not possible to define which
method is best because each can achieve the desired result.
However, it is possible to narrow the choice of methods when they
are considered on a cost and time basis.

It is also notable that proper material selection, heat treating, and

allow control can have a large effect on the runout condition of a
rotor. If rotors are to be replaced or rebuilt, it is far more cost-
effective to detect and correct glitch at the earliest stages of
machine assembly.

(a) Degaussing- Residual magnetism in a shaft, caused as

a result of magnetic particle crack detection or by working in
a magnetic field, can produce very serious electrical runout.
It is therefore prudent to measure residual magnetism in the
probe area of every shaft before attempting glitch removal.
A localized residual magnetism of field strength 5 gauss on
a rolling shaft can give an electrical runout in the order of
0.5 mil. Thus any shaft which exhibits residual magnetism
in excess of 2.0 gauss, or variations greater than 1 gauss,
should be degaussed. This is generally not a complete
glitch removal process on its own, but it does help to ensure
that glitch readings do not change as a result of a shaft
losing residual magnetism in service.

(b) Diamond Burnishing - This API recommended method

has a high success rate of reducing glitch to within
acceptable limits. The probe area is rolled under a diamond
burnishing tool to work a shaft surface to a uniform finish. In
effect, this procedure produces an even work-hardened
surface which requires no additional treatment. This method
is undoubtedly the easiest with very little skill required.

(c) Further Machining - Should a probe area be outside

 acceptable limits, the target area can be reground, and
should be degaussed following grinding. This can be a hit
and miss method that could lead to even more
unacceptable results.

(d) Polishing or Stoning - Similar process to (c) but not

quite so drastic. However, the problem of making matters
worse still exists.

(e) Sleeving - Shrinking a sleeve onto the shaft has been

used, but it is rather an expensive way of producing results
as unpredictable as (c) and (d) above. As with both of those
methods, further treatment may well be necessary.

(f) Plasma Spray Finishing - Although not recommended

by API, metal coating has been used effectively as a
solution for stubborn runout problems. Aluminum - Nickel
flame spray coatings are currently being used very
successfully in industry to reduce glitch problems, often to
less than 1/2 mil pp.

The Metco® process consists of machining a 1-1/2 inch

wide, 0.060 deep grove around the shaft. The groove is
then grit blasted, and coated with a bond coat of Metco 447
to about 0.010 inch thick. The remainder of the grove is
then flame sprayed filled with Metco 52C aluminum silicon
composite and machined flat with the surface of the shaft.
Since the electrical properties are different than 4140 steel,
a specially calibrated Proximitor® must be used at these
locations.

Protection of Probe Areas The removal of glitch by one or a combination of the treatments
discussed above is essential for customer acceptance of levels of
recorded vibration. Having treated the probe areas, it is
essential to protect these areas to prevent corrosion damage,
scratching and indiscriminate "cleaning up". During production
and testing, this protection is afforded by the care taken by skilled
craftsmen who appreciate the sensitive nature of this surface.

Between acceptance testing and startup much less care can be

exercised. There have been many field reports of probe areas
being damaged by rust, scratches or dents. Cleaning up the
damage by stoning or polishing may give the desired appearance,
 but the original glitch recordings will have been completely altered
and undoubtedly taken beyond acceptable limits. Glitch removal
then has to be carried out in the field by selective micropeening,
and considering the adverse conditions, the results are invariably
inferior to those which can be achieved in the shop. It is therefore
recommended that after glitch treatment and recordings are
taken in the shop, the probe areas be given a coating of non-
metallic epoxy resin which can remain in position for the life
of the machine. This coating will not affect probe readings but will
protect the probe area from corrosion and all minor mechanical
damage.

Compensation Obviously, shaft treatment to remove the source of runout is the

most desirable procedure. If this is done, there is no reason to
have to "account for'' runout in subsequent vibration signals.

However, the occasional shaft material or forging may not respond

well to the standard shaft treatment methods. The shaft may have
a "permanent" bow, or it may be impossible or impractical to treat
the shaft surface before a time when vibration data is required on
a given machine. If it is impractical to treat the shaft surface or
remove the shaft bow, an electronic method may be used. The
following is an explanation and discussion of this application.

Vector Nulling - Digital Also called slow roll compensation, this system is an integral part
Vector Filter of the Digital Vector Filter. It provides a means for nulling a slow
roll vector. It should be noted that the nulling operation is a true
vector subtraction (phase and amplitude) and not merely a voltage
suppression circuit. The nulling circuit operates on the filtered
vibration waveform (the vector information in the DVF). Since the
filter employed in the DVF-3 is tuned to the rotational (1X rpm)
frequency of the rotor, the nulling circuit eliminates that portion of
shaft runout which is coincident with the rotational frequency.
Typically, it is used to eliminate a 1X component of runout such as
a bow in the shaft or a nonconcentric (egg-shaped) shaft condition
at the probe measurement plane. All higher orders of runout (non
1X components, such as scratches, metallurgical irregularities,
etc.) are eliminated through the filter action of the DVF-3.

Once the initial slow-roll vector has been nulled, it is automatically

subtracted from all future dynamic signals. This system provides
the means for properly examining the mechanical response
and impedance of a system, definition of the balance
resonances (critical speeds), and amplification factors, over
 the operating speed range. Vector nulling also allows for the
compensation of the residual unbalance vector after a balance
resonance, and for observation of a higher balance resonance
response.

It is possible, and even probable on larger machinery, that nominal

axial position changes and differential expansion up to running
speed will cause a vibration probe to observe a "new'' lateral
location on the shaft. When considering the overall runout pattern,
this "new'' shaft location may be significantly different than the
overall pattern observed with the machine at slow-roll.

Vector nulling does not, however, deal with the overall runout
pattern; because of the filter in the system, only the 1X runout
vector is considered. The once-per-turn runout vector is not likely
to change from slow-roll to operating speed and temperature. In
this regard, vector nulling offers a distinct advantage over any
other type of digital runout compensation. Vector nulling also
offers the capability of nulling the residual vector of shaft motion
after passing through a resonant speed region to observe the
action of the next higher resonance when Bode plots (amplitude
vs. rpm and phase vs. rpm) are made.

Recommendations The above considerations lead to the following conclusions and

recommendations by Bently Nevada:
1. Glitch often can be controlled at its source (the shaft) to a
level acceptable for monitoring purposes and in most cases
to levels usable for machine acceptance testing and
diagnostic purposes. Every reasonable attempt should be
made to correct the runout problem at its source.

2. The use of electronic runout compensation for

continuous machine monitoring should be avoided except
in rare cases (e.g. a damaged shaft that cannot be
corrected until the next turnaround). Bently Nevada does
not recommend the use of electronic runout
compensation in a vibration monitor system wired for
automatic machine shutdown.

3. Nulling - compensation for an initial 1X vector - can be

accomplished with the Digital Vector Filter 3.

4. When reproducing vibration data from magnetic tape,

special care should be taken to ensure proper
 synchronization of the signal. Most tape recorders provide a
function whereby one channel can be dedicated as a
synchronizing signal for tape flutter compensation.

5. When runout compensation is used, it should be used as

a "last resort''. In all cases, both the original transducer
signal and the compensated signal must be available for
observation on external instruments.

References "New Techniques in overcoming Electrical Runout" by Dale W.

Beebe, Turbodyne Corporation, Hydrocarbon Processing, August
1976.

Electrical Runout and Eddy Current Displacement Proximity

Transducers, by Biggs, David H., ASME Paper, September 1975.
(Bently Nevada Literature No. L0360)

Elliot-Wiedeke paper

API 670, Second edition, Section 4.1.2: Machine Shaft

Requirements for Electrical and Mechanical Runout