Vous êtes sur la page 1sur 11

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/293795081

Support Vibration Diagnostics and Limits in Gas Turbines

Conference Paper · June 2016


DOI: 10.1115/GT2016-56548

CITATIONS READS

0 875

3 authors, including:

Marcin Bielecki
General Electric
11 PUBLICATIONS   13 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Marcin Bielecki on 23 June 2016.

The user has requested enhancement of the downloaded file.


Proceedings of the ASME Turbo Expo 2016: Turbomachinery Technical Conference and Exposition
GT2016
June 13-17, 2016, Seoul, South Korea

GT2016-56548

SUPPORT VIBRATION DIAGNOSTICS AND LIMITS


IN GAS TURBINES

Marcin Bielecki
Senior Engineer Salvatore Costagliola
GE Oil & Gas - Engineering Design Center GT Engineering Manager
Warsaw, Poland GE Oil & Gas – Nuovo Pignone
Florence, Italy
Piotr Gebalski
Senior Engineer
GE Oil & Gas - Engineering Design Center
Warsaw, Poland

ABSTRACT INTRODUCTION

The paper deliberates vibration limits for non-rotating parts in Support vibration become an issue in daily operation, once
application to industrial gas turbines. As a rule such limits detected by seismic probes or velocitimeters installed on bearing
follow ISO 10816-4 or API616, although in field operation it is housing, casing or support frame [2,4,10,11,28,40].
not well known relationship between these limits and failure Conventionally they are expressed by velocity in mm or inch per
modes. In many situations, the reliability function is not well- second (mm/sec, ips) as well as optionally by integrated
defined, and more comprehensive methods of determining the deflection in mils or microns. Of course majority of the heavy
harmful effects of support vibrations are desirable. duty gas turbines have also installed non-contact probes to
measure the shaft vibrations relative to a Fluid Film Bearing
In the first part, the undertaken approach and the results are (FFB), nevertheless the support vibrations are chosen most
illustrated based on the field and theoretical experience of the frequently to set up the alarm and trip limits, while non-contact
authors about the failure modes related to alarm level of probes installed as optional and mostly reporting only alarms.
vibrations. Here several failure modes and diagnostics From reliability point of view the non-roting parts vibrations are
observations are illustrated with the examples of real-life data. most important here, because most directly determine acceptable
In the second part, a statistical approach based on correlation of or non acceptable machine conditions, if compared with the
support vs. shaft vibrations (velocity / displacement) is industrial standards. Conditional monitoring for aero-derivative
demonstrated in order to assess the risk of the bearing rub. The turbines with roller element bearings calls almost exclusively for
test data for few gas turbine models produced by General vibrations taken usually at supporting frames (optionally casing),
Electric Oil & Gas are statistically evaluated and allow to draw due to the minute shaft-to-bearing movement, what is an effect
an experimentally based transfer function between vibrations of a small bearing clearance as well as their high stiffness.
recorded by non-contact and seismic probes. Then the vibration Considering above, these vibrations are surprisingly not as
limit with objectives like bearing rub is scrutinized with aid of popular topic of the research and scientific papers as other
probabilistic tools. In the third part, the attention is given to a aspects of rotordynamics. Due to the fact that the support sensors
few examples of the support vibrations – among other gas rarely measure vibrations caused by aerodynamic forces in flow
turbine with rotors supported on flexible pedestals and path, blade-tip rub and pressure pulsation in combustor; the
baseplate. Here there is determined a transfer coefficient paper pays attention mostly to rotor-bearing-support interactions.
between baseplate and bearing vibrations for specific foundation
configurations. Based on the test data screening as well as VIBRATION LIMITS IN DESIGN CODES
analysis and case studies thereof, the conclusions about more
specific vibration limits in relation to the failure modes are While the turbomachinery condition evaluation started to take
drawn. form in the 1960’s, the first documented chart was the Rathbone

1
Chart (fig. 1) which was first published in 1939 [27]. Although duty GT [23], while for aeroderivative GT is roughly around 108
the chart plots vibration amplitude versus frequency, it was N/m, so that one magnitude is lower than for heavy duty GT. To
supposed to be applicable for seismic measures from the bearing address this problem, it would be reasonable in next edition of
cap or machine structure. Later there was proposed a modified API616, to differentiate acceptance criteria for class 3 (rigid
chart ([5], fig. 1), which could be customized to specific machine supports) and 4 (soft supports).
and conditions by so-called Service Factor Sf with example
values for gas turbines [10, 25]: Typically the support stiffness in horizontal direction is smaller
• 1.12 filtered vibrations near running speed for GT on than on vertical direction, hence the support vibrations in
flexible supports horizontal direction are likely to be higher than on vertical, if the
• 1.36 unfiltered vibrations near running speed bearing stiffness is same on both directions (e.g. for roller
• 3.5 high frequency blade vibration sources element or load-between-pad tilted pad bearings). This can one
• 3.3 unstable or self-excited vibration sources of the reasons, why seismic probes or velocitmeters are installed
• 0.7 rigid rotors on rigid foundations on vertical direction preferably. However most of the properly
Allowable limits as on chart (fig. 1) should be compared with loaded FFB show 2 to 5 times lower stiffness in horizontal
vibration measured on the bearing housing and increased by the direction than vertical. Assuming similar support stiffness in
above service factors. both directions (as frequently happens for rigid supports), the
bearing reactions on vertical direction are higher. In such case
Nowadays, the support vibration limits in application to installation of the seismic probes on vertical direction is more
industrial gas turbines (IGT) are not any more restrictive and conservative approach.
usually follow the ISO 10816-4:2009 “Mechanical vibration -
Evaluation of machine vibration by measurements on non-
rotating parts, Part 4: Gas turbine sets with fluid-film bearings”.
Also at present the API616 standard of American Petroleum
Institute for gas turbine industry is compliant with the ISO
10816-4. The IGT should have vibrations as defined for class 3
of ISO 10816 i.e. by default its supports are regarded as rigid.
Presumption that the supports are rigid is not always correct. As
a matter of fact, an overhang support configurations are typical
for compressor inlets and power turbines (i.e. low pressure),
which are preferred in a hot-end-drive IGT. This kind of
configuration, almost always should classify the support in
category “flexible” [30,36,40]. Also oil & gas industry mounts
the gas turbines on offshore platforms, what requires light
baseplates and flexible frame supports. Although excluded from
ISO 10816 part 4 the aero-derivative gas turbines (utilizing
REB) almost always require flexible support, because originally
have been designed to be mounted under the wings and still they
are by default suspended from the enclosure frame. Due to the
fact that the REB are of very high stiffness and small clearance,
the rotordynamic systems are always enhanced with an elastic
support structures and additional squeeze film dampers.

Deliberating the issue of the support flexibility in detail, a


vigilant reader can find differentiation between rigid and flexible
support in ISO 10816 part 3: “These support conditions are
determined by the relationship between the machine and
foundation flexibilities. If the lowest natural frequency of the
combined machine and support system in the direction of
measurement is higher than its main excitation frequency (this is
in most cases the rotational frequency) by at least 25 %, then the Fig. 1. Vibration criteria for rotating machinery [5]. The
support system may be considered rigid in that direction. All vibrations measured on gas turbines bearing housings should be
other support systems may be considered flexible.” The compared with the chart after multiplication by service factor Sf
screening of the IGT supports of the General Electric fleet (see (values in the text).
table 5 in the Appendix) shows that actually around 70% of the
bearing – support systems could fall in category “flexible Summary of the limits can be found in the table 1. In general, the
support”, what practically means that almost each turbine has limits for factory tests and commissioning are established based
flexible support in some form. The support stiffness, if the on expected satisfactory behavior of FFB and acceptable lateral
bearing centerline is directly at the pedestal center (i.e. no response i.e. preferred dominant synchronous response,
overhung configuration), reaches 109 N/m to 1010 N/m for heavy avoidance of fluid film instability or rub on bearings and seals.

2
Alarm - Zone C Trip - Zone D likely mix of synchronous and non-synchronous signals). While
RPM 3000 10000 3000 10000 in case of the non-harmonic response including short pulsating
Rathbone chart >9.1 >7.3 >17.3 >15.2 signals the proportion of zero-peak value to RMS can be
Blake chart incl. significantly higher [6,7,28]. This observation is well known in
GT Service Fac tors 2.9 - 8.7 2.2 - 5.1 8.7 - 24.9 5.1- 9.0
diagnostics of roller element bearings, where the crest factor CF
API 616 & ISO 10816-4 9.3 - 14.7 >14.7
can reach level of more than 10 for malfunctioning bearings with
Table1. Support vibration ranges [mm/s RMS, unfiltered] for
progressing damage. Actually the use of the crest factor in REB
IGT measured at bearing housing as recommended by
diagnostics is simple method to measure how “peaky” is the
[1,2,5,10,25,27].
signal. The idea behind is, that most of defects in REB are
causing impacts, impacts are making the signal peaky and this
VIBRATION DIAGNOSTICS
increases the ratio of the zero-peak value to RMS value. So, if
somebody is trending the crest factor, than is getting some
As mentioned in the Introduction, the bearing housing, support
warning on impulses.
frame or casing vibrations are detected by velocity transducer or
seismic probes [10,11,20,40,41,42]. The first is an electro-
Diagnostics with the aid of the crest factor is not popular
mechanical transducer, typically of inertial design, able to
measurement for FFB. It is because, that even if there is a
measure vibration in frequency range 1 to 2000 Hz. The second -
vibration problem, which manifests itself by generation of
and less popular in gas turbines - contains a piezoelectric crystal
impulses - for instance partial rub - there are more interesting
(accelerometer) with a built-in amplifier and integrator with
and direct techniques available like orbit / time-base shape, shaft
capability to measure up to 20000 Hz. Typical the vibrations are
centerline position, full spectrum and bearing metal temperature
taken with filtered signal during RMS processing and high
than observation of the crest factor. The author experience with
frequency signals are muted [11,41,28]. This approach is
FFB shows that the conditions like rub, whirl or whip can only
compliant with ISO 10816-4 or API616 requirements, which
partly be detected by support vibrations based on the proportion
states in point 4.7.5.2 "The RMS measurement is broadband
of RMS to 0-peak values. Actually only practical application of
vibration over a frequency range from 10 Hz to at least 500 Hz
CF seems be the situation, when non-contact probes are not
or six times the maximum normal operating speed, whichever is
installed and entire diagnostics must use every piece of
greater". Most of aerodynamic excitations are of high
information from the data acquisition system.
frequencies, even up to 15 kHz [38,39] i.e. usually well above
sixth harmonic. This kind of vibrations can be more effectively
CAUSES OF HIGH VIBRATIONS ON SUPPORTS
measured by pressure transducers or microphones [38,39,42].
Other sensors are installed on the compressor rear frames to
The failure modes related to non-rotating parts vibration can be
monitor pressure pulsations in low emission combustion systems
split in cause or effect phenomena – table 2. Most often causes
(“Humming Detection”). The frequency of such pulsations is
are, if neglected aeromechanic forces and airfoil rub, by some
usually not any direct proportion of rotation speed, but due to the
means related to rotor vibrations, bearing malfunction or
fact that is usually in frequency range 10 – 600 Hz they can
unexpected support modes. The presumption is here, that the
overlap rotor induced vibrations [3,28].
high bearing reactions generate forces shifting the support and
subsequently the support vibrations signal as forced response.
The measured signal is processed to the direct amplitude (i.e.
Such interaction can be more precisely detected by non-contact
zero – peak value) or the Root Mean Square (RMS) values. Here
probes, if they are installed. A comprehensive review of lateral
one should mention up-front some possible linguistic issue in
rotor behavior affecting support vibration can be found in [4, 10,
interpretation about the abbreviation DA, which is in frequent
21,28,31,41; with summary in left column of table 2].
use. For probably most people in the turbo-machinery industry it
means the Direct Amplitude i.e. zero-peak value, but for some
Assuming that the vibration forcing function originates in the
(like engineers from General Electric and Bently-Nevada) it is
rotor (e.g. unbalance), the contingency of the support vibration
the Double Amplitude i.e. peak-to-peak value. Hence discussing
depends first of all how much rotor vibration is transmitted
the DA, it is always better ask about its meaning. The
trough the bearings. Of course this effect is determined by how
relationship between zero-peak value and RMS is expressed with
well the machine is mounted to the foundation. It can change, if
aid of so-called crest factor CF, that can be an auxiliary
the turbine support weakens e.g. by crack propagation or loosens
diagnostic feature:
i.e. machine develops a “soft foot”. Other scenario can be, when
a rotor is at high eccentricity ratio. In such case the very high
V0− pk fluid-film stiffness of the bearing more effectively couples the
CF = (1)
rotor to the bearing. Thus, a misaligned machine or under
VRMS
increased radial load, the unit may experience higher than
normal stator vibrations. Shaft relative vibration, due to the
In general, when vibration are generated in linear system with increased constraint on the rotor, may decrease as more of the
harmonic response (sinus function of time domain), the RMS vibration energy is transmitted to the support. Concurrently the
value is proportional to zero-peak value divided by crest factor misalignment frequently generates super-synchronous (2X)
CF = 21/2 = ~1.414. Typical range of CF values is 1.0 to 1.8 for signal with two times higher velocity than synchronous (1X), if
supports with FFB in good conditions and as long CF is below with same amplitude of deflections. As results the misalignment
3.0, the prevailing character of vibrations is sinusoidal (most is effectively observable by the seismic probes. Next cause,
3
which affects support vibration, is a rub around the rotor (seals,
bearing or casing distortion). A heavy rub can severely constrain Recently a test performed on one gas turbine of General Electric
the rotor and reduce the synchronous (1X) shaft vibration, while has proved (results not published yet) that the turbine was able to
results in more energy transfer to the machine casing (at rub operate at vibrations as high as 56 mm/sec RMS for a longer
location, what can be as well as out of the bearing support), time period (i.e. generating at least 107 vibratory cycles to the
causing an increase in non-synchronous stator vibrations with turbine support structure). The test was set-up on fully equipped
significant share of harmonics 2X and 1/2X [4, 33,37,41], turbine and created artificial conditions of very high unbalance,
because of different stiffness in horizontal and vertical directions equivalent to the loss of few blades in medium stages of the high
for horizontal split casings or Morton effect. pressure compressor. This caused the HP rotor unbalance and
support vibrations at steady state on levels indicated above. Peak
Causes Effects vibrations 56 mm/sec RMS reached 87% of trip threshold and
- Rotor unbalance - Wear and high cyclic fatigue of the most likely approached allowable deflection on a squeeze film
- Shaft line misalignment hardware installed at casing and support
damper. The test was successful and proved full operability of
- Rub at bearing or seals - Crack propagation in the support or
- Coupling failure baseplate
the gas turbine in all loading conditions with minor issues
- Looseness of the support - Foundation distress, baseplate reported in post-inspections:
("soft foot") settlement, support stiffness reduction • Brackets of some piping attached to casings showed wear and
- Casing distortion (next speed margine reduced, created one bolt was missed.
- Rotating stall in compressor misalignment or "soft foot") • Electric connector of auxiliary hardware had broken pin
- Fluid film instability - Static structure born noise
• Accelerated damage of wrongly assembled connector
- Blade tip rub - Human exposure to vibrations
- Combustion pressure - Performance loss due to rub, open The inspection has not revealed neither signs of bearing wear nor
pulsation clearance and seal leakages oil contamination. Also the compressor after so heavy unbalance
- Lost calibration of instruments showed no signs of rubs or seals degradation. Practically
Table 2. Typical causes of excessive stationary parts vibrations observed change in turbine performance (which can be
and its likely effects. considered marginal) was accounted much more to blade fouling
than the vibration-driven damages.
FAILURE MODES RELATED TO HIGH STATOR
VIBRATIONS The failure modes caused by high vibrations are related mostly
to the high cyclic fatigue (HCF) or wear of the equipment
Independently of what are the origins of support or casing installed on the casing or turbine support. Many of the piping
vibrations, they can be harmful to the turbine systems and items of the auxiliary system in oil & gas as well as process
components – see the examples in the right column of table 2. industries are allowed for continuous field operation by one of
The reliability study [20] on aero-derivative turbines in operation below targets:
by US Navy listed that the effects of long and short term engine • API618 [16] limits to 11.3 mm/s RMS for frequency range
vibration can include: 10 – 200 Hz
• Power turbine case vane rail wear. • VKI 3842 [17] limits to 10 mm/sec RMS at 10 Hz with
• Turbine mid-frame liner clocking. ramp to 50 mm/sec for 200 Hz
• Fuel manifold Y-tube breakage. • ASME 76-PET-18 [18] limits to 9 mm/sec RMS at 10 Hz
• High pressure turbine oil tube looseness. with ramp to 39.5 mm/sec for 200 Hz
• Wear and breakage of the many engine mounted clamps and It is anticipated that these values should not cause the piping
tubes. distress by HCF [24], although they indirectly assume, that the
• Worn or lose accessory gearbox support links. excitation is displacement driven. It is correct assumption for
• Worn variable stator vane components. many items, but not the best one for components fully or
• Worn or damaged combustor components. dominantly installed to the casing like cooling piping connecting
• Higher than desired structure-borne noise. compressor with turbine or combustors. For such configurations
That kind of failures or functional issues were observed before the vibrations are virtually acceleration driven and the
the allowable limits on gas generator were lowered to: alarm acceptance criteria should be rather expresses in acceleration
18.8 mm/s RMS at 5000 RPM to 42.3 mm/sec RMS at 9000 unit. Nevertheless, the velocity criteria look as most universal.
RPM, trip 7 mils pk-pk = 178 microns pk-pk. Then further effort The study [19] of failure of auxiliary tubing show that above
in maintenance procedures and improvement in field balancing piping limits are genuinely rational for gas turbines, because root
process vastly eliminated most of above issues. Noticeable, that causes of fatigue here were externally induced vibrations via
none of these listed malfunctions was related to high rotor supports, when auxiliary piping was connected to casing surfaces
vibrations originating rub, seal degradation or loss of with velocity above ~15 mm/sec RMS for frequency in range 80
performance. The cause-and-effect like this of course can take – 150 Hz. The special vibrations limits can be also applied on
place, but most the seals on flow path in gas turbines is designed mounting points of the combustor system, which as thin walled
the way that they can accommodate large relative rotor-stator structures are vibration sensitive. Because combustion process
movement due to the thermal growth (2 mm is not surprising dynamics is a basic excitations for combustor liners, they
value), so that additional vibrations like 100 microns do not open undergo dedicated design for vibration-driven fatigue.
clearances a lot and actual change in leakage is hardly In vast majority of cases, the HCF excited by unbalance driven
observable in performance. vibrations (i.e. 1X) is at low risk and does not put in jeopardy the

4
support integrity. It is because, supports and casings are designed E.g. military ships should generate as low as possible noise in
to survive the blade out event (one or more airfoils), what calls order to avoid identification by enemy [29] or some sensitive
for meeting yield criteria in emergency load case. This allows diagnostics equipment installed in hospital and laboratories can
alongside to reach high design margins for fatigue as design require very low ground vibrations, what can be challenging, if a
shaking load is 5-10% of blade-out load, even if the pedestal power plant is less than 300 meters away [32]. Dependently on
nominal load includes dead load, turbine thermal growth and regulation, such special applications can require the operator to
torque generated on vanes and nozzles. A unpublished authors keep ground vibration (areas where workers can stay or sit) to
study for a typical fabricated support under the GT compressor less than 1 mm/sec RMS, although in most cases 2.5 mm/sec
inlet showed, that the bearing dynamic reaction could reach up to RMS will be requested, if the workers are allowed to stay up to
3.8 times of static rotor load on the bearing due to the risk of 30 min around the operating turbine. Also the WBV regulations
crack initiation and propagation. At the same time, such high may allow that horizontal vibration limits are three time less
level of the unbalance would generate extreme shaft vibration - stringent as for vertical direction.
exceeding a rub limit and triggering the trip due to the high shaft
or support vibrations. On the other hand more prone to failure Comparison of some discussed above functional limits is shown
under extreme vibrations can be oil pipe connections as reported on fig. 2. Because all these limits are indirect, they can be
in [43]. Other internal GE study on a fatigue analysis of a thin considered as basis for set up of alarm level vibrations as
walled compressor casing also revealed negligible possibility of measured on turbine by condition monitoring system. As visible
HCF. Here the HPC rotor unbalance force is transferred through in many cases the baseplate synchronous vibration is expected to
the bearing between HP compressor and turbine to the casing stay below ~2 mm/sec RMS for large GT (with operating speed
and generates its bending. The high stress areas, sensitive to 3000 – 3600 RPM) in land application and ~3.5 mm/sec RMS in
bending moment, were features like flanges, borescope port and most of offshore application, where the operating speed is 6000
bleed air extraction ports. Comparison of alternating stresses RPM and more. Also it is visible, that velocity limits due to the
with endurance limit for casted casing, demonstrates that the EHS can be more demanding for sub-synchronous signal and
casing vibration could be almost 200 mm/sec RMS to initiate less restrictive for super-synchronous signals.
high cycle fatigue cracking. Beside these observations, some
customers operating turbines in very cold conditions concern 15
14
about risk of fatigue and crack propagation in very low ambient 13
temperature. From this reason they may want to reduce trip limit 12

during cold start-up to lower level for example from 12.7 to 5.6 11
Baseplate velocity limits .

10 Whole-body vibrations - 1/2 hr exposure (ISO 2631)


mm/sec RMS like in the study presented in [8].
[mm/sec RMS]

9 Whole body vibrations - 8 hr exposure (ISO 2631)


8 Habitability on ships - working area (ISO 6954)
Foundation integrity - OEM limit based on API
FUNCTIONAL VIBRATION LIMITS 7
6
Foundation integrity - steel frame (DVN RP-C203)
Foundation integrity - concrete (civil eng. codes)
5 Auxiliary piping support design limits

Because a bearing forced response can travel also through the 4


3
baseplate and foundation to adjacent more sensitive hardware 2
(e.g. oil tanks, lube system piping, control system), most of 1

foundations have assigned specific limitations due to design 0


0 20 40 60 80 100 120 140 160 180 200
limits (concrete block, steel frame, possible interaction with soil, Frequency [Hz]

etc.; [24]). This can force the turbine operator sometime to limit Fig. 2. Comparison of some design targets applicable for
vibration on the top of baseplate to 5 – 20 microns [21 - 26], baseplate or foundation vibrations.
with usually lower value for concrete foundations and higher for
steel frames. Most of OEMs try to make baseplate and NON-SYNCHRONOUS VIBRATIONS LIMITS
foundation stiff enough to reduce vibration here to value e.g.
several time less than the standard design limit for shaft relative A common question in rotor system diagnostics is how much
vibrations per API: Non-Synchronous Vibration (NSV) is acceptable? The API 616
says that “any non-synchronous discrete vibration shall not
12000 exceed 20 % of the synchronous vibration magnitude”. However
U = 25.4 ⋅ [microns pk-pk] (2) it should be point out, that this recommendation is compelling, if
N a support is in category “rigid”, what as already mentioned is not
applicable for most of IGT. There are also published
Other functional requirements for vibrations transferred to the recommendations (e.g. [11]) about sub-synchronous magnitude
baseplate, foundation and adjacent ground can be set up by the limits for shaft and support vibrations specific to percentage of
EHS regulation for occupational safety [41]. Examples of the the synchronous (optionally total) signal magnitude. This
limitations can be a whole-body vibrations (WBV) - problematic approach is compatible with implementation of the service
in frequencies below 80 Hz [15,41] - as well as structure-born factors Sf [10,25] discussed in previous section. List of
noise in audible range, which in case of synchronous frequency recommended limits for NSV is shown in table 3. Such stringent
is difficult to insulate. The impact of vibration on workers is limits anticipate the fact that the NSV are caused either by fluid
addressed in general standards like ISO 2631 [13], ASTM film instability, which actual response and amplification factor
F1166 [12] (less restrictive than ISO 2631) or industry specific are very sensitive and erratic to the operating conditions (e.g. oil
like ISO 6954 for ships [14] – the last most restrictive of them.
5
temperature, bearing static load) or the rub i.e. journal and pad 0.0775 ⋅VRMS
U pk − pk 0.434
for 50% ( average) trendline
are worn and the major damage is approaching. =  (3)
2 ⋅ Cb  0.1861⋅VRMS
0.434
for 99% (upper ) trendline 
Relative shaft Absolute casing
Band Alarm Trip Alarm Trip Considering statistical variation of both parameters one can
Subsynchronous <1X 0.20 0.30 0.20 0.25 expect that the shaft vibration should stay below rub limit. Let
Synchronous 1X 0.85 0.85 0.85 0.85 simplify such limit to the assembly clearance reduced by the
Supersynchronous 2X 0.35 0.35 0.40 0.40 eccentricity at nominal speed. Practically and on average for
Supersynchronous 3-5X 0.30 0.30 0.50 0.50 seven bearings evaluated here, the rub conditions should here
Table 3. Summary of band alarm recommendation [10,11,25] limit the relative shaft vibration to about 77% of nominal
excluding support modes. Numbers show acceptable share of the clearance. In such case the maximum support vibration could
band in overall (direct) vibration breakdown. reach 26.4 mm/sec RMS to reach this operational alarm with
99% probability trend line for entire population as presented on
It should be pointed out, that beside the non-synchronous modes fig 3 (i.e. only 1% would run in rub). Also looking for worst case
visible on support as result of rotor-bearing interaction, it is for 1 of 7 recorded bearings and assuming the relationship for
possible, that the natural mode for support itself can generates a the 50% trend line, the rub alarm could be set at 29.7 mm/sec
NSV signal. It can be detected on a waterfall plot, as frequency RMS. These numbers are established in very simply but also
independent on rotor speed – see also Example 1. This mode is conservative statistical model. Indeed it can be expected that
most likely observable in sub-synchronous range for flexible actual signs of rub will be observed much above the
support, because designers avoid leaving such mode close to recommended trip limit per ISO 10816 – at least for normal
maximum operating speed, due to the high energy transferred to operation conditions (around 100% speed). It allows to conclude
the support system. It can be easily detected by comparison of – essentially due to the moderate correlation coefficient 0.4 of
displacement amplitudes of relative shaft and absolute support both signals - that vibration limits as in ISO10816 or API616
vibrations. If second is similar or exceeds the first, the support have not direct effect on the acceptable conditions of FFB, what
mode is most likely present and further phase angle and vector also confirms observations reported in [28].
change can confirm this probability.
25%

CORRELATION BETWEEN BEARING AND SHAFT


Max shaft relative vibrations to .

VIBRATIONS FOR FLUID FILM BEARINGS


average bearing clearance [%]

20% 99% trend


line

The basic input for a statistical assessment of the support 15% 50% trend line

vibration versus rotor-bearing conditions can be results of


mechanical running tests. 600 of such tests were performed on 10%
four types of new or refurbished gas turbines produced by
General Electric. It is important here, that they are carried out in 5%
controlled conditions i.e. oil temperature is almost same, rotor
Turbine and bearing types A-1 A-4 B-1 B-2 C-1 C-4 D-1
and bearings are in good conditions, supports are identical, kept 0%
same speed (either 100 or 105%), negligible non-synchronous 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
vibrations. Virtually the turbines are different by residual Casing vibrations [mm/sec RMS]

unbalance and bearing clearance. Tested turbines with FFB and Fig. 3. Comparison of shaft to bearing cap vibrations during the
cover power range 5 - 32 MW and speed range 4900 - 11100 mechanical running tests.
RPM – their description in the Appendix, table 5. Following data
have been collected for each bearing: 1) max direct relative shaft EXAMPLE 1: STATOR MODE INTERPRETATION BY
vibrations, 2) absolute vibrations measured on bearing cap. The BODE PLOTS
test results presented in the figure 3 on Y axis are non-
dimensional proportion of the maximum relative shaft vibration Example of a hurdle with interpretation of NSV signal on the
to assembly diametric bearing clearance i.e. Upk-pk/(2*Cb). This support vibration is illustrated on fig. 4. The Bode plots show
non-dimensional parameter was proposed in [2], in order to vibrations taken on the turbine D-1 from table 5 with the
quantify the risk of bearing rub under vibration i.e. Upk-pk/(2*Cb) elliptical bearing located on compressor inlet with overhang
= 0.0 means no risk of rub induced by vibrations, while = 1.0 configuration of the bearing support (i.e. flexible support). The
represents extreme vibration conditions where rub occurs with identified problem – from machine reliability point of view - was
100% likelihood at bearing (if neglected thermal change of fluid film instability in form of the whip, which developed in hot
clearance and hydrodynamic driven eccentricity). Red points oil conditions along with high bearing clearance or pad
represent the supports with elliptic bearings, blue ones with looseness. The whip mode here was of high amplification factors
tilting pad bearings. The plotted trend lines represent 13 (peak-peak value in range 50-80 microns) seen at ~40% of
approximation of the recorded data points with following best-fit synchronous frequency for the nominal turbine speed. Also this
functions (velocity in mm/sec RMS): mode crosses with synchronous signal during the shut down. It is
visible, that the whip mode manifests itself in the read-out of the
shaft vibrations at 5200 RPM, but is barely observable by the
seismic probes. Concurrently support mode excited by rotor
6
unbalance (synchronous 1X signal) dominates the support particular case, the rotor balancing performed with aid of seismic
vibration read-out at 4500 RPM. Note that the support vibration probes would face more difficulties than with aid of non-contact
mode here of value 0.355 in/sec 0-pk =9.0 mm/sec 0-pk is probes, if thermal conditions of the turbine vary (e.g. same
equivalent to 38 microns pk-pk and that very similar value is speed, but different power load and thermal load or too short
recorded by shaft relative vibrations. This means that the non- turbines runs to establish stable influence vector). This kind
contact probes at 4500 RPM actually measure bearing observation, where bearing reaction are more sensitive than rotor
movement, while absolute journal vibrations are almost nil. The vibration to operating conditions is common observation in the
amplification factors for this mode are: ~12 for support vibration industrial practice and among others was also reported in
signal and ~21 for shaft vibration signal. If only the support [30,45].
vibration signal is available in the diagnostics, the fluid film
instability mode would not be detected. This is other evidence, 5000 14
that the support vibration should not be trusted as reliable Compr. brg - start-up

Bearing oscialating reaction [N 0-pk] .


diagnostic tool for verification of the rotordynamic system Compr. brg. - shut down 12

Journal vibrations [microns pk-pk] .


4000
conditions. Turbine brg. - start-up
10
Turbine brg. - shut down
3000
8
Support mode Shaft vibration
6
2000

1000
Fluid film 2
instability mode
0 0
Casing vibration 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Rotor speed [RPM]

Fig. 5. Bearing reactions vs. shaft (journal) vibrations in


simulated start-up and shut down conditions. Solid lines –
journal vibrations with values on right axis, dotted lines –
bearing reactions with values on left axis.

EXAMPLE 3: BEARING CAP VS. BASEPLATE


Fig. 4. Example of Bode plot for shaft and casing vibration in
VIBRATIONS
small heavy duty GT. Start-up shown by green dots, shut down
by red dots.
As discussed in previous sections, sometimes the train operators
are supposed to limit baseplate or foundation vibration due to the
EXAMPLE 2: HIGHER VARIABILITY OF SUPPORT
specific functional limits for these components or EHS
THAN SHAFT VIBRATIONS
regulation. Usually these limits are expressed for operating
frequency of the train, so that vibrations of interest would be
One of plausible explanation of the moderate correlation
synchronous first of all and most likely generated by the rotor
between support and shaft vibrations is high sensitivity of
unbalance gained during operation e.g. from damage of
support vibrations to bearing conditions like lube oil temperature
compressor or turbine airfoil after FOD. Obviously the
or assembly clearance. A simulation of a HP rotor-bearing
relationship between the bearing cap and baseplate vibrations
system for a heavy duty gas turbine (model F-1/2 in table 5 from
baseplate depends mainly on static structure stiffness. This
the appendix) illustrates such sensitivity. Here the support is
example investigates the customer’s concern, what should be
rigid, so that the bearing reactions would be enough to represent
recommended bearing cap limits to avoid baseplate vibrations on
forced response of the support. The simulation is carried out for
example of large power turbine (coded J in the table 5 with GT
two types of conditions representing:
parameters).
• Start-up – bearing clearance as assembled, oil temperature
40 C (initial during cold start-up) The standard lateral response analysis was extended with
• Shut-down after steady state operation – turbine side bearing prediction of bearing cap and baseplate vibrations. The model –
has reduced clearance due to the thermal growth of journal similar to one described in [9] - was built for an overhang rotor
(rotor on turbine end is ventilated with air from compressor with explicit support structures, which are split in several items
discharge), oil temperature 55C (nominal setting) like: bearings, pedestals, baseplate and others of known stiffness
Results for both runs are shown on the fig. 5. There are plotted: and mass i.e. support is of multi-DOF type. It is considered that
shaft (journal) vibrations by solid lines and bearing reactions by both supports are working independently each other i.e. “cross-
dots. It is visible, that in most cases the rotor crossing its natural taking” effect is negligible [34]. The model was validated
modes has lower differences (relative, in percent) in peak shaft among other by the full speed / full load test – similar way as
vibrations than for bearing reactions. described in [33,35]. The turbine can be installed on concrete
foundations (i.e. practically as rigid) or steel frame (several time
This means that the support vibrations would show here higher lower stiffness) like in offshore application.
sensitivity to bearing conditions than the shaft vibrations. In this
7
Bearing J3 - horizontal Vbearing
Bearing J3 - vertical VTC = (4)
Relative shaft [mils pk-pk] .

Bearing J4 - vertical
Vbaseplate
Bearing J4 - horizontal

Brg J3 - vert. Brg J3 - horiz.


Brg. J4 - vert. Brg J4 - horiz.
10.0%

Shaft vibr. / nominal clearance [%] .


9.0%
50% trend line from MRT
8.0%
0 500 1000 1500 2000 2500 3000 3500 4000
Rotor speed [RPM] 7.0%

Baseplate J3 - vert. 6.0%


Casing vibrations [in/sec RMS] .

Baseplate J3 - horiz.
Baseplate J4 - vert. 5.0%
Baseplate J4 - horiz.
Bearing J3 - horiz. 4.0%
Bearing J3 - vert.
Bearing J4 - vert. 3.0%
Bearing J4 - horiz.
2.0%
1.0%
0.0%
0.00 0.05 0.10 0.15 0.20 0.25
0 500 1000 1500 2000 2500 3000 3500 4000 Bearing cap absolute vibrations [in/sec RMS]
Rotor speed [RPM]
Fig. 7. Shaft vs. bearing cap vibrations for operating speed
Fig. 6. Relative shaft (top) and absolute casing vibration range. Arrows show speed increase.
(bottom) response for the coupling unbalance. Power turbine
supported on concrete foundation. Obtained values are collected in the table 4, rows 2-5. The
transfer coefficient is in range 1.6 to 4.9 for the concrete
A demanding customer expects to keep vibrations on baseplate foundation and 1.1 to 2.9 for the steel frame. One reference here
no more than 3.5 mm/sec RMS for steel frame mounted on can be verification performed on a gas turbine installed in a
offshore platform or 2.0 mm/sec RMS for concrete foundation in navel vessel [29], which showed that steel baseplate support
onshore location. The synchronous 1X excitation comes from the practically reduces the vibrations with VTC = 1.3 - 2.5 for
unbalance due to the coupling misalignment, which is on a level configuration without anti-vibration dampers. The lower value
5 times higher than design values per API616. The synchronous for light baseplate on steel frame is not surprising, because their
analysis reveals that the Bode plots (fig. 6) for the shaft relative combined stiffness is similar or even lower than the stiffness of
and bearing absolute vibrations are much different and is in other the pedestal with the bearing housing. The last row of table 4
form presented on fig. 7. The overall observation is that, the lists the maximum response on bearing cap in operating range
relationship between support and shaft vibrations is on no more for the coupling unbalance. Basic finding is that bearing
than moderate level. Practically peaks at support vibration are vibration would be ~50% higher for the steel frame in
either natural vibration of the support itself or a forced response comparison to the concrete plate. Moreover the results show, that
generated from the unbalance force around MCOS. This the concrete foundation with lower acceptance limit 2.0 mm/sec
particular turbine configuration has a very low level of the shaft RMS makes the satisfaction the vibratory criteria more easy than
vibrations along the moderate level of the support vibrations. for the steel frames and light baseplates with higher limit at 3.5
Fig. 6 displays only 6% of nominal bearing clearance at support mm/sec RMS, because transference of the bearing vibration
vibrations around 6.3 mm/sec RMS. This would locate the peak down to the baseplate is high.
response much on the right of the reference data from fig. 3.
(also compare with the average trend line (3) copied from figure Concrete Steel
3). Concluding, the support vibration itself is a poor prognostic Foundation type
plate frame
tool of the shaft vibration – even in as simple situation like J3 Vert. 3.6 1.8
synchronous response in as design conditions Vibr. transfer
support Horiz. 2.3 1.1
coefficients
J4 Vert. 1.6 1.2
Other investigated information was a velocity transfer coefficient VTC
support Horiz. 4.9 2.9
(VTC) between the bearing and baseplate vibrations (at same Vibration limits Baseplate 2 3.5
support and direction) due to the need to control the baseplate [mm/sec RMS] Bearing 9.8 10.3
vibrations by assignment of the bearing vibration limits. The Max bearing response [mm/sec RMS] 6.22 9.63
velocity transfer coefficient VTC is here a simple proportion of Table 4. Sensitivity of the support structure vibration to the
both signals (0-peak) at nominal speed (with neglected phase lag foundation type.
between them):

8
CONCLUSIONS GG – gas generator i.e. high pressure (HP) compressor & turbine
HCF – high cycle fatigue
The paper has presented a comprehensive overview of the IGT - industrial gas turbines
support vibration diagnostics and limits in application to IGT. HPC – high pressure compressor
Such vibrations – as recorded on bearing cap, support frames, MCOS - Maximum continuous operating speed
casings, pedestals or baseplates - are still one of basic diagnostic NSV – non-synchronous vibrations e.g. 2X, 1/2X
tool with important reliability function. Discussion of specific RMS – root mean square value of the signal
cases and experience from simulations as well as field Sf – service factor
observations allows to draw following conclusions: Upk-pk - vibratory displacement, peak-to-peak value
• The moderate correlation between the sets of on shaft WBV – whole-body vibrations
(journal) and support (bearing cap) vibrations recorded in VRMS – RMS value of non-rotating parts velocity
independent measurements and simulations indicates that V0-pk – zero-peak value (amplitude) of the vibration velocity
the support vibrations is poor predictor of shaft vibrations VTC – velocity transfer coefficient
and conditions in case of fluid film bearings. The support 1X – synchronous vibrations, whose frequency is same as
vibrations as high as for zone D threshold per API616 and rotational speed
ISO 10816-4 should not mean by default, that the vibration 2X – super-synchronous signal, whose frequency is two times of
level is close to rub limit. the rotational speed
• Bearing cap vibration limits as in per API616 and ISO 1/2X – sub-synchronous signal, whose frequency is ½ of the
10816-4 are conservative and assumes among others that the rotational speed.
gas turbine rotors have rigid supports. As shown on example 2*Cb - Bearing assembly clearance (diametric)
of many gas turbines produced by General Electric and
Alstom, the most of bearing supports should be considered References
as flexible, what means higher bearing cap vibrations with 1. ISO 10816 “Mechanical vibration - Evaluation of machine
same bearing reactions transferred from the rotor to the vibration by measurements on non-rotating parts”
support. Consequently some of low frequency vibrations are 2. Standard of American Petroleum Industry API616 “Gas
not result of rotor-fluid system instability, but crossing of Turbines for the Petroleum, Chemical, and Gas Industry
the unbalance force the natural mode of the support. Services”
• Bearing cap (absolute) vibration caused for forced response 3. S. Prakash “Acoustic Based Rapid Blowout Mitigation in a
of the unbalanced rotor are more sensitive to bearing and Swirl Stabilized Combustors” Proceedings of ASME Turbo
lubrication conditions (e.g. variation in bearing clearance, 2005, paper GT2005-68589
oil viscosity etc.) than the journal relative deflection. During 4. D. E. Bently, Ch. T. Hatch “Fundamentals of Rotating
balancing the rotor based on seismic probe signal, it is more Machinery Diagnostics” by Bently Pressurized Bearing
important than for balancing with aid of non-contact probes, Press
to identify the influence vector in very same thermal 5. M. P. Blake "New Vibration Standards for Maintenance"
conditions of the turbine. Hydrocarbon Processing and Petroleum Refinery, 43, pp
• The paper illustrated the usefulness of support vibration .111- 114 , (January 1964)
measurement and limits as a prevention method for 6. S. J. Lasey “An Overview of Bearing Vibration Analysis”
reduction of failures for the equipment attached to casings. by www.maintenanceonline.co.uk
• If the bearing vibrations stays in the zone A to C per API616 7. S. Sabin “Understanding Discrepancies in Vibration
and ISO 10816-4 i.e. below 14.7 mm/sec RMS = 0.58 ips Amplitude Readings Between Different Instruments” Orbit
RMS, the excitation to standard equipment installed on the Magazine, Vol .26 No.1 2006
casings or supports is low enough that such equipment can 8. M. Blomstedt at al “Innovative Starting Procedure for
successfully operate in high cycle fatigue condition, even if Siemens SGT-600 in Cold Climate Conditions” Proceedings
its natural modes are in resonance with excitation from the of ASME Turbo Expo 2011, paper GT2011-46409
casing. 9. M. Bielecki “Non-linear Behavior and Loading Capability
• As proven by long term operation of the aero-derivative GT, of the Spring Support in Lateral Response” paper on 9th
which traditionally bear higher vibrations than heavy duty IFToMM International Conference on Rotor Dynamics
GT, an exclusively designed equipment with aid of modal or 10. A. Lifson, H.R. Simmons “Vibration Monitoring of
harmonic analysis can survive during long-term operation as Turbomachinery” Tutorial on 19th Turbomachinery
for vibrations even 50 mm/sec rms = 2 ips rms. Acceptance Symposium
of such conditions should, however, require a test validation 11. K. R. Guy “Triage for Your Database – Part 2: Setting
on component, sub-system or turbine level covering entire Meaningful Alarms” paper of Vibration Institute www.vi-
operation speed range. institute.org
12. American Standard ASTM F1166 - 07(2013) “Standard
NOMENCLATURE AND ABBREVIATIONS Practice for Human Engineering Design for Marine
Systems, Equipment, and Facilities”
CF – crest factor 13. ISO 2631 “Mechanical vibration and shock - Evaluation of
DA – direct amplitude or double amplitude human exposure to whole-body vibration”.
FFB – fluid film bearing

9
14. ISO 6954 “Mechanical vibration — Guidelines for the 34. E. Meli at al "Development and Preliminary Validation of a
measurement, reporting and evaluation of vibration with New Strategy to Model the Interaction Between Rotating
regard to habitability on passenger and merchant ships” Machines and Elastic Supporting Structure" paper on 9th
15. H. W. Paschold, A.G. Mayton “ Whole-Body Vibrations. IFToMM International Conference on Rotor Dynamics
Building Awarness in SH&E” in Professional Safety Apr 35. G. Vannini at al "Calibrating a Large Compressor’s
2011 Rotordynamic Model: Method and Application" published
16. Standard of American Petroleum Industry API618 at 7th IFToMM-Conference on Rotor Dynamics.
“Reciprocating Compressors for Petroleum, Chemical, and 36. A.J. Volponi "Gas Turbine Engine Health Management -
Gas Industry Services” Past, Present and Future Trends" Proceedings of ASME
17. German Standard VKI 3842 “Vibrations in piping systems” Turbo Expo 2013, paper GT2013-96026
18. J. C. Wachel, C.L. Bates “Techniques for Controlling 37. G. Jacquet-Richardet at al "Rotor to stator contacts in
Piping Vibration and Failures” ASME Paper 76-PET-18 turbomachines - Review and application" in Mechanical
19. K.C.Upreti, Dwaipayan Banerjee “Preventing Small Bore Systems and Signal Processing 40 (2013) 401–420
Piping / Tubing Failures in Reciprocating Compressors” 38. K. Mathioudakis at al "Casing Vibration and Gas Turbine
paper GT2013-94535 on ASME Turbo Operating Conditions" Proceedings of ASME Turbo Expo
20. B. D. Thompson, B. Wainscott, J. Mulcahy “Understanding 1989, paper 89-GT-78
& Minimizing the Vibratory Response of LM2500 Gas 39. E. Loukis “Combination of Different Unsteady Quantity
Turbine” ASME Turbo paper 2000-GT-600 Measurements for Gas Turbine Blade Fault Diagnosis”
21. M. I. Friswell “Dynamics of Rotating Machines” Proceedings of ASME Turbo Expo 1991, paper 91-GT-201
Cambridge University Press 40. M. Maalouf “Gas Turbine Vibration Monitoring” published
22. Varanasi Rama Rao “Machine Foundations in Oil and Gas in ORBIT Journal [Vol.25 No.1 2005]
Industry” 41. “Vibration Monitoring by Bruel&Kjear”
23. H. E. Wettstein “Consequential Damage of Blade Failures in http://www.bksv.com/doc/br0094.pdf
Single Shaft Gas Turbines” Proceedings of ASME Turbo 42. Product specifications on Bently-Nevada web page
2012, paper GT2013-94221 www.gemeasurement.com
24. Norwegian Standard DNV-RP-C203 “Fatigue Design of 43. S. S. Florjanic “Mechanical Behavior of an Industrial Gas
Offshore Steel Structures” Turbine under Fault Conditions” Inter. Journal of Rotating
25. “Foundations for Dynamic Equipment” American Concrete Machinery 1999, Vol. 5, No. 3, pp 219-230
Institute report ACI 351.3R-04
26. P. Nawrotzki, G. Huffmann, T. Uzungolu “Static and APPENDIX.
Dynamic Analysis of Concrete Turbine Foundations”
Structural Engineering International 2/2008 Gas turbine models of General Electric discussed in the
27. T.C. Rathbone “Vibration tolerances”, Power Plant paper
Engineering, 1939
28. M. Maalouf , T. Eldridge “Gas Turbine and Driven Bearing Support Speed
Model Rotor
Machinery Management and Diagnostics” Proceedings of type stiffness [RPM]
ASME Turbo 2011, paper GT2011-45759 A-1 HP rotor Tilting pad Flexible 11000
29. Hyung-Suk Han, Ki-Yong Choi “Analysis of Reducing A-4 LP rotor Tilting pad Rigid 8300
Tonal Noise of the Gas Turbine Generator in order to B-1 Single shaft Lemon Flexible 5100
Reduce Underwater Radiated Noise of a Naval Vessel” B-2 Single shaft Lemon Flexible 5100
Transactions of the Korean Society for Noise and Vibration C-1 HP rotor Lemon Flexible 5100
Engineering 01/2009 C-4 LP rotor Tilting pad Rigid 4900
30. M. Bielecki, S. Costagliola “Lateral Rotordynamic D-1 HP rotor Lemon Flexible 11100
Response of Low Pressure Gas Turbine Rotors” E-2 HP rotor REB Flexible 9900
Proceedings of the ASME Turbo Expo 2013, paper E-3 HP rotor REB Flexible 9900
GT2013-94265 E-4 LP rotor Tilting pad Flexible 6100
31. T.C. Allison, H.R. Simmons “ Application of an Interactive F-1 HP rotor Tilting pad Flexible 7450
Balancing Procedure for Gas Turbines and other F-2 HP rotor Tilting pad Flexible 7450
Turbomachinery” Proceedings of ASME Turbo Expo 2009, J-3 LP rotor Tilting pad Flexible 3600
paper GT2009-60134 J-4 LP rotor Tilting pad Rigid 3600
32. Ch. Himmel “Two Healthcare Buildings Get the Power, the Table 5. Basic information about discussed GT models.
Noise, and the Vibration from a Gas Turbine Generator
System”
http://acoustics.org/pressroom/httpdocs/159th/himmel.htm
33. J.J. Yu “The Necessity of a Third Balance Plane for
Generator Rotor Field Balancing” Proceedings of the ASME
Turbo Expo 2014, paper GT2014-25706

10

View publication stats

Vous aimerez peut-être aussi