Experimental and Numerical Investigation of Rotating Bladed Disk Forced Response Using Under-Platform Friction Dampers

Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Proceedings of GT2007

ASME Turbo Expo 2007: Power for Land, Sea and Air
May 14-17, 2007, Montreal, Canada

GT2007-27307

EXPERIMENTAL AND NUMERICAL INVESTIGATION OF ROTATING BLADED DISK


FORCED RESPONSE USING UNDER-PLATFORM FRICTION DAMPERS
Ibrahim A Sever Evgeny E Petrov David J Ewins

Rolls-Royce plc Imperial College London Imperial College London


PO Box 31, Derby, DE24 8BJ, Mechanical Engineering Dept., Mechanical Engineering Dept.,
UK London, SW7 2AZ, UK London, SW7 2AZ, UK
Tel: +44 1332 2 47214 Tel: +44 20 7594 7068 Tel: +44 20 7594 7068
Fax: +44 1332 2 46746 Fax: +44 20 7584 1560 Fax: +44 20 7584 1560
ibrahim.sever@rolls-royce.com e.petrov@imperial.ac.uk d.ewins@imperial.ac.uk

ABSTRACT many years. There have been a number of experimental studies


In this paper we present a methodology and results from an aimed at understanding of characteristics of friction damping as
experimental investigation of forced vibration response for a well as its effects on system response. An experimental
bladed disk with fitted under-platform ‘cottage-roof’ friction evaluation of effectiveness of different damping mechanisms
dampers, together with the corresponding numerical on a rotating bladed disk was given by Jaiswal and Bhave [1].
predictions. A carefully-designed and constructed rotating test Here two groups of 7 blades, each covering 90 degree sectors
rig is used to make precise measurements which involve only of the bladed disk circumference were considered. Several
the phenomena of interest. For this purpose, the measurement configurations of damping pins, lacing wires and damping wire
rig is operated under vacuum to eliminate aerodynamic effects were investigated where the latter is found to be the most
on the rotating blisk and non-contact excitation and effective. However the way blades were arranged and fixed to
measurement techniques are employed so as not to modify the the bladed disk did not allow the effect of coupling between
bladed disk dynamics. The experimental data measured are blades on the damping characteristics of overall bladed disk to
used for validation of multi-harmonic balance-based prediction be observed. This is highlighted in a paper presented by
tools developed at Imperial College. Predictions are carried out Hollkamp and Gordon, [2, 3], where they carried out an
both with and without taking inherent mechanical mistuning experimental work to identify damping in jet engine blades.
into account, which is identified from measured data. Measured They showed that the common test practice of damping only a
and predicted response curves are compared with each other few of blades in a bladed disk and keeping the remaining ones
and the degree of correlation is discussed. un-damped in identification of damping values was valid only
if the damped blades’ response was localised. Berruti et al [4]
INTRODUCTION designed a test rig to investigate friction damping effects on a
Increased demand for higher performance of gas turbine stator low pressure stage blade segments where the friction
engines makes the safer option of leaving all resonances out of damping was introduced through contact of neighbouring
the operating range almost impossible. As a result, bladed disks surfaces. They also studied behaviour of an asymmetric
may be subjected to unacceptably high vibration amplitudes for underplatform damper [5] in a two-blade setup. D’Ambrosio et.
considerable time, thus threatening their structural integrity. al. investigated forced response of shrouded bladed disks [6] by
Energy dissipation devices are commonly used in such cases to first considering a simple cantilever beam and then by
bring high levels of vibration down to acceptable limits. This is assembling a number of these beams, linked by simple shrouds,
often achieved by using so-called under-platform friction around a disk. Koh and Griffin used a blades/damper test setup
dampers where the vibration energy in the blades is dissipated to analyse dynamic behaviour of spherical friction dampers [7].
by rubbing of the mating damper and bladed disk surfaces. They concluded that strain concentrations in contact region
The problem of controlling and predicting the vibration caused significant reductions in effective stiffness of dampers
amplitudes of bladed disks has been a primary concern for compared with damper bulk stiffness from FE analysis. Sextro

1 © 2007 by ASME and Rolls-Royce plc

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 05/20/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


[8] published a paper in which he presented a contact model for necessary simplifications. To this end, the bladed disk was
shrouded blades with friction contact and experimentally designed as an integral piece (i.e. a blisk) to eliminate contact
demonstrated its effectiveness. A blade with two non-Hertzian interfaces in blade root joints, a common feature of bladed
contact surfaces was used in experiments and normal contact disks, and therefore friction damping in these joints and
force was simulated by dead weights. Additional studies uncertainties in their modelling. Special effort was made during
concerning two-blade-and-a-damper arrangement are reported design of the test piece to have natural frequencies of the first
in [9] and [10]. In the former, Sanliturk et. al. sought to validate flap (1F) family of modes sufficiently well separated, so that,
predictions of vibration response of under-platform dampers the tuned modes could be clearly identified in measurements,
for turbine blades using a test setup comprising 2 turbine blades and that these modes could be excited by steady-state Engine
and a damper in between. In the latter, Stanbridge et.al. aimed Order (EO) forcing at speeds which would be achievable under
at identification of damping characteristics at high temperatures laboratory conditions (less than 4000 rev/min). Platforms were
up to 1000 C using non-contacting measurements via a Laser introduced at 1/3 of blade height to provide a means of
Doppler Velocimeter (LDV). A similar study is given by Filippi incorporating inter-blade, under-platform friction dampers. The
et. al. in [11]. damper sockets in these platforms were designed to allow
Most of the experimental studies deal with the problem of dampers of various types to be accommodated. Blades were
friction damping in turbomachinery on simple test setups such staggered, so that 1F vibration mode shapes have roughly
those comprising two blades and a damper in between, at rest. similar components both in the plane of the disk (to give
Performing experiments with underplatform dampers at rest reasonably relative under-platform damper motion) and out-of-
requires introducing a special system to simulate action of plane (to give sufficient blade-disk modal coupling). A picture
centrifugal forces that are inherent in realistic blade disks. of the machined test piece and an FE model of a single bladed
Adequate simulation of action of centrifugal forces is usually sector are shown in Figure 1.
difficult to achieve and, moreover, such systems can change
dynamic properties of a structure measured. To the best
authors’ knowledge there have been no experimental studies of
bladed disks with dampers in rotating conditions. In this study
we present a methodology and results from an experimental
investigation using a rotating test rig where a specially
designed integral bladed disk (blisk) is fitted with a full set of
under-platform friction dampers and operated under rotation
with an aim to provide necessary reference data for validation
of corresponding numerical non-linear models developed by
Imperial College [12-15].

EXPERIMENTAL SETUP (a)


Experimental investigations, especially aimed at validation of
numerical predictions of forced response for a bladed disk with
fitted friction dampers would be extremely difficult if they
were attempted to be made on a real gas turbine engine. This is
because in the engine there are many factors which affect the
bladed disk vibration response and yet remain unknown and
not included in the mathematical model. In most cases there is
not enough room in a real engine to incorporate an adequate
measurement system to acquire all the necessary data required
for analysis.
As the numerical model is only as accurate as the input
data, these measurements should be made on a carefully- (b)
constructed experimental setup, which exhibits only the (c)
phenomena under investigation. The following sections give
details of such experimental setup for achieving the objectives
of study at hand.
Figure 1. Manufactured blisk (a) and its cyclic-sector
Test piece FE model (b) used to calculate 1F family natural
Main considerations in the design of test piece were to capture frequencies (c).
the representative dynamic characteristics of a real bladed disk
and at the same time to reduce complexity by introducing

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 05/20/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Test rig Measurement system
The rig designed for the investigation presented here was Making successful response measurements on rotating
operated in vacuum to eliminate aerodynamic effects. The blisk machinery is always a challenge. In real gas turbine engines
was mounted on a rotating shaft as overhang and this is often done via strain gauges or at times using tip timing
accommodated in a cylindrical casing. The shaft was kept fairly technology. These techniques, although effective, are very
short to avoid critical speeds within operation range. A separate demanding in terms of instrumentation and/or data acquisition
disk was also introduced to enable attachment of balancing and processing. Because of this, in these tests an LDV is used
masses in case needed. The rig was driven by an electric motor and which has advantages of ease of use and being non-
which provided a rotational speed range of 0-4000 rev/min. A intrusive.
tacho signal was acquired via an optical sensor by observing a Generally, this is done by using the deflection mirrors in
high-reflectivity mirror mounted on the drive shaft to measure the LDV head and by synchronising the signals driving these
the actual rotor speed. This signal was also used as a trigger mirrors with the rotational speed of the rig. However, as the
signal for vibration measurements. The drive shaft was coupled rotation speed is increased, due to inertia of the mirrors and the
to the motor by a drive belt. Comparison of speed values from motors driving them, the laser beam gets delayed by a certain
the motor encoder and the tacho signal revealed differences of amount. This results in a phase shift with respect to the input
up to 0.1%, confirming no significant slippage in the belt drive. signal, and this shift becomes significant as speed increases. In
Vibration response data were acquired via a data acquisition order to overcome this spurious phase shift, a system of
PC. Through a code written in Matlab, the rotational speed, vibration response measurement using an LDV which is
magnet position and the blade to be measured could all be independent of rotational speed was designed. This so-called
controlled, and vibration data, force and tacho signals could be “self-tracking” technique is based on the idea proposed by
acquired and saved for a range of rotational speeds Lomenzo et.al. in [16] but it is modified considerably, by
automatically, without any intervention from the user. A picture introducing a 45-degree conical mirror and a 45-degree centre
of the rig is given in Figure 2-a. mirror fixed to the rotor axis. A schematic of the self-tracking
measurement system is given in Figure 2-b. These mirrors
allow the laser beam to be directed to a blade tip at a right
angle without moving the LDV head. Moreover, 45-degree
centre mirror rotating with the blisk (see Figure 2-b) is
mounted on a special mechanism which provides a means of
selecting the blade to measure. Each operation of this
mechanism moves the centre mirror, and thus the laser beam,
from one blade to the next, enabling all blades to be measured
without stopping the rig. More details on the measurement
system can be found in [17]. In addition, circular Continuous
Scanning LDV (CSLDV) method developed in [18] was used
(a) to measure circumferential Operating Deflection Shapes (ODS)
on stationary and rotating blisk.

Excitation
Most of the vibration-related problems in bladed disks in real
gas turbine engines are caused by steady forces, introducing
excitation at multiples of engine rotation speed (i.e. EOs). An
excitation system, therefore, is required to produce EO type
excitation and by non-contacting means. This can be done in a
number of ways such as by holding the test piece and the
exciters (electromagnets or speakers) stationary and by phasing
exciters accordingly [19, 20], or, by again holding the test piece
stationary but to mount the exciter(s) on a rotating fixture.
However, since simulating realistic operation conditions of
friction dampers require the rotation of the test piece itself, both
solutions are inapplicable in the present investigation. In our rig
(b) excitation was simulated, in its simplest form, by employing a
permanent magnet fixed to the rig. The rotation of test piece
provides realistic forcing and inclusion of centrifugal effects
Figure 2. Test rig and measurement system. enables friction dampers to operate.

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 05/20/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


The provision was also made to incorporate up to 6 amplitude could be obtained. The dampers were represented by
magnets, equally distributed around the blisk. The magnitude their equivalent stiffnesses, obtained from FE analysis. As the
of steady force could be adjusted via a stepper motor which relation between blade tip amplitude and the force in the
controlled the magnet-blade gap. Ideally, the use of stiffness elements (representing the dampers) was of interest, a
electromagnets would be preferable as the frequency of simple model, such as the one shown in Figure 3 (a), could be
excitation could be varied independently of rotational speed. used. The resultant forces in the springs representing the
However, the force produced via these magnets was so small dampers were then fed into the free body diagram of Figure 3
that they could not be used. (b) and the forces on the damper faces were balanced.
As a result of adopted excitation scheme, in normal
operation conditions the bladed disks with this arrangement are
being excited by phased pulses generated by the magnet, each
time a blade rotates past it. The excitation force fluctuates from
near-zero when the magnet is between blades, to a maximum
when the magnet is centred on a blade tip, hence the variable
excitation load has many harmonic components. Moreover, in
the present case, the magnet can only attract the blades;
therefore, all the force values are positive and the excitation
load has also a constant component. The amplitudes of the
multiharmonic components of excitation force were calculated
by performing Fourier analysis. Each such component excites a
different travelling wave EO excitation and the amplitude of
n − th harmonic therefore gave the amplitude of n − th EO
excitation.

UNDER-PLATFORM FRICTION DAMPERS


One of the most important choices in the planning of tests with
friction dampers is to determine the speed range in which the
dampers will operate successfully. A damper has a certain mass
value and the centrifugal load exerted on it increases with
increasing speed. Usually at a certain point this centrifugal load
becomes so high that it prevents dampers from slip and hence
from energy dissipation. Dampers in such a state are said to be Figure 3. (a) Simple FE model used in damper design,
locked-up. In this situation, the neighbouring blades are (b) free body diagram of a damper
interconnected through dampers which lead to substantial
stiffness increases in the assembly, thus resulting in the natural Assuming a friction coefficient of 0.3 and a roof angle of
frequencies being raised. Often the shifts in natural frequencies 150°, the mass of the CR damper allowing slip was calculated
are considerably higher so that exciting the “new” resonances as a function of rotational speed, Ω , and was identified as 4.5
may not be possible within the speed range of the test setup. grams at the maximum rig speed of 4000 rev/min. This meant
In order to reduce the possibility of such a situation, it is that they would work for even smaller response amplitudes at
essential that certain parameters such as damper mass, lower rotational speeds. The simplifying assumptions
excitation forces, operating speed range etc. are chosen necessitated further measures to be taken to ensure the success
accordingly. It was decided to use Cottage-Roof (CR) dampers of the test campaign. For this purpose it was decided to have
in the experiments since they were commonly used in real another set of dampers made from a lighter material in case the
engines and relatively cheap to manufacture. A rather simple original ones, which were to me made from steel, proved too
methodology was adopted to calculate a rough estimate for the heavy to be effective. This second set was made from titanium
damper mass. The idea was to work out the relationship which reduced the damper mass to 2.2 grams.
between the blade tip response amplitudes and the forces
applied to the dampers. Therefore, assuming a blade-tip IDENTIFICATION OF BLISK CHARACTERISTICS
amplitude response, the mass of a damper which could ensure When predictions of a particular model are to be compared
damper slip could be calculated from forces applied to a with those of the real component they simulate, it is important
damper. The dampers were considered in their stuck position that the model is validated to make sure that it is a good
(i.e. no relative motion between the dampers and blades), so representation of the actual specimen. For this purpose the
that the system behaved linearly. To this end, a 0.5 mm blade natural frequencies and mode shapes or more precisely ODSs
tip amplitude was assumed and the amplitude of the applied of the blisk were measured. Non-trivial amount of split in
harmonic force was adjusted so that the assumed response double modes of the 1F family were measured with highly

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 05/20/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


irregular mode shapes, which indicated significant level of mistuned and multitude of resonances were observed in the
mistuning in the blisk. In an attempt to obtain a tuned reference frequency range studied. This, it was believed, would provide
blisk, a tuning process was undertaken. This was achieved by enough data to gauge the degree of correlation between
removing metal from tip of each blade so that all the first measurements and numerical simulation. For example, the
family of natural frequencies were as near identical as possible. results of measurements for two of blisk blades are compared
This process required individual blades to be isolated so that with the predicted ones in Figure 4. Given the fact that a small
their natural frequencies could be measured, uninfluenced by frequency range was considered and that inherent mistuning
the other blades on the blisk. Only the first blade-alone natural was not identified accurately, the general behaviour of the
frequencies were tuned. response curves is well predicted. Deviations in the locations of
ODS for 1F modes were re-measured after tuning using resonances were generally around 0.3%. Frequency splits
circular scanning LDV so that the tuned state of the blisk could shown in predicted curves could not be captured in
be assessed. Measured ODSs were found to be regular measurements for some of the modes. This is mainly because of
deflection shapes for simple ND mode shapes up to ND=7, as the insufficient sweep rate and frequency resolution achieved in
would be expected from a tuned blisk. However, they were measurements. The excitation frequency in measurements was
distorted for higher ND numbers, eventually becoming highly controlled by rotational speed and the motor controller used did
localised for 9 to 12 ND modes. In fact this was not entirely not allow finer speed resolution to be used. Also due to this
surprising since there are many ways in which the first blade- reason and very lightly damped nature of the blisk (damping
alone frequencies can be tuned. However, not all of them can value), the resonance peak amplitudes could not be measured
yield the desired assembly modes. For a successful tuning accurately. In addition to frequency response curves, some
process one may need to target a number of blade-alone ODSs under rotation were also measured for various EO
frequencies rather than the fundamental one alone. resonances and compared with the predictions. Reasonable
After the tuning process the effort was focused on correlation was obtained for lower ND modes which vibrated
identification of the residual mistuning. To this end the FMM- in the form of travelling waves and for higher ND modes which
ID method proposed in [21] was used to identify blisk exhibit vibrations in the form of standing waves as they were
mistuning. The FMM ID uses vibratory response measurements affected more by inherent mistuning.
on a whole bladed disk to estimate existing frequency deviation Blade #1 Blade # 19
30 30
of sectors using a 1 DOF per sector bladed disk model. As this
Amplitude (dB) re 100 mm/s

20
method uses system mode shapes and natural frequencies, it is 20

suitable for blisks where not only blade mistuning but also 10
10

mistuning of disk sectors can be accounted for. The FMM ID 0


0
method was first validated on a nominally tuned blisk by -10
introducing a known mistuning and then by identifying it back. -10
-20
More details on this can be found in [22]. The validation of -20 -30
mistuning identification was performed by applying this pattern 286 288 290 292 286 288 290 292
Frequency(Hz) Frequency(Hz)
to blisk FE model. Natural frequencies obtained from this
Figure 4. Comparison of 2 blade responses to 9EO
model were compared with the measured with most deviations
excitation. —Measured, —Predicted.
under 0.3%. Similar correlation of measured mode shapes with
those from FE model, on the other hand, proved to be less
Damped rotating blisk measurements
striking with some of the higher modes exhibiting sizeable
The friction dampers are intended to reduce resonant vibration
deviations.
by energy dissipation via the friction forces occurring at contact
surfaces of the blisk and dampers, as the latter are pressed to
MEASUREMENTS ON ROTATING BLISK
blade platform by centrifugal forces. In the case of low
excitation and, accordingly, low response amplitudes the
Undamped rotating blisk analysis dampers will lock up and will not operate: i.e. will not slip and
Here, the undamped analysis refers to the case of no additional dissipate energy. Since, in such a situation, there is no relative
damping other than material damping. It was important to motion between the dampers and the blisk, the individual
validate predictions of the linear model prior to introduction of blades will be permanently coupled through the dampers,
complexity through non-linear friction elements. This required which leads to substantial increases in the natural frequencies
mistuned response predictions to be conducted. The tool used and producing little if any damping. Realistic reduction of
in these predictions, program code MISTRES, (see [15]), was resonance amplitudes via friction damping is achieved if
validated in [23]. Therefore likely deviations of forced amplitude level allowing dampers to slip is exceeded. This
response from measurements would, in effect, indicate threshold amplitude level depends on the ratio of the excitation
inaccuracies in mistuning identification or in its representation. forces to the static load applied to damper surfaces and is
All the blades in the blisk were measured for a speed range dependent on centrifugal forces. The latter is proportional to
around 9EO-9ND resonance as this mode was significantly

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 05/20/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


the damper mass and the square of the rotational speed. At
higher amplitudes, the energy dissipation does not rise in
proportion to the blisk’s vibration energy, and the effective
damping decreases again. The effectiveness of the dampers
therefore depends, amongst other factors, on the rotational
speed of the rotors, the damper masses and the strength of the
exciting force. The latter two can be adjusted to make the
dampers work in the speed range of interest. However, the
forcing magnitude can not be increased in very wide ranges due
to the danger of producing a fatigue failure in the test blisk.
Overall forced vibration response of blisk was mapped via Figure 6. Undamped response of blade 1 to (a) 19EO,
measurements taken on a blade in 0-2000 rev/min speed range. and (b) 17EO excitations
As mentioned before, two sets of dampers were manufactured:
one from steel and one from titanium. Despite all the attempts Damped individual blade measurements
made at increasing the excitation force, it was not possible to At the beginning of tests, forced response for a selected blade
make steel dampers operate successfully. They proved to be too was monitored over a period of time to ensure that the dampers
heavy and locked-up for most of speed range. Therefore all the had bedded in and that the measurements were repeatable.
effort was directed to investigation of titanium dampers. The Following repeatability tests, all blade responses were
results of measurements of blisk response with these dampers measured in 800-1050 rev/min speed range. Measured forced
are shown in an interference or z-mod diagram given in Figure response curves demonstrated reasonable consistency in terms
5. Examining the plot in detail, three groups of modes may be of overall shape and frequency corresponding to maximum
distinguished: (i) those excited by 20 EO and above, in which amplitude, as seen in Figure 7a. However, there was significant
the damper forces are too small to produce significant damping, variation in the level of maximum amplitude of each blade.
(ii) those excited by 12-19 EOs, which appear to be 'spread', These levels showed a general trend of decreasing amplitudes
signifying that there is significant damping, and (iii) those with increasing blade number (which, since blade
excited by 11 EO and below, which have apparently measurements were made sequentially from first blade to last
disappeared from the picture because the dampers have 'locked blade in the blisk, also corresponds to increasing rig running
up', shifting the natural frequencies out of the measured range. time and hence damper contact surfaces were subjected to
Strong EO lines in the z-mod picture at 4 and 8EO are probably different time of rubbing and were not exactly identical), and
due to misalignment in the LDV measurement laser beam, were considerably lower than those obtained through ODS
rather than to true vibration. measurements at the beginning of the measurement session,
Attention was directed, primarily, at a speed range of 800- Figure 7 (b).
1050 rev/min which covered the transition region for operation Some of the blades were re-measured after the given set of
of dampers. This range contained 17, 18, 19 and 20 EO lines, blade measurements was completed and large variations in
exciting, due to spatial aliasing, 7, 6, 5 and 4 ND modes, amplitudes from run to run were observed. To locate the source
respectively. Additionally, modes excited in the range were of their inconsistency, the response of blade 1 at the 19EO/5ND
mostly tuned, so that the uncertainties involved in mistuning resonance was recorded over 20 minutes of continuous rig
identification would have a minimal effect. Undamped running and the recorded response is given in Figure 8, top
measurements and predictions (with and without mistuning) inset.
corresponding to 19 and 17 EO are compared in Figure 6. A
very good correlation was obtained. Resonances missed in
measurements were simply because of low damping and
insufficient sweep rate, as mentioned before.

III II I
Figure 7. (a) Response of all blades, and, (b)
maximum amplitude variation to 19 EO excitation

Figure 5. Blisk response with fitted Ti CR damper

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 05/20/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Despite a clear steady-state value, the vibration response However, the response curves were significantly different from
presented jumps at random intervals. In fact, further those of the first 3 installations, both in terms of amplitudes and
investigation showed that the conditions at the damper-insert shape. Blade 1 response around 19EO/5ND resonance is given
interface were changing with the time of running. Close in Figure 9 for all installations, together with undamped
inspection of dampers and blade inserts, after shutting down, response curve for the same blade. Despite consistency
revealed fretting debris over part of the damper contact observed from the first three installations, the fourth one
surfaces. Some of the dampers were jammed on one side and presents a substantial shift in the resonance frequency with
thus were unable to operate, possibly because of fretting debris much reduced amplitudes. Since the excitation force was kept
which had accumulated between the contacting surfaces of the same in all installations, the deviation can probably be
inserts and dampers. This situation worsened as the operation attributed to a substantial change in friction coefficient.
of the rig continued, resulting in big discrepancies between Dampers were further away from stuck condition in first three
blade response curves. A photograph of some of the dampers installations, as evident from high amplitudes. As a result no
after shutting down is given in Figure 8. Inspection of used and additional stiffness was brought into the system and the natural
unused damper contact surfaces under the microscope revealed frequencies were very close to that of the undamped case.
no significant visual differences in surface conditions. Since the However, in the case of the last installation, the dampers were
blade inserts accommodating the dampers were made of steel, it much closer to the stuck condition, and as a result they
is believed that during the operation of the rig, relatively soft increased the stiffness of the system as well as the damping.
titanium dampers were taking asperities off the harder steel Short samples of time histories of the first 6 blades are also
surfaces of the inserts. This so-called “bedding-in” period given in Figure 9. b,c, and these were measured after the first
continued until smooth insert surfaces were obtained, at the and last installations of the dampers, respectively.
cost of the presented variability in response measurements

Figure 9. (a) Response of blade 1 to 19EO excitation


for all installations, and time histories @ 884 rev min
for (b) installation 1 and (c) installation 4.

Figure 8. Blade 1 tip response @ 19EO/5ND The low frequency oscillations are due to laser beam-
resonance (inset-top), jammed damper, and a damper rotation axis misalignment. The fact that each blade received
surface after shutting down (inset-bottom). the magnetic force at a slightly later time was perfectly
manifested by the given time histories, which exhibited larger
In an attempt to improve the consistency of the response amplitudes at magnet passing times. Although the blisk as a
amplitudes, the dampers were taken out and both steel insert whole was excited at all times, enough damping was produced
and damper surfaces were cleaned thoroughly. However, the at corresponding rotational speed that the vibrations on a
amplitudes again varied with the rig running time. This process particular blade were much reduced and in some cases
had to be repeated 3 times, resulting in 4 different installations completely damped out before the time of next magnet passing
of dampers in the blisk. Each time, care was taken to put the causing a regular excitation force impulse. However, in the
individual dampers back in those slots where they were close proximity of the shown resonance peak, the first three
removed from. Measurements reported for the first installation installations showed no clear sign of decay in vibration
were repeated for each new installation of dampers. For the last amplitudes.
installation during measurements the speed range of interest
was limited to an even shorter time interval in an attempt to Variation of response with amplitude of excitation
finalise measurements quickly, to ensure that same/similar force
conditions were attained for all blade measurements. The The main goal in this set of measurements was to measure the
results of the last installation, unlike the first 3, showed response of a particular blade for various magnet-blade gaps, in
consistent results throughout the entire measurement session. order to observe changes in the response characteristics as a

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 05/20/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


result of different excitation levels. Figure 10a shows the numerical predictions of the nonlinear forced response of a
variation of the peak-to-peak amplitude of the force signal with blisk with CR dampers.
rotational speed for 4 different magnet-blade distances. The
corresponding response curves are given in Figure 10b. The Modelling of friction dampers
black curve in Figure 10a reveals significant force magnitude Figure 11 shows the way the friction dampers are modelled.
modulation around the resonance. This is because around a The forced response analyses for the whole blisk were
resonance blade vibrations cause a change in the gap between performed using a cyclic sector of the blisk. An FE model of
the magnet and the blades. As the clearance was increased, this one sector used in calculation contains 21555 DOFs. Nodes of
effect became negligible. It is clearly seen from Figure 10b that finite element mesh of adjacent blades were connected by
the damping and apparent natural frequency increased with a friction interface elements (see Ref. [13]), as shown in the
decrease in the excitation force, as the dampers moved closer to Figure. The interaction of all sectors is accounted for by
the 'stuck' condition. applying cyclic conditions at boundaries where adjacent sectors
are connected: i.e. at the sector planes cutting the disk (shown
in Figure 11 in blue). In order to include the elasticity of the
damper body, an estimate of the contact stiffness of the friction
contact elements was determined from damper FE model. This
stiffness was determined from solution of elastic deformation
problem for the FE damper model subjected to a uniform load
on one CR damper contact face while fixing the other contact
face. Two contact stiffnesses were calculated by application of
this load in horizontal and vertical directions in turn and then
by establishing a relationship between average displacement
over the loaded face and the applied load.

Figure 10. (a) Variation of force signal amplitude with


rotational speed, (b) response of blade 1 to EO20
excitation at various blade-magnet gaps

CORRELATION OF PREDICTIONS WITH


MEASUREMENTS

The prediction tool


New, non-Coulomb friction models developed for modelling of
friction contact interaction take into account effects of variable
normal loads on friction forces and on time instants of slip- Figure 11. Modeling of friction dampers and
stick transitions. Unilateral contact interaction is allowed for introduction of mistuning (inset)
including cases of temporary and partial separation of contact
surfaces. Multiharmonic balance method is used for Forced response predictions
formulation of equations of motion in frequency domain. The inset of Figure 11 shows nodes where forced response is
Nonlinear equation of steady-state forced response is calculated calculated. The nodes where mistuning identified in earlier
using Newton-Raphson method in conjunction with solution sections was introduced are also displayed here.
continuation techniques. Expressions for nonlinear contact Validation of the damped vibration response predictions
interaction forces and tangent stiffness matrices of contact was carried out by using a subset of the experimental results as
interfaces are derived analytically which provides very fast and presented in previous section. Owing to the variability
numerically accurate calculation of forced response even for observed in measured response characteristics between
strongly nonlinear contact interactions, which occur in bladed different installations, the predicted results were compared with
disks with underplatform dampers. Effectiveness of the friction all installations whenever the corresponding experimental data
contact interactions and special accurate condensation were available. It was observed that the first 3 installations
techniques allows using large-scale finite element models for exhibited similar dynamic behaviour which was different from
bladed disks. that of the last installation. There were no verified data about
Mode detailed description of the models and methods can friction coefficient of the contact interface and in order to help
be found in Refs. [12-14]. These theoretical developments were explain this variation, the predictions were made for different
implemented in a program code FORSE (Force Response contact parameter values.
SuitE) developed at Imperial College which is used here for

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 05/20/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Numerical analyses showed that the damped blisk response vibration amplitudes were sufficiently large, the dampers slid
was insensitive to contact stiffness variations and sensitive to on the blade insert faces freely. Therefore, their contribution to
normal load and the coefficient of friction. Since the normal the system stiffness was minimal. The consideration of variable
load was known accurately from centrifugal load acting on normal load and of, especially, multi-harmonic vibrations,
dampers of known masses and location, the observed variation produced better predictions, particularly at positions away from
between the different installations was likely to be caused by the main resonance. The solid black curve in Figure 12 follows
variation in the friction coefficient. The predictions were the measurements reasonably, both qualitatively and
carried out first for a friction coefficient, µ , of 0.45. This value quantitatively. The predictions given in the plots were
was measured in an earlier study for the titanium alloy used and performed in the presence and absence of mistuning. No
was expected to yield close predictions for the results of the significant difference was observed between the tuned and
first 3 installations. The preliminary predictions showed that mistuned cases. This is not surprising as the mode excited in
varying the coefficient of friction from 0.45 to 0.6 allowed the this case (5ND) was found to be a regular ND mode, unaffected
variability observed in the vibration response measured to be by inherent mistuning in the blisk.
covered. Therefore, predictions were performed for both values Despite the mentioned similarities between the
of friction coefficient and measurement results obtained from measurements of the first 3 installations and the predictions
all installations were included in comparison plots. given in Figure 12, the resonance amplitudes predicted were
Main effort in measurements was directed to 19EO/5ND much higher. Close inspection of measured time data revealed
resonance though 17, 18, and, in some cases, 20 EO excitations that the signals were truncated as they have exceeded data
were also considered. The 19EO response of blade 1 for all 4 acquisition board’s limits and that they were in fact higher than
installations is compared to the corresponding predictions recorded levels. Also, the resonance peaks were very sharp that
( µ=0.45 ) in Figure 12. The predictions were performed for two even the finest allowable steep steps (i.e. achievable frequency
cases: (i) monoharmonic excitation, when excitation only by resolution) in measurements could not capture resonance peaks
19EO is applied, and (ii) multiharmonic excitation, when accurately. In this particular case, the resonance peak amplitude
simultaneous excitation by 17, 18, 19, 20 and 21 harmonics was determined mainly by the available structural damping,
was considered. The results of both analyses are shown in the which was very low ( η = 7.5e-5 ). Nonetheless, the predictions
Figure. In the case of multi-harmonic force, 5 harmonics (17 to showed that compared with the undamped case, a 70%
21) were considered. Additionally, although the normal load reduction in the resonance peak amplitude was achieved by the
was taken as constant in the mono-harmonic force case, its dampers.
variation with rotational speed was taken into consideration for Response curves measured for 17 EO and 18 EO
the multi-harmonic case. excitations were also correlated with predictions, as shown in
Figure 13. The speed range used for the 4th installation was
smaller as it was intended to cover the 19EO/5ND resonance
only. Although not available for entire frequency range of other
curves, the extracted response of 18EO excitation for this last
installation was given as an indication of the departure in the
behaviour of different installations. Only the predictions with
variable normal load and multi-harmonic force are shown.
Despite the unavailability of the complete response curve, the
behaviour of the last installation is, again, considerably
different from the others. As for the 17EO excitation response,
a different trend was observed between the predictions and the
measurements of first 3 installations. Here, the dampers were
predicted to be working more effectively thanks to the higher
Figure 12. Comparison of measured and computed normal load. In fact, the increase in the normal load (i.e. due to
damped response on blade #1: 19EO, µ =0.45 increasing rotation speed) was around 10%, which
demonstrated that the performance of the dampers was very
It is clearly seen from the Figure that the predictions sensitive to the normal load. Measurement data from the last
correlate much better with the first 3 installations than the last installation was not available and therefore is not given. For all
one. The apparent natural frequency of the damped system other installations, the response curves obtained for blade 1
obtained from the first 3 installations and the predictions was were similar and somehow dissimilar to the predicted one.
almost the same as that of the undamped system. (Note that the Particularly near the resonance peak, the amplitudes as well as
undamped curve given in the figure was obtained by the apparent natural frequency were observed to be different.
amplitude-scaling of the original curve, which was measured in Recalling the variability encountered in the measurements of
the presence of a much smaller input force). This meant that the the 1st installation, the response of the minimum responding
dampers did not increase the stiffness of the blisk. As the blade (blade # 18) is also given in Figure 13 (b) to indicate the

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 05/20/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


range of variation. The measured amplitude response of this coefficient was not available. The vibration response data for
blade agreed better with the predicted one. the 17EO-7ND resonance was not available for the last
installation and only the 18 and 19 EO excitation responses
were compared. The comparisons obtained are given in Figure
14.
The predictions obtained in this case were clearly better
correlated with the measurements of the last installation, as
expected. It is seen from the figure that the increase in the
coefficient of friction from 0.45 to 0.6 resulted in a significant
reduction in response amplitudes, covering the range of
amplitude variation observed at various stages of the
measurements. This finding strengthens the explanation that the
mentioned response variation between the first 3 installations
and the last was due to variation of coefficient of friction.
Despite the incompleteness of the data, Figure 14a shows very
good agreement between the measurements of the last
installation and the predictions. Although the amplitudes are
not predicted as accurately in Figure 14b, a reasonable
qualitative agreement is evident. However, in both Figure 14a
and Figure 14b, resonance frequency is apparently a little
under-predicted. Since the predictions for the undamped blisk
for this resonance frequency were demonstrated to be closer,
the discrepancy here is probably due to the modelling of the
under-platform dampers or discrepancy between actual friction
Figure 13. Measured and computed damped coefficient value and used in the numerical simulations.
response on blade#1 to (a) 18 and (b) 17EOs; µ =0.45. At this stage it is worth mentioning that although the
predictions are given for two extreme values of friction
coefficient the actual values of this parameter in reality were
probably changing even within the same measurement session.
By giving predictions for different values of coefficient of
friction, it was aimed to show that provided that the correct
values of this parameter were available at corresponding stages
of the measurements, reasonable predictions could be obtained.
The evidence presented in this section demonstrated that the
prediction tool was capable of providing reliable estimates for a
blisk integrated with under-platform friction dampers once the
correct input parameters were supplied.

CONCLUDING REMARKS
A forced response investigation is performed for an integral
bladed disk with fitted under-platform friction dampers in
rotating conditions. A carefully-designed and constructed
rotating test rig is used and a methodology is presented for
acquisition of reliable experimental data for correlation with
corresponding predictions. These data are collected through
non-intrusive measurement and excitation methods.
The modelling of blisk with under-platform friction
dampers was achieved by considering a single sector and by
making use of cyclic symmetry. Predictions were carried out by
Figure 14. Measured and computed damped a multi-harmonic forced response solver. Performed
response on blade#1 to (a) 18 and (b) 19EOs; µ =0.6. preliminary sensitivity analysis showed that the contact
stiffness was found to have less impact compared with the
The predictions were repeated by using a friction friction coefficient which, together with the normal load, was
coefficient of 0.6. This case was expected to simulate the last demonstrated to have played a greater role in forced response.
installation, although an exact actual value of the friction Since the normal load was accurately known, the friction

10

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 05/20/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


coefficient was varied to account for the variability observed [10] Stanbridge, A.B., Sanliturk, K.Y. & Ewins, D.J.,
during the measurements. Given the complicated nature of “Measurement and Analysis of High-Temperature Friction
measurement data (provided that the variability in input data Damper Properties” 4th US International Turbine Engine
was accounted for), obtained correlation between High Cycle Fatigue (HCF) Conference, 1999
measurements and predictions showed that numerical tool [11] Filippi, S., Rodrigues, E.B., Gola, M.M., “Experimental
captured response characteristics under such realistic characterization of contact hysteresis at high
conditions reasonably well. temperatures”, Proc. of ASME Turbo Expo 2006, May 8-
To the best of the authors’ knowledge, an experimental 11, Barcelona, Spain.
investigation of a rotating bladed disk fitted with a full set of [12] Petrov, E.P. and Ewins, D.J., “Generic friction models for
friction dampers under such realistic conditions is reported here time-domain vibration analysis of bladed discs,” Trans.
for the first time. ASME: J. of Turbomachinery, 2004, Vol.126, January,
pp.184-192.
ACKNOWLEDGMENTS [13] Petrov, E., Ewins, D. J., “Analytical formulation of
The authors gratefully acknowledge the support of the friction interface elements for analysis of nonlinear multi-
European Union for this work within ADTurB II project, harmonic vibrations of bladed disks”, Journal of
G4RD-CT2000-00189. We are also grateful to all the project Turbomachinery, Vol. 125, April 2003, pp. 364-371.
partners for their support. [14] Petrov, E.P., “A method for use of cyclic symmetry
properties in analysis of nonlinear multiharmonic
REFERENCES vibrations of bladed discs,” Trans. ASME: J. of
[1] Jaiswal, B.L. and Bhave, S.K., “Experimental evaluation Turbomachinery, 2004, Vol.126, January, pp.175-183.
of damping in a bladed disk model”, J. of Sound & [15] Petrov, E.P., Sanliturk, K.Y. and Ewins, D.J., “A new
Vibration, 1994, Vol.177, No. 1, pp.111-120. method for dynamic analysis of mistuned bladed discs
[2] Hollkamp, J.J. and Gordon, R.W., “Modal test experience based on exact relationship between tuned and mistuned
with a jet engine fan model”, J. of Sound & Vibration, systems”, Trans. ASME: J. of Eng. for Gas Turbines and
2001, Vol. 248, No. 1, pp.151-165. Power, 2002, Vol.122, July, pp.586-597.
[3] Hollkamp, J. J. and Gordon, R. W., “An experimental [16] Lomenzo, R.A., Barker, A.J., Wicks, A.L., “Laser
investigation of non-uniform damping in bladed disks” Vibrometry System for Rotating Bladed Disks”,
IAA-98-3747, 1998. Proceedings of the 18th IMAC, 277-282, 1999.
[4] Berruti, T., Goglio, L., Filippi, S., Gola, M. M., “A test rig [17] Sever, I.A., Stanbridge. A.B., and Ewins, D.J,
for frictionally damped bladed segments”, Proc. Of “Turbomacinery blade measurements with tracking LDV
ASME Turbo Expo, 2000, Munich Germany. under rotation”, 7th AIVELA Conf., 2006, Ancona, Italy.
[5] Berruti, T., Gola M.M., Zucca, S., De Poli, G.P., [18] Stanbridge, A.B., Marterelli, M., Ewins, D.J., “Modal
“Underplatform damper performance with turbine testing using a scanning laser Doppler vibrometer”,
blades”, Proc. of NATO AVT-121 Symposium on Mechanical Systems and Signal Processing, Volume
Evaluation, Control and Prevention of High Cycle Fatigue 13, Number 2, pp. 255-270(16), 1999.
in Gas Turbine Engines, Granada, Spain, 3-7 October [19] Kruse, M. J., and Pierre, C., “An Experimental
2005. Investigation of Vibration Localization in Bladed Disks,
[6] D’Ambrosio, F., Chatelet, E., Ravoux, J., Jacquet- Part I: Free Response”, Proc. 42nd ASME Gas Turbine
Richardet, G. “Forced response of shrouded bladed disc and Aeroengine Cong., 1997, Orlando, Florida, USA.
assemblies: a jointed experimental numerical approach”, [20] Jones, K. and Cross, C., “A traveling wave excitation
Proc. of ASME Turbo Expo 2004, June 14-17, Vienna, system for bladed disks”, 43rd AIAA /ASME /EASCE
Austria. /AHS /ASC Structures, Structural Dynamics and Mat.
[7] Koh, K-H, Griffin, J.H., “Dynamic behavior of spherical Con, AIAA 2002-1531, 2002, Denver, Colorado.
friction dampers and its implication to damper contact [21] Feiner, D.M., and Griffin, J. H., “Mistuning Identification
stiffness”, Proc. of ASME Turbo Expo 2006, May 8-11, of Bladed Disks Using a Fundamental Mistuning Model –
2006, Barcelona, Spain. Part I: Theory”, ASME Paper GT2003-38648, Int. Gas
[8] Sextro, W., “The calculation of the forced response of Turbine Institute Turbo Expo, Atlanta, Georgia.
shrouded blades with friction contacts and its [22] Sever, I. A., “Experimental Validation of Turbomachinery
experimental verification”, Proc. of ASME TurboExpo, Blade Vibration Predictions”, PhD Thesis, 2004, Imperial
2000, Munich Germany. College London, England.
[9] Sanliturk, K.Y., Ewins, D.J. & Stanbridge, A.B., [23] Sever, I.A., Stanbridge. A.B., and Ewins, D.J,
“Underplatform dampers for turbine blades: theoretical “Experimental Validation of Mistuned Bladed Disk
modeling, analysis and comparison with experimental Vibration Predictions”, 7th IFToMM Rotor Dynamics
data”, ASME Journal of Engineering for Gas Turbines and Conference, 25-28 Sep. 2006, Vienna, Austria.
Power, v.123, pp.919-929.

11

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 05/20/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy