T. M.

Cameron

J. H. Griffin
An Integrated Approach for
Department of Mechanical Engineering, Friction Damper Design
Carnegie-Mellon University,
Pittsburgh, PA 15213 A procedure is outlined for determining the optimal design of friction dampers for
high-speed turbomachinery blading. The procedure includes: An integration of
bench test results with finite-element analysis and a single-mode blade model to en-
R. E. Kielb sure accuracy of the analytical model and improve reliability of the friction damper
NASA Lewis Research Center, design; an extension of the single-mode blade model to predict the engine behavior
Cleveland, OH offriction dampers; and a new way of viewing analytical and experimental results in
terms of a damper performance curve to determine optimal design parameters, when
the levels of excitation and damping in the system are unknown. Unique experiments
T. M. Hoosac are performed on a test disk in order to demonstrate and verify the accuracy of the
Department of Mechanical Engineering, design procedure.
Carnegie-Mellon University,
Pittsburgh, PA 15213

1 Introduction
For a number of years, considerable effort has been made Airfoil
toward understanding how friction affects the dynamic
response of structures. For example, see Plunkett's (1980)
review of friction damping research; more recent work is
discussed in Section 2.2. One motivation for this research is
Platform.
the application of friction dampers in gas turbine engines and
high-speed turbopumps (both will be referred to as engines).
Friction dampers are used to reduce resonant stresses in tur-
bine blades by providing sliding contact between points ex-
periencing relative motion due to vibration, thereby
dissipating resonant vibratory energy (e.g., blade-to-ground
damper, Fig. 1). This paper presents a procedure which in-
tegrates analytical friction models, finite-element structural
models, and experimental data into a systematic approach for
developing an accurate model of friction damping. The
friction-damping model, developed with corroborating bench
test and finite-element data, is used to predict the engine per- Ground,
formance of the damper without requiring additional engine
or spin rig tests. In addition, a new way of viewing the damper
optimization problem is introduced in terms of a damper per-
formance curve, which results in a damper design that is in-
dependent of the levels of excitation and nonfriction damping Fig. 1 Turbine blade with idealized blade-to-ground damper
that occur in the engine. Thus optimal parameters for friction-
damping devices may be determined in applications for which predicting the damper's performance in a rotating environ-
there are no engine test data from which excitation or damping ment. The accuracy of the procedure is then assessed by com-
levels may be determined.1 paring results of the prediction with spin rig test data.
First, the new procedure for developing the friction- The spin pit tests are unique in that the normal load on the
damping model and optimizing the friction joint is presented. friction damper is applied using electromagnets. This allows
The approach is demonstrated by applying it to a test disk and the damper loads to be varied continuously during the tests
which results in a far more complete set of test data than can
be obtained using conventional tests methods.
This approach is especially important for new designs since there is no test
data from which the excitation or damping can be inferred. This also is the case
for applications such as the high-pressure fuel pumps on the space shuttle since 2 Design Procedure
it is not feasible to instrument them because of potential leakage problems.
Contributed by the Technical Committee on Vibration and Sound for The goal of the design procedure is to determine the normal
publication in the JOURNAL OF VIBRATION AND ACOUSTICS. Manuscript received load on the friction interface between the damper and the
at ASME Headquarters, August 30, 1988. blade which will keep the blade stress below some specified

Journal of Vibration and Acoustics APRIL 1990, Vol. 1 1 2 / 1 7 5

Copyright © 1990 by ASME
Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
 durability limit for the largest possible amplitude of excita- agrees with bench test data, and predicting the damper's per-
tion. The difficulty in finding this optimal normal load lies in formance under rotating (engine) conditions.
the fact that friction is nonlinear and that a normal load which
works well under one set of excitation conditions may cause 2.1 Phase 1: Modal Analysis. The objective of the first
the friction joint to lock up and not dissipate any energy under phase of the design is to obtain an accurate modal model of
other conditions. When the damping and excitation in the the blade. Typically one mode, the mode most strongly excited
system are known, the approach presented by Griffin (1980) in the engine, is of greatest concern. A finite-element model of
may be used to find the optimal normal load. However, when the blade (ignoring the friction damper) provides natural fre-
these input parameters are not known, the performance of the quencies, mode shapes, modal masses, modal stiffnesses, and
damper must be optimized differently. The approach pro- modal stresses of the blade. The finite-element model is re-
posed here is to choose the normal load so that the resonant fined until it predicts frequencies and mode shapes that agree
response is less than some specified maximum value for as with bench test results on a single blade. If only one mode
large an excitation force as possible. While this is an obvious causes concern that mode is selected for testing. In particular,
objective, under the circumstances, the interesting fact is that the modal displacement of the blade at the damper location (in
this optimization procedure results in a damper normal load the mode of concern) is critical in the friction calculations and
that is insensitive to the amount of damping that is available must be checked.
from other sources that may be present in the operating en-
vironment, e.g., aerodynamic and material damping. This is 2.2 Phase 2: Analytical Model of Friction Effects. The
not true for the procedure proposed by Griffin (1980). Conse- goal of Phase 2 is to develop an analytical model of the com-
quently, with this new approach a unique damper design may bined blade and friction-damper system that accurately
be developed in an application for which no engine test data characterizes the effect of the friction damping on the
exist. response of the blade. The accuracy of the model is assessed
Geometric considerations typically constrain the shape and by comparing actual bench test data to the results of simula-
allowable sizes of the friction damping devices to lie within tions. One difference between single-blade bench tests and
certain ranges. For a given damper configuration, the pro- engine conditions is that the damper is usually tested in a
cedure described here determines an optimal damper normal blade-to-ground (B-G) configuration; whereas, in the engine,
load, which is proportional to the damper's mass (since the it may be employed in a blade-to-blade (B-B) configuration. 2
contact pressure in the rotating environment is usually induced Consequently, one purpose of the analytical model may be to
by centrifugal loading). In addition, as the damper's mass take this difference in configuration into account in ex-
changes, its stiffness also may change, affecting the damper trapolating to engine operating conditions. This is discussed in
efficiency. Thus, in practice, the theoretical optimum may not Section 2.3.
be achievable for a given design configuration. Consequently, No single friction model adequately characterizes all
a very important aspect of the design process, which is not ex- friction-damping systems so this section begins with a discus-
plored in this paper, is finding friction-damper configurations sion of the variety of models available and the importance of
in which the range of normal loads is such that the optimum experimental corroboration. The discussion is followed by
can be achieved. While prior experience with similar blades descriptions of the bench tests and simulations used to verify
may indicate a feasible configuration, it is necessary typically the model. The Phase 2 discussion concludes with the presen-
to consider several alternative damper configurations. The tation of a new way to view simulation and test data in order
procedure presented in this paper may be used to evaluate dif- to optimize the design of friction dampers when excitation and
ferent configurations and optimize the most promising one. damping conditions are unknown.
The first step in the design process is to develop an accurate A number of analytical models have been proposed in the
dynamic model of the blade/damper system. This is ac- literature for calculating the steady-state response of a fric-
complished by calibrating the dynamic model so that it agrees tionally damped system. For example, see Beards (1983);
with bench tests conducted on a single blade and damper. Dowell and Schwartz (1983); Muszynska and Jones (1983);
Therefore, it is assumed that some general configurations of Jones (1980); Muszynska, et al. (1981); Griffin (1980); Menq,
the damper have been selected and that a prototype is et al. (1986 a, b, c). One reason for the variety is that different
available for each configuration. Additional facilities required models are required under different circumstances. For exam-
are: An instrumented prototype blade, a shaker table or other ple, the last four papers cited provide models for a single-
blade excitation device, and a finite-element program capable mode structure with Coulomb friction, which may be ade-
of dynamic analysis. quate for simple systems; a variable normal load model, which
The design procedure consists of three phases: Developing a may be required if the friction interface is not parallel to the
finite-element model of the blade, developing an analytical
model of the effect of the friction damper on the blade B-G dampers connect a vibrating point on the blade to a relatively fixed
response which predicts a damper performance curve that point on the coverplate or disk. B-B dampers connect neighboring blades.

Nomenclature

A = maximum damped k = modal stiffness of blade natural frequency of blade

amplitude of blade vibration [k = mwj] with friction damper stuck
a = interblade phase angle kd = damper stiffness natural frequency of blade
(b, a) = coordinates of design point, m = modal mass of blade with friction damper absent
P (see Fig. 3) M = coefficient of friction ! = viscous damping (fraction
c = viscous damping constant N = friction normal load of critical) [f=c/(2/w«/)]
4> = ratio of blade tip to damper Q = magnitude of harmonic ex-
contact nodal displacements citation on blade Nondimensional Variables
in the mode-shape vector frequency at which max- e = kd/£k+kd)
g = acceleration due to gravity imum damped blade vibra-
(32.2 fps2) tion (A) occurs

176/Vol. 112, APRIL 1990 Transactions of the ASME

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 /////////////////////

(A/N) Increasing C

Increasing C

(Q/N)

Fig. 4 Viscous damping effect on response (idealized)

the precise frequency and response characteristics (stress levels

and amplitude of tip vibration) when the response is a
maximum.
Fig. 2 Single-mode model of turbine blade with friction damper
An algorithm for determining the maximum response of a
fricitionally damped blade, based on Griffin's (1980) single-
degree-of-freedom (sdf) blade model (Fig. 2), is presented in
the Appendix. The modal inputs to the model that describe the
blade come from Phase 1. They are blade mass and stiffness
for the mode of concern, and mode-shape information—the
(A/N) displacements of the blade tip and damper contact points and
the modal stress at critical locations. Levels of damping and
i/[2 k a excitation are obtained from the bench tests, as discussed
previously. Typically, the coefficient of friction, /x, and the
damper stiffness, kd, are chosen to correlate best with the
bench test data.3
Both bench test data and computer modeling results are
plotted in the normalized form of Fig. 3—(A/N) versus
(Q/N). The resulting curve is referred to as the damper perfor-
mance curve and is first introduced in this paper. The advan-
tage to this representation of the blade response is that it pro-
(Q/N) vides a design point which is independent of both excitation
and viscous damping levels in the engine. At the low end of the
Fig. 3 Idealized response curve response curve, the friction damper is stuck, since the excita-
tion is too small to cause it to slip, and the response is linear.
direction of vibratory motion; and a microslip model of the At the high end of the curve, the large excitation causes the
friction interface, which may be required to properly simulate friction damper to slip freely and the response is linear again.
high damper loads. Since friction damping is nonlinear, an im- In the middle section of the response curve, the excitation is
portant check on the validity of the model is whether it can, in large enough to cause slip, but is still low relative to the nor-
fact, match bench test data over a range of excitation frequen- mal load (Q/Nsmall). In this range, friction damping controls
cies and amplitudes. The model proposed by Griffin (1980) is the response, rendering viscous damping effects negligible.
used here for its simplicity and computational efficiency. (The dependence of the overall response on viscous damping is
More complex models may be required in some circumstances. depicted in Fig. 4.) In the midrange, where friction
Again, the key test of the appropriateness of a model is damping is effective, the curve is flat, indicating that the
whether or not it properly correlates with the bench test data. response also is independent of the level of excitation. The in-
If it does not, then either the experimental data are incorrect teresting feature to observe is that it is the same point, " P "
or one or more assumptions underlying the model are not be- (Fig. 3), at which the response begins to depend on both dam-
ing satisfied and a more accurate model of the physical system ping and excitation.
is required.
The point P is selected to correspond to the maximum
The bench tests supply input data for the friction-damper allowable blade response (discussed later) and, consequently,
model and provide a means of calibrating and verifying the the maximum excitation that can be sustained by the present
model. The level of extraneous damping in the bench tests blade. If the actual excitation exceeds this maximum value
(represented as viscous damping in the model) is inferred from then the entire blade must be redesigned since friction damp-
half power measurements of the blade response, when the fric- ing cannot keep the blade response (or stress level) below
tion damper is not in contact with the blade. The level of ex- allowable limits. Using this approach provides a friction
citation is either measured directly (e.g., it is proportional to damper that is optimal—in that the stresses in the blade will be
the base acceleration on a shaker table or the current level in a acceptable—for as large an excitation as possible. Thus the
magnetic exciter) or it is inferred from the response of the un-
damped blade before the friction damper is applied. The
bench tests should be performed over a wide range of damper
normal loads and amplitudes of excitation. At each In the response curve (Fig. 3), /x primarily affects the horizontal scale and kd
amplitude, the frequency of excitation should be varied con- control the effectiveness of the damper, i.e., once /t is fixed, kd determines the
tinuously while the blade response is monitored to determine height and length of the horizontal segment, L-P, of the curve.

Journal of Vibration and Acoustics APRIL 1990, Vol. 112/177

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 optimal design is independent of the excitation (for values B-B dampers may be calculated using a B-G damper analysis
lower than the maximum allowable excitation) and is insen- provided the effective values of the normal load and damper
sitive to variations in viscous forms of damping. stiffness are used (Griffin and Sinha, 1985)
Momentarily ignoring engine effects, the optimal normal Neff = 2A/sin(a/2) (4)
load may be found from the ordinate value of the design point 2
P ("a," Fig. 3) and the maximum allowable blade displace- kdim=4kdsin (a/2) (5)
ment, v4design. 04desjgn is the displacement of the blade tip that Finally, temperature effects should be taken into account to
corresponds to the maximum allowable blade stress. It is properly model engine conditions. High engine operating
found by multiplying the maximum allowable stress at the temperatures affect material moduli producing further
critical node by the ratio of the modal displacement of the changes in blade and damper stiffnesses. Using temperature
blade tip to the modal stress of the critical node.) From these corrected moduli, the finite-element code may be used to ob-
values, the optimal normal load is tain modified stiffnesses. The coefficient of friction between
the damper and blade also may be affected by temperature. A
No»^=~^ (1) corrected coefficient of friction may be found by further
testing or from tables, when they are available.
The maximum allowable excitation, to keep blade stresses With the appropriate modifications made to the input
within tolerance, is obtained from "b," the abscissa of P, and parameters, the procedure in Phase 2 (using the algorithm
the optimal normal load presented in the Appendix or another friction-damping model)
=
may be repeated to determine the optimal normal load for the
t:max "^optimal (*•)
damper under engine conditions.
2.3 Phase 3: Predicting Damper Performance at Engine Engine or spin rig tests may be used to further refine the
Conditions. The initial phases of the design procedure have model and improve the friction-damper design. However, if
focused on development and refinement of a blade and the model developed in Phase 2 accurately simulates bench test
friction-damper model that accurately characterizes their results and the parameters are correctly modified in Phase 3,
dynamic behavior on a shaker table. All the parameters for the the improvement from engine testing should be small.
model have been determined. The final phase consists of
modifying certain input parameters to account for differences
between shaker table and engine conditions. Engine condi- 3 An Application
tions differ due to centrifugal effects of the rotating environ- In order to illustrate the design procedure and experimental-
ment, multiple-blade interactions from phase differences in ly verify the accuracy of the predictions, a special test blade,
the blade vibrations, and temperature effects. The parameters disk, and damper assembly were designed, fabricated, and
that may require modification are blade stiffness and mass, tested. A description of some of the problems encountered in
damper stiffness, normal load, coefficient of friction, and experimentally verifying friction-damping models is provided
viscous damping. An educated guess should be made regard- first in order to motivate the description of the experimental
ing the level of viscous damping in the engine (based on the assembly which follows. The design procedure is illustrated
authors' experience, 1 percent of critical is typical for tur- using the experimental blade, and, finally, spin rig tests results
bines). This is not crucial if the damper is designed effectively, for verifying the predictions of Phase 3 are presented.
as was shown in Phase 2. The problem in providing experimental verification of fric-
The predominant effect of the rotating engine environment tion damping models (as required in Phase 2) is typically due
is centrifugal stiffening of the blade. To a lesser extent there is to the difficulty in varying the normal load on the damper. In
a change in the effective modal mass which "softens" the one configuration, which has been employed successfully in
blade. Accounting for temperature effects (discussed later), bench tests (despite its lack of elegance), the normal load is ap-
which also soften the blade, the effective mass and natural fre- plied by means of fishing line with weights hanging from one
quency at engine speed may be obtained from a finite element end, and fish hooks pulling on the damper at the other end. A
or beam code. If codes with these capabilities are not pulley system provides the proper orientation. Varying the
available, the mass effect may be neglected and the natural normal load is even more complicated in an engine since the
frequency may be estimated from prior experience, or from normal load is applied centrifugally, and is determined by the
beam theory, to approximate the blade stiffness. mass of the damper. Typically, the entire engine and disk
Multiple-blade interactions arise due to circumferential assembly would require dismantling and reassembly with an
pressure variations in the engine. The excitation frequencies entire set of new dampers to test each normal load. (This also
on the blades, as they rotate through the pressure field, are in- may change the damper's stiffness which further complicates
teger multiples of the engine speed. These integer multiples of the results.)
the engine speed are called "engine orders" (the engine order In order to avoid these complications a unique
is the integer n, denoted by "nE," which multiplies the engine blade/disk/damper assembly was devised in which the damper
rpm). Due to the dynamics of the blades and disk, one engine normal load would be applied by using electromagnets to pull
order tends to dominate causing the blades to vibrate at the a damper bar into contact with the blade platform (Fig. 5).
same frequency but with a fixed phase relative to one The electromagnets were implanted in the body of the disk
another.4 The phase difference between two contiguous blades under the damper plate. A simple two-bladed disk was chosen
is defined as the "interblade phase angle," a, and is given by to demonstrate the new electromagnetic damper concept in a
a = 27r(Engine order of excitation)/ rotating environment. With this configuration, the normal
load could be adjusted easily over a continuous range of
(Number of blades on stage) (3) values, using a remote power supply to the magnets, and in the
The interblade phase angle has no effect on the performace of same manner for both bench and spin rig tests. Initial studies
B-G dampers, but changes the effective normal load and stiff- were conducted to determine the precise relationship between
ness of B-B dampers. The dynamic response of stages with current to the magnets and damper normal load on the blade.
Full details of the work may be found in Cameron (1985).
For Phase 1 of the design procedure, a NASTRAN finite-
The approach discussed here strictly applies to perfectly "tuned" stages in element model of an experimental blade was developed from
which each blade has exactly the same natural frequency. However, Griffin and
Sinha (1985) show that blade mistuning does not significantly affect how fric- the design drawings. The instrumented bladed disk was tested
tion dampers should be optimized, so the mistuned case is not considered here. on a shaker table to determine the natural frequency of a

178/Vol. 112, APRIL 1990 Transactions of the AS ME

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 NASA LEWIS RESEARCH CENTER

Fig. 5(a) Fig. 6 Frequency response of blade with constant amplitude base ac-
celeration (0.4g) and different normal loads (response is in terms of
stress at center of airfoil base)

Phase 2: Bench Test Data end Simulation

0.030

0.025 |

0.020 — Simulation
• l.Og Base Excitation
A/N 0.015 ° 0.4 g Base Excitation

[ inches]
lb, J 0.010

0.005
A_aO&&.
0.000 J*
0.00 0.02 0.04 0.06 0.08 0.10 0.12
Fig. 5(b) Q/N IV*,]

Fig. 7 Bench test response: data and simulation

blade through resonance. Figure 6 shows a typical family of

curves produced by frequency sweeps at various normal loads,
with the amplitude of excitation held constant (0.4g base ac-
celeration). The peak response amplitudes at "resonance" (us-
ing the term loosely for both linear and nonlinear cases) are
the data points (after dividing by the normal load) for the
response curves, e.g., Fig. 7.
The tests at 0.4g and l.Og base accelerations were used to
"calibrate" the analytical model. The unknown parameters in
the analytical model—coefficient of friction, /x, and damper
stiffness, kd—were obtained by a trial-and-error process of
getting the simulations based on the analytical model to match
the 0.4g and 1.0g test data. The final values of the parameters,
used in the single-mode model, are given in Table 1. The cor-
Fig. 5(c) relation between the analytical model and the bench test data
Fig. 5 Experimental two-bladed disk with electromagnetic dampers was very good (Fig. 7). The value obtained for [i is comparable
to tabulated values for the dynamic coefficient of friction be-
selected blade, the equivalent viscous damping (using the half tween soft steel surfaces (both the blade and damper were
power method), and the modal component of the damper con- 1080 cold-rolled steel). A further verification was peformed by
tact point. These data were used to refine the NASTRAN using the analytical model developed from the other base ac-
model. celerations to predict the blade response at 0. lg base accelera-
For Phase 2, a sdf model of the blade was developed from tion. Again, the correlation was reasonably good (Fig. 8).
the first-mode parameters of the NASTRAN model (cf. Grif- In comparisons, between bench test and simulation results,
fin, 1980) in order to employ the algorithm in the Appendix. the correlation was better for lower (Q/AO-values. This effect,
For the specially designed disk, it was easier to vary the nor- which becomes more apparent in the case of the engine predic-
mal load than the excitation so only three excitation levels tions, resulted from the use of very low normal loads (rather
were used (O.lg, 0.4g and l.Og base accelerations), and the than large excitations which may have caused the blade to fail)
normal load was varied for each excitation. For each excita- to generate the higher (Q/N) part of the response curve. At
tion amplitude and normal load setting, the frequency of ex- lower normal loads, chatter would occasionally dislodge the
citation was increased slowly (0.5 octaves/minute) to take the damper producing unreliable results. All tests were repeated

Journal of Vibration and Acoustics APRIL 1990, Vol. 112/179

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Table 1 Bench test simulation parameters Engine Performance: 2E Excitation
Predicted and Actual
Parameter YallJS

m 0.000228 Ibf-sec 2 / inch 0.08

0.07
k 106.7 lb, 0.06

0.0S
kd 74.6 Ibf A/N
004
(inches/lbf)

0.51 0.03

0.02

t 0.0047 (1,0g base acceleration) 0.01

0.00
0.147 Ibf (1.0g base acceleration)
0.00 0.35

Phase 2: Model Verification

Fig. 9(a)

0.0026 Engine Performance: 4E Excitation

Predicted and Actual
0.0024 — Simulation

0.0022 • O.I g Base Excitation

0.0020
0.0018 •
A/N
0.0016
[ Inchest
lb
i J 0.0014

0.0008

Fig. 8 Bench test model verification: data and prediction

0.35
several times to assess repeatability and accuracy; tests at very
low normal loads were distinctly less repeatable. (The data Fig. 9(b)
shown are average values from all tests.) However, the
repeatability and correlation were very good in the range of in- Engine Performance: 6E Excitation
terest, up to the design point. Predicted and Actual
To obtain modified parameters for Phase 3, only cen-
trifugal effects were considered since the dampers on the ex-
0.16 ,
perimental disk were B-G and temperature differences were •
not a factor. Also, since spin rig tests were performed to verify 0.14
the engine predictions, some of the difficulty of Phase 3 was 0.12
avoided by obtaining damping and natural frequency data 0.10
directly from the spin rig tests. The stiffness of the blade was A/N
0.08
estimated using the original modal mass and the natural fre- (inches/lbf) • y^
quency obtained from the spin tests. The performance of the 0.06
friction damper under engine conditions was predicted using 0.04 • S^
— 6E Simulation
modified damping and blade stiffness values in the Appendix
algorithm, with other parameters remaining the same. 0.02 •^ • 6EData

0.00
Spin rig tests were conducted to check the accuracy of the 0. DO 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.1
predictions. The tests consisted of exciting the blades and Q / N (Ibf/lbf)
monitoring stresses as the disk was spun in an evacuated spin
chamber. [A description of the spin rig and its utility may be Fig. 9(c)
found in Brown, et al., (1984)]. The blades were excited in the
plane of rotation by nitrogen jets—introduced through tubes Fig. 9 Spin rig verification: data and prediction

in the wall of the spin chamber—impinging on the blades. Dif-

ferent levels of excitation were achieved by spinning the disk at damper performance was very good for low (JQ/N), but not as
three different speeds to excite different engine orders (2E, 4E, good at the higher end of the response curve for reasons
6E). The excitation levels were deduced by relating the blade discussed previously. (Neglecting the change in the modal
stresses caused by the nitrogen jets at each engine order to the mass in the rotating environment also introduced some error.)
stresses induced on the shaker table, in the absence of friction As in the bench tests, however, the correlation was very good
damping. Strain gauge readings and current to the elec- at least through the design point, so an optimal friction
tromagnetic dampers were both transmitted through a slip damper design could be obtained realiably from the engine
ring. The predictions of Phase 3 and the corresponding test simulations (Fig. 10).
data are shown in Fig. 9. Figure 10 shows the simulated Detailed descriptions of the test facilities, experiments,
predictions for all three engine orders together, illustrating the NASTRAN model, simulations (including FORTRAN source
uniqueness of the design point and the value of viewing the code) and results may be found in Cameron (1985). The
data in terms of a damper performance curve of this type. NASTRAN simulations and all tests—shaker table and spin
The correlation between the spin rig test data and predicted rig—were performed at NASA Lewis Research Center.

180/Vol. 112, APRIL 1990 Transactions of the AS ME

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Predicted Engine Performance Menq, C.-H., Griffin, J. H., and Bielak, J., 1986a, "The Influence of
Variable Normal Load on the Forced Vibration of a Frictionally Damped Struc-
ture," ASME Journal of Engineering for Gas Turbines and Power, Vol. 108,
0.14 pp. 300-305.
Menq, C.-H., Bielak, J „ and Griffin, J. H., 19866, "The Influence of
0.12 Microslip on Vibratory Response; Part I: A New Theoretical Model," Journal
•*• 2E Simulation of Sound and Vibration, Vol. 107, pp. 279-293.
0.10 Menq, C.-H., Griffin, J. H., and Bielak, J., 1986c, "The Influence of
-*- 4E Simulation
Microslip on Vibratory Response; Part 2: A Comparison With Experimental
A/N 0.08 •*- 6E Simulation
Results," Journal of Sound and Vibration, Vol. 107, pp. 295-307.
(inches/lbf) Muszynska, A., and Jones, D. I. G., 1983, "On Tuned Bladed Disk
0.06
Dynamics: Some Aspects of Friction Related Mistuning,'' Journal of Sound and
0.04 •• Vibration, Vol. 86, pp. 107-128.
Muszynska, A., Jones, D. I, G., Lagnese, T., and Whitford, L., 1981, "On
0.02 Nonlinear Response of Multiple Blade Systems," in the Shock and Vibration
Bulletin, Vol. 51, Part 3, pp. 88-110.
0.00 Plunkett, R., 1980, "Friction Damping," in Damping Applications for
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 Vibration Control, AMD Vol. 38, Torvik, P. J., Ed., ASME, New York.
Q / N (Ibf/lbf)

APPENDIX
Fig. 10 Spin rig predictions
The following algorithm, based on Griffin's model (1980),
4 Summary and Conclusions determines the maximum response of a blade subject to fric-
Minimization of resonant stresses in turbine blades is a ma- tion damping. Griffin's model applies to structures excited in
jor concern in turbine engine and pump design. A procedure one predominant mode, in which the mode shape is not
has been developed for determining the normal load of a fric- significantly affected by the application of friction damping.
tion damper so that the maximum stress experienced by any The friction also is Coulomb and the nonfriction damping
blade will lie below some designated maximum value. The pro- must be either viscous or small enough to be modeled as
cedure provides a way to integrate analytical friction models viscous without significant error. The dimensional parameters
with finite-element computer models and experimental data to of the blade and friction damper are converted to nondimen-
ensure the accuracy of the model development and the sional parameters in order to conform to Griffin's model and
friction-damper design. A damper performance curve—a nor- use his results directly.
malized plot of A/N versus Q/N— is introduced as a key 1 Select a normal load (N) and form the nondimensional
representation of the response of a frictionally damped blade parameters e and f((f is the fraction of critical damping in the
because it provides a design point that is independent of system when the friction damper is not present)
engine excitation and damping levels—two quantities which
are difficult to determine, especially for new designs. At the " kd
(A-l)
same time the design point optimizes the friction damper for
as large an excitation as possible. Finally, the design procedure (A-2)
describes how to predict the engine behavior of a frictionally
damped blade, and how to design the damper accordingly, 2 Set the excitation (Q) to the lowest level for which the
without spin rig or engine tests. A unique test fixture was damper will begin to slip (shown next) and form the non-
devised which used electromagnets to alter the damper loads. dimensional parameter /
Because the current in the magnets could be varied while the
spin pit test was being conducted, this testing approach pro- Q = JL-*L (A-3)
vided an extensive assessment of damper performance in a
rotating environment and provided experimental corrobora-
tion of the design approach.
/= Q (A-4)
lxN<i>
Acknowledgments 3 The nondimensionalized peak response of the blade, and
This research was supported by NASA, Grant Number the nondimensional natural frequency, are given by the
NAG-3-367. Special thanks are due to Erwin Meyn, of NASA following nonlinear algebraic equations which may be solved
Lewis Research Center, who conducted the spin rig tests and iteratively
assisted in all experimental aspects of the research. 2e
•Am =r+4
— r 2 +- (A-5)
References *iiwff
Beards, C. F., 1983, "The Damping of Structural Vibration by Controlled In- 0 + O.5sin(20)] (A-6)
terfacial Slip in Joints," ASME JOURNAL OF VIBRATION, ACOUSTICS, STRESS, 11
IT
AND RELIABILITY IN DESIGN, Vol. 105, pp. 369-373.
Brown, G. B., Kielb, R. E., Meyn, E. H., Morris, R. E., and Posta, S. J., where
1984, "Lewis Research Center Spin Rig and Its Use in Vibration Analysis of
Rotating Systems," NASA TP-2304.
Cameron, T. M., 1985, "Spin Rig Tests of a Turbine Blade With Friction e(f )
•K
Damping: Extending Bench Test Results to a Rotating Environment," Master (A-7)
of Engineering Report, SM 85-28, Carnegie-Mellon University, Pittsburgh, PA.
Dowell, E. H., and Schwartz, H. B., 1983, "Forced Response of a Cantilever
Beam With a Dry Friction Damper Attached. Part I: Theory, Part II: Experi-
ment," Journal of Sound and Vibration, Vol. 91, No. 2, pp. 269-291.
Griffin, J. H., 1980, "Friction Damping of Resonant Stresses in Gas Turbine ) = COS-'(l—-T~) O<0<TT (A-8)
Engine Airfoils," ASME Journal of Engineering for Power, Vol. 102, pp.
329-333. Start the iterations with an initial guess of wra. Allowable
Griffin, J. H., and Sinha, A., 1985, "The Interaction Between Mistuning and
Friction in the Forced Response of Bladed Disk Assemblies,'' ASME Journal of ranges for the parameters are Am>\, and ( l - e ) < u m 2 < l .
Engineering for Gas Turbines and Power, Vol. 107, pp. 205-211. Convergence is quite rapid.
Jones, D. I. G., 1980, "Vibrations of a Compressor Blade With Slip at the
Root," Air Force Wright Aeronautical Laboratories Report AFWAL-
4 The maximum amplitude of vibration and the frequency
TR-80-4003. at which it occurs are given by

Journal of Vibration and Acoustics APRIL1990, Vol.112/181

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 A - A |~'t^l (\ Q\
5
Increment Q and return to Step 2 (just recalculate/).
~~ m L kd J (A_y
) Continue with Steps 2-5 until the full range of possible excita-
tion levels has been covered.
i, ,— 6 The entire algorithm should be repeated for different
o) = wm I - (A-10) values of N. The results of the simulations are plotted in the
m
normalized form of Fig. 3; (A/N) versus (Q/N).

.Readers of_
The Journal of Vibration and Acoustics
Will Be Interested In;
DE-Vol. 18-3
Structural Vibration and Acoustics
Editors: T.S. Sankar, T.C. Huang, A.V. Srinivasan, R.B. Bhat, S.V. Hoa, V. Ramamurti
This publication serves as a valuable source of new knowledge and information in the field of
vibration engineering. Areas addressed are structural vibration analysis, structural vibrations-
composite applications, bladed-disk assembly, acoustics —modeling, and noise control.
1989 Order No. H0508C 337 pp. ISBN No. 0-7918-0364-3
$90 List / $45 ASME Members

To order, write ASME Order Department, 22 Law Drive, Box 2300, Fairfield, NJ 07007-2300
or call 1-800-THE-ASME (843-2763) or FAX 1-201-882-1717.

182/Vol. 112, APRIL 1990 Transactions of the ASME

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use