RMember AIAA, AHS.

Journal of Sound and <ibration (2002) 250(3), 519}539

doi:10.1006/jsvi.2001.3935, available online at http://www.idealibrary.com on
PASSIVE DAMPING AUGMENTATION OF A VIBRATING
BEAM USING PSEUDOELASTIC SHAPE MEMORY ALLOY
WIRES
F. GANDHIR AND G. CHAPUIS
Department of Aerospace Engineering, he Pennsylvania State ;niversity, 229 Hammond Building,
;niversity Park, PA 16802, ;.S.A. E-mail: [email protected]
(Received 18 December 2000, and in ,nal form 25 July 2001)
This paper examines the e!ectiveness of pseudoelastic shape memory alloy (SMA) wires
for passive damping of #exural vibrations of a clamped}free beam with a tip mass. A
"nite-element model of the system is developed and validated with experimental results. The
SMA behavior is modelled using amplitude-dependent complex modulus. Numerical
simulations indicate that the damping introduced by the SMA wires will increase for higher
excitation-force amplitudes that produce higher strain levels in the SMA wires. Increasing
the wire cross-section area provides more damping at low force-excitation amplitudes, but
reduced damping at higher amplitudes. The angle between the beam and the SMA wires is
an in#uential parameter, and a value in the 10}203 range was found to introduce maximum
damping. The underlying physical mechanisms are examined in detail. System damping
depends only mildly on the SMA wire length, and is una!ected by the tip mass.
2002 Academic Press
1. INTRODUCTION
Nickel}titanium (Ni}Ti) shape memory alloys (SMAs) are known to exhibit stress/strain
pseudoelastic hysteresis associated with stress-induced austenite}martensite phase
transformations, when subjected to cyclic stress/strain loadings above a critical temperature
(the austenite "nish temperature). Distinct from the more commonly used shape memory
e!ect (whereby temperature-induced phase transformations are exploited for use in
low-frequency actuators), pseudoelastic hysteresis of SMAs can potentially be exploited for
passive damping augmentation in various mechanical, civil, and aerospace applications.
A damper can be designed with SMA wires pre-stressed (or pre-strained) to a level
somewhere in the middle of the pseudoelastic range [1]. Cyclic variations in stress and
strain around the baseline, as the system undergoes vibration, will then result in energy
dissipation due to hysteresis in the SMA.
If SMAs are integrated into structures, they would have several potential advantages over
traditional damping materials such as elastomers. For example, the energy dissipated per
unit volume (a measure of damping capacity) is much higher than that provided by
commonly used elastomeric materials [2]. Additionally, available damping from elastomers
can reduce signi"cantly at low temperatures, compromising damper performance in
critical systems. Since SMA damper temperature can be controlled by electrical heating,
degradation in performance due to temperature variation can be avoided. Clearly, high
0022-460X/02/080519#21 $35.00/0 2002 Academic Press
damping capacity and controllability of SMA-based damping systems make them attractive
candidates for structural damping augmentation. In recent years, some studies have already
been reported demonstrating the use of pseudoelastic SMAs for passive structural damping
augmentation (see for example references [3}8]). However, many of these studies have
assumed an overly simpli"ed SMA hysteresis behavior (often due to lack of suitable
models), and do not focus on several fundamental issues such as the impact of the SMA
damper con"guration or excitation amplitude on system damping.
In recent years, there have also been continued e!orts in SMA material characterization
and development of constitutive models suitable for structural dynamics applications. In
particular, there has been emphasis on experimentally determining the SMA pseudoelastic
hysteresis behavior over a range of excitation frequencies. With hysteresis data previously
reported only at very low strain rates (quasi-static conditions), unsuitable for vibration
damping applications, the "rst author and co-workers experimentally characterized
the SMA hysteresis behavior up to frequencies of 10 Hz [2, 9]. There has also been
an emphasis on development of simple yet accurate SMA constitutive models that can
be easily integrated into structural "nite-element analyses. From this standpoint,
thermodynamic-based SMA constitutive models [10}13] that require calculation of
martensite}austenite phase fractions and transformations as a function of stress and
temperature, have limited appeal. A better approach is to characterize SMAs using the
complex modulus. The complex modulus approach, frequently used in structural dynamics
for characterizing damping materials [14}16], comprises a storage modulus (representative
of sti!ness) and a loss modulus (representative of damping), both of which are
frequency- and temperature-dependent. In addition, for non-linear materials the complex
modulus also varies with excitation amplitude [17}19]. A complex modulus representation
of SMA behavior, as presented in reference [20], can be easily integrated into structural
analyses, and in#uence of the SMA on structural response or modal damping can then be
readily calculated.
The detailed characterization of SMA hysteresis behavior and availability of complex
modulus representations thereof, provides a tremendous opportunity to develop
"nite-element analyses and conduct parametric studies to provide detailed physical insight,
as well as optimization studies to evaluate the most e!ective SMA design con"gurations for
structural damping augmentation.
2. OBJECTIVES AND APPROACH OVERVIEW
In the present study, a clamped}free vibrating beam with a tip mass and symmetrically
mounted pseudoelastic SMA wires (schematic sketch shown in Figure 1) is considered to
obtain a fundamental understanding of the in#uence of SMA damper con"guration and
systemparameters, as well as operational parameters such as excitation force amplitudes on
the performance of the SMA-based structural damping treatment.
The beamis subjected to a harmonic excitation force near the tip, and the bending response
of the beam to this input excitation is calculated. An amplitude-dependent complex-modulus
characterization of SMA hysteresis behavior is used, based on experimental data from
references [9, 21]. On account of the non-linear behavior of the SMA wires, the beam periodic
response is calculated iteratively, and damping in the fundamental mode of vibration is
examined. After verifying the accuracy of the "nite-element analysis through experimental
validation, the excitation-force amplitude, SMA wire cross-section area, wire length, angle
between the wires and the beam, and the tip mass are varied, and the in#uence of these
variations on system damping is evaluated and physically explained.
520 F. GANDHI AND G. CHAPUIS
Figure 1. Schematic sketch of clamped}free beam with tip mass and symmetrically mounted pseudoelastic SMA
wires to introduce damping of #exural vibrations.
3. ANALYTICAL MODEL AND SOLUTION SCHEME
To examine the e!ectiveness of the symmetrically mounted pseudoelastic SMA wires in
introducing passive damping in the clamped}free beam with tip mass, a "nite-element
discretization of the system is carried out (Figure 1). The beam is assumed to undergo pure
bending deformations with no extension or transverse shear (Euler}Bernoulli beam theory).
The SMA wires are subjected to a pre-tension load, . The angle between the wires and the
undeformed beam is 0. The SMA behavior is described using a complex modulus. With the
beam oscillating in the horizontal plane, the gravitational force on the tip mass is not
considered. Section 3.1 describes the development of the "nite-element equations of motion.
This is followed by a description of the procedure used to calculate the system response
under a tip harmonic excitation, P"P

eSR, in section 3.2 and a description of the method

used to calculate the damping in section 3.3.
3.1. FINITE-ELEMENT EQUATIONS OF MOTION
In the "nite-element analysis, the Euler}Bernoulli beam is spatially discretized into
several elements, each having four degrees of freedom: w

and w

representing the transverse

displacements at the left and right nodes, and w'

and w'

representing the slope at the two

nodes. The transverse displacement within any element is then assumed to vary as a cubic
function of the non-dimensional local co-ordinate, (0))1), and is related to the
elemental degrees of freedom as
w"(1!3#2)w

#(!2#)lw'

#(3!2) w

#(!#)lw'

, (1)
where l is the length of the element. Using the above-assumed displacement function, the
element sti!ness and mass matrices can be obtained from expressions of the element strain
energy and kinetic energy respectively.
3.1.1. Beam element sti+ness matrix
The element bending sti!ness matrix under the in#uence of a constant tensile force, F, is
derived from the element strain energy, ;:
;"
1
2
j
CJCKCLR
JCLERF
EI(w") dx#
1
2
j
CJCKCLR
JCLERF
F(w') dx"
1
2
q2 K

q#
1
2
q2 K

q, (2)
SHAPE MEMORY ALLOY WIRES 521
where
K

"
EI
l

12 6l !12 6l
6l 4l !6l 2l
!12 !6l 12 !6l
6l 2l !6l 4l

, K

"
F
l

6/5 l/10 !6/5 l/10

l/10 2l/15 !l/10 !l/30
!6/5 !l/10 6/5 !l/10
l/10 !l/30 !l/10 2l/15

(3)
and q"[w

w'

w'

]. (4)
In equation (3), the #exural sti!ness, EI, would assume di!erent values for the portions of
the beam with and without the tip mass. The longitudinal force, F, is due to the pre-tension,
, in the SMA wires and can be shown to be F"!2cos 0 (for portions of the beam
inboard of the point of attachment of the wires). Then, the total element sti!ness matrix is
K"K

#K

for the portion of the beam under pre-compression, and K"K

for the
portion of the beam outboard of the point of attachment of the wires.
3.1.2. Beam element mass matrix
The element mass matrix, M, is obtained from the element kinetic energy :
"
1
2
j
CJCKCLR
JCLERF
m(wR ) dx#
1
2
j
CJCKCLR
JCLERF
b(uR ) d<. (5)
In equation (5), m is the mass per unit length, is the mass density and b is the beam width.
The "rst term represents the element translational kinetic energy due to transverse motion,
and the second term represents the element rotational kinetic energy due to longitudinal
motion. The element mass matrix due to translational kinetic energy, M

, and the
rotational kinetic energy, M

, are given by
M

"
ml
420

156 22l 54 !13l

22l 4l 13l !3l
54 !13l 156 !22l
!13l 3l !22l 4l

,
M

"
:
30

36/l !3 !36/l !3
!3 4l 3 !l
!36/l 3 36/l 3
!3 !l 3 4l

. (6)
It should be noted that the mass per unit length, m, and the rotational interia, :, assume
di!erent values for the portion of the beam without the tip mass. The total element mass
matrix is M"M

#M

.
3.1.3. In-uence of SMA wires
In section 3.1.1, the pre-tension in the SMA wires was shown to in#uence the sti!ness
matrix by subjecting the beam to a steady compressive force. In addition, the SMA wires
undergo cyclic variations in strain due to the oscillations of the beam. The corresponding
522 F. GANDHI AND G. CHAPUIS
Figure 2. Schematic of SMA wire in deformed con"guration.
changes in the SMA wire force directly act on the beam at the point of attachment. This
section develops the equations to represent this e!ect.
Figure 2 shows one of the SMA wires (wire 1) in a deformed con"guration due to the
beam transverse de#ection and rotation at the point of attachment. From the "gure, the
deformed length is
(#)"(h!w
I
)#(X!w'
I
t
@
/2). (7a)
Further, the undeformed wire length can be expressed as
"h#X. (7b)
Subtracting equation (7b) from equation (7a) and neglecting the terms in , w
I
, and
w'
I
(which are of higher order),
2L"!2hw
I
!2Xw'
I
t
@
/2. (8a)
Equation (8a) can be rewritten to obtain the net change in length:
"!
h

w
I
!
X

t
@
2
w'
I
"!sin0 w
I
!cos 0
t
@
2
w'
I
. (8b)
Thus, the extensional strain in wire 1 associated with the beam de#ection is
c

"!
sin0

w
I
!
cos 0

t
@
2
w'
I
. (9)
Similarly, it can be shown that the extensional strain in wire 2 is c

"!c

. Using the
complex modulus representation of SMA material behavior, the corresponding stresses in
wire 1 and 2 are o

"(G'#jG")c

and o

"(G'#jG")c

"!o

. The forces in wire 1 can

then be expressed as
F

"A(G'#jG") c

"!
A(G'#jG")

(sin0 w
I
#cos 0
t
@
2
w'
I
) (10)
and the force in wire 2 is F

"!F

. From the forces in individual wires, the net transverse

force and moment exerted by both SMA wires on the beam is obtained. From Figure 3, the
net transverse force is
F
8
"F

sin0!F

sin0"2F

sin0. (11)
SHAPE MEMORY ALLOY WIRES 523
Figure 3. Force and moment exerted on the beam by the SMA wires at the point of attachment.
Introducing equation (10) into equation (11) yields
F
8
"!
2A(G'#jG")

sin 0 w
I
#sin0 cos 0
t
@
2
w'
I

. (12)
From Figure 3, the net bending moment, M, is
M"F

cos 0(t
@
/2)!F

cos 0(t
@
/2)"2F

cos 0(t
@
/2). (13)
Introducing equation (10) into equation (13) yields
M"!
2A(G"#jG")

(sin0 cos 0(t

@
/2)w
I
#cos 0(t
@
/2) w'
I
). (14)
3.1.4. Global equations of motion
The element sti!ness and mass matrices are assembled to obtain the global sti!ness and
mass matrices. After application of geometric boundary conditions the global "nite-element
equations of motion are expressed as
M%qK %#K%q%"F1+#FCVR, (15)
where K% is the global sti!ness matrix, M% is the global mass matrix, q% is the vector of
global degrees of freedom, F1+ represents the loads exerted on the beamby the SMAwires,
and FCVR represents the external excitation on the system. In the present study, the external
excitation comprises only a concentrated force acting near the beamtip (as seen in Figure 1).
The only non-zero elements in F1+ are the force F
8
(equation (12)) and moment M
(equation (14)) at the node point where the SMA wires are attached to the beam.
Equations (12) and (14) can be represented in the form

F
8
M
"!
2A(G'#jG")

sin 0
sin0 cos 0(t
@
/2)
sin0 cos 0(t
@
/2)
cos 0(t
@
/2)
w
I
w'
I

, (16)
Since the loads exerted by the SMA wires on the beam are motion-dependent, they will have
an in#uence on the global sti!ness matrix. It should be noted that in the 2;2 matrix above,
only the (1, 1) term retains signi"cance (while the (1, 2), (2, 1) and (2, 2) terms become
524 F. GANDHI AND G. CHAPUIS
Figure 4. Flowchart depicting the iterative scheme used to calculate the system response.
negligible) for very slender beams (except for values of 0P0). The global "nite-element
equations of motion are then written as
M%qK %#(K'#jK")q%"FCVR. (17)
The augmented global sti!ness matrix is now complex due to the complex modulus of the
SMA wires.
3.2. SOLUTION OF EQUATIONS OF MOTION USING HARMONIC BALANCE
Since the complex modulus approach is used for characterization of the SMA behavior,
the system is analyzed in the frequency domain. The response to the transverse excitation
force, P"P

eSR, acting close to the beam tip, is calculated using the harmonic balance
method. The applied force and the resulting response in equation (17) can be expressed as
FCVR"F

eSR and q%"(q

P
!jq
G
) eSR. (18)
Using equation (18) in equation (17), it can be shown that

K'!cM%
K"
K"
!(K'!cM%)
q
P
q
G

"

0
or

q
P
q
G

"

K'!cM%
K"
K"
!(K'!cM%)
\

0
. (19)
After calculating the system response (q
P
and q
G
) at any excitation frequency, c, the
amplitude, q

, of the response at some desired nodal degree of freedom, k, near the beam tip
can then be obtained as
q

"(q
P
(k)#q
G
(k). (20)
SHAPE MEMORY ALLOY WIRES 525
It should be emhasized that since G' and G" of the SMA wires are dependent on strain
amplitude (which implies that K' and K" in equation (17) and (19) depend on response
amplitude), q
P
and q
G
have to be evaluated iteratively at any frequency. An estimated value of
G' and G" is used to "rst calculate q
P
and q
G
. With the beam displacement and slope then
available at the point of attachment of SMA wires, the strain amplitude in the SMA wires
can be calculated using equation (9). Based on this strain amplitude, an updated value of G'
and G" is now introduced and the process is repeated until convergence. A #owchart of this
iterative solution procedure is shown in Figure 4. The compliance response function of the
system, H(c)"q

(c)/p

, can then be obtained by calculating the response q

over a range of
excitation frequencies.
3.3. EVALUATION OF SYSTEM DAMPING
After calculating the system response (as described in section 3.2), the modal damping can
be evaluated. One measure of damping is the modal damping ratio calculated by applying
the half-power bandwidth method to the compliance response function, H(c). However, the
half-power bandwidth method and the modal damping ratio are rigorously applicable only
for linear systems. Alternatively, the modal loss factor, p, which represents a ratio of
dissipated energy, D, to the maximum energy stored in a cycle, ;
K?V
, can be calculated. For
a linear system, the modal loss factor at resonance condition is simply twice the modal
damping ratio. In the present study, the modal loss factor at resonance is used as
a quantitative measure of damping. It can be shown that the maximum stored energy in
a cycle is given by [21]
;
K?V
"
1
2
(q2
P
K'q
P
#q2
G
K'q
G
) (21)
and the energy dissipated in a cycle, D, is given by [21]
D"(q2
P
K"q
P
#q2
G
K"q
G
). (22)
Alternatively, since energy is only dissipated in the SMA wires, the dissipated energy can be
obtained by calculating the area inside the hysteresis loop:
D"( j o dc) <"G" c

<, (23)
where < is the wire volume and c

is the SMA strain amplitude. The loss factor is then

obtained as
p"
1
2
D
;
K?V

c"c
P
. (24)
In equation (24), ;
K?V
is the sum of the energy stored in the SMA wires (;
1+
) and in the
beam. Noting that ;
1+
"

G'c

<, and using the de"nition for SMA material loss factor,
p
1+
"G"/G', it can be shown that the modal loss factor of the system can be expressed in
the form
p"(;
1+
/;
K?V
)
c"c
P
p
1+
. (25)
4. SMA AMPLITUDE-DEPENDENT COMPLEX MODULUS
In section 3, the complex modulus approach is used to represent the SMA behavior in the
"nite-element formulation. The complex modulus, which is strongly amplitude-dependent
526 F. GANDHI AND G. CHAPUIS
Figure 5. Compilation of data showing SMA complex modulus variation versus strain amplitude.
Figure 6. (a) Approximation of SMA complex modulus variation versus strain amplitude; (b) SMA wire
(material) loss factor versus strain amplitude.
for SMAs, is calculated from hysteresis data obtained through cycling pre-strained SMA
wires in an MTS machine. These tests, over a range of strain amplitudes and frequencies, are
described in references [9, 21]. Figure 5 shows the variation of the storage modulus, G', and
the loss modulus, G", versus strain amplitude for several excitation frequencies. It is seen
that storage modulus decreases for higher strain amplitudes, while the loss modulus
displays a maximum at moderate strain amplitudes. Further, both storage and loss moduli
show relatively small variations with changes in frequency (between 0)5 and 10 Hz).
Therefore, for simplicity, the complex modulus is assumed to be invariant with the
excitation frequency in this study. A &&master-curve'' (Figure 6(a)) is then generated using
a least-square polynomial approximation [21] of the data in Figure 5. The variation of the
SMA loss factor, p
1+
"G"/G', versus strain amplitude, is shown in Figure 6(b).
SHAPE MEMORY ALLOY WIRES 527
Figure 7. Schematic of experimental set-up.
5. RESULTS AND DISCUSSION
The "nite-element analysis and the response and damping calculation methods outlined
in section 3 are "rst validated against experimental measurements (section 5.1). Thereafter,
variation in damping with excitation amplitude, as well as parameters such as SMA wire
area, wire length, wire angle relative to the beam, and tip mass are numerically examined.
The pre-tension force in the SMA wires is selected to allow the wires to be strained inside
the hysteresis loop, while ensuring that the beam does not buckle.
5.1. DESCRIPTION OF EXPERIMENT AND VALIDATION WITH TEST DATA
The experimental set-up used to validate the "nite-element formulation is depicted
schematically in Figure 7. The base of the set-up is a 1" thick aluminum plate, "xed on
a sturdy table. A net of holes is drilled on the surface of the plate to "x clamps and
mountings. An additional plate is welded on the side to "x a shaker. As seen in the "gure,
two wire clamps, one beam clamp, and a system of pulleys are installed on the plate. The
beam clamp holds an aluminum beam that has a mass attached at its end. The tip mass
comprises steel slabs screwed on the beam. The SMA wires are clamped between the beam
and the pieces of steel constituting the tip mass. The wires pass through the wire clamps
installed on the base plate. These clamps can be tightened, holding the wires in place. After
passing through the system of pulleys, the ends of the wires are attached to adjustable tip
masses used to introduce pre-tension in the wires.
528 F. GANDHI AND G. CHAPUIS
Figure 8. Frequency response function, validation of FEM simulation (**) with experimental data (} } },
random noise; , sine dwell).
The shaker is connected to the beam tip by a &&stinger'' of low #exural sti!ness and high
axial compressive sti!ness to decouple the structure from the rotary inertia of the shaker.
Aload cell is installed between the stinger and the structure to monitor the force input to the
system. An accelerometer is placed on the other side of the beam, at the same axial location
to monitor the acceleration near the tip of the beam in the transverse direction. The
accelerometer and load cell signals are monitored using a digital signal analyzer and
recorded for processing.
The system properties are given below: the aluminumbeamhas a length of 41 cm, a width
of 48)4 mm, and a thickness of 5)05 mm. The SMA wires are 12)8 cm in length, 0)51 mm in
diameter, are mounted at an angle of 19)13 relative to the beam, and have a pre-tension of
87 N. The tip mass is 1)22 kg and extends along the beam length from 35 to 38 cm. The
shaker, load cell and accelerometer are connected 1 cm from the beam tip. The SMA wires
were pre-cycled to approximately 400 stress/strain cycles (as suggested in reference [9]) at
strain amplitudes corresponding to the maximum tip displacement available with the
shaker, to stabilize the hysteresis behavior before collecting data.
Random noise tests as well as sine dwell tests were carried out. The amplitude of the
compliance frequency response function is shown in Figure 8. The correlation between
the "nite-element predictions and experimental results is very good. Due to limitations
of the shaker, large excitation input could not be achieved so results in Figure 8 do not
display non-linear behavior associated with the amplitude-dependent modulus of the
SMA wires.
For the correlation results in Figure 8, the values of SMA storage and loss modulus used
in the "nite-element analysis were G'"4)15;10 psi and G""0)15;10 psi (instead of
G'"3)85;10 psi and G""0)11;10 psi, as seen in Figure 6(a)). The higher value for G'
was used because of the lower ambient temperature during the beam vibration test
(reference [9] suggests that G' at 70F is higher than that at 90F). The larger value of G" used
is meant to account for factors such as friction at the wire clamps, etc., in an approximate
sense.
SHAPE MEMORY ALLOY WIRES 529
Figure 9. (a) Compliance response functions for increasing values of excitation amplitude, P

; (b) variation of
modal damping versus excitation force amplitude, P

; (c) variation of SMA wire strain level at resonance versus

excitation amplitude, P

.
5.2. VARIATION IN EXCITATION-FORCE AMPLITUDE
Figure 9(a) shows compliance response functions corresponding to di!erent
excitation-force amplitudes. For the simulations in this section and in section 5.3, the
following system properties are used: the beam length is 50 cm, width is 80 mm, and
thickness is 3)5 mm. The SMA wires are 12)8 cm in length, have a cross-sectional area of
0)203 mm, are mounted at an angle of 203 relative to the beam, and have a pre-tension of
87 N. The tip mass is 14 kg and extends along the beam length from 40 to 47 cm. From
Figure 9(a), it is seen that for larger excitation forces, the compliance response function
clearly displays the characteristics of a non-linear softening system (since the SMA storage
modulus decreases, Figure 6(a), implying that the wires &&soften'' at higher strain
amplitudes), and the peak response is observed at lower frequencies. Figure 9(b) shows the
variation in modal damping versus force amplitude. Damping is seen to increase upto
a force amplitude of 4)6 N. Over this force amplitude range (0}4)6 N) the peak-to-peak
strains in the SMA increase upto a level 2)7% (see Figure 9(c)). Recall from Figure 6(b) that
the SMA wire loss factor increases signi"cantly upto a peak-to-peak strain level of 2)7%,
and decreases thereafter. Thus, when the force amplitude exceeds 4)6 N and peak-to-peak
530 F. GANDHI AND G. CHAPUIS
Figure 10. (a) Compliance response function for increasing SMA cross-section area; (b) variation of modal
damping versus SMA wire cross-section area; (c) variation of SMA wire strain level versus SMA wire cross-section
area; P

"0)01 N.
strains in the SMAwires exceed 2)7% the system damping reduces due to the decrease in the
SMA wire loss factor. The increase in damping seen in Figure 9(b) (up to 4)6 N) causes
reductions in the peaks of the compliance response function in Figure 9(a). The decrease in
damping beyond 4)6 N causes the response peaks in Figure 9(a) to increase again.
5.3. INFLUENCE OF PARAMETER VARIATIONS ON MODAL DAMPING
5.3.1. In-uence of SMA wire cross-section area
Figures 10(a) and 11(a) show compliance response functions, at low and high
excitation-force amplitudes, respectively, for varying values of SMA wire cross-sectional
area. In both "gures, the system natural frequency is seen to increase as the wire area
increases. This is reasonable as the system sti!ness increases linearly as the wire area
increases (see equation (16)).
Figure 10(b) shows the variation in modal damping versus SMA wire area (normalized by
the baseline wire area) for low excitation-force amplitude, and Figure 10(c) shows
SHAPE MEMORY ALLOY WIRES 531
Figure 11. (a) Compliance response function for increasing SMA cross-section area; (b) variation of modal
damping versus SMA wire cross-section area; (c) variation of SMA wire strain level versus SMA wire cross-section
area; P

"1)0 N.
the corresponding variation in SMA peak-to-peak strain. Similar results for high
excitation-force amplitude are shown in Figures 11(b) and 11(c). It should be recalled that
the modal loss factor depends on the product of the SMA wire loss factor, p
1+
, and the
strain energy ratio, ;
1+
/;
K?V
(equation (25)). At low force amplitudes, the SMA
peak-to-peak strain shows little absolute variation with increasing wire area (Figure 10(c)).
Thus, the SMA wire loss factor, p
1+
, is relatively unchanged, and it can be deduced that the
increase in modal loss factor in Figure 10(b) is due to a corresponding increase in strain
energy ratio. At a higher force amplitude, the SMA peak-to-peak strains sharply decrease
between 3 and 0)75%, with increasing wire area (Figure 11(c)). A decrease in strain over this
range corresponds to a sharp decrease in SMA wire loss factor, p
1+
(see Figure 6(b)). Then,
the overall variation in modal damping seen with increasing wire area (Figure 11(b)) is
a combination of an increase due to increasing strain energy fraction and a decrease due to
a reduction in SMA wire loss factor.
5.3.2. In-uence of SMA wire length
Figures 12(a) and 13(a) show compliance response functions, at low and high excitation-force
amplitudes, respectively, for varying values of SMA wire length. In both "gures, the system
532 F. GANDHI AND G. CHAPUIS
Figure 12. (a) Compliance response function for increasing SMA wire length; (b) variation of modal damping
versus SMA wire length; (c) variation of SMA wire strain level versus SMA wire length; P

"0)01 N.
natural frequency is seen to decrease as the wire length increases. This is consistent with
equation (16), which indicates that the in#uence of SMA wires on system sti!ness is
inversely proportional to the wire length.
Figures 12(b) and 13(b) show that for both low as well as high excitation-force
amplitudes, the system damping has relatively low sensitivity to SMA wire length.
The peak-to-peak strains in the SMA wire (Figures 12(c) and 13(c)) do not show much
sensitivity to wire length either. Consequently, the SMA wire loss factor, p
1+
(which
depends on the wire strain), does not vary signi"cantly with wire length. Since the strain
energy ratio, ;
1+
/;
K?V
, also has low sensitivity to wire length, it can be deduced from
equation (25) that the overall system damping would not be highly sensitive to the SMA
wire length.
5.3.3. In-uence of angle, 0, between SMA wires and beam
Figures 14(a) and 15(a) show compliance response functions, at low and high excitation-
force amplitudes, respectively, for varying values of the angle, 0, between the beam and the
wires. From equation (16) it is seen that the in#uence of the wires on the system sti!ness is
dominated by the sin 0 term (the other terms are smaller since the beamis slender). Thus, as
SHAPE MEMORY ALLOY WIRES 533
Figure 13. (a) Compliance response function for increasing SMA wire length; (b) variation of modal damping
versus SMA wire length; (c) variation of SMA wire strain level versus SMA wire length; P

"4)0 N.
0 increases, the system sti!ness increases, resulting in the increase in natural frequency seen
in Figures 14(a) and 15(a).
Figures 14(b) and 15(b) show the variation in system damping versus 0, for low and
high excitation-force amplitudes respectively. The corresponding peak-to-peak strain
amplitudes in the SMA wire are shown in Figures 14(c) and 15(c). The strain levels in the
wire are seen to decrease asymptotically with increasing 0, and show little change for values
of 0'30 or 403. Thus, the wire loss factor, p
1+
, is unchanged over this range, and it can be
deduced that the decrease in system damping seen in Figures 14(b) and 15(b) must be
associated with a reduction in the strain energy ratio, ;
1+
/;
K?V
(equation (25)). Figure 16
shows the mode shapes of the beam for wire angles of 15, 30 and 903 respectively. Clearly, as
0 increases in the range of 303}903, the signi"cant change in the mode shape (more curvature
along the beam length for the same displacement at the point of attachment of the SMA
wire) would result in a larger fraction of the strain energy being stored in the beam, and
a smaller fraction in the SMA wires. At very low values of 0, the system damping could
again decrease because the strain in the SMA wires per unit tip displacement of the beam is
small (equation (9)). This is clearly seen at low force amplitudes (Figure 14(b)). At higher
force amplitudes (Figure 15(b)), however, this e!ect appears to be compensated by the
534 F. GANDHI AND G. CHAPUIS
Figure 14. (a) Compliance response function for increasing angle, 0; (b) variation of modal damping versus
angle, 0, between wire and beam; (c) variation of SMA wire strain level versus angle 0; P

"0)01 N.
increase in SMA wire loss factor, p
1+
, due to increased strain amplitude levels. An angle of
around 0"153 between the wire and the beam appears to be optimal.
5.3.4. In-uence of tip mass
Figures 17 and 18 show compliance response functions, at low and high excitation-force
amplitudes, respectively, for increasing values of tip mass. In each of these "gures, the system
natural frequency is seen to decrease as the tip mass increases but the peak (resonant)
response amplitude is independent of the tip mass. Thus, the tip mass has no in#uence on the
operating peak-to-peak strain levels in the SMA wires or the operating wire loss factor,
p
1+
(for a given force-excitation amplitude). Consequently, the system damping is also
independent of the tip mass. For the two force-excitation amplitudes considered, the resonant
peak-to-peak strains in the SMA wire are 0)026 and 2)15%, respectively, and the corresponding
system loss factor values were 2)6 and 31)5%, respectively, irrespective of the tip mass.
6. SUMMARY AND CONCLUDING REMARKS
A "nite-element analysis is developed of a clamped}free vibrating beam with a tip mass,
and symmetrically mounted pseudoelastic SMA wires to introduce passive damping. The
SHAPE MEMORY ALLOY WIRES 535
Figure 15. (a) Compliance response function for increasing angle, 0; (b) variation of modal damping versus
angle, 0, between wire and beam; (c) variation of SMA wire strain level versus angle 0; P

"4)0 N.
Figure 16. Mode shapes of the beam for di!erent values of the angle, 0, between beam and wire.
536 F. GANDHI AND G. CHAPUIS
Figure 17. Compliance response function for increasing values of tip mass, P

"0)01 N.
Figure 18. Compliance response function for increasing values of tip mass, P

"4)0 N.
"nite-element results showed excellent correlation with experimental results. The SMA
behavior is represented using an amplitude-dependent complex modulus (based on
previously reported experimental data). The beam is subjected to a harmonic excitation
force near the tip, and its bending response is calculated using the harmonic balance
method. The response calculation is iterative on account of the non-linear behavior of the
SMA wires. Damping in the fundamental mode of vibration is examined by considering the
system loss factor at resonance. The e!ects of excitation-force amplitude, SMA wire
cross-section area, wire length, the angle between the SMA wire and the beam, and the
tip mass, on available damping, were examined in detail. The following observation were
made:
SHAPE MEMORY ALLOY WIRES 537
(1) The SMA wire loss factor increases with strain amplitude up to about 2)7%
peak-to-peak strain. Thus, in general, larger amount of damping is available at higher
excitation}force amplitudes that produce operational strain amplitudes in the SMA
wire around this range. In the present study, it was seen that modal loss factors varied
from around 2)5% at low excitation-force amplitudes, to over 30% for high excitation
levels.
(2) In addition to the operational SMA wire loss factor (which depends on the strain
amplitude in the wire), system damping depends on the fraction of the strain energy in
the SMA wire (relative to the total strain energy in the system). Higher values of this
strain energy ratio result in better damping performance.
(3) Dependence on SMA wire cross-section area: at low excitation levels, more damping
is available for larger wire cross-section area. However, at higher excitation levels, the
damping may decrease with increasing cross-section area. This is because even though
the strain energy ratio increases, there is a signi"cant reduction in the SMA wire loss
factor at the lower strain amplitudes obtained in the thicker (sti!er) wires.
(4) Dependence on SMA wire length: the damping level is only mildly sensitive to this
parameter.
(5) Dependence on the angle between the SMA wire and the beam: the available damping is
strongly dependent on this parameter, with the optimal value being in the range of
15}203. For larger angles, the strain energy ratio reduces due to a signi"cant change in
the mode shape of the beam.
(6) The damping introduced in the system by the SMA wires is independent of the tip mass.
REFERENCES
1. D. E. HODGSON 1988 ;sing Shape Memory Alloys. Sunnyvale, CA: Shape Memory Applications,
Inc.
2. D. WOLONS, F. GANDHI and B. MALOVRH 1998 Journal of Intelligent Material Systems and
Structures 9, 116}126. Experimental investigation of the pseudoelastic hysteresis damping
characteristics of shape memory alloy wires.
3. J. A. INAUDI, J. M. KELLY, W. TANIWANGSA and R. C. KRUMME 1993 Proceedings of Damping 193,
San Francisco, HAA-1}HAA-20. Analytical and experimental study of a mass damper using
shape memory alloys.
4. P. R. WITTING and F. A. COZZARELLI 1993 Proceedings of Damping 193, San Francisco,
ECC-1}ECC-19. Design and seismic testing of shape memory structural dampers.
5. E. RIVIN and L. XU 1994 Proceedings of the ASME, Materials for Noise & <ibration Control,
NCA-Vol. 18/DE-Vol. 80, 35}41. Damping of NiTi shape memory alloys and its application for
cutting tools.
6. P. THOMPSON, G. I. BALAS and P. H. LEO 1995 Smart Materials and Structures 4, 36}41. The use
of shape memory alloys for passive structural damping.
7. Y. C. YIU and M. E. REGELBRUGGE 1995 Proceedings of the 36th AIAA/ASME/ASCE/AHS/ASC
Structures Structural Dynamics and Materials Conference, 3390}3398. Shape memory alloy
isolators for vibration suppression in space applications.
8. P. W. CLARK, I. D. AIKEN, J. M. KELLY, M. HIGASHINO and R. C. KRUMME 1995 Proceedings of
SPIE on Smart Structures and Materials 2445, 241}251. Experimental and analytical studies of
shape memory alloy dampers for structural control.
9. D. WOLONS 1997 M.S. hesis in Aerospace Engineering, he Pennsylvania State ;niversity. An
experimental investigation of the pseudoelastic damping characteristics of NiTi shape memory
alloy wires.
10. K. TANAKA 1990 Journal Pressure <essel echnology 112, 158}163. A phenomenological
description on thermomechanical behavior of shape memory alloys.
11. C. LIANG and C. ROGERS 1990 Journal of Intelligent Material Systems and Structures 1, 207}234.
One-dimensional thermomechanical constitutive relations for shape memory materials.
538 F. GANDHI AND G. CHAPUIS
12. L. C. BRINSON 1993 Journal of Intelligent Material Systems and Structures 4, 229}242.
One-dimensional constitutive behavior of shape memory alloys: thermomechanical derivation
with non-constant material functions and rede"ned martensite internal variable.
13. J. G. BOYD and D. C. LAGOUDAS 1994 Proceedings of the ASME, Mechanics of
Phase ransformations and Shape Memory Alloys, AMD-Vol. 189/PVP-Vol. 292, 159}177.
A constitutive model for simultaneous transformation and reorientation in shape memory
materials.
14. B. J. LAZAN 1969 Damping of Materials and Members in Structural Mechanics Oxford: Pergamon
Press; "rst edition.
15. A. D. NASHIF, D. I. G. JONES and J. P. HENDERSON 1985 <ibration Damping New York: John
Wiley & Sons.
16. W. FLUGGE 1967 <iscoelasticity. New York: Blaisdell.
17. F. GANDHI and I. CHOPRA 1994 Journal of the American Helicopter Society 39, 59}69. An
analytical model for a nonlinear elastomeric lag damper and its e!ect on aeromechanical stability
in hover.
18. F. GANDHI and I. CHOPRA 1996 Journal of the American Helicopter Society 41, 267}277. Analysis
of bearingless main rotor aeroelasticity using an improved time-domain nonlinear elastomeric
damper model.
19. F. GANDHI and I. CHOPRA 1996 Smart Materials and Structures 5, 517}528. A time-domain
non-linear viscoelastic damper model.
20. F. GANDHI and D. WOLONS 1999 Smart Materials and Structures 8, 49}56. Characterization of the
pseudoelastic damping behavior of shape memory alloy wires using complex modulus.
21. G. CHAPUIS 1999 M.S. hesis in Aerospace Engineering, he Pennsylvania State ;niversity.
Use of pseudoelastic shape memory alloy wires for passive damping augmentation of a vibrating
beam.
SHAPE MEMORY ALLOY WIRES 539