Sie sind auf Seite 1von 9

Proceedings Third National Congress on Computational Mechanics

N. Aravas & J. T. Katsidadelis, Eds., 137144. Volos, Greece, 1999.


BENDING OF VISCOELASTIC BEAMS UNDER
PIEZOELECTRIC CONTROL ANALYTICAL AND
MASSIVELY PARALLEL COMPUTATIONAL SIMULATIONS
1
Cristina E. Beldica, Harry H. Hilton and Seid Koric
Aeronautical and Astronautical Engineering Department
National Center for Supercomputing Applications
Center for Simulation of Advanced Rockets
University of Illinois at Urbana-Champaign
104 South Wright Street, Urbana, IL 61801-2935 USA
1. SUMMARY
Analytical and numerical solutions are presented for viscoelastic beams with piezoelec-
tric devices. Probabilities of failure are calculated. The ability of current ABAQUS
TM
programs were examined and found accurate in handling viscoelastic meshes of dierent
sizes and/or dierent viscoelastic materials.
2. INTRODUCTION
Piezoelectric eects are of increasing importance in the control and sensing of structural
element deections. Elastic piezoelectric theory has been extensively developed (Lei-
bowitz & Vinson 1993). Hilton et al. (1997) have presented an extended bibliography
and discussed past contributions as well as formulated an analytical nonlinear theory
of anisotropic piezoelectric-thermo-viscoelasticity. Only a limited number of analyti-
cal and experimental publications on linear piezo-viscoelastic materials may be found
in the literature (Holloway & Vinogradov 1997, Vinogradov & Holloway 1997). Yi et
al. (1996) have formulated nite element solutions to demonstrate viscoelastic material
damping. Hilton et al. (1997, 1998a, b) and Beldica et al. (1998a, b, c) have applied
this theory to a number of linear viscoelastic bending and aero-viscoelastic problems,
such as creep utter and torsional divergence and the response of piezo-viscoelastic
structures to aerodynamic noise. Yi et al. (1997) have formulated and carried out nite
element solutions for a number of piezoelectric viscoelastic problems.
In the present paper, the eects of combined bending and shear stresses are analyzed
using parallel nite element analyses with existing ABAQUS
TM
subroutines. The formu-
lation of nonuniform beams leads to solutions which can only realized through massively
parallel computations. Viscoelastic piezoelectric devices are attached to the upper and
lower surfaces of the beam (Fig. 1). The numerical simulations lead to determina-
tions of states of stress and deformation as functions of position and time, from which
failure probabilities can be determined through the use of Weibull probability density
functions.
3. ANALYSIS
3.1 Bending of Beams Small Deformations
Consider a constant temperature nonhomogeneous variable cross section beam in a
Cartesian coordinate system x = (x
i
) with i = 1, 2, 3 under a loading F
3
(x
1
, t) resulting
in small deformations and strains with linear viscoelastic properties. Viscoelastic piezo-
electric devices are attached to the upper and lower surfaces of the beam at x
3
= h
b
/2
(Fig. 1). Equilibrium in bending is given by two surface integrals over the cross sectional
area A
1
This research is supported by NCSA DoD Grant No. DAHC94-96-C-0005, High Performance Com-
puting Modernization Program (HPCMP-PET) and by CSAR DoE Grant No. B341494, Accelerated
Strategic Computing Initiative (ASCI).
1

_
A

11
(x
1
, x
3
, t) dA = 0
_
A

11
(x
1
, x
3
, t) x
3
dA = M
2
(x
1
, t) (1)
where M
2
is the applied bending moment. Shear due to bending can be derived from
purely equilibrium considerations and in thin walled beams of thickness b(x
1
, x
3
) with

23
= 0, it is given by

13
(x
1
, x
3
, t) =
1
b(x
1
, x
3
)
_
h
x
3

11
(x
1
, x

3
, t)
x
1
b(x
1
, x

3
) dx

3
(2)
where h = h
b
/2 + h
a
. For Euler-Bernoulli beams with plane sections remaining plane,
the normal strains
11
become

11
(x
1
, x
3
, t) = K(x
1
, t) x
3
=
w(x
1
, t)
x
1
x
3
(3)
where w is the beam neutral axis deection in the x
3
-direction dened as

2
x
2
1
_
_

I(x
1
)
t
_

11
(x
1
, t, t

3
w(x
1
, t

)
x
2
1
t

dt

_
_
= F
3
(x
1
, t) (4)
and where the moment of inertia

I(x
1
) is

I(x
1
) =
h
b
/2
_
h
b
/2
b(x
1
, x
3
) x
2
3
dx
3
+ 2 h
2
c
h
a
b(x
1
, h
c
) with h
c
= (h
b
+h
a
)/2, h
a
h
b
/2
(5)
A degenerate form of Eq. (4) with restricted boundary conditions may be expressed as

I(x
1
)
t
_

11
(x
1
, t, t

)

3
w(x
1
, t

)
x
2
1
t

dt

= M
2
(x
1
, t) (6)
After manufacturing, the in-service generalized isothermal small deformation linear
isotropic or orthotropic viscoelastic constitutive relations without thermal expansions
for a composite are given by (Hilton et al. 1997)

11
(x
1
, x
3
, t) =
3

k=1
_
_
_
t
_

kkij
(x
1
, x
3
, t t

)

kk
(x
1
, x
3
, t

)
t

dt

+ H(x
3
.5 h
b
)
t
_

E
kkij
(x
1
, h, t t

)
E
kk
(x
1
, t

)
t

dt

_
_
_
(7)
and

13
(x
1
, x
3
, t) =
t
_

1313
(x
1
, x
3
, t t

)

13
(x
1
, x
3
, t

)
t

dt

+ H(x
3
.5 h
b
)
t
_

E
1313
(x
1
, h, t t

)
E
13
(x
1
, t

)
t

dt

(8)
where H(x) is the Heaviside step function. For this beam 0 |x
3
| .5h
b
+h
a
= h and

22
=
33
=
12
=
23
= 0. The presence of any or all of the
E
kkl
and
E
13
integrals
2

will depend on the orientation of the piezoelectric devices and on nonzero values of the
electric potentials E
kk
and E
13
. In the simplest 1-D case, Eq. (7) reduces to

11
(x
1
, x
3
, t) =
t
_

1111
(x
1
, x
3
, t t

)

11
(x
1
, x
3
, t

)
t

dt

+ H(x
3
.5 h
b
)
t
_

E
1111
(x
1
, h, t t

)
E
11
(x
1
, t

)
t

dt

(9)
The relaxation functions for the beam are expressible in terms of Prony series, such
that

ijkl
(x
1
, x
3
, t) =

ijkl
(x
1
, x
3
) +
N
ijkl

n=1

ijkln
(x
1
, x
3
) exp
_
t

ijkln
(x
1
, x
3
)
_
(10)
where the underscore indicates no summation over so designated indices and where the

ijkl
,
ijkl
and N
ijkl
are material property parameters, to be experimentally determined.
Similar relaxation functions, but each with distinct parameters, describe the behavior
of the tape and wire of the piezoelectric devices.
The voltage V
kl
(x
1
, t) in the viscoelastic piezoelectric devices is given by
V
ij
(x
1
, t) = h
a
t
_

E
ijkl
(x
1
, h
a
, t, t

)

kl
(x
1
, h
a
, t

)
t

dt

(11)
where
E
are the creep functions of the piezo devices.
3.2 Multiaxial Failure Simulations
Hilton & Ariaratnam (1994) have formulated an empirical invariant combined stress
failure law, which is applicable to both deterministic and stochastic conditions. These
probabilistic formulations have been applied to column creep delamination buckling and
to linear viscoelastic service and manufacturing delamination onset predictions (Hilton
et al. 1993, 1996a, b, 1997a, b). Hiel al. (1991) have experimentally shown that uniaxial
composite delamination failures obey Weibull distributions (Lin 1967).
Consider applied random stress invariants dened by

J
1
=
ii

J
2
=
ij

ij

J
3
=
ij

ik

jk

J
1
=

F
ii

J
2
=

F
ij

F
ij
(12)
with similar expressions for the uniaxial random failure stresses

F
ij
and their means F
ij

J
2
=

F
ij

F
ij

J
3
=

F
ij

F
ik

F
jk
(13)
Functions with a

indicate stochastic quantities, while those without are their corre-
sponding mean values. Mean uniaxial viscoelastic delamination strengths are shown in
Fig. 2 (Dillard & Brinson 1983).
These failure criteria are cast in the forms of
1
3
3

i=1
_

J
i
(x, t)
J
i
(x, t)
_
c
i
=

V (x, t) and
1
3
3

i=1
_

J
i
(x, t)
J
i
(x, t)
_
c
i
= v(x, t) (14)
and local failure occurs whenever
u(x, t) =

V (x, t) v(x, t) 0 (15)
Uniaxial failure data by Hiel et al. (1991) for epoxy/ber composite delaminations
indicate Weibull distributions (Hilton & Yi 1993, Lin 1967), consequently for combined
3

loads the local failure probability

P
L
and the failure probability

P for any structural
component then become

P
L
(x, t) = 1 exp
_
_
_

_
u(x, t)

_
A
_
_
_
and

P(t) = max
_

P
L
(x, t)
_
(16)
where A and are experimentally determined material parameters.
For the present beam conguration, the invariants are

J
1
=
11

J
2
=
2
11
+ 2
2
13
+ 2
11

13

J
3
=
3
11
+
3
13
+ 3
11

2
13
+ 3
2
11

13
(17)
with similar expressions for the failure stress invariants

J
i
in terms of

F
11
and

F
13
. A
typical deterministic invariant failure surface is shown in Fig. 3, where delamination
onset occurs for combination values of J
1
, J
2
and J
3
on or above the depicted surface.
3.3 Numerical Analysis Finite Element / Finite Dierence Formulations
Yi & Hilton (1993, 1994) have formulated FE / FD protocols based on Newmarks
average acceleration method for the analysis of linear viscoelastic media and have shown
that a recursion expression can be derived based on the two previous time steps which
leads to faster CPU times, because larger time steps can be executed for the same
computational accuracy although requiring twice the storage. Symbolically, the FE
relations for the displacements U(x, t) can be written as
M(x)
d
2
U(x, t)
dt
2
. .
inertia
+
t
_

k[x, t, t

, T(x, t

), M(x, t

)]
U(x, t

)
t

dt

. .
nite element viscoelastic stiness
= F(x, t)
. .
nodal forces
(18)
which leads to the recursive relations
U
n
(t
p
) = F [U
n
(t
p1
), U
n
(t
p2
), t
p
, M
mn
(t
p
), k
mn
(t
p1
), k
mn
(t
p2
), F
m
(t
p
)] (19)
(m, n = 1, 2, 3, N
dof
)
with N
dof
the number of nodal degrees of freedom and where all the functions on the
r.h.s. of Eq. (19) are known for the current time t
p
as well as the two immediate past
ones t
p1
and t
p2
. Material moisture content is represented by M(x, t), whereas M(x)
is the local beam mass.
Yi et al. (1993a, 1995) have conducted performance evaluations on computers with a
modest number of parallel processors using the above protocols. Their work serves as
a starting point for the present massively parallel study. In the present paper, the rst
illustrative example is chosen as the analysis of a prismatic beam in order to validate
the computational protocol by comparing its solution to the analytical ones obtained by
Hilton et al. (1997, 1998a, c). For prismatic beams, all state variables and material and
geometric properties become independent of the axial position x
1
and lead to results
which are at most functions of x
3
and t.
The parallel computations were performed using ABAQUS
TM
with no additional sub-
routines to handle the two distinct viscoelastic materials.
4. DISCUSSION
4.1 Results
4

When these loading and material conditions are applied to viscoelastic composite beams,
delamination failures are possible in time. Typical deterministic uniaxial viscoelastic
delamination strengths are shown in Fig. 2 (Dillard & Brinson 1983). Additionally,
experimental failure data by Hiel et al. (1991) for epoxy/ber composite delaminations
indicate Weibull distributions. Invariant deterministic and probabilistic stress failure
theories have been developed by Hilton & Ariaratnam (1994). A representative deter-
ministic invariant failure surface is shown in Fig. 3, where delamination onset occurs
for combination values of the stress invariants J
1
, J
2
and J
3
on or above the depicted
surface.
Figs. 4 and 5 show stresses and strains in the piezo strip and in the immediately adjacent
beam bers. This simulation was selected because an analytical solution exists allowing
comparison (Hilton et al. 1997) and these results were found to compare well with the
corresponding analytical ones. The dimensions of the simulation are h
a
= .01, h
b
= 9.5
and L = 500 and with viscoelastic material properties E
o
= 10000 for both beam and
piezo strip, relaxation times
P
= 5 and
B
= 1. A constant end load of 100 was applied,
resulting in longitudinal total forces which are time independent, while the beam and
piezo stresses vary dierentially in time due to the distinct material properties.
Fig. 6 exhibits the contributions of two viscoelastic piezo devices with distinct voltages.
The horizontal line represents beam bending stresses without piezo devices under a
constant end load, which eventually meets in time the failure stresses of Fig. 2 causing
delamination. The application of piezo voltages raises the stresses in the early time
frame when failure stresses have not yet degraded and then attempts to stay behind the
degradation rate. The constant voltage scheme is unsuccessful in the long time period,
however application of a linearly increasing voltage with time is able to retard the onset
of failure until full relaxation takes place. It must, of course, be remembered that if the
beam is part of a ight structure of limited time duration, such as a missile, it needs
to survive only for the duration of the ight and not for ever. The creep pattern of the
neutral axis deections at span three locations can be seen in Fig. 7 and they are seen
to stabilize as the stresses become fully relaxed.
These analytical and computational simulation results are applied through the use of
invariant failure theories of Hilton & Ariaratnam (1994) to the present problem and
produce delamination onset probabilities as shown in Fig. 8. The eects of controlling
deformations by extracting electrical energy through resistors or by the application of
voltages to the piezoelectric devices is also depicted in these curves. In eect, the piezo
devices while not eliminating failures, help to postpone them. The explanation of the
PVM eect on delaying viscoelastic failure can be seen in the uniaxial simulation of
Fig. 9. The constant bending stress on a cantilever beam due to a time independent
end load is reduced by a constant voltage input, thus delaying but not eliminating even-
tual failure in time. Such failures could be negated by increasing voltages in time which
in turn would reduce bending stresses at rates which are faster than the degradation of
failure stresses.
4.2 Computational Issues
Version 5.6-7 of ABAQUS
TM
was used on UIUC NCSAs HP-Exemplar computer with
modications to simulate a relatively deep beam with thin piezo devices (Fig. 1) as well
as their dissimilar viscoelastic properties. The nite element mesh consisted of 201,000
DOF. Fig. 11 shows the results of performance studies. The modest maximum num-
ber of 16 CPU is due to internal constraints imposed by ABAQUS
TM
, which does not
accept requests for more than 16 processors. Additional programming by the vendor
is needed to allow for a larger number of parallel processors to be executed simulta-
neously. Fig. 11 depicts the results of computational performance evaluation studies,
indicating that the total computer time does not signicantly decrease when 16 proces-
sors are engaged as compared to 8. The speed up rises from 2.45 to 2.6 for the same
number of CPU doublings. The lack of substantial speed up after 8 processors is due to
increased communication time between the larger number of processors and indicates
that signicant programming improvements are needed.
5

5. REFERENCES
Beldica, Cristina E., Harry H. Hilton and Sung Yi (1998a) A sensitivity study of
viscoelastic, structural and piezo-electric damping for utter control, Proceedings
39th AIAA/ASME/ ASCE/AHS/ASC Structures, Structural Dynamics and Mate-
rials Conference, AIAA Paper 98-1848 2:1304-1314.
Beldica, Cristina E., Harry H Hilton and Sung Yi(1998b) Viscoelastic damping and
piezo-electric control of structures subjected to aerodynamic noise, Proceedings of
the 4th AIAA/ CEAS Aeroacoustics Conference, AIAA Paper 98-2343 2:805-815.
Beldica, Cristina E. and Harry H. Hilton (1998c) Piezo-electric and viscoelastic con-
trol of lifting surface aerodynamic derivatives and aileron reversal, Submitted to
Proceedings of the 40th AIAA/ASME/AHS/ASC Structures, Structural Dynamics
and Material Conference, St. Louis, MO., April 19-23, 1999.
Dillard, D. A. and H. F. Brinson (1983) A numerical procedure for predicting creep
and delayed failures in laminated composites, Long-Term Behavior of Composites,
ASTM STP 813, T. K. OBrien (ed.), 2337, ASTM, Philadelphia.
Hiel, C. C., M. Sumich and D. P. Chappell (1991) A curved beam test specimen for
determining the interlaminar tensile strength of a laminated composite, Journal
of Composite Materials 25:854868.
Hilton, Harry H. and S. T. Ariaratnam (1994) Invariant anisotropic large deformation
deterministic and stochastic combined load failure criteria, International Journal
of Solids and Structures 31:32853293.
Hilton, Harry H., Jack R. Vinson and Sung Yi (1997) Anisotropic piezo-electro-
thermo-visco- elastic theory with applications to composites, Proceedings of the
11th International Conference on Composite Materials (ICCM11) VI:48814890,
Gold Coast, Australia.
Hilton, Harry H., Sung Yi, Jack R. Vinson and Chong Hin Koh (1998a) Probabilistic
structural integrity of piezoelectric viscoelastic composite structures, in press Pro-
ceedings of the ONR US - Pacic Rim Workshop on Composite Materials for Ship
and Oshore Structures Honolulu.
Hilton, Harry H. and Sung Yi (1998b) Creep divergence of nonlinear viscoelastic lifting
surfaces with piezo-electric control, in press Proceedings of the Second International
Conference on Nonlinear Problems in Aviation and Aerospace Daytona Beach, FL.
Hilton, Harry H., Sung Yi, Jack R. Vinson and Chong Hin Koh (1998c) Anisotropic
nonlinear piezo-electro-chemo-thermo-viscoelasticity: characterization, simulations
and probabilistic structural integrity, submitted to Journal of Reinforced Plastics
and Composites.
Holloway, Frank and Aleksandra Vinogradov (1997) Material characterization of thin
lm piezoelectric polymers, Proceedings of the 11th International Conference on
Composite Materials (ICCM-11) VI:474482, Gold Coast, Australia.
Leibowitz, Moti M. and Jack R. Vinson (1993) The coupling between generally anisotro-
pic composite structures and piezoelectric materials, Proceedings of the Eighth An-
nual Conference of the American Society for Composites 151161. Cleveland, OH.
Lin, Y. K. (1967) Probabilistic Theories of Structural Dynamics. McGraw-Hill Book
Co., New York.
Vinogradov, Aleksandra and Frank Holloway (1997) Electro-mechanical properties of
of the piezoelectric polymer PVDF, in press Journal of Ferroelectrics.
Yi, Sung and Harry H. Hilton (1993a) Anisotropic viscoelastic nite element analysis
of mechanically and hygrothermally loaded composites, International Journal of
Composite Engineering 3:123135.
6
X_1
h_a
h_b
h_a
L X_3
Fig . 1 - LO NG ITUDINAL VIEW O F BEAM
Yi, Sung, Harry H. Hilton and M. Fouad Ahmad (1993b) Performance evaluations of
viscoelastic nite element supercomputer algorithms, Applications of Supercom-
puters in Engineering III (C. A. Brebbia and H. Power, eds.) Elsevier, London,
498-510.
Yi, Sung and Harry H. Hilton (1994) Dynamic nite element analysis of viscoelastic
composite plates in the time domain, International Journal for Numerical Methods
in Engineering 37:40814096.
Yi, Sung, Harry H. Hilton and M. Fouad Ahmad (1995) Performance evaluations of
viscoelastic nite element analyses on coarse grained and massively parallel super-
computers, High Performance Computing in Engineering, Applications to Partial
Dierential Equations (H. Power and C. A. Brebbia, eds.), Computational Mechan-
ics Publications, London, 2:167-198.
Yi, Sung, M. Fouad Ahmad and Harry H. Hilton (1996) Dynamic responses of plates
with viscoelastic damping treatment, ASME Journal of Vibration and Acoustics
118:362-374.
Yi, Sung, Shih Fu Ling, Ming Ying, Harry H. Hilton and Jack R. Vinson (1997) Fi-
nite element formulation for anisotropic coupled piezo-hygro-thermo-viscoelasto-
dynamic problems, Technical Report UILU-ENG 97-0511, University of Illinois
UIUC, Urbana, IL. In press International Journal of Numerical Methods in Engi-
neering.
7

1.5 10
7
2 10
7
2.5 10
7
3 10
7
3.5 10
7
4 10
7
4.5 10
7
5 10
7
- 2 - 1 0 1 2 3 4
Fig . 2 - DELAM INATIO N UNIAXIAL C REEP STRENG TH
TENSIO N M =0%
TENSIO N M =1.6%
SHEAR M =0%
SHEAR M =1.6%
U
N
I
A
X
I
A
L


D
E
L
A
M
I
N
A
T
I
O
N


S
T
R
E
S
S


(
P
a
)
LO G ( TIM E)
0
0.2
0.4
0.6
0.8
1
0.4
0.5
0.6
0.7
0.8
0.9
1
0
0.2
0.4
0.6
0.8
1
J
1
/ J
1max
Fig. 3 INVARIANT DETERMINISTIC FAILURE SURFACE
J
2
/ J
2max
J
3

/

J
3
m
a
x
1 0 0 0
1500
2000
2500
3000
3500
4000
4500
- 2 5
- 20
- 15
- 10
- 5
0
5
1 0
- 4 - 3 - 2 - 1 0 1 2
Fig . 4 C ANTILEVER BEAM TO P FIBER STRESSES

11
PIEZO

11
BEAM

12
PIEZO

12
BEAM
B
E
N
D
I
N
G


S
T
R
E
S
S
E
S
S
H
E
A
R


S
T
R
E
S
S
E
S
LO G ( TIM E)
0
0.2
0.4
0.6
0.8
1
1.2
- 4 - 3 - 2 - 1 0 1 2
Fig . 5 TO P FIBER BENDING STRAINS

11
PIEZO

11
BEAM
B
E
N
D
I
N
G


S
T
R
A
I
N
S
LO G ( TIM E)
0
0.5
1
1.5
2
- 3 - 2 - 1 0 1 2
Fig . 6 BENDING WITH & WITHO UT PIEZO C O NTRO L
BENDING O NLY
C O NST. VO LTAG E
INC REASING VO LTAG E N
O
R
M
A
L
I
Z
E
D


S
T
R
E
S
S
E
S
LO G ( TIM E)
8
Fig. 10 - BEAM WITH PIEZOELECTRIC VISCOELASTIC DEVICES
- 2000
- 1500
- 1000
- 5 0 0
0
- 1 0 0 1 0 20 30 40
Fig . 7 NEUTRAL AXIS DEFLEC TIO NS DUE TO BENDING
w @.25 L
w @.5 L
w @L
D
E
F
L
E
C
T
I
O
N
S
TIM E
- 2 5
- 20
- 15
- 10
- 5
0
- 0.2 0 0.2 0.4 0.6 0.8 1
Fig . 8 PRO BABILITY O F DELAM INATIO N O NSET
PIEZO ELEC TRIC C O NTRO L
NO C O NTRO L
L
O
G

(
D
E
L
A
M
I
N
A
T
I
O
N


P
R
O
B
A
B
I
L
I
T
Y
)
TIM E ( min)
0
0.5
1
1.5
2
2.5
3
- 2 - 1.5 - 1 - 0.5 0 0.5 1 1.5 2
FIG . 9 VISC O ELASTIC UNIAXIAL APPLIED AND
FAILURE STRESSES
BENDING STRESS
FAILURE STRESS
FAILURE STRESS
BENDING + PIEZO
STRESSES
S
T
R
E
S
S
E
S
LO G ( TIM E)
0
1 0 0 0
2000
3000
4000
5000
6000
0
0.5
1
1.5
2
2.5
3
0 5 1 0 15 20
Fig . 11 PERFO RM ANC E EVALUATIO NS FO R 201,000 DO F
TO TAL TIM E
SPEED UP
T
O
T
A
L


T
I
M
E


(
s
e
c
)
S
P
E
E
D

U
P
NUM BER O F C PU
9

Das könnte Ihnen auch gefallen