Vous êtes sur la page 1sur 12

Journal of Sound and Vibration (1996) 193(2), 453–464

A FINITE ELEMENT TIME DOMAIN MODAL


FORMULATION FOR LARGE AMPLITUDE
FREE VIBRATIONS OF BEAMS AND PLATES
Y. S  C. M
Department of Aerospace Engineering, Old Dominion University, Norfolk,
VA 23529-0247, U.S.A.

(Received 30 May 1995, and in final form 20 September 1995)

A finite element time domain modal formulation is presented for the large amplitude free
vibration of plates. The procedure of deriving the non-linear modal equations of motion
is simple and general. Accurate frequency–maximum deflection relations can be obtained
for the fundamental and higher non-linear modes. The percentage of participation from
each linear mode determines the number of modes to be included for accurate frequency
results. Numerical examples on beams and square plates are given and compared with
classical continuum analytical solutions. A plate of complex planform and boundary
conditions is also given to show the versatility of the presented modal formulation.
7 1996 Academic Press Limited

1. INTRODUCTION
The problem of large amplitude vibrations of beams and plates has interested many
investigators [1] since the first approximate solutions for simply supported beams by
Woinowsky–Krieger [2] and for rectangular plates by Chu and Herrmann [3], respectively.
Singh et al. [4] gave an excellent review of various formulation and assumptions, including
the finite element method for large amplitude free vibration of beams. Srirangarajan [5]
recently presented two alternative solutions, based on the method of multiple scales
(MMS) and the ultraspherical polynomial approximation (UPA) method, for the large
amplitude free vibration of a simply supported beam. The frequency ratios for the
fundamental mode, v/vL , at the ratio of maximum beam deflection to radius of gyration
of 5·0 (wmax /r = 5·0) are 3·3438 and 3·0914, using the MMS and the UPA method,
respectively. Eleven frequency ratios including nine from reference [4] were also given (see
Table 1 of reference [5]). It is rather surprising that the frequency ratio for the fundamental
mode at wmax /r = 5·0 for a simply supported beam varied in a such wide range: from the
lowest of 2·0310 to the highest of 3·3438, and with the elliptic function solution by
Woinowsky–Krieger [2] giving 2·3501. Similar results exist for the vibration of plates. Rao
et al. [6] presented a finite element element method for the large amplitude free flexural
vibration of unstiffened and stiffened plates. For the simply supported square plate,
non-linear frequency ratios from six different approaches were reported (see Table 1 of
reference [6]). The frequency ratio at wmax /h = 1·0 varied from a low of 1·2967 to a high
of 1·5314, with Chu and Herrmann’s analytical solution [3] at 1·4023.
This paper presents a finite element time domain modal formulation for the large
amplitude free vibration of isotropic plates. Beams can be treated as special cases of the
plate. The convergence of the fundamental frequency ratio is investigated for a simply
453
0022–460X/96/220453 + 12 $18.00/0 7 1996 Academic Press Limited
454 .   . 
supported beam and a simply supported square plate with a varying number of finite
elements and a varying number of linear modes. Accurate frequency ratios for fundamental
and higher modes at various maximum deflections, and percentages of participation from
various linear modes, are obtained for beams and plates. The presented finite element
time domain modal formulation is general and it can be extended to laminated composite
plates [7].

2. FORMULATION

2.1. –   


The von Kármán strain–displacement relations for an isotropic plate undergoing
extension and bending at any point z through the thickness are the sum of membrane and
change of curvature strain components:

89 8 9 8 9 8 9
ox u,x w,x2 /2 −w,xx
{o} = oy = v,y + w,y2 /2 + z −w,yy
gxy u,y + v,x w,x w,y −2w,xy

= {om0 } + {ob0 } + z{k} = {o 0} + z{k}, (1)


where {om0 } and {ob0 } are the membrane strain components due to in-plane displacements
u and v and the transverse deflection w, respectively. The stress resultants, membrane force
and bending moment are related to the strain components as follows:

6 7 $ %6 7
N A 0 o0
= , (2)
M 0 D k

where the elastic in-plane and flexural rigidity stiffness matrices are defined by

& ' & '


1 g 0 1 g 0
Eh Eh 3
[A] = g 1 0 , [D] = g 1 0 , (3a, b)
1−g 2
12(1 − g )
2
0 0 (1 − g)/2 0 0 (1 − g)/2

in which E is the elastic modulus and g is Poisson’s ratio.

2.2.  ,   


Proceeding from this point, the displacements in equation (1) are approximated over a
typical plate element, e.g., rectangular [8] or triangular [9], using the corresponding
interpolation functions. The in-plane displacements and the linear strains are interpolated
from nodal values by

67
u
= [Hm ]{wm } (4a)
v

and

{om0 } = [Bm ]{wm }, (4b)


-   455

where [Hm ] and [Bm ] denote the displacement and strain–displacement interpolation
matrices, respectively, and {wm } is the in-plane nodal displacement vector. The transverse
displacements, slopes and curvatures are interpolated from nodal values by

6 7
1w/1x
w = [Hb ]{wb }, = [G]{wb }, {k} = [Bb ]{wb }, (5a–c)
1w/1y

where [Hb ], [G] and [Bb ] denote the bending displacement, slope and curvature
interpolation matrices, respectively, and {wb } denotes the nodal transverse displacements
and slopes. Through the use of Hamilton’s principle and finite element expression, the
equations of motion for a plate element undergoing large amplitude vibration may be
written in the form

$ %6 7 0$ % $ % $ %16 7
[mb ] 0 {ẅb } [kb ] 0 [k1]Nm [k1bm ] [k2b ] 0 {wb }
+ + + = 0,
0 [mm ] {ẅm } 0 [km ] [k1mb ] 0 0 0 {wm }

(6)

where [m] and [k] are constant matrices and represent the element mass and linear stiffness
matrices, respectively; [k1] and [k2] are the first order and second order non-linear stiffness
matrices, respectively; [k1]Nm (see equation (A.5) in Appendix A) depends linearly on the
unknown membrane displacements ({Nm } = [A]{om0 } = [A][Bm ]{wm }) and represents the
bending stiffness induced by the linear membrane strains {om0 }; [k1bm ] (see equation (A.6))
depends linearly on the unknown plate slopes and represents coupling between membrane
and bending displacements; and [k2b ] (see equation (A.7)) depends quadratically on the
unknown plate slopes and represents the bending stiffness induced by the non-linear
membrane strains {ob0 }. The element mass, linear stiffness and non-linear stiffness matrices
are all symmetrical and are defined in Appendix A.

2.3.  


After assembling the individual finite elements for the complete plate and applying the
kinematic boundary conditions, the system equations of motion for the plate can be
expressed in the form

$ %6 7 0$ %$ %$ %16 7
[Mb ] 0 {Wb } [Kb ] 0 [K1]Nm [K1bm ] [K2b ] 0 {Wb }
+ + + =0.
0 [Mm ] {Wm } 0 [Km ] [K1mb ] 0 0 0 {Wm }

(7)

The system equations of motion for large amplitude free vibration of an isotropic plate
of arbitrary shape and support boundary conditions are thus obtained. Most of the finite
element large amplitude free vibration results for plates and beams (a special case of the
plate) in the literature, e.g., references [4–6] and others, were based on equation (7) using
an iterative scheme and certain particular approximate procedures.
A finite element time domain modal method is presented in what follows for the large
amplitude free vibration of plates. Accurate frequency versus maximum deflection and
modal participation can be determined.
456 .   . 
3. TIME DOMAIN MODAL METHOD
The system equations of motion presented in equation (7) are unsuitable for numerical
integration because (1) the non-linear stiffness matrices are functions of the unknown
displacements and (2) the number of degrees of freedom (DOF) of the system nodal
displacements {W}(={{Wb }, {Wm }}T) is too large. Therefore, equation (7) has to be
transformed into a set of modal equations of constant non-linear modal stiffness matrices
with rather small DOF. This is accomplished by the modal transformation

n
{Wb } = s qr (t){fb }(r) = [F]{q}, (8)
r=1

where the number of retained linear modes n is much less than the system DOF, {fb }(r)
is the normal mode (normalized with the maximum component to unity) from the linear
flexural vibration solution of the system and vLr is the corresponding frequency:

vLr
2
[Mb ]{fb }(r) = [Kb ]{fb }(r) . (9)

By neglecting the membrane intertia term for the lower few frequencies, equation (7)
can be partitioned and written as two separate equations. Solving the partitioned equations
for {Wm } leads to the following reduced system equations in terms of transverse
displacements {Wb }:

[Mb ]{Wb } + ([Kb ] + [K1]Nm − [K1bm ][Km ]−1 [K1mb ] + [K2b ]){Wb } = 0 (10)

and

{Wm } = −[Km ]−1 [K1mb ]{Wb }. (11)

3.1. -   


The non-linear stiffness matrices [K1bm ] and [K2b ] in equation (10) can be expressed in
terms of modal co-ordinates as sum of the product of modal co-ordinates and non-linear
modal stiffness matrices:

n n n
[K1bm ] = s qr [K1bm ](r) , [K2b ] = s s qr qs [K2b ](rs) . (12, 13)
r=1 r=1 s=1

The non-linear modal stiffness matrices [K1bm ](r) and [K2b ](rs) are evaluated with the
corresponding element components {wb }(r) (see equations (A.9) and (A.11)) obtained from
the known system linear normal mode {fb )(r):

[K1bm ](r) = s [[k1bm ](r)], [K2b ](rs) = s [[k2b ](rs)]. (14, 15)
all elements all elements
+bdy. conds. +bdy. conds.

Thus, the non-linear modal stiffness [K1bm ](r) and [K2b ](rs) are constant matrices.
The non-linear stiffness matrix [K1]Nm is linearly dependent on the membrane
displacements. From equation (11), {Wm } can be expressed in terms of modal co-ordinates
as
n n n n
{Wm } = −[Km ]−1 s s qr qs [K1mb ](r){fb }(s) = s s qr qs {fm }(rs) , (16)
r=1 s=1 r=1 s=1
-   457
where
{fm }(rs) = −[Km ]−1 [K1mb ](r){fb }(s) . (17)
Thus, the non-linear stiffness matrix [K1]Nm can be expressed as

n n
[K1]Nm = s s qr qs [K1](rs)
Nm . (18)
r=1 s=1

The non-linear modal stiffness matrix [K1](rs)


Nm is a constant matrix and is evaluated with
the corresponding element components {wm }(rs) (see equations (A.5), (A.8) and (A.10))
obtained from the system mode {fm }(rs) as

[K1](rs)
Nm = s [[k1](rs)
Nm ]. (19)
all elements
+bdy. conds.

3.2. -  


With the modal transformation of equation (8) and the non-linear modal stiffness
matrices in equations (12), (13) and (18), the system equations of motion (10) are thus
transformed into the well-known Duffing equations in reduced modal co-ordinates as
 b ]{q̈} + ([K
[M b ] + [Kqq ]){q} = 0, (20)
where the modal mass and modal linear stiffness matrices are diagonal, as
 b ], [K
([M b ]) = [F]T([Mb ], [Kb ])[F], (21)
and the cubic terms in {q} are

n n
[Kqq ]{q} = [F]T s s qr qs ([K1](rs) (r)
Nm − [K1bm ] [Km ]
−1
[K1mb ](s) + [K2b ](rs) )[F]{q}. (22)
r=1 s=1

All the modal matrices in equation (20) are constant matrices. With a given initial
conditions, the response of modal amplitudes can be determined from equation (20) using
any numerical integration scheme such as Runge–Kutta or Newmark–b method. The most
common initial conditions are specified amplitudes and zero modal velocities. The plate
response consists of bending and membrane components. The bending nodal
displacements are obtained from equation (8), and the membrane nodal DOF are
determined from equation (16). The maximum plate deflection time history, and also the
period and frequency, can then be determined. The minimum number of modes, n, used
in the Duffing equations to obtain a converged frequency – maximum deflection relation
has to be investigated. In the derivation of the Duffing equations, the only approximation
was the neglecting of the in-plane inertia. Since the bending frequency of vibration is much
lower than the in-plane vibration frequency, the approximation is reasonable and the
converged frequency is accurate.
Once the converged frequency – maximum deflection is achieved, the participation from
each linear mode to the maximum deflection can be calculated from
max =qi =
participation from mode i = . (23)
n
s (max =qr =)
r=1
458 .   . 
T 1
Convergence of the fundamental frequency ratio at wmax /r = 5·0
for a simply supported beam
Number of elements Number of modes
and three modes (v/vL )1 and 20 elements (v/vL )1
10 2·3539 1 2·3526
15 2·3537 2 2·3526
20 2·3526 3 2·3526
4 2·3526

This modal participation will indicate whether a single mode or multiple modes should
be employed.
In addition, the procedure in deriving the Duffing equations shown in equation (20) is
simple and straightforward. This is especially convenient for plates of arbitrary planform
and complex boundary conditions using the present finite element procedure, however, it
is quite difficult, if not impossible, using the classical continuum Galerkin’s approach. For
derivation of the non-linear modal equations for symmetrical or unsymmetric laminated
composite plates, readers are referred to reference [7].

4. RESULTS AND DISCUSSION


4.1. 
The beam is simply treated as a special case of the plate. The conventional beam element
having six (two axial and four bending) DOF is used in the examples. A simply supported
uniform beam with immovable ends is studied first. The convergence of the fundamental
and the second frequency ratios at wmax /r = 5·0 is investigated with different numbers of
beam elements and numbers of modes. A half-beam is modelled with 10, 15 or 20 elements,
and one to four symmetrical linear modes are used in the non-linear modal equations of
motion. In Table 1 are shown the frequency ratio values for the fundamental mode; it is
clearly indicated that a 20-element and three-mode model will yield converged and accurate
results. This certainly indicates that the MMS and the UPA method [5] will not give
accurate frequency results. A convergence study of the second lowest frequency ratio is
shown in Table 2. The participation from each linear mode is also given. Results are
obtained using a 20-element and three antisymmetric modes model, and a 40-element and
five antisymmetric modes model in half of a beam. The difference in frequency ratios in
using three and five modes is only 0·3%, and the modal participation values reveal that

T 2
Convergence of the second lowest frequency ratio at wmax /r = 5·0 for a simply supported
beam
Modal participation (%)
ZXXXXXXXXXXXCXXXXXXXXXXV
Modal (v/vL )2 q2 q4 q6 q8 q10
20 elements
and three modes 2·7342 91·89 1·87 6·25 — —
40 elements
and five modes 2·7259 90·86 1·48 6·78 0·24 0·64
-   459
T 3
Frequency ratios and modal participations for a simply supported beam
Present FEM
ZXXXXXXXXXXCXXXXXXXXXXV
Modal participation (%)
Elliptic integral [2] ZXXXXXXXCXXXXXXXV
wmax /r (v/vL )1 (v/vL )1 q1 q3 q5
0·2 1·0038 1·0045 99·99 1·0E − 4 2·2E − 5
0·4 1·0150 1·0155 99·99 3·5E − 4 4·0E − 5
0·6 1·0331 1·0338 99·99 2·8E − 4 3·3E − 5
0·8 1·0580 1·0587 100·00 2·2E − 6 3·9E − 5
1·0 1·0892 1·0900 99·99 3·6E − 4 3·7E − 5
2·0 1·3178 1·3189 99·99 5·1E − 3 3·9E − 5
3·0 1·6257 1·6270 99·99 6·7E − 3 3·5E − 5
4·0 1·9760 1·9779 99·93 7·0E − 2 3·9E − 5
5·0 2·3501 2·3526 99·98 1·8E − 2 2·6E − 5
wmax /r (v/vL )2 q2 q4 q6
0·2 1·0056 99·97 2·3E − 2 1·2E − 2
0·4 1·0222 99·83 8·2E − 2 8·8E − 2
0·6 1·0490 99·63 0·19 0·18
0·8 1·0850 99·35 0·32 0·33
1·0 1·1295 99·01 0·48 0·51
2·0 1·4361 97·73 0·80 1·47
3·0 1·8339 96·12 0·41 3·47
4·0 2·2316 94·41 0·96 4·64
5.0 2·7342 91·89 1·87 6·25

at least three modes are needed for an accurate frequency solution for a simply supported
beam at wmax /r = 5·0.
In Table 3 are shown the fundamental and the second frequency ratios and their
corresponding modal participation values for various values of wmax /r for a simply
supported uniform beam (20 elements in a half-beam). The modal participation values
demonstrate that a single mode will certainly yield accurate fundamental frequencies (also
see Table 1). The small difference between the present frequency ratios and the elliptic
integral solution [2] (e.g., 2·3526 versus 2·3501 at wmax /r = 5·0) is because of the difference
of in-plane forces in the two approaches. The FEM gives a non-uniform distribution of
the in-plane force due to the assumed linear in-plane displacement field in each element,
although the average values of the in-plane force for each element are the same as the one
in the classical approach [2]. The modal participation information also indicates that the
vibration deflection shape of the second or higher modes at high values of wmax /r is the
sum of multiple linear modes. Multiple modes are needed for the second or higher
frequency ratio at high values of wmax /r.
The fundamental frequency ratios and modal participations for a clamped uniform
beam with immovable ends are shown in Table 4. The half-beam is modelled with 20
elements, and three symmetrical modes are used in the Duffing equations. The results show
that at least two modes are needed for wmax /r e 3·0.

4.2. 
A simply supported square plate of uniform thickness with immovable edges,
u(0) = u(a) = v(0) = v(a) = 0, is studied. Convergence of the fundamental frequency
460 .   . 
T 4
Fundamental frequency ratios and modal participations for a clamped beam
Modal participation (%)
ZXXXXXXXXCXXXXXXXXV
wmax /r (v/vL )1 q1 q3 q5
1·0 1·0221 99·84 0·15 1·5E − 2
1·5 1·0484 99·61 0·34 4·3E − 2
2·0 1·0854 99·59 0·33 7·9E − 2
2·5 1·1339 99·17 0·59 0·25
3·0 1·1842 98·66 1·17 0·16
3·5 1·2439 98·12 1·50 0·30
4·0 1·3020 98·00 1·66 0·34
4·5 1·3721 97·73 1·87 0·40

ratios at wmax /h = 1·0 and 1·4 is investigated. A quarter of the plate is modelled with 6 × 6,
10 × 10 and 16 × 16 mesh sizes and one, three or four symmetrical linear modes are
used. The C1 conforming rectangular plate element [8] having 24 (8 membrane and 16
bending) DOF is used. Results shown in Table 5 indicate that for converged and accurate
frequencies the 16 × 16 (or 256 elements) mesh size and four-mode quarter-plate model
should be employed.
In Table 6 are shown the frequency ratios and modal participation values for the lowest
three modes at various wmax /h values for a simply supported square plate (16 × 16 mesh
size in a quarter plate). The results indicate that at least two linear modes are needed for
accurate fundamental frequency prediction at wmax /h = 1·0. The low (1·2967) and high
(1·5314) frequency ratios given by Rao et al. [6] are not accurate.
A plate of L-shaped platform with complex boundary conditions, shown in Figure 1,
is presented next. The plate is modelled with 75 elements. The fundamental frequency
ratios at wmax /h = 1·4 with different number of linear modes are listed in Table 7; it shows
that three modes will give accurate frequency predictions. The fundamental frequency
ratios and modal participations at various wmax /h are given in Table 8. The relationship
between the frequency ratio and the maximum deflection is given in Figure 2. Node A is
the location of the maximum deflection.

5. CONCLUSIONS
A finite element formulation in the time domain and modal co-ordinates is presented
for large amplitude free vibration of beams and plates. The procedure for deriving the
non-linear modal equations is simple and it is advantageous as compared to the classical

T 5
Convergence of fundamental frequency ratios for a simply supported square plate (g = 0·3)
(v/vL )1 at wmax /h (v/vL )1 at wmax /h
Mesh sizes ZXXCXXV Number of modes and ZXXXCXXXV
and four modes 1·0 1·4 16 × 16 mesh size 1·0 1·4
6×6 1·4938 1·8499 1 1·4352 1·7497
10 × 10 1·4453 1·7712 3 1·4361 1·7552
16 × 16 1·4363 1·7560 4 1·4363 1·7560
-   461
T 6
The first three frequency ratios and modal participations for a simply supported square plate
(g = 0·3)
Present FEM
ZXXXXXXXXXXXXXCXXXXXXXXXXXXXV
Elliptic Modal participation (%)
integral [3] ZXXXXXXXXXXCXXXXXXXXXXV
wmax /h (v/vL )11 (v/vL )11 q11 q13 q31 q33
0·2 1·0195 1·0218 99·95 2·7E − 2 6·0E − 3 1·8E − 2
0·4 1·0757 1·0832 99·68 0·21 4·0E − 2 6·7E − 2
0·6 1·1625 1·1774 99·32 0·47 9·1E − 2 0·13
0·8 1·2734 1·2973 98·92 0·67 0·12 0·28
1·0 1·4024 1·4363 98·36 1·08 0·20 0·35
1·2 1·5448 1·5898 97·84 1·39 0·26 0·51
1·4 1·6933 1·7560 97·15 1·82 0·36 0·66
wmax /h (v/vL )21 q21 q23 q41 q43
0·2 1·0300 99·84 7.8E − 2 7·9E − 2 1·7E − 3
0·4 1·1136 99·11 0·53 0·35 4·0E − 3
0·6 1·2194 97·93 1·98 8·7E − 2 3·6E − 3
0·8 1·3902 97·38 2·45 0·16 8·9E − 3
1·0 1·5600 97·01 2·81 0·15 2·9E − 2
1·2 1·7446 96·24 3·05 0·53 0·18
1·4 1·9387 95·44 3·46 0·70 0·40
wmax /h (v/vL )22 q22 q24 q42 q44
0·2 1·0216 99·96 1·8E − 2 7·3E − 3 9·9E − 3
0·4 1·0833 99·18 9·1E − 2 3·6E − 2 4·0E − 2
0·6 1·1782 99·60 0·22 8·8E − 2 9·3E − 2
0·8 1·2985 99·27 0·38 0·15 0·19
1·0 1·4375 98·64 0·55 0·21 0·59
1·2 1·5844 97·91 0·16 1·79 0·13
1·4 1·7381 97·61 1·60 0·49 0·30

Figure 1. The geometry and support conditions of the L-shaped plate.


462 .   . 
T 7
Convergence of fundamental frequency ratio at wmax /h = 1·4 for the
L-shaped plate
Number of modes (v/vL )1
1 1·55694
2 1·52474
3 1·48706
4 1·47703
5 1·47310

T 8
Fundamental frequency ratios and modal participations for the L-shaped
plate
wmax /h (v/vL )1 q1 q2 q3 q4
0·2 1·015 99·89 0·01 0·02 0·08
0·4 1·056 99·15 0·52 0·31 0·02
0·6 1·119 98·52 0·10 1·37 0·01
0·8 1·196 99·51 0·07 0·01 0·42
1·0 1·284 98·31 0·74 0·95 0·01
1·2 1·377 98·70 0·16 1·05 0·09
1·4 1·477 98·78 0·68 0·44 0·10

continuum Galerkin’s method, especially for plates of arbitrary planform and complex
boundary conditions.
The frequency results obtained by the present finite element time domain modal
method are accurate and compare well with the approximate single mode solutions by
Woinowsky–Krieger for the simply supported beam and by Chu and Herrmann for the
simply supported square plate. Frequency ratios for higher non-linear modes can also be
determined. The convergence of frequency value at certain maximum deflection is
investigated using varying numbers of elements and varying numbers of linear modes. The

Figure 2. Maximum deflection versus frequency for the L-shaped plate.


-   463
minimum number of linear modes needed for an accurate frequency result can be
determined from the participation percentage value of each mode.

ACKNOWLEDGMENTS
The authors gratefully acknowledge the support by NASA Langley Research Center
NAG1-1684, AF Wright Laboratory F33615-91-C3205 and the Aerospace Engineering
Department.

REFERENCES
1. M. S 1987 Applied Mechanics Review 40, 1553–1561. Nonlinear vibrations
analysis of plates: a review and survey of current developments.
2. S. W-K 1950 Journal of Applied Mechanics 17, 35–36. The effects of an axial
force on the vibration of hinged bars.
3. H. N. C and G. H 1956 Journal of Applied Mechanics 23, 523–540. Influence of large
amplitudes on free flexural vibrations of rectangular elastic plates.
4. G. S, A. K. S and G. V. R 1990 Journal of Sound and Vibration 142, 77–85. Large
amplitude free vibration of beams–a discussion of various formulations and assumptions.
5. H. R. S 1994 Journal of Sound and Vibration 175, 425–427. Nonlinear free
vibrations of uniform beams.
6. S. R. R, A. H. S and M. M 1993 Journal of the Acoustical Society of
America 93(6), 3250–3257. Large-amplitude finite element flexural vibration of plates/stiffened
plates.
7. R. C. Z, D. Y. X and C. M 1994 American Institute of Aeronautics and Astronautics
Journal 32, 2044–2052. Finite-element time domain-modal formulation for nonlinear flutter of
composite plates.
8. F. K. B, R. L. F and L. A. S 1966 Proceedings of Conference on Matrix Methods
in Structural Mechanics, AFFDL-TR-66-80, Wright–Patterson Air Force Base, Ohio, October
1965, 397–444. The generation of interelement compatible stiffness and mass matrix using the
interpolation formulas.
9. A. T and T. J. R. H 1985 Computer Methods in Applied Mechanics and Engineering
50, 71–101. A three-node Mindlin plate element with improved transverse shear.

APPENDIX A: ELEMENT MATRICES

[mm ] =
gV
[Hm ]Trh[Hm ] dV, [mb ] =
gV
[Hb ]Trh[Hb ] dV, (A1, A2)

[kb ] =
gV
[Bb ]T[D][Bb ] dV, [km ] =
g V
[Bm ]T[A][Bm ] dV,

[k1]Nm = 12
g V
[G]T[Nm ][G] dV, (A3–A5)

[k1bm ] = 12
g
V
[G]T[C]T[A][Bm ] dV, [k2b ] = 12
gV
[G]T[C]T[A][C][G] dV, (A6, A7)
464 .   . 
with

& '
1w/1x 0
$ %
Nmx Nmxy
[Nm ] = , [C] = 0 1w/1y , (A8, A9)
Nmxy Nmy
1w/1y 1w/1x

and

8 9
Nmx
6 7
1w/1x
{Nm } = Nmy = [A]{em0 } = [A][Bm ]{wm }, = [G]{wb }. (A10, A11)
1w/1y
Nmxy

Vous aimerez peut-être aussi