Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
SUMMARY
In this paper a unied nite element model that contains the EulerBernoulli, Timoshenko and simplied
Reddy third-order beam theories as special cases is presented. The element has only four degrees of freedom,
namely deection and rotation at each of its two nodes. Depending on the choice of the element type, the
general stiness matrix can be specialized to any of the three theories by merely assigning proper values to
parameters introduced in the development. The element does not experience shear locking, and gives exact
generalized nodal displacements for EulerBernoulli and Timoshenko beam theories when the beam is
homogeneous and has constant geometric properties. While the Timoshenko beam theory requires a shear
correction factor, the third-order beam theory does not require specication of a shear correction factor.
An extension of the work to axisymmetric bending of circular plates is also presented. A stiness matrix
based on the exact analytical form of the solution of the rst-order theory of circular plates is derived.
# 1997 by John Wiley & Sons, Ltd.
Commun. Numer. Meth. Engng, 13, 495510 (1997)
KEY WORDS classical deformation; shear deformation; axisymmetric circular plates; unied nite elements
1. INTRODUCTION
The nite element models of the EulerBernoulli beam theory (EBT) and Timoshenko beam
theory (TBT) are now standard (see Reddy1 ). A number of Timoshenko beam nite elements
have appeared in the literature.1 5 They dier from each other in the choice of interpolation
functions used for the transverse deection w0 and rotation f. Some are based on equal inter-
polation and others on unequal interpolation of w0 and f.
The Timoshenko beam nite element with linear interpolation of both w0 and f is the simplest
element. However, it is very sti in the thin beam limit, i.e. as the length-to-thickness ratio
becomes large (say, 100). Such behaviour is known as shear locking.2 5 The locking is due to the
inconsistency of the interpolation used for w0 and f. To overcome the locking, one may use equal
interpolation for both w0 and f but use a lower-order polynomial for the shear strain,
gxz dw0 =dx f. This is often realized by using selective integration, in which reduced-order
integration is used to evaluate the stiness coecients associated with the transverse shear
Correspondence to: J. N. Reddy, Department of Mechanical Engineering, Texas A & M University, College Station,
TX 77843-3123, U.S.A.
strain, and all other coecients of the stiness matrix are evaluated using full integration. The
selective integration Timoshenko beam element is known to exhibit spurious energy modes.2;5
Pratap and Bhashyam5 discussed using a consistent interpolation of the variables to alleviate
locking.
Since the transverse shear strain in the Timoshenko beam theory6;7 is represented as a constant
through the beam thickness, a shear correction factor is introduced to calculate the transverse
shear force that would be equal in magnitude to the actual shear force. Since the actual shear
stress distribution through beam thickness is quadratic, Jemielita,8 Levinson,9 Bickford10 and
Reddy11;12 developed third-order beam theories to capture the true variation of the shear stress.
The displacement eld of these third-order theories accommodates quadratic variation of the
transverse shear strain and stresses, and there is no need to use shear correction factors in a third-
order theory. The Levinson third-order beam theory has the same equations of equilibrium as the
Timoshenko beam theory but the force and moment resultants contain higher-order strain terms.
Bickford10 used Levinson's displacement eld and developed variationally consistent equations
of motion of isotropic beams, while Reddy11;12 developed a variationally consistent third-order
theory of laminated composite plates.
Heyliger and Reddy13 used the third-order laminated plate theory of Reddy (RBT) to develop
a beam nite element and study bending and vibrations of isotropic beams. The element is based
on Lagrange linear interpolation of the rotation f and Hermite cubic interpolation of w0 , as they
are the minimum requirements imposed by the weak forms of the third-order theory (also see
Phan and Reddy14 and Reddy15 ).
The objective of the present paper is to develop a unied beam nite element that contains the
nite element models of the EulerBernoulli, Timoshenko and rened third-order beam theory
(RBT). The derivation of the unied element is based on the exact relationships developed by
Wang16 and Reddy et al.17 between various theories. The relationships allow interdependent
interpolation of w0 and f, and any redundancy of degrees of freedom is removed, resulting in an
ecient and accurate locking-free nite element for the analysis of beams according to classical
as well as rened beam theories.
d2 M Exx
qx 1
dx2
d2 wE0
M Exx Dxx 2
dx2
dM Txx dQTx
QTx ; qx 3a; b
dx dx
COMMUN. NUMER. METH. ENGNG, VOL. 13, 495510 (1997) # 1997 by John Wiley & Sons, Ltd.
UNIFIED FINITE ELEMENTS 497
dfT T dwT0
M Txx Dxx ; QTx Axz K s f 4a; b
dx dx
dM R dPxx
xx
QR
x a bRx 5a
dx dx
dQRx dRx d2 Pxx
qx b a 5b
dx dx dx2
R
dfR df d2 wR
MR
xx D xx aF xx 0
6a
dx dx dx2
R dwR
QRx Axz f
0
6b
dx
R
dfR df d2 wR
0
Pxx F xx aH xx 6c
dx dx dx2
R
dw0
Rx D xz fR 6d
dx
where quantities with superscript `E' refer to the EulerBernoulli beam theory, `T' refers to the
Timoshenko beam theory K s denotes the shear correction factor) and `R' refers to the rened
third-order beam theory. In RBT, Pxx and Rx denote the higher-order stress resultants, and the
bending stiness Dxx , shear stiness Axz and parameters appearing in the above equations are
dened below: Z Z
Dxx E x z2 dA E x I 2
yy ; F xx E x z4 dA E x I 4
yy
A A
Z Z
H xx E x z6 dA E x I 6
yy ; Axz Gxz dA Gxz A 7a
ZA ZA
Dxz Gxz z2 dA Gxz I 2
yy ; F xz Gxz z4 dA Gxz I 4
yy
A A
D xx Dxx aF xx ; Fxx F xx aH xx
A xz Axz bDxz ; D xz Dxz bF xz 7b
D^ xx D xx aFxx ; A^ xz A xz bD xz
4 4
a ; b 3a 8
3h2 h2
where I i
yy denotes the ith area moment of inertia about the y-axis:
Z
I i
yy zi dA 9
A
# 1997 by John Wiley & Sons, Ltd. COMMUN. NUMER. METH. ENGNG, VOL. 13, 495510 (1997)
498 J. N. REDDY, C. M. WANG AND K. Y. LAM
dw0
Primary variables : w0 ; 10a
dx
Secondary variables : Qx ; M xx 10b
dw0
Primary variables : w0 ; ; f 12
dx
Secondary variables : V x; Pxx ; M xx aPxx 13
where
dPxx
V x Qx bRx a 14
dx
The specication of a primary variable constitutes a geometric boundary condition, whereas the
specication of a secondary variable constitutes a force boundary condition. One should note
that the third-order theory requires the specication of both f and dw0 =dx, and the eective
shear force in RBT is V x .
dwE0 x2
Dxx fT x Dxx C1 C2 x C3 16
dx 2
M Txx x M Exx x C 1 x C2 17
QTx x QEx x C1 18
where C 1 , C 2 , C 3 and C 4 are constants of integration, which are to be determined using the
boundary conditions of the particular beam. For example, for hingedhinged beams all C i are
COMMUN. NUMER. METH. ENGNG, VOL. 13, 495510 (1997) # 1997 by John Wiley & Sons, Ltd.
UNIFIED FINITE ELEMENTS 499
zero, for xedfree beams all Ci except C4 M Exx 0=Dxx Axz K s are zero, and for the clamped
hinged case they are given by
3OT
C1 M E 0; C 2 C 1 L; C 3 0; C 4 OT M Exx 0L2 19
1 3OT L xx
where OT Dxx =Axz K s L2 and L is the length of the beam. The values of the constants for
additional boundary conditions can be found in Wang.16
d2 QR hq i
2
x
l2 Q R
x m
0
L 2x C 1 20a
dx 2
where
^ xz Dxx
A A xz D xx
l2 ; m 20b
aF xx Dxx Fxx Dxx
aF xx Dxx Fxx Dxx
The solution to this dierential equation is
m E
QR
x x C 5 sinh lx C 6 cosh lx Qx C1 21
l2
Equations for force and moment resultants, rotation and deection of RBT in terms of the
EBT solutions are given by17
dPxx dM R
VR R
x x Qx bRx a xx
QEx x C 1 22
dx dx
MR E
xx x M xx x C 1 x C 2 23
dwE0 F xx R x2
Dxx fR x Dxx a Qx C 1 C 2 x C 3 24
dx A xz 2
Z
Dxx x3 x2
Dxx wR x D wE
xx 0 x Q R
Z dZ C1 C2 C3 x C4 25
A xz
0 x
6 2
The six constants of integration, C1 C6 , are determined using the six boundary conditions
available in the third-order theory. The boundary conditions for various types of supports are
# 1997 by John Wiley & Sons, Ltd. COMMUN. NUMER. METH. ENGNG, VOL. 13, 495510 (1997)
500 J. N. REDDY, C. M. WANG AND K. Y. LAM
dened below, consistent with the primary and secondary variables of RBT:
dPxx
freeF: QR
x bRx a 0 26
dx
MR
xx aPxx 0; Pxx 0 27
simply supportedS: wR
0 0; MR
xx aPxx 0; Pxx 0 28
dwR
clampedC: wR
0 0; fR 0; 0
0 29
dx
For example, for a hingedhinged beam under a uniformly distributed load of intensity q0 , the
constants are given by
q0 m Dxx
C1 C2 C3 0; C4 30a
l4 A xz
q0 m q0 m lL
C5 3 ; C6 3 tanh 30b
l l 2
For a cantilever beam under a uniformly distributed load, the constants have the values
Dxx q0 m 1 lL sinh lL
C 1 C 2 C 3 0; C 4 31a
A xz l4 cosh lL
q m 1 lL sinh lL q Lm
C 5 03 ; C6 0 2 31b
l cosh lL l
dfR d2 wR d2 wE0
D xx aF xx 0
Dxx C1 x C2 32
dx dx2 dx2
COMMUN. NUMER. METH. ENGNG, VOL. 13, 495510 (1997) # 1997 by John Wiley & Sons, Ltd.
UNIFIED FINITE ELEMENTS 501
dwR dwE x2
D xx fR aF xx 0 Dxx 0 C1 C 2 x C 3 33
dx dx 2
From 5a and 6b; d, we have
R
dM xx dM R
xx dPxx ^ R dwR0
a Axz f 34
dx dx dx dx
so that
1 dM R dwR
fR xx
0
35
^ xz dx
A dx
# 1997 by John Wiley & Sons, Ltd. COMMUN. NUMER. METH. ENGNG, VOL. 13, 495510 (1997)
502 J. N. REDDY, C. M. WANG AND K. Y. LAM
Now, we wish to simplify the RBT by neglecting the second-derivative term in (42). This
amounts to reducing the order of the theory from sixth to fourth. We obtain
! " ! #
D^ xx x3 ^ xx
D
R E E
Dxx w0 x Dxx w0 x M xx x C1
A^ xz 6 ^ xz
A
" !#
x2 ^ xx
D
C2 C3 x C4 43
2 ^ xz
A
VR E
x x Qx x C 1 44
MR E
xx x M xx x C 1 x C 2 45
dwE0 x2
Dxx yR x Dxx C1 C2 x C3 46
dx 2
3 2
x x
Dxx wR
0 x D xx wE
0 x AM E
xx x C 1 Ax C 2 B C3 x C4 47
6 2
where
8
>
> 0 for EBT 8
>
> > 0 for EBT
>
> D >
>
>
< xx for TBT >
< 0 for TBT
A Axz K s B 48
>
> >
> ^ xx
D
>
> ^ >
>
>
> D : for simplified RBT
> xx
: for simplified RBT ^ xz
A
^ xz
A
w0 0 D1 ; y0 D2
50
w0 h D3 ; yh D4
COMMUN. NUMER. METH. ENGNG, VOL. 13, 495510 (1997) # 1997 by John Wiley & Sons, Ltd.
UNIFIED FINITE ELEMENTS 503
Figure 1. Geometry and sign convention for the generalized displacements and forces of a beam element: (a) generalized
displacements and generalized forces; (b) interpretation of the stiness coecients as forces required to produce unit
generalized displacements
where yx denotes the slope (positive clockwise), which has dierent meaning in dierent
theories, as dened below:
8
> dw0
>
> dx for EBT
<
yx 51
>
> fT x
for TBT
>
: R
y x for simplified RBT
where x denotes the element co-ordinate, the origin of which is located at node 1 of the element.
Next, let Q1 and Q3 denote the shear forces (i.e. values of V Rx at nodes 1 and 2, respectively;
# 1997 by John Wiley & Sons, Ltd. COMMUN. NUMER. METH. ENGNG, VOL. 13, 495510 (1997)
504 J. N. REDDY, C. M. WANG AND K. Y. LAM
similarly, let Q2 and Q4 denote the bending moments (i.e. values of M R xx ) at nodes 1 and 2,
respectively. Figure 1(a) shows the sign convention used for the generalized displacements and
forces.
The stiness matrix for the unied element is derived using the traditional method to calculate
stinesses in structural analysis. The method involves imposing a unit generalized displacement,
while all other generalized displacements are zero, and determining the generalized forces
required to keep the beam in equilibrium (i.e. equivalent to using the unitdummydisplacement
method). The formulation utilizes the relationships between EBT, and TBT and RBT. This
amounts to using Hermite cubic interpolation for the transverse deection and a dependent
interpolation for the slope. The procedure is outlined briey here.
To obtain the rst column of the element stiness matrix, we set [see Figure 1(b)]:
dwE0
at x 0: wE0 wT0 wR
0 D1 ; yT y R 0 52a
dx
dwE0
at x h: wE0 wT0 wR
0 0; yT yR 0 52b
dx
where
1 A
m ; O 55
1 12O h2
COMMUN. NUMER. METH. ENGNG, VOL. 13, 495510 (1997) # 1997 by John Wiley & Sons, Ltd.
UNIFIED FINITE ELEMENTS 505
This completes the derivations of the stiness coecients of the rst column of the stiness
matrix. The same procedure can be repeated, with dierent generalized displacements set to
unity, to obtain the remaining stiness coecients. The complete nite element model is given by
2 3
12Dxx 6Dxx 12Dxx 6Dxx
6
6 h3 h2 h3 h2 77
6 78 9 8 9 8 9
6 6D 4D 6D e
2D 7 q1 > Q1 >
6 xx xx
l xx xx
x 7>
> D1 >
> >
> > >
> >
6 h2 h h2 h 7< = < = < =
6 7 D2 q2 Q2
m6 7 56
6 12Dxx 6D 12D 6D 7>
> D3 >
> >
> q3 >
> >
> Q3 >
>
6 xx xx xx 7: ; : ; : ;
6 h3 h2 h3 h2 7 D4 q4 Q4
6 7
6 7
4 6Dxx 2Dxx 6Dxx 4Dxx 5
2 x 2
l
h h h h
where
l 1 3O; x 1 6O 57
and
Z h
qei qx'i x dx 58
0
dw0
ur r; z z
dr 59
uz r; z w0 r
where w0 is the transverse deection of the point r; 0 on the midplane (i.e. z 0) of the plate.
The rst-order shear deformation plate theory is based on the displacement eld
ur r; z zfr
60
uz r; z w0 r
where f denotes rotation of a transverse normal in the plane y constant. The bending
equations of equilibrium and stress resultantdisplacement relations of the two theories are
summarized below for constant material and geometric properties.
# 1997 by John Wiley & Sons, Ltd. COMMUN. NUMER. METH. ENGNG, VOL. 13, 495510 (1997)
506 J. N. REDDY, C. M. WANG AND K. Y. LAM
d d
rQC
r rq; rQC
r rM C C
rr M yy 61a; b
dr dr
where
d2 wC 1 dwC
MC
rr D11
0
D 12
0
62a
dr2 r dr
d2 wC 1 dwC
MC
yy D12
0
D 22
0
62b
dr2 r dr
and QCr denotes the shear force in the classical plate theory.
First-order shear deformation plate theory (FST):
d
rM Frr M Fyy rQFr 0 63a
dr
d
rQFr rq 63b
dr
where
dfF 1
M Frr D11 D12 fF 64a
dr r
dfF 1
M Fyy D12 D22 fF 64b
dr r
dwF0
QFr A44 K s fF 64c
dr
COMMUN. NUMER. METH. ENGNG, VOL. 13, 495510 (1997) # 1997 by John Wiley & Sons, Ltd.
UNIFIED FINITE ELEMENTS 507
and
1 d dMC
r q 66b
r dr dr
We can also establish the following equality using denition (65) and equations 62a; b:
dMC d
r rM C C C
rr M yy rQr 67
dr dr
Similarly, we have
dfF 1 F 1 d
F
M D f D rfF 68
dr r r dr
and
1 d dMF
r q 69
r dr dr
dMF d
r rM Frr M Fyy rQFr 70
dr dr
From 61a; b and 63b it follows that
rQFr rQC
r c1 71
dMF dMC
r r c1
dr dr
or
MF MC c1 log r c2 72
dwC c1 r c2 r c3
fF 0
2 log r 1 73
dr 4D 2D rD
Finally, from 64c, (71) and (73), we obtain
dwF0 1 C c1
fF Qr 74
dr GhK s r
C
and noting that QC
r dM =dr, we have
c1 r 2 c1 c2 r2 c3 log r c4 MC
wF0 wC
0 1 log r log r 75
4D K s Gh 4D D D K s Gh
# 1997 by John Wiley & Sons, Ltd. COMMUN. NUMER. METH. ENGNG, VOL. 13, 495510 (1997)
508 J. N. REDDY, C. M. WANG AND K. Y. LAM
The four constants of integration are determined using the boundary conditions of the
problem. Note that, for a solid circular plate, for r 0, (71) gives c1 0.
1 h c3 i
fF r c1 r2 log r 1 2c2 r 4
4D r
c^3 1
2^c2 r c^ 4 r1 2 log r G 77
r r
or
fDg Hf^cg 78
Q1 2prQFr r ra 8pD^ c4
1 n 1 n
Q2 2prM Frr r ra 2pD 21 nra c^ 2 c^ 3 La G c^ 4
ra ra
Q3 2prQFr r rb 8pD^ c4
1 n 1 n
Q4 2prM Frr r rb 2pD 21 nrb c^ 2 c^ 3 Lb G c^ 4
rb rb
or
COMMUN. NUMER. METH. ENGNG, VOL. 13, 495510 (1997) # 1997 by John Wiley & Sons, Ltd.
UNIFIED FINITE ELEMENTS 509
and K GH 1 is the unied element stiness matrix. The classical plate element stiness
matrix is obtained from K by setting G 0. Note that the stiness matrix derivation presented
herein is based on the form of the exact solution, which diers from the conventional polynomial
based stiness matrix. Of course, the present element gives exact nodal values.
5. CONCLUSIONS
In this paper, a unied nite element model of the EulerBernoulli, Timoshenko and simplied
Reddy third-order beam theories is developed. The bending stiness matrix of the unied nite
element is derived. The development is based on the exact relationships between the bending
solutions of the EulerBernoulli beam theory (EBT), Timoshenko beam theory (TBT) and
simplied Reddy third-order beam theory (RBT). The relationships provide an interdependent
interpolation of the deection and rotation of the form
X
4 X
4
wR
0 x Dj 'j x; yR x Dj cj x 81
j 1 j 1
where cj are quadratic interpolation functions related to 'j . Consequently, the element stiness
matrix is of the same order as that for the EulerBernoulli beam element 4 4, and it gives
exact nodal values of the generalized displacements for EulerBernoulli and Timoshenko beams
with uniform cross-section and homogeneous material properties (i.e. the element has a
superconvergence property). An independent interpolation of the form
X
4 X
3
wR
0 x Dj 'j x; yR x Yj cj x 82
j 1 j 1
would have led to a stiness matrix of order 7 7 for the same accuracy as the element derived
herein. Thus, shear deformable nite elements based on the Timoshenko and third-order beam
theories can be included in any computer program by simply replacing the stiness matrix of the
existing EulerBernoulli beam nite element with that given in (56). The element proves to be
very ecient for use in frame structural analysis (after including axial displacement degrees of
freedom) because only one element for the structural member is sucient to obtain exact
displacements and forces from element equilibrium equations. The traditional, equal but linear
interpolation element with selective reduced integration will not give the same accuracy unless
two or more elements per structural member are used. Further, the present element includes shear
deformation and therefore the user never has to know if the shear deformation is signicant
enough to include in the analysis.
# 1997 by John Wiley & Sons, Ltd. COMMUN. NUMER. METH. ENGNG, VOL. 13, 495510 (1997)
510 J. N. REDDY, C. M. WANG AND K. Y. LAM
A unied element stiness matrix for axisymmetric bending of a circular plate is also
developed. It includes the classical and rst-order shear deformation theories. The element
stiness matrix is unconventional in that it is based on the form of analytical solution, which
contains logarithmic terms and yields exact nodal values, whereas the conventional polynomial
based element yields only approximate nodal values. One must be careful about such an element
because it does not contain the rigid body mode.
Extension of the present formulation to a general curved beam, with dynamic and non-linear
eects, is desirable to make it a more practical element. Similar ideas may prove to be useful in
developing locking-free plate elements, although the two-dimensional nature and arbitrary
geometry of a plate element makes it dicult to readily extend the present thinking.
REFERENCES
1. J. N. Reddy, An Introduction to the Finite Element Method, 2nd edn, McGraw-Hill, New York, 1993.
2. R. C. Averill and J. N. Reddy, `On the behaviour of plate elements based on the rst-order theory', Eng.
Comput., 7, 5774 (1990).
3. R. E. Nickell and G. A. Secor, `Convergence of consistently derived Timoshenko beam nite elements',
Int. j. numer. methods eng., 5, 243253 (1972).
4. A. Tessler and S. B. Dong, `On a hierarchy of conforming Timoshenko beam elements', Comput.
Struct., 14(34), 335344 (1981).
5. G. Prathap and G. R. Bhashyam, `Reduced integration and the shear exible beam element', Int. j.
numer. methods eng., 18, 195210 (1982).
6. S. P. Timoshenko, `On the correction for shear of the dierential equation for transverse vibrations of
prismatic bars', Philosophical Magazine, 41, 744746 (1921).
7. S. P. Timoshenko, `On the transverse vibrations of bars of uniform cross section', Philos. Mag., 43,
125131 (1922).
8. G. Jemielita, `Techniczna teoria plyt srednieej grubbosci' (Technical theory of plates with moderate
thickness), Rozprawy Inzynierskie (Engineering Transactions), Polska Akademia Nauk, 23(3), 483499
(1975).
9. M. Levinson, `A new rectangular beam theory', J. Sound Vib., 74, 8187 (1981).
10. W. B. Bickford, `A consistent higher order beam theory', Dev. Theor. Appl. Mech., 11, 137150 (1982).
11. J. N. Reddy, `A simple higher-order theory for laminated composite plates', J. Appl. Mech., 51,
745752 (1984).
12. J. N. Reddy, Energy and Variational Methods in Applied Mechanics, Wiley, New York, 1984.
13. P. R. Heyliger and J. N. Reddy, `A higher-order beam nite element for bending and vibration
problems', J. Sound Vib., 126(2), 309326 (1988).
14. N. D. Phan and J. N. Reddy, `Analysis of laminated composite plates using a higher-order shear
deformation theory', Int. J. numer. methods eng., 12, 22012219 (1985).
15. J. N. Reddy, Mechanics of Laminated Composite Plates: Theory and Analysis, CRC Press, Boca Raton,
Florida, 1997.
16. C. M. Wang, `Timoshenko beam-bending solutions in terms of EulerBernoulli solutions', J. Eng.
Mech.,, ASCE, 121(6), 763765 (1995).
17. J. N. Reddy et al., `Exact relationships between the bending solutions of the EulerBernoulli beam
theory and shear deformable beam theories', Int. J. Solids & Struct. (1997), in press.
18. C. M. Wang and K. H. Lee, `Deection and stress-resultants of axisymmetric Mindlin plates in terms of
corresponding Kircho solutions', Int. J. Mech. Sci. 38(11), 11791185 (1996).
19. R. Szilard, Theory and Analysis of Plates, Prentice-Hall, Englewood Clis, NJ, 1974.
COMMUN. NUMER. METH. ENGNG, VOL. 13, 495510 (1997) # 1997 by John Wiley & Sons, Ltd.