Académique Documents
Professionnel Documents
Culture Documents
449/91
$3.00 + 0.00
Printed in Great Britain. Pergamon Press plc
Abstract-The C’ continuous element based on the Euler-Bernoulli theory of bending and the Co
continuous element based on Timoshenko’s theory of bending are compared for their computational
efficiency and economy for the analysis of inelastic frames subjected to transient dynamic loads. Moreover,
the size of a Co mesh (with only linear elements) is evaluated, which is computationally equivalent to C’
Hermitian element. Also the ranges of slenderness ratio, in which both these elements are effective and
the limits of aspect ratio below which they fail are also determined through the numerical studies.
(4 to use Co elements for transient dynamic analysis U(x, Y) = u(x) -Y&(X), (2)
of inelastic and elastic frames and compare their
computational efficiency and economy with C’ Ux, Y) = r+), (3)
elements,
(b) to find the size of Co mesh, consisting of linear
elements, which must be comparable in perform-
659
660 T. KANT and S. R. MAWR
1
-65 (l-3<) 65 (1+35) If the corrected stress can be represented as
cb(C,4)=$ 7 - --~ where the superscript c stands for the
L L2 L
Vi
[ 4r,sXi+,,
corrected quantity, then the incremental corrected
4
stress at t = tn+,, can be expressed as
H
{Bb(t> rl)jr{&)> (8)
VI
3.4. Internal resisting force vector
ej
The incremental axial internal resisting force vector
where d is the depth of the cross-section and {db} is can be given as
the vector of bending displacements of the element.
The strain at any generic point (5, q) at t = t, + , can
{APP,I,+, = {WAaK rlX+ I do, (14)
be expressed as v,
1
stress, the cross section can be said to have become
x MT, II):+, drl G (17)
completely plastic.
A comparative study of Co and C’ elements 661
u =2&u, (19b)
I---.
F(t) i 5000 Ibs
F(l)
(20)
1.0
WI = 24EI
FlL
riiL4
2
96EI (24)
AL
XP
2
24EI
rf2L’
?FiL4
24
96EI
1
(18) L FflL’
662 T. KANT and S. R. MARUR
and the highest frequency corresponding to flexural where 4 is the rotation due to transverse shear
displacements is given from the matrix given by eqn deformation.
(24) as Using the two noded, linear [5-71 element, the
displacements can be expressed as
1
192EI 192c2r2B
&Z-.-C (25) 2
riiL4 [ L4
u= 1 N,u,, (274
,=I
where the radius of gyration of the cross-section
rg = I/A and c2 = E/p.
The higher value of the two, given by the eqns (23) (27b)
and (25) is used in eqn. (20) to evaluate (At),., for C’
continuous elements.
0 = i N,B,, (27~)
2 -.
have become plastic, while the rest of the layers 2
continue to remain in their respective states of stress.
With the increase in strain, the stresses at the mid- The axial component of axial strain is given as
points of outer most layers reach the value of yield
stress first and gradually the stresses at the midpoints
t, = [$ z]{::}= {B,}r{d:J. (29)
of interior layers also attain the yield value. When all
the layers reach the yield stress, the section is said to
have become plastic. The gradual plastification of the Any generic point in the frame can be denoted by
cross-section is modelled more realistically by this (<,I), where 5 is a Gauss point along the length of
concept and with the increase in the number of layers, the member and ,’ is the distance between the mid-
modelling of plastification becomes very close to the point of a layer and the neutral axis of the cross-
reality. section. The bending component of axial strain at
(5, j) is given by
4.2. Displacement field and strain displacement
relationship
(30)
The axial and transverse displacements are ex-
pressed as given by eqns (2) and (3). and the rotation
of the cross-section inclusive of that due to transverse which when expressed in matrix notation becomes as
shear deformation can be expressed as
e,(x) = F + c#J,
(32)
Fft)b
0.02
Time (see)
and expressed in matrix form as
4
6NODESj9ELEMENTS
C’- DISCRETIZATION
Yts =
[
alv,
dx - Nl
aN2
--
ax N2
Iii
The incremental axial and shear strains at t = tn+ ,
4
v2
02
= {&(wv:I.
(33)
can be expressed as
Co-DISCRETIZATION
+ {4,<<, Y))‘{W)n+ I (34)
and
24w0.5 6 0.729
1 &(t]n+, = {&<&]YW%.~. (35)
LAYER DIMENSIONS
(36)
BY C’ ELEMENTS
---- BY TRILINEAR MESH
-.- BY BILINEAR MESH
TIME (SEC)
Fig. 3. Comparison of solutions by C’ and Co elements through the variation of third floor displacements
with time [17].
664 T. KANT and S. R. MARUR
where E, is the Young’s modulus of the layer. When and after applying the layer concept
materially nonlinear analysis is to be carried out, the
total elastic stress at t = I, + , is given by
{APP,)n+, = ’ {B,(t)}’
-I
a(&.X+, = Ae(5, jX’+ , + d5, L’)n, (37)
{ApP,).+, = L d2
SI0 -dZ I b
{&Jr
zY
x W&1): + I dz dy dx (40)
F . 4000 Ibr
w . 1.257 rod/w
{APP,A+I = ’ {Bo}’
s -I 6y = 36,OOOpsi (for lrtt column and beam)
1
ThL A x 6.49 in2
I/
cross-section, and IJI for the determinant of the
Jacobian. E, * (0.1 x10’ psi )
6, -.-
The incremental bending force vector is given by
LEGEND
1.6 -
- [Ial
--- BY C’ ELEMENTS
-1.6
t
Fig. 5. Time history of elastic horizontal displacement of top right node of the frame.
{AA).+ I =
s
I,14W}T and [Nq is the matrix of shape functions of an
element. The consistent mass matrix for the element
1I4dt.
is given by
W,Ad5)1,+ I (45)
AL 6lL
-
While evaluating the shear forces, the shear rigidity 3 6
GA is replaced by GA, where the area A’ is given by
A/V. The parameter v is a correction factor to allow AL CiL
for cross-sectional warping, which is taken to be 1.2 3 6
for rectangular sections. Moreover, {Ap,} is evaluated
AM’ ALdZ
using reduced integration to cure the shear locking,
when slender members are analysed. When the vectors 36 72
WI =
I&,1, {&+,I and &,I are appropriately assembled, PFIL +iL
the internal resisting force vector is obtained. -
6 3
4.5. Specially lumped mass matrix
ML tiL
The simultaneous use of central difference oper-
-
6 3
ators and the Co formulation, requires the consistent
mass matrix to be converted into a diagonal one. The ALd2 niLd2
procedure to achieve this objective, known as ‘special 72 36
lumping’ outlined by Hinton et al. [ 16) is adopted here
for the linear element.
The consistent mass matrix can be expressed as A scaling factor, defined as the ratio of the
total mass of the element to the sum of the
masses (in the diagonal elements) corresponding
to any one translational degree of freedom, is
calculated as
in which
1
pbd scaling factor = (&!,) (48)
pbd
Im7= pbd’ WW
When all the diagonal elements are scaled by this
12 factor and the offdiagonal elements are made equal
666 T. KANT and S. R. MAWR
to zero, the specially lumped mass matrix is obtained and the frequency corresponding to flexural displace-
as ment is given by
AL
+&[;+~]=~~+!S$]. (53)
2
AL
2 The higher one of the two given by eqns (23) and (53)
is to be used in the condition (20) to evaluate the
rX.d’
critical time step.
24
WI = - .
AL
- 5. NUMERICAL EXAMPLES
2
AL The examples are chosen such that both Co and C’
-
2 elements are used for frames with slender members in
AM2 both elastic and inelastic conditions and in deep
beams under going elastic and plastic deformations.
24
All the computations were carried out on CYBER
(49) I 80/840computer in SINGLE PRECLSIOK with 16 signifi-
and the damping matrix is expressed as in CL formu- cant digits as word-length.
lation as
5.1. Co mesh formation
[Cq = r[M’]. (50) Every frame element, discretized by a C’ element
is also discretized by THREE Co elements, denoted as
4.6. Critical time step
TRILINEAR MESH, by TWO Co elements represented as
Expressions for evaluating (At),, for Co elements, BILINEAR MESH and a Co element known as LINEAR
following Belytschko’s expressions for C’ elements MESH. The objective of such a discretization is to
are derived. For axial displacements, the frequency obtain meshes equivalent to quadratic and cubic
for Co elements is the same as that of C’ elements [as elements with simple linear elements, as more number
given by eqn (23)]. of lower order elements are computationally econ-
The stiffness matrix for the linear element using omical than few higher order elements. These dis-
reduced integration for shear terms is given by cretizations form the basic ground for comparison of
results of one mesh with another, with C’ results and
[&I= with those available in the literature.
GA 5.2. Examples
2 Example 1. Biggs [ 171has analysed a three storeyed
frame with nodal dynamic loads as shown in Fig. 1
using modal superposition method. The same frame
is analyzed using C’ elements, first with rigid girders
and then with flexible girders.
Table 1 gives the displacements of flexible girders
L while Table 2 displays the maximum displacements of
GA rigid girders. In both the cases, the close agreement
-- between the results by C’ elements and Biggs can be
2
observed.
G;A 2
- --
L PiiL
GA 2
--
L rilL
(52)
A comparative study of Co and C’ elements
Fig. 6. Time history of elastopiastic strain hardening horizontal displacement of top right node of the
frame.
sin (wt)
l&L h
E.$XlofPSi
E ~0.36, IO’psi
C’-OiscWiration
i
Linear Mesh Bilinear Mesh Trilinear Mesh
4 Nodes, 3 Elwncnts 7 Nodes, 6 Elements 10 Nodes, 9 Ekrrwntr
Co- DlSCREllZAllON
J
Fig. 8. Comparison of Co and C’ elements through the horizontal (strain hardening) displacements of top right node of
the frame.
F(t)
While using Co elements, all the cross-sections are
split into six layers and all the three types of meshes
are adopted as shown in Fig. 2. The response history
by trilinear and bilinear meshes along with that
produced by C’ are shown in Fig. 3.
The response by linear mesh was too small to be
plotted in the same graph, as the first order poly-
nomial shape functions are not adequate enough to
predict the actual response. The trilinear mesh pro-
duces results very close to that of C’ and bilinear
results are slightly stiffer.
The computational costs are presented in Table 3
and it can be observed that Co meshes are compara-
tively time consuming.
Example 2. Hilmy and Abel [ 181 had considered a
single storey portal frame as shown in Fig. 4 and is
discretized by four C’ elements. The elastic and
Tim8
elastoplastic responses are obtained and the displace-
STRESS STRAIN DIAGRAM OF STRAIN HARDENING ment histories are plotted in Figs 5 and 6 which depict
MATERIAL (6~ s 0.66 E5 psi 1 the agreement of C’ results with the reference outputs.
Fig. 9. Rectangular frame with lateral load and strain While adopting Co meshes as shown in Fig. 7, the
hardening material model [I]. cross-section is split into six layers. The nonlinear
A comparative study of Co and C’ elements 669
LEGEND
F-u F-
7
-_[‘I
---BY C’ ELEMENTS
,‘.\ I-1 f
’ ‘1
: ‘,
A
I
\
I
I
I I
I 1
I \
I
I
I
I
:
I
: I
\
I
I I
0.1 0.0
TIME (SEC)
10 NODES, 9 ELEMENTS
C’ DISCRETIZATION
i
LINEAR MESH
db BILINEAR MESH TRILlNEAR MESH
4
10 WOES ,9 ELEMENTS 19NODES,19 ELEMENTS 29NODES, 27 ELEMENTS
Co_ DISCRETIZATION
LAYER DIMENSIONS
Legend
111
TRILINEAR MESH
BILINEAR MESH
LINEAR MESH
0 1.0 20 3.0
TIME (SEC)
Fig. 12. Comparison of solutions by Co and C’ elements through the variation of horizontal displace-
ments with time.
response histories are given in Fig. 8 and the compu- closely follow the displacement pattern of the refer-
tational costs are shown in Table 4. The results by ence output. Of all the three, trilinear mesh is very
trilinear mesh are closer to C’ but slightly stiffer. close to the reference curve while other two meshes
Bilinear mesh results are still stiffer while those by yield slightly stiffer response histories. Here also the
linear mesh are very small to be plotted. Computa- Co meshes were computationally time consuming.
tionally C’ elements are economical. Example 4. The frames of the previous examples
Example 3. Toridis and Khozeimeh [1] had studied contained slender members where the transverse
a single storeyed frame with the masses lumped at shear deformation is negligible. The shear deformable
eight points as shown in Fig. 9. The actual lumped Co meshes are adopted here to study the dynamic
masses are multiplied by a factor of 625 to make the behaviour of deeper beams with predominant trans-
fundamental frequency of the frame very close to that verse shear strain.
of actual buildings. The frame has been discretized A simply supported beam analysed by Bathe
with nine C’ elements, keeping the lumped mass et al. [19] and later on by Liu and Lin [20] is con-
positions of the reference frame as the nodes, for the sidered here for studying the efficiency of C’ and Co
elements. The elastic response is shown in Fig. 10 meshes. The length to depth ratio of the beam is 15
along with reference response. and is subjected to a uniformly distributed load of
The use of Co meshes with the layered cross-section 0.75po, where p. is the static collapse load and
are shown in Fig. 11 while the response history is in perfectly plastic model with a yield stress of
Fig. 12 along with the computational costs in Table 5. 0.5E05 psi is adopted as shown in Fig. 13.
In this particular example, as the number of ele- The beam has been discretized by six C’ elements
ments are comparatively very high, (nine for linear and four trilinear meshes. The elastic and elastoplas-
mesh, eighteen for bilinear mesh and twenty seven for tic response histories are plotted in Fig. 14. Bathe et
trilinear mesh) all the meshes produce results, which al. had made use of a eight-noded plane stress
rqltl
11111111111
I
--_-_-_-_-_--
4-3
30’
r- I
t=3x10CKip/lnZ Y ‘03
q(t)
tJy- 30 KLp/d p- 0’733~10-~
lb m/in’
Pm= STATIC COLLAPSELOAD
1tnt
STEP PRESSURE -0’3.0 P. lb/ II?
q(t)-0625 -m-
-0.75 -ln-
7WODtS, 6 tLEMtNTS
cl_ DlSCRETlZATlON
Fig. 13. A simply supported beam with suddenly applied uniformly distributed load 1191.
ELASTOPLASTIC
I- /RESPONSE BY ,
C’- ELEMENTS
O
,_
$=50,000 psi
0 0 0 : ELASTICSOLUTKI
BY C’ ELEMENT
/ + + + :ELASTlC SOLUTIOI
ELASTOPLASTIC SOLUTKIN BY Co ELEMENT!
+++:PolNTS COMMON
BOTH Co AND C
ELEMENTS IN
‘?\
ELASTIC SOLUTlOh
ELASTIC SOLUTIO l l m : ELASTOPLASTIC ’
SOLUTION BY /
,L4, , \ P
15 30 45 60 75
TIME (SEC X10-’ )
Fig. 14. Elastic and clastoplastic midspan deflections of a simply supported beam, for L/D = 15.
672 T. KANTand S. R. MAIKJR
LEGEND
- [201
0 0 o BY Co ELEMENTS
A 6 BV C’ ELEMENTS
0.54’
TIME
0.c
10 20 30 LO 50 $0 70
TIME (SEC X10 1
Fig. 15. Elastic midspan displacements of a simply supported beam for L/D = IS.
two-dimensional element and von Mises yield centre of its span is discretized with four trilinear
criterion to analyse the beam. It can be observed that meshes and six C’ elements. Making use of the
though both C’ and Co elements yield results closer symmetry of the geometry and loading, one fourth of
to that of Bathe et al.‘s for elastic conditions, for the same beam is also analysed as a plane stress
elastoplastic conditions C’ predictions have higher problem by using six numbers of eight-noded
peaks and periods, while Co is very close to that of serendipity element, for comparing the results of both
Bathe et af.‘s. Co and C’ elements.
For another loading of 0.5po, the elastic displace- The details of the beam and loading are
ment history is given in Fig. 15. Here both Co and C’ shown in Fig. 17. The transient response of the
elements agree very closely with the reference values. beam, for different L/D ratios are plotted in
For the same loading, but for elastoplastic analysis, Figs 2&24.
as shown in Fig. 16, the deviation of C’ results is It can be observed that for slenderness ratio less
quite high in comparison with Co output, which is than or equal to two, neither Co nor C’ element
very close that of Liu and Lin. predicts the response closer to that of serendipity
For yet another loading of 0.625~~. the variation element.
of C’ results in the elastoplastic analysis with When L/D ratio lies in the range of three to
respect to Co and the reference results is shown in eight, trilinear mesh yields results very closer to
Fig. 17. the two-dimensional elements than the C’ elements.
The strain hardening has been incorporated in As shear deformation would be predominant, in this
the material model for the same loading, and the region of L/D ratio, Co elements perform better (as
response history is plotted in Fig. lg. For the this deformation has been taken into account in the
value of /I equal to 0.25, the deviation of C’ predic- formulation) than C’ elements.
tions are more compared to the reference output or As the aspect ratio reaches the value of nine
Co results. and above, both the elements can be seen to yield
Example 5. Another simply supported beam results closer to those obtained from the plane stress
with a suddenly applied concentrated load at the solution.
LEGEND
LEGEND
- f201
- L.201 o o o BY Co ELEMENTS
0 0 o BY co ELEMENTS 626~~----- --A--b BY C’ ELEMENTS
-*--A- BY C’ ELEMENTS TIME
cy = 5Ovooopsl
0.3
0.3 )-
2
2 0.2 5-
E
y 0.2 )-
8
4
a 0.1 j-
6
f
g 0.N ,-
f
0.0 5-
I , , I 1 L , I , ,
0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70
TIME (SEC X 10-4) TIME (SEC X 10-4 1
Fig. 16. Elastoplastic midspan displacements of a simply supported beam for Fig. 17. Elastoplastic midspan displacements of a simply supported beam for
L/D = 15. LID = 15.
614 T. KANT and S. R. MARUR
LEGEND
- [201
G o o o BY C- ELEtltNlS
----- BY C’tLtHENTS
035-
0.30
\
\
\
\
\
;i
uo.25 - \
z \
z
g
zo.20
‘\\
4
\\
0’ \
v)
B
,04 5
h
::
f
0.10.
Fig. 18. Elastoplastic midspan displacements of a simply supported beam with strain hardening effect for
LID = 15.
N
d
(S3H3NI 1 SNOl1331d30 NVdSOlW
678 T. KANT and S. R. MARUR