You are on page 1of 12

Finite Elements in Analysis and Design 39 (2002) 4354

www.elsevier.com/locate/nel
Finite element analysis of steel members under cyclic loading
Su Mingzhou

, Gu Qiang, Guo Bing


School of Civil Engineering, Xian University of Arch. & Tech., Xian, China
Received 19 December 2000; accepted 17 September 2001
Abstract
A new nite element formulation for steel members under cyclic loading was presented. The formulation
was based on degenerated shell element, together with the updated Lagrange formulation to deal with the
geometric nonlinearity, and the material was assumed to follow mixed hardening law. Then a program based
on the formulation was completed. Finally, by performing series of examples including post-buckling, strain
reserve loading, and cyclic loading, numerical experiments were carried out to verify its eciency. In general,
there was a good agreement compared with other analytical and experimental results. ? 2002 Elsevier Science
B.V. All rights reserved.
Keywords: Finite element analysis; Shell element; Updated Lagrange formulation; Mixed hardening law; Cyclic loading
1. Introduction
The hysteretic behavior of structural members under cyclic loading is very important in investigat-
ing the dynamic response of members against repeated loading, e.g., earthquake and wind motion.
Generally, it is directly related to the collapse of structures under major dynamic motions. Several
experiments and numerical analysis have been carried out on steel members under cyclic loading
since 1950s, only remarkable works are listed below.
Fukumoto and Kusama [1] conducted an experiment on steel box-section columns under uniaxial
cyclic loading to nd out the inuence of local buckling on load-deection curve. As a result, the
ultimate bearing capacity of plates was suggested.
Billio and Calado [2] performed experiments on the steel bent sections under cyclic loads, and
carried out a numerical analysis based on small displacement assumption to compare.
Castiglioni and Palma [3] presented a number of quasi-static cyclic tests on specimens built with
rolled commercial shapes of type HE and IPE. In this study, the author suggested that the inuence
of depth-to-thickness ratio of webs needed further investigation.

Corresponding author.
E-mail address: sumingzhou@263.net (S. Mingzhou).
0168-874X/02/$ - see front matter ? 2002 Elsevier Science B.V. All rights reserved.
PII: S0168- 874X(02)00060- 4
44 S. Mingzhou et al. / Finite Elements in Analysis and Design 39 (2002) 4354
Hao Jiping [4] carried out some tests and analysis on I-section beams and beam-columns. By
adopting the plastic hinge theory with damage, the author analyzed the characteristics of the hysteretic
behaviors of specimens, but some main parameters in the analysis depended on the experimental
results due to lack of enough data, so the general untested members could not be predicted.
Most of the investigations mentioned above were very simple and could only be used in a limited
range, comprehensive knowledge of steel members under seismic was still lacking, and the design
criterion could not avoid structural damage, so it is urgent to nd a high accuracy and an ecient
analytical model.
In this paper, a formulation based on degenerated shell element was presented, together with up-
dated Lagrange formulation to deal with the geometric nonlinearity, and the material was assumed to
follow mixed hardening law. Then a program based on the formulation was completed by the author
[5]. After that, by performing series of examples including post-buckling, strain reserve loading, and
cyclic loading, numerical experiments were carried out to verify its eciency. In general, there was
a good agreement compared with other analytical and test results.
2. Method of analysis
2.1. Geometric properties
The behavior of the degenerated shell element is based on the assumption that straight lines dened
by the nodal director vectors remain straight during the element deformations and that no transverse
normal stress is developed in the directions of the director vectors.
The element geometry is dened by the displacements at nodal points:
{u} = [N]{u}
e
=
8

k=1
[N]
(k)
{u}
(k)
(1)
in which {u} is the displacements at any point, {u}
(k)
is the displacements at the nodal point k, and
[N] is the element shape function given by [N] =[N]
(1)
[N]
(2)
[N]
(8)
, where [N]
(k)
(k =1, 8)
is the element shape function at the nodal point k,
[N]
(k)
= N
k
_

_
1 0 0
t
[
11k

t
[
12k
0 1 0
t
[
21k

t
[
22k
0 0 1
t
[
31k

t
[
32k
_

_
in which
[
t
[
k
] =
_

_
t
[
11k
t
[
12k
t
[
21k
t
[
22k
t
[
31k
t
[
32k
_

_
=
h
k
2
[
t

J
k
1

t

J
k
2
],
is the natural coordinate in the director vector, and
t

J
k
1
,
t

J
k
2
are unit vectors orthogonal to the
nodal director vector.
S. Mingzhou et al. / Finite Elements in Analysis and Design 39 (2002) 4354 45
The generalized strain components {
t
E} may be separated into two parts, one is linked with
linear functions of displacements ({
t
E
l
}), and the other is with quadratic functions of displacements
({
t
E
nl
}):
{
t
E} = {
t
E
l
} + {
t
E
nl
} (2)
in which
{
t
E
l
} = [
t
t
B
l
]{u}
(e)
,
{
t
E
nl
} = [
t
t
B
nl
]{u}
(e)
, (3)
where [
t
t
B
l
] is the linear straindisplacement matrix, [
t
t
B
nl
] is the nonlinear incremental strain
displacement matrix.
In an updated Lagrange formulation, all quantities are referred to deformed conguration. The
principle of virtual displacements may be written as
_
J
(o
i)
+ S
i)
)o

E
i)
dJ = R
t+t
(4)
in which
R
t+t
=
_
J
(
t
+ )ou dJ
_
A
(
t
q + q)ou dA,
where E and o are vectors of Green strain and Cauchy stress, respectively, and S are the vectors of
Kirchho stress increment. The term on the right side of Eq. (4) represents the work done by body
force and surface traction q.
For simplicity, the second- and higher order terms relating the vector of residual forces to the
vector of displacement increments are omitted, then the formulation may be expressed as
([
t
K
0
]
e
+ [
t
K
o
]
e
){u}
e
= {R
t+t
}
_
J
[
t
t
B
l
]
T
{o} dJ (5)
in which [
t
K
0
]
e
is the element stiness matrix taking account of the changes of geometry and [
t
K
o
]
e
of the eect of internal stresses, [
t
D] is the tangent modulus as shown in the next section:
[
t
K
0
]
e
=
_
J
[
t
t
B
l
]
T
[
t
D][
t
t
B
l
] dJ,
[
t
K
o
]
e
=
_
J
[
t
G]
T
[M][
t
G] dJ
in which
[M] =
_

_
o
11
I s,m
o
12
I o
22
I
o
13
I o
23
I o
33
I
_

_
46 S. Mingzhou et al. / Finite Elements in Analysis and Design 39 (2002) 4354
is the matrix of Cauchy stress, {o} is the vector of Cauchy stress, and I = I
33
:
[
t
t
B
l
]
(k)
=
_

_
t
N
k, x
1
0 0
t
[
11k
t
a
k, x
1
t
[
12k
t
a
k, x
1
0
t
N
k, x
2
0
t
[
21k
t
a
k, x
2
t
[
22k
t
a
k, x
2
0 0
t
N
k, x
3
t
[
31k
t
a
k, x
3
t
[
32k
t
a
k, x
3
t
N
k, x
2
t
N
k, x
1
0
t
[
11k
t
a
k, x
2
+
t
[
21k
t
a
k, x
1
t
[
12k
t
a
k, x
2
+
t
[
22k
t
a
k, x
1
0
t
N
k, x
3
t
N
k, x
2
t
[
21k
t
a
k, x
3
+
t
[
31k
t
a
k, x
2
t
[
22k
t
a
k, x
3
+
t
[
32k
t
a
k, x
2
t
N
k, x
3
0
t
N
k, x
1
t
[
31k
t
a
k, x
1
+
t
[
11k
t
a
k, x
3
t
[
32k
t
a
k, x
1
+
t
[
12k
t
a
k, x
3
_

_
,
[G]
(k)
=
_

_
t
N
k, x
1
0 0
t
[
11k
t
a
k, x
1
t
[
12k
t
a
k, x
1
0
t
N
k, x
1
0
t
[
21k
t
a
k, x
1
t
[
22k
t
a
k, x
1
0 0
t
N
k, x
1
t
[
31k
t
a
k, x
1
t
[
32k
t
a
k, x
1
t
N
k, x
2
0 0
t
[
11k
t
a
k, x
2
t
[
12k
t
a
k, x
2
0
t
N
k, x
2
0
t
[
21k
t
a
k, x
2
t
[
22k
t
a
k, x
2
0 0
t
N
k, x
2
t
[
31k
t
a
k, x
2
t
[
32k
t
a
k, x
2
t
N
k, x
3
0 0
t
[
11k
t
a
k, x
3
t
[
12k
t
a
k, x
3
0
t
N
k, x
3
0
t
[
21k
t
a
k, x
3
t
[
22k
t
a
k, x
3
0 0
t
N
k, x
3
t
[
31k
t
a
k, x
3
t
[
32k
t
a
k, x
3
_

_
,
{o} = {o
11
o
22
o
33
o
12
o
23
o
31
}
T
in which
t
N
k, x
i
= cN
k
}c
t
x
i
,
t
a
k, x
i
= cN
k
}c
t
x
i
+ N
k
c}c
t
x
i
.
2.2. Mixed hardening model
Mixed hardening model proposed by Axelsson and Samuelsson [6] is adopted to describe the
properties of steel, which links the isotropic hardening and kinematic hardening linearly with a
parameter M. According to this model, the yield surface is allowed to expand (contract) and translate.
So dierent degrees of Baushingers eect can be simulated, and some common models may be
obtained by varying M from 1 to 1 (1 included).
If the virgin material is of von Mises type, the loading criterion corresponding to mixed hardening
can be written in the following form:
[ = F(o
i)
:
i)
) o
s
(c

i)
) = 0 (6)
in which o
i)
is the Cauchy stress tensor, :
i)
denotes the translation of the center of the yield surface
in the stress eld, and o
s
is a function governing the isotropic expansion or contraction of the yield
surface.
The rate of plastic strain can be simply split into two colinear components:
dc

i)
= dc
(i)
i)
+ dc
(k)
i)
, (7)
dc
(i)
i)
= Mdc

i)
, dc
(k)
i)
= (1 M) dc

i)
, (8)
where dc
(i)
i)
is associated with the expansion of the yield surface and dc
(k)
i)
with the translation of
the yield surface.
S. Mingzhou et al. / Finite Elements in Analysis and Design 39 (2002) 4354 47
According to Pragers [7] assumption, the increment of translation of the yield surface takes place
in the direction of the normal to the surface in the stress point, i.e.
d:
i)
= C dz
_
c[
co
i)
_
(k)
= C dc
(k)
i)
. (9)
Together with Hookes law and consistency condition, the tangent modulus under small deection
can be expressed as
{do} = [D]
e
{dc} (10)
in which
[D]
e
= [D]
e
[D]

,
[D]

=
[D]
e
cF
c{o}
_
cF
c{o}
_
T
[D]
e
_
cF
c{o}
_
T
[D]
e
cF
c{o}
+ C(1 M)
_
cF
c{o}
_
T
cF
c{o}
+ 2HM o
s
_
2
3
_
cF
c{o}
_
T
cF
c{o}
_
1}2
, (11)
where C = 2H}3.
For metals, as elastic strain is slight compared with plastic strain under large deection, the tangent
modulus in the updated Lagrange formulation can be expressed as
[
t
D] = [
t
D]
e
[t
d
], (12)
where [t
d
] is the Cauchy stress matrix
[t
d
] =
_

_
2o
11
0 2o
22
s,mmetrica!
o
12
o
12
1
2
(o
11
+ o
22
)
0 o
23
1
2
o
13
1
2
o
22
o
13
0
1
2
o
23
1
2
o
12
1
2
o
11
_

_
. (13)
2.3. Determination of strain hardening modulus H
The strain hardening modulus H denotes the tangent slope of stressstrain curve in uniaxial tension.
According to RambergOsgood equation, H can be written as
H =
do
dc

=
o
m
_
o
0
o
_
m
, (14)
where m and o
0
are constant numbers, and can be determined by tting the uniaxial tensile test data
by the least-square method.
48 S. Mingzhou et al. / Finite Elements in Analysis and Design 39 (2002) 4354
Input basic data
Form the data for the calculation
Initial the working variables
Determine the loading method
LIM DIM MLM
Determine the coefficient of each step
Form the stiff matrix and solve the equations
Iterate in terms of loading method
Proceed until all the steps finished
Analyze the results
End
Fig. 1. Main procedures.
3. Complementation of nite element program
3.1. Procedure to perform nite element program
Fig. 1 gives the main procedure to perform the nite element program.
3.2. Strategies in the procedure
3.2.1. Loading methods
Three loading methods are adopted for the sake of easy control, i.e., load incremental method
(LIM), displacement incremental method (DIM), and mixed loading method (MLM). The last
one combines the former two, so it can be controlled both by load and by displacement in one
calculation.
3.2.2. Uniform displacements
The condition of uniform displacements in the edges of members may often exist, e.g. strong
stiener in the edge or loading by rigid end plate. The pretreatment technique is adopted here to
satisfy the condition.
3.2.3. Stress increment associated with a given strain increment
The stress increment during plastic ow must be such that the plastic ow rule and the yield
surface are satised within sucient accuracy. Because of the nonlinearity of geometry and material
S. Mingzhou et al. / Finite Elements in Analysis and Design 39 (2002) 4354 49
constitutive law, the true stress increment must be evaluated by
{S}
i
= {S}
i1
+
_
{E+E}
i
{E}
[D(S)] d{E}. (15)
This equation can be integrated by sub-incremental technique described as follows:
During the iteration, convergence may not often exist because of ctitious numerical unloading.
On the basis of Nyssens [8] assumption of incremental reversibility for plastic behavior, the sum of
plastic strain {1
P
} is calculated during each sub-increment at each integrating point, if {1
P
} {0},
unloading is deemed to take place, and the stress may be obtained by elastic law.
The following equations may be used to calculate plastic strain:
{1
P
} =

{1
P
}, (16)
{1
P
} = dz

2
3
_
cF
c{o}
_
T
cF
c{o}
. (17)
The stress incremental vector may be calculated by the following procedure:
(1) For points in plastic state in the former increment, assume that the portion of elastic strain
increment to total strain increment equals zero ([ = 0), go directly to step (3). Or the stress
increments will be evaluated on the basis of Hookes law:
{S}
i
= [D]
e
{E}
i
. (18)
The total current stresses are
{
t
t+t
S}
i
= {
t
t
S}
i
+ {S}
i
. (19)
(2) Substitute the stresses obtained by Eq. (19) into Eq. (6) to see whether the yield function is
satised or not. If the function [ in Eq. (6) is negative, then the point is still in elastic state, go
to step (5); if [ is positive or zero, then the point yields under the current load level. Assume
the elastic portion of stress increments is [{S}
i
, then stresses {S}
i1
+ [{S}
i
must satisfy
Eq. (6), and the value of [ can be obtained by
[ =
a
2
+
_
a
2
2
a
1
a
3
a
1
, (20)
where
a
1
= S
2
11
+ S
2
12
S
11
S
22
+ 3(S
2
12
+ S
2
23
+ S
2
31
),
a
2
= o
11
S
11
+ o
22
S
22

1
2
( o
11
S
22
+ o
22
S
11
) + 3( o
12
S
12
+ o
23
S
23
+ o
31
S
31
),
a
3
= o
2
11
o
11
o
22
+ o
2
22
+ 3( o
2
12
+ o
2
23
+ o
2
31
) o
2
S
, (21)
where o
i)
is the deviator Cauchy stress, o
i)
= o
i)
:
i)
.
50 S. Mingzhou et al. / Finite Elements in Analysis and Design 39 (2002) 4354
(3) The equivalent plastic strain increment

E
P
(absolute value) may be calculated by the following
Equation:


E
P
=
_
2
3
(1 [)
2
_
E
2
11
+ E
2
22
+ (E
11
+ E
22
)
2
+
1
2
[(2E
12
)
2
+ (2E
23
)
2
+ (2E
31
)
2
]
_
.
(22)
Then the number of sub-increments may be evaluated as
N
Sb
=

E
P
}c
0
(23)
in which c
0
generally takes a value of 0.0002 for the sake of accuracy.
(4) Start sub-increments. For mth sub-increment, the incremental strain is
{dE}
m
=
(1 [)
N
Sb
{E}
i
. (24)
(a) For the point unloading at (m1)th, the incremental stress can be evaluated by elastic law,
substitute the total stress into Eq. (6), if function [ is negative, the point is still in elastic
state, go to step (d); or [
m
may be calculated in terms of Eq. (20). The total elastic stress
vector is
{
t
t+t
S}
m
= {
t
t
S}
m1
+ [
m
{S}
m
. (25)
The incremental plastic strain is
{dE}
m
=
(1 [)(1 [
m
)
N
Sb
{E}
i
. (26)
(b) Calculate the relevant stress increments, and evaluate cumulative plastic increments {1
p
}.
If {1
p
} is negative, then the point begins to unload elastically in this sub-increment, go
to step (d). If {1
p
} is positive or zero, the point remains in plastic state, then renew the
yield stress vector, evaluate and add the stress increments obtained by using a second-order
RungeKutta method, and calculate the magnitude of translation of yield center {d:}
m
:
o
,m
= o
,m1
+ H(o
m1
) d1
m
, (27)
{do}
m
= 0.5{D
e
(o
m1}2
) + D
e
(o
m1
)}{dE
m
}, (28)
{o}
m
= {o}
m1
+ {do}
m
, (29)
{d:}
m
=
2
3
H(1 M){dE}
m
, (30)
{:}
m
= {:}
m1
+ {d:}
m
. (31)
(c) Substitute the total stresses into Eq. (6) to see whether the yield function is satised or not.
If it is not satised, the stresses must be rened by the following procedure:
Total strain vector due to rening is zero:
{oc
e
} + {oc

} = 0. (32)
Linearizing the yield function yields
F( o
i)
+ o o
i)
) = ( o
s
+ o o
s
)
2
. (33)
S. Mingzhou et al. / Finite Elements in Analysis and Design 39 (2002) 4354 51
B
L
Q
Crisfield
this paper
R=25400mm
L=B=508mm
E=3.105kN/mm
2
=0.3
t=12.7mm
Q
c

(
k
N
)
w
c
/t
0.0 0.5 1.0 1.5 2.0 2.5
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
C
Fig. 2. Post-buckling path.
The rening factor oz is given by substituting Eqs. (32) and (33) into Eq. (6) in the
following equation:
oz =
F( o
i)
) o
2
s
_
cF
c{o}
_
T
[D]
e
cF
c{o}
+
2
3
H(1 M)
_
cF
c{o}
_
T
cF
c{o}
+ 2 o
s
HM
_
2
3
_
cF
c{o}
_
T
cF
c{o}
. (34)
If oz is positive, the stress vector due to rening is
{oS} = oz[D]
e
cF
c{o}
. (35)
(d) For unloading points, the sub-incremental stresses may be calculated by elastic law.
(5) Stop sub-increment looping until m equals N
Sb
.
4. Examples
4.1. Buckling of plates and shells
Fig. 2 gives the post-buckling path of a shallow shell with pinned and nonmoveable straight edges
and free curved edges. A concentrating load acts on its center. The geometric and material characters
are listed in Fig. 2, and Criselds [9] solution is given for comparison. Good agreement can be
seen from this gure.
4.2. Strain reverse loading
To verify the eciency of the mixed hardening law, the result of a thick aluminum ring subjected
to a pair of tensile forces P with various mixed hardening coecients M are given in Fig. 3.
Axelsson and Samulssons [6] analysis are given for comparison, and Owens [10] test results are
also given.
52 S. Mingzhou et al. / Finite Elements in Analysis and Design 39 (2002) 4354
P
P
0.005 0.01
100
200
E=87.6GPa
=0.3237
s=192.5MPa
H=3220MPa

0
.
1
7
8
m
0
.
1
2
7
m
0.0
Mixed hardening M=0.2
Isotropic hardening
Kinematic hardening (Ref. 6)
Mixed hardening M=0.2 (Ref. 6)
Isotropic hardening (Ref. 6)
Test (Ref. 10)
Kinematic hardening
P
(
k
N
)
u(m)
-0.015 -0.010 -0.005 0.0 0.005 0.010
10
20
30
40
-10
-20
-30
-40
-50
-60
Fig. 3. Behavior of a thick aluminum ring.
-1 0 1 2 3 4 5 6
-0.2
0.2
0.4
0.6
0.8
1.0
-1 0 1 2 3 4 5 6
1.0
0.8
0.6
0.4
0.2
-0.2
/
y

y
/
y
(a) (b)
Fig. 4. Average stressstrain curves of specimen P40-1. (a) Numerical curve, (b) Experimental curve.
4.3. Uniaxial cyclic loading of box-section steel strut
Cyclic loading curves of box-section steel strut are given in Figs. 46, and compared with Fuku-
moto and Kusamas [1] experimental results. Curves in Fig. (a) are analytical ones of this paper and
those in Fig. (b) are experimental ones in Ref. [1]. For mild steel, the yield plateau is comparatively
long, and isotropic hardening is very slight, so analysis results give a good agreement with test ones
when M = 0.0.
S. Mingzhou et al. / Finite Elements in Analysis and Design 39 (2002) 4354 53

y
/
y

y
/
y
-1 0 1 2 3 4 5 6
-0.2
-0.1
0.1
0.2
0.3
0.4
0.5
0.6
-0.2
-0.1
0.1
0.2
0.3
0.4
0.5
0.6
-1 0 1 2 3 4 5 6
(a) (b)
Fig. 5. Average stressstrain curves of specimen P60-1. (a) Numerical curve, (b) Experimental curve.
0.1
y
0.2
0.3
0.4
0.5

0
-0.1
/
y
0.1
y
0.2
0.3
0.4
0.5

0
-0.1
/
y
-1 1 2 3 4 5 -1 1 2 3 4 5
(a) (b)
Fig. 6. Average stressstrain curves of specimen P80-1. (a) Numerical curve, (b) Experimental curve.
5. Conclusion
The above analysis and comparison indicate that the program has high accuracy when used in
problems of large deection, buckling, strain reserve loading and cyclic loading, etc. So it can be
used to analyze hysteretic behavior of steel members under cyclic loading.
References
[1] Y. Fukumoto, H. Kusama, Local instability tests of plate elements under cyclic uniaxial loading, J. Struct. Eng.
ASCE 111 (5) (1985) 10511067.
[2] G. Ballio, L. Calado, Steel bent sections under cyclic loads: experimental and numerical approaches, Costruzioni
Metalliche 1 (1986) 97114.
[3] C.A. Castiglioni, N.Di. Palma, Experimental behavior of steel members under cyclic loading, Costruzioni Metalliche
23 (1989) 123.
[4] H. Jiping, Test and theoretical study on the local buckling and low cycles fatigue of steel structures under cyclic
loading, Ph.D. Thesis, School of Civil Engineering, Xian University of Arch. & Tech., Xian, China, 1995 (in
Chinese).
54 S. Mingzhou et al. / Finite Elements in Analysis and Design 39 (2002) 4354
[5] S. Mingzhou, Interactive buckling failure mechanism and designing criterion of box-section steel columns and beam
columns under cyclic loading, Ph.D. Thesis, School of Civil Engineering, Xian University of Arch. & Tech., Xian,
China, 1999 (in Chinese).
[6] K. Axelsson, A. Samuelsson, Finite element analysis of elasticplastic materials displaying mixed hardening, Int. J.
Numer. Methods Eng. 14 (1979) 211215.
[7] W. Prager, A new method of analyzing stresses and strains in work hardening solids, J. Appl. Mech. Trans. ASME
23 (1956) 493496.
[8] C. Nyssen, An ecient and accurate iterative method, allowing large incremental steps, to solve elasto-plastic
problems, Comput. Struct. 13 (1981) 6371.
[9] M.A. Criseld, Accelerating and damping the modied NewtonRaphson method, Comput. Struct. 8 (3) (1984)
359407.
[10] D.R.J. Owen, A. Prakash, O.C. Zienkiewicz, Finite element analysis of non-linear composite materials by use of
overlay system, Comput. Struct. 4 (1974) 12511267.

You might also like