This action might not be possible to undo. Are you sure you want to continue?
www.elsevier.com/locate/engstruct
Limit and shakedown analysis of 3D steel frames
Hoang Van Long
∗
, Nguyen Dang Hung
LTAS Fracture Mechanics, University of Li` ege, Chemin des Chevreuils, 1, B 52/3, 4000 Li` ege, Belgium
Received 13 July 2007; received in revised form 6 December 2007; accepted 6 December 2007
Available online 22 January 2008
Abstract
This paper presents an efﬁcient algorithm for both limit and shakedown analysis of 3D steel frames by the kinematic method using linear
programming technique. Several features in the application of linear programming for rigidplastic analysis of threedimensional steel frames
are discussed, as: change of the variables, automatic choice of the initial basic matrix for the simplex algorithm, direct calculation of the dual
variables by primal–dual technique. Some numerical examples are presented to demonstrate the robustness, efﬁciency of the proposed technique
and computer program.
c 2007 Elsevier Ltd. All rights reserved.
Keywords: Limit analysis; Shakedown analysis; Plastic hinges; Space frames; Linear programming
1. Introduction
The fundamental theory of plastic analysis of the frame
structures was pointed out ﬁfty years ago. This technique
consisting of an application of the mathematical programming
is widely exposed in the literature [1–6]. When the linearized
condition of plasticity admissibility is adopted, the plastic
analysis problem can be reduced to a linear programming (LP)
problemwhere simplex technique is largely used. This direction
has been deeply exploited in the years 1970–1990, and some
interesting computer programs have been developed [7–11].
Unfortunately, for the last two decades, the research in this area
has been sporadic and limited; practical engineering has not yet
responded.
The major advantage of the shakedown analysis is
solely applicable for the arbitrary loading histories (often
in the practices). However, under geometrical nonlinearity
conditions, usually considered for steel structures, difﬁculties
have generally appeared in this kind of problem (shakedown
analysis).
In recent years, numerous authors have concentrated their
efforts on the advanced nonlinear analysis of 3D steel
frames [12–17]. Modern analysis must take into account the
∗
Corresponding author.
Email address: VanLong.Hoang@student.ulg.ac.be (H. Van Long).
interrelated effects of material inelasticity and geometrical
nonlinearity in furnishing an adequate response to the problems
of structural systems and their components. The complicated
3D steel structures may be solved in this direction. However,
this development is based on the step by step methods that may
contain a lot of difﬁculties in considering the cases of arbitrary
loading histories.
Consequently, the parallel development of both step by
step and direct methods (by mathematical programming) is
necessary. They give a better view of the behaviour of the
real structure and also they may mutually make up for
their deﬁciencies. It is the fundamental motivation of our
work: the development of the theoretical foundations and the
practical software useful for inelastic structures analysis and
optimization. In the following, one can ﬁnd a brief presentation
of a computer program, namely CEPAO:
This package had been developed in the Department of
Structural Mechanics and Stability of Constructions of the
University of Li` ege by NguyenDang Hung et al. in the
1980s [9–11]. Indeed, CEPAO was a uniﬁed package devoted to
automatically solving the following problems for 2D frames:
Elastic analysis, limit rigidplastic analysis with proportional
loadings; step by step elastic–plastic analysis; shakedown
analysis with variable repeated loadings; optimal plastic design
with ﬁxed loading; optimal plastic design with choice of
discrete proﬁles and stability checks; shakedown plastic design
01410296/$  see front matter c 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2007.12.009
1896 H. Van Long, N. Dang Hung / Engineering Structures 30 (2008) 1895–1904
with variable repeated loadings; shakedown plastic design with
updating of elastic response in terms of the plastic capacity.
With the CEPAO, efﬁcient choice between statical and
kinematic formulations is realized leading to a minimum
number of variables; also there is a considerable reduction
of the dimension of every procedure. The basic matrix
of LP algorithm is implemented under the form of a
reduced sequential vector which is modiﬁed during each
iteration. An automatic procedure is proposed to build up
the common characteristic matrices of elastic–plastic or rigid
plastic calculation, particularly the matrix of the independent
equilibrium equations. Application of duality aspects in the LP
technique allows direct calculation of dual variables and avoids
expensive reanalysis of every problem.
At this time, the CEPAO is extended to the case of space
steel frames with the following problems: rigidplastic analysis
and design by LP; elastic–plastic analysis by the stepbystep
method in both ﬁrst and second order (Pdelta effects). The limit
and shakedown analysis is based on the upper bound theorem
while the rigidplastic design is supported by the lower bound
theorem. It is important to indicate here that in the case of
frame structures both kinematic and statical methods lead to
the same one and only solution. The geometrical nonlinearity is
completely ignored in this direct method (by LP).
Because of the limitation of an article, the present work
describes only details of the module of limit and shakedown
analysis where some original contributions are indicated in the
Section 3. We hope that other contents will be presented in our
next papers.
2. Assumptions and modelling plastic hinges
The following assumptions have been made:
– Loading is quasistatic and service load domain is speciﬁed
by linear constraints;
– The torsional stiffness and the effect of the shear forces are
negligible;
– The rigid — perfectly plastic material is used in the limit
analysis; the elastic — perfectly plastic material is applied
in the other problems.
– Plastic hinges are located at critical sections.
Modelling plastic hinges
Since the effect of both shear forces and torsional moments
are ignored, the condition of plastic admissibility at the critical
sections becomes Φ(N, M
y
, M
z
) ≤ 0, with N is the normal
force and M
y
, M
z
are respectively bending moments about to
y and z axes. The plastic hinge modelling is described by the
choice of net displacement (relative)  force relationship at the
critical sections. In present work, the normality rule is adopted.
¸
¸
∆
θ
y
θ
z
¸
¸
¸
= λ
¸
¸
∂Φ/∂ N
∂Φ/∂ M
y
∂Φ/∂ M
z
¸
¸
¸
,
or, symbolically:
e
i
= λ
i
N
i
C
, (1)
where λ
i
is the plastic deformation magnitude; e
i
is the vector
of longitudinal displacement and two rotations of i th section;
N
i
C
is a gradient vector of the yield surface Φ.
The application of the LP techniques requires that the
nonlinear yield surfaces must be linearized. In civil engineering
practices, for bisymmetrical wideﬂange shapes, several
Standards replace the curvilinear yield surface by a polyhedron
sixteenfacet:
α
1
N
N
P
+ α
2
M
y
M
Py
+ α
3
M
z

M
Pz
= 1 for
N
N
P
≥ α
0
; (2a)
α
4
N
N
P
+ α
5
M
y
M
Py
+ α
6
M
z

M
Pz
= 1 for
N
N
P
< α
0
; (2b)
where: M
Py,
M
Pz
are the plastic moment capacity with respect
to y and z axis, N
P
is the squash load, 0 ≤ α
0
< 1, α
1
, . . . α
6
are the dimensionless coefﬁcients. The Eqs. (2a) and (2b) may
also be written
a
1
N + a
2
M
y
+ a
3
M
z
 = S
0
for
N
N
P
≥ α
0
; (3a)
a
4
N + a
5
M
y
+ a
6
M
z
 = S
0
for
N
N
P
< α
0
; (3b)
with S
0
is a referential value, and a
1
, . . . a
6
are the nonzero
coefﬁcients.
At the i th critical section, the plastic admissibility deﬁned
by Eqs. (3a) and (3b) has the following form:
Y
i
s
i
≤ s
i
0
, (4)
where matrix Y
i
contains the coefﬁcients a
1
, . . . a
6
; s
i
collects
the vector of internal forces; the column matrix s
i
0
contains the
corresponding terms S
0
.
According to the deﬁnitions of the matrices N
C
, Y, we can
see that
N
i
C
= Y
iT
. (5)
The detailed form of matrix N
C
is belowmentioned by Eq.
(13).
3. Application of LP
A systematic treatment of the application of LP in plastic
analysis can be found in [4,5]. In the present work, we restrict
ourselves to describing some practical aspects of the CEPAO
package applied to the case of 3D steel frames. They are:
the further reduction of the kinematic approach (Sections 3.2.2
and 3.3.2), and the direct calculation of the internal force (or
residual internal force) distribution (Sections 3.2.3 and 3.3.3).
3.1. General formulation
In the CEPAO, the canonical formulation of the LP is
considered:
Min π = c
T
xWx = b (6)
where π is the objective function; x, c, b are respectively
the vector of variables, of costs and of second member. W
H. Van Long, N. Dang Hung / Engineering Structures 30 (2008) 1895–1904 1897
is called the matrix of constraint. For the sake of simplicity,
the objective function has a state variable, and the matrix
formulation is arranged such that the basic matrix of the initial
solution appears clearly as follows:
¸
−c
T
1
1 −c
T
2
W
1
0 W
2
¸
¸
¸
x
1
π
x
2
=
¸
0
b
¸
. (7)
The basic matrix of the initial solution is
X
0
=
¸
1 −c
T
2
0 W
2
¸
.
Eq. (7) can then be written under a general form
W
∗
x
∗
= b
∗
. (8)
The matrices W
∗
, x
∗
, b
∗
and X
0
for both limit and shakedown
analysis problems will be accurately calculated in the following
sections:
3.2. Limit analysis by kinematic method
3.2.1. Kinematic approach
A kinematically admissible state is deﬁned by a collapse
mechanism that satisﬁes the condition of compatibility. It leads
to a positive external power supplied by the reference loading.
Based on the upper bound theorem of limit analysis, the
kinematic formulation of limit analysis can be stated as a LP
problem.
Min φ = s
T
0
λ
N
C
λ − Bd = 0
f
T
d = ξ
λ ≥ 0.
(9)
The safety factor will be obtained by
µ
+
= φ/ξ.
In Eq. (9), λ is the vector of the plastic deformation magnitude;
B is the kinematic matrix deﬁned in Appendix A; d, f are
respectively the vector of independent displacements and the
vector of external load; ξ is a constant (generally, one takes
ξ = 1).
3.2.2. Further reduction of the kinematic approach
In the kinematic method, the unknowns are the plastic
deformation magnitude, λ, and the independent displacement,
d (negative or positive). In LP procedure we need nonnegative
variables so that we adopt the change of the variables as in the
following:
d
= d + d
0
so that d
≥ 0.
The way to ﬁx the value of d
0
, which depends on the real
structure, such that d
are always nonnegative is explained in
the Appendix C. Now, the problem of Eq. (9) becomes
Min φ = s
T
0
λ
N
C
λ − Bd
= −Bd
0
f
T
d
= ξ + f
T
d
0
λ, d
≥ 0.
(10)
Therefore, the vector of variables, matrix of constraint, vector
of second member corresponding to the problem of Eq. (8) for
limit analysis are given below.
x
∗T
=
¸
π x
T
η
¸
=
¸
π d
T
λ
T
η
¸
b
∗T
=
¸
0 b
T
¸
=
¸
0 −Bd
0
ξ + f
T
d
0
¸
W
∗
=
¸
¸
1 0
T
−s
T
0
0
0 −B N
C
0
0 f
T
0
T
1
where η is an artiﬁcial variable which must be taken out of the
basic vector in the simplex process.
Because of the use of the simplex technique, ﬁnding an
initial admissible solution such that the initial value of any
variable (except the objective function) must be nonnegative
is needed. To satisfy this requirement, it appears that the
following arrangement leads to good behaviour of the automatic
calculation.
The linearized condition of plastic admissibility for the i th
section (Eq. (4)) may be expanded as follows:
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
a
i
1
−a
i
2
−a
i
3
a
i
1
a
i
2
−a
i
3
a
i
1
a
i
2
a
i
3
−a
i
1
a
i
2
a
i
3
−a
i
1
−a
i
2
a
i
3
−a
i
1
−a
i
2
−a
i
3
a
i
1
−a
i
2
a
i
3
−a
i
1
a
i
2
−a
i
3
a
i
4
−a
i
5
−a
i
6
a
i
4
a
i
5
−a
i
6
a
i
4
a
i
5
a
i
6
−a
i
4
a
i
5
a
i
6
−a
i
4
−a
i
5
a
i
6
−a
i
4
−a
i
5
−a
i
6
a
i
4
−a
i
5
a
i
6
−a
i
4
a
i
5
−a
i
6
¸
¸
¸
¸
N
i
M
i
y
M
i
z
¸
¸
¸
¸
¸
≤
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
S
i
0
S
i
0
S
i
0
S
i
0
S
i
0
S
i
0
S
i
0
S
i
0
S
i
0
S
i
0
S
i
0
S
i
0
S
i
0
S
i
0
S
i
0
S
i
0
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
. (11)
The Fig. 1 describes the projection of 16 planar facets of the
polyhedral stressresultant yield surface corresponding to the
16 inequalities numbered on the Eq. (11).
According to Eqs. (4) and (5), we see that Eq. (11) can be
written under symbolic formulation
N
iT
C
s
i
≤ s
i
0
. (12)
Put
˜
N
i
C
=
¸
¸
a
i
1
a
i
1
a
i
1
−a
i
2
a
i
2
a
i
2
−a
i
3
−a
i
3
a
i
3
.
Let us note that:
˜
N
i
C
is always nonsingular because a
i
1
, a
i
2
, a
i
3
are certainly positive.
1898 H. Van Long, N. Dang Hung / Engineering Structures 30 (2008) 1895–1904
Fig. 1. Projection of the yield surface on the plan M
y
OM
z
.
Then, matrix N
i
C
in Eq. (12) may be decomposed into three
submatrices:
N
i
C
=
¸
˜
N
i
C
−
˜
N
i
C
¯
N
i
C
¸
, (13)
with
¯
N
i
C
is the rest of N
i
C
after deducting
˜
N
i
C
and −
˜
N
i
C
.
The decomposition of matrix N
i
C
leads then to the following
form:
s
iT
0
=
¸
˜ s
iT
0
˜ s
iT
0
¯ s
iT
0
¸
=
¸
S
i
0
· · · S
i
0
¸
;
λ
iT
=
¸
˜
λ
iT
˜
λ
iT
+3
¯
λ
iT
¸
,
where
˜
λ
iT
=
¸
λ
i
1
λ
i
2
λ
i
3
¸
;
˜
λ
iT
+3
=
¸
λ
i
4
λ
i
5
λ
i
6
¸
;
¯
λ
i
=
¸
λ
i
7
. . . λ
i
16
¸
.
Let now S
i
be a diagonal matrix, such that
S
i
= diag [1 x sign of ((
˜
N
i
C
)
−1
b
i
)],
with
b
iT
=
¸
b
3(i −1)+1
b
3(i −1)+2
b
3(i −1)+3
¸
.
Let E be a unity matrix of dimension 3 × 3.
And consider now the new plastic deformation magnitude
distribution:
λ
i
T
=
¸
˜
λ
i
T
˜
λ
i
T
+3
¯
λ
iT
¸
,
in which
˜
λ
i
= 0.5(E + S
i
)
˜
λ
i
+ 0.5(E − S
i
)
˜
λ
i
+3
;
˜
λ
i
+3
= 0.5(E − S
i
)
˜
λ
i
+ 0.5(E + S
i
)
˜
λ
i
+3
.
With the mentioned arrangement, and if the case of initial basis
of variables is
¸
˜
λ
1
T
˜
λ
2
T
. . .
˜
λ
n
s
T
¸
, the initial basic matrix
may be determined as follows:
X
0
=
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
1 −˜ s
1T
0
−˜ s
2T
0
. . . −˜ s
n
s
T
0
0
0
˜
N
1
C
S
1
0 . . . 0 0
0 0
˜
N
2
C
S
2
. . . 0 0
.
.
.
.
.
.
.
.
. . . .
.
.
.
.
.
.
0 0 0 . . .
˜
N
n
s
C
S
n
s
0
0 0
T
0
T
. . . 0
T
1
,
in which, n
s
is the number of critical sections.
Easily, we may demonstrate that the initial solution
x
0
= X
−1
0
b
is certainly nonnegative.
3.2.3. Direct calculation of the internal force distribution
The strain rate at critical sections is chosen as variables
in kinematical approach. The collapse factor and mechanism
are given as output. To obtain the internal force distribution
while avoiding the static approach, the dual properties of LP
are used. The physical signiﬁcance of the dual variables may be
established as follows:
The canonical dual form of the LP problem of Eq. (6) is
Max (b
T
y + 0
T
h)
W
T
y + h = c
h ≥ 0,
(14)
in Eq. (14), y
T
=
¸
s
T
µ
−
¸
,
and h are the nonnegative slack variables:
h
T
=
¸
0
T
h
1T
h
2T
. . . h
n
s
T
¸
, with:
h
iT
=
¸
˜
h
iT
˜
h
iT
+3
¯
h
i
¸
.
It may be seen from the equality (14) that the internal forces are
related to the slack variables h:
s
i
= (
˜
N
iT
C
)
−1
(˜ s
i
0
−
˜
h
i
). (15)
H. Van Long, N. Dang Hung / Engineering Structures 30 (2008) 1895–1904 1899
It can be shown that the slack variables h are identiﬁed exactly
as the reduced costs ¯ c of the primal problem (8):
h = ¯ c = (X
 1
op
(1, :))W
∗
where X
−1
op
(1, :) is the ﬁrst row of the inverses basic matrix at
optimal solution.
The reduced costs ¯ c necessary for the convergence test of
the simplex algorithm are variable in the output of the primal
calculation. The automatic computation by (15) of the internal
forces distribution is independent of the type of collapse:
partial, complete or overcomplete.
3.3. Shakedown analysis by kinematic method
3.3.1. Kinematic approach
Based on the upper bound theorem of shakedown analysis,
the safety factor can be determined by minimizing the
kinematically admissible multiplier. Since the service load
domain is speciﬁed by linear constraints, the kinematic
approach leads to a LP problem:
Min φ = s
T
0
λ
N
C
λ − Bd = 0
s
T
E
N
C
λ = ξ
λ ≥ 0,
(16)
where s
E
is the envelope of the elastic responses of the
considered loading domain.
The safety factor will be obtained by
µ
s+
= φ/ξ.
3.3.2. Further reduction of the kinematic approach
As in the limit analysis, by an appropriate choice of d
0
such
that:
d
= d + d
0
≥ 0,
and by using the new plastic deformation magnitude
distribution, the vector of variables, matrix of constraints and
vector of second member corresponding to the problem of
Eq. (8) for shakedown analysis have the following form:
x
∗T
=
¸
π d
λ η
¸
;
b
∗T
=
¸
0 −Bd
0
ξ
¸
;
W
∗
=
¸
¸
1 0
T
−s
T
0
0
0 −B N
C
0
0 0
T
s
T
E
N
C
1
.
With initial basic matrix:
X
0
=
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
1 −˜ s
1T
0
−˜ s
2T
0
. . . −˜ s
n
s
T
0
0
0
˜
N
1
C
S
1
0 . . . 0 0
0 0
˜
N
2
C
S
2
. . . 0 0
.
.
.
.
.
.
.
.
. . . .
.
.
.
.
.
.
0 0 0 . . .
˜
N
n
s
C
S
n
s
0
0 s
1T
E
˜
N
1
C
S
1
s
2T
E
˜
N
2
C
S
2
. . . s
n
s
T
E
˜
N
n
s
C
S
n
s
1
And the initial basic variables:
¸
˜
λ
1
T
˜
λ
2
T
. . .
˜
λ
n
s
T
¸
.
The problems of Eqs. (10) and (16) are similar except for the
choice of the initial admissible point in the permissible domain
and the shakedown analysis requires preliminary calculation of
elastic responses.
3.3.3. Direct calculation of the residual internal force
distribution
Again the dual form of Eq. (16) is written similarly to Eq.
(14) with:
y
T
=
¸
ρ
T
µ
s−
¸
; h
T
=
¸
0
T
h
1T
h
2T
. . . h
n
s
T
¸
where ρ is the residual internal force vector, h
iT
=
¸
˜
h
iT
˜
h
iT
+3
¯
h
i
¸
.
From Eq. (14), the residual internal forces are related to the
slack variables h as the following relation:
ρ
i
= (
˜
N
iT
C
)
−1
(˜ s
i
0
− µ
s
˜
N
i
C
s
E
−
˜
h
i
).
As h is identiﬁed to be the reduced costs of the primal problem
(16), the distribution of residual internal force is directly
obtained without performing a second static approach.
4. Numerical examples and discussions
The presentation of the two following examples aims at a
comparison of the CEPAO results with those of some other
authors, and the comparison of the ultimate states of the frames
ﬁned by different models in CEPAO. Therefore, we present
not only the results given by limit and shakedown analysis
but also those calculated by the stepbystep method (a brief
presentation is presented in Appendix B).
In those examples, with the elastic–plastic analysis by hinge
byhinge method (ﬁrst and second order), the plastic interaction
function proposed by Orbison [18] for compact wideﬂange
sections is introduced in the CEPAO.
Φ = 1.15n
2
+ m
2
y
+ m
4
z
+ 3.67n
2
m
2
y
+ 3n
6
m
2
z
+4.65m
2
z
m
4
y
− 1 = 0,
in which, n = N/N
p
is ratio of the axial force to the squash
load, m
y
= M
y
/M
py
and m
z
= M
z
/M
pz
are the ratios of
the majoraxis and minoraxis moments to the corresponding
plastic moments. This yield surface is already used in several
references [12–14] that we consult to compare with our results.
In the direct analysis by LP, the plastic strength of cross
sections used in the AISC [19] is installed in the CEPAO, with
the value of a
1
, . . . , a
6
and α in the Eqs. (3a) and (3b) are:
a
1
= S
0
/N
p
; a
2
= 8S
0
/9M
yp
, a
3
= 8S
0
/9M
zp
, a
4
= S
0
/2N
p
,
a
5
= S
0
/M
yp
, a
6
= S
0
/M
zp
, α = 0.2.
Example a — Sixstorey space frame: Fig. 2 shown Orbison’s
sixstory space frame. The yield strength of all members is
250 MPa and Young’modulus is 206.850 MPa. Uniform ﬂoor
pressure of 4.8β
1
kN/m
2
; wind loads are simulated by point
loads of 26.7β
2
kN in the Ydirection at every beam–column
joint. In which, β
1
, β
2
are the factors that deﬁne the loading
domain.
1900 H. Van Long, N. Dang Hung / Engineering Structures 30 (2008) 1895–1904
Fig. 2. Example a— Sixstorey space frame ((a) perspective view, (b) plan
view).
Example b — Twentystorey space frame: Twentystorey space
frame with dimensions and properties shown in Fig. 3. The
yield strength of all members is 344.8 MPa and Young’modulus
is 200 MPa. Uniform ﬂoor pressure of 4.8β
1
kN/m
2
; wind
loads = 0.96β
2
kN/m
2
, acting in the Y direction.
Concerning the loading domain (for two examples), two
cases are considered for shakedown analysis: (a) 0 ≤ β
1
≤
1, 0 ≤ β
2
≤ 1 and (b) 0 ≤ β
1
≤ 1, −1 ≤ β
2
≤ 1. For ﬁxed
or proportional loading, we obviously must have: β
1
= β
2
= 1.
The uniformly distributed loads are lumped at the joints of
frames.
Diverse models have been adopted by some research
to capture both the material inelasticity and geometrical
nonlinearity [12–17], the corresponding load ratios are well in
accord (see Table 1). In which, the secondorder plastichinge
model has been used in the CEPAO, the large deﬂection is
ignored.
The results analysed by CEPAO with different methods
shown on the Table 2, Figs. 4 and 5 point out:
– An expectable coincidence of results calculated by limit
analysis and elastic–plastic analysis ﬁrst order, it allows us
to deduce the good convergence between the dual methods
Fig. 3. Example b— Twentystorey space frame ((a) perspective view, (b) plan
view).
Table 1
Comparison of results (elastic–plastic 2nd order)
Author Model Load multiplier
Example a Example b
Liew JYR2000 [12] Plastic hinge 2.010 –
Kim SE2001 [13] Plastic hinge 2.066 –
Chiorean CG2005 [14] Distributed
plasticity
2.124
(n = 30)
1.062 (n = 30)
Chiorean CG2005 [14] Distributed
plasticity
1.998
(n = 300)
1.005 (n = 300)
Cuong NH2006 [15] Fiber plastic
hinge
2.040 1.003
Liew JYR2001 [16] Plastic hinge – 1.031
Jiang XM2002 [17] Fiber
element
– 1.000
CEPAO2007 Plastic hinge 2.033 1.024
H. Van Long, N. Dang Hung / Engineering Structures 30 (2008) 1895–1904 1901
(a) Example a.
(b) Example b.
Fig. 4. Deformation at limit state given by CEPAO (From the left to the right: Elastic–plastic ﬁrst order, Elastic–plastic second order; Limit analysis; Shakedown
analysis, load domain a; Shakedown analysis, load domain b. The points on Fig. (a) indicate the plastic hinges.).
Fig. 5. Load–deﬂection results at point A (Figs. 2 and 3) given by CEPAO.
in the CEPAO (kinematic and static methods) and the good
correlation between the Orbison’yield surface and this in AISC
LRFD.
– In the case of symmetrical horizontal loading (seismic load
or wind load), the load multipliers determined by shakedown
analysis are the smallest (alternating plastic occurs).
1902 H. Van Long, N. Dang Hung / Engineering Structures 30 (2008) 1895–1904
Table 2
Ultimate strengths of the frames given by CEPAO with different analysis
Method Load multiplier Limit state
Example
a
Example
b
Hingebyhinge, ﬁrst order 2.489 1.689 Formation of a mechanism
Hingebyhinge, second
order
2.033 1.024 Unstableness
Limit analysis 2.412 1.698 Formation of a mechanism
Shakedown analysis,
domain load a
2.311 1.614 Incremental plasticity
Shakedown analysis,
domain load b
1.670 0.987 Alternating plasticity
5. Conclusions
From the performed work, we can draw the following
conclusions:
It appears that the canonical formulas in both limit and
shakedown analysis using LP for 3D steel frames may be
reduced by a special change of the variables and by a natural
choice of the initial basic matrix useful for the simplex
algorithm. The distribution of the internal forces may be
directly calculated by the application of duality aspects in
the LP technique. This allows avoiding expensive reanalysis
of the primal problem. The abovementioned techniques are
very suitable for automatic computation; consequently, they
have been completely implemented in CEPAO package. By the
way, the problem of ultimate strengths of the largescale 3
D steel frames under ﬁx or repeated loading, in the sense of
respectively limit and shakedown analysis, can be solved now
by the CEPAO package in an automatic manner look like any
ﬁnite element algorithm devoted to 3D frame structures. This
paper shows also that the simplex technique still is a necessary
tool in the automatic plastic analysis of 3D steel frameworks
after a less eventful period of the application of LP in the
analysis of frame structures.
Appendix A. Compatibility relation
Let e
T
k
=
¸
∆
A
θ
y A
θ
z A
∆
B
θ
yB
θ
z B
¸
be the vector
of the axial displacement and the net rotation of the member
ends (Fig. A.1(a)). Assemble for the frameworks (system of the
elements) we have the vector e.
Let d
T
k
= [d
1
d
2
d
3
d
4
d
5
d
6
d
7
d
8
d
9
d
10
d
11
d
12
d
ek
]
be the vector of the member independent displacements in
the global coordinate system OXYZ, as shown in Fig. A.1(b).
Assembled for the frameworks we obtain the vector d.
In the sense of limit analysis, we may think that: d
1
, d
2
,
d
3
, d
7
, d
8
, d
9
are the displacements corresponding to the
deﬂection mechanisms (beam and sideways); d
4
, d
5
, d
6
, d
10
,
d
11
, d
12
are the displacements showing the joints mechanisms;
d
ek
displacement in the longitudinal direction of the element,
describes the bar mechanisms (the bar translates along this
axis). Since the torsional stiffness of the elements is negligible,
we must eliminate the degree of freedom that only provokes
pure torsion in the bars.
The compatibility relation is deﬁned as
e = Bd,
where B namely the kinematic matrix that is determined by
B =
¸
k
A
k
T
k
L
k
. (A.1)
In Eq. (A.1), L
k
is a localization Boolean matrix of member k;
and
A
k
=
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
1 0 0 0 0 0 0 0 0 0 −1
0 0 −
1
l
k
1 0 0 0
1
l
k
0 0 0
0 −
1
l
k
0 0 −1 0
1
l
k
0 0 0 0
0 0 0 0 0 1 0 0 0 0 −1
0 0
1
l
k
0 0 0 0 −
1
l
k
−1 0 0
0
1
l
k
0 0 0 0 −
1
l
k
0 0 1 0
,
T
k
=
¸
¸
¸
¸
¸
C
k
C
k
C
k
C
k
1
,
with:
l
k
is the length of element k;
C
k
=
¸
c
11
c
12
c
13
c
21
c
22
c
23
c
31
c
32
c
33
¸
is the matrix of direction cosines of
element k;
C
k
=
¸
c
21
c
22
c
23
c
31
c
32
c
33
¸
.
Appendix B. Hingebyhinge method
In the CEPAO, the step by step method is used for the
increment nonlinear elastic–plastic analysis. After each step,
a new plastic hinge occurs, the elastic–plastic constitutive
equation is then updated and it replaces the elastic constitutive
in the elastic analysis. The other procedures are identical to
those of the elastic analysis, even when the Pdelta effect is
taken into account. Therefore, in a brief presentation, we only
present the construction of the elastic–plastic matrix.
Let e
C
and e
R
be the vectors of relative displacement
increments at the yielded sections (plastic hinges), and at the
elastic sections. Elastic constitutive equation (the Hooke’s low)
for the structure may be written as follows:
¸
s
R
s
C
¸
=
¸
D
RR
D
RC
D
T
RC
D
CC
¸ ¸
e
R
− 0
e
C
− e
p
C
¸
, (B.1)
in which, e
p
C
are the plastic strain increments at the plastic
hinges, s
R
and s
C
are the vectors of internal force
increments at the elastic sections and the plastic hinges.
Based on the Drucker’s normality rule, plastic strain
increments are normal to the yield surface and orthogonal to
H. Van Long, N. Dang Hung / Engineering Structures 30 (2008) 1895–1904 1903
(a) Relative displacements at critical sections. (b) Member’s independent displacements (global
axis).
Fig. A.1. Member k.
the internal force increments. Thus, for the sections (element
ends) in the plastic state, we have
e
pT
C
s
C
= 0 and, (B.2a)
e
p
C
= N
C
λ, (B.2b)
(λ and N
C
are deﬁned in Eq. (1)).
From Eqs. (B.2a) and (B.2b) and noting that λ is arbitrary,
we obtain
N
T
C
s
C
= 0. (B.3)
Using (B.1) and (B.2b) and (B.3), the plastic deformation
magnitude may be deduced
λ =
¸
(N
T
C
D
CC
N
C
)
−1
N
T
C
D
T
RC
(N
T
C
D
CC
N
C
)
−1
N
T
C
D
CC
¸
¸
e
R
e
C
¸
,
or:
¸
0
λ
¸
=
¸
0 0
R
1
R
2
¸ ¸
e
R
e
C
¸
. (B.4)
Eq. (B.2b) may be rewritten in the following form:
¸
0
e
p
C
¸
=
¸
0 0
0 N
C
¸ ¸
0
λ
¸
. (B.5)
Substituting (B.4) in (B.5), one obtains
¸
0
e
p
C
¸
=
¸
0 0
N
C
R
1
N
C
R
2
¸ ¸
e
R
e
C
¸
. (B.6)
From (B.1) and (B.6), one ﬁnally obtains the elastic–plastic
constitutive relation:
¸
s
R
s
C
¸
=
¸
D
RR
− D
RC
N
C
R
1
D
RC
− D
RC
N
C
R
2
D
T
RC
− D
CC
N
C
R
1
D
CC
− D
CC
N
C
R
2
¸ ¸
e
R
e
C
¸
.
Appendix C. Determination of the value of d
0
Suppose that
¯
d is the real displacement ﬁeld (the real
mechanism), in which,
¯
d
max
is the largest (absolute value). In
case of limit analysis, the safety factor is determined by the
equilibrium between the internal power and the external power.
µ
+
= s
T
0
λ/f
T
d = φ/ξ (C.1)
where the symbols are deﬁned in Eq. (9).
Based on the upper bound theorem of limit analysis, we have
µ
+
≤ µ
∗
, (C.2)
with µ
∗
is a load factor of any licit mechanism. By giving any
licit displacement ﬁeld d
∗
(for example, only one component
equals unity, and all other components are nil), µ
∗
may be
easily obtained.
From the Eqs. (C.1) and (C.2), one has
φ/ξ ≤ µ
∗
. (C.3)
On the point of view of geometry (kinematic), with the real
mechanism,
¯
d, there is at least a plastic deformation component,
¯ e, such that:
¯ e ≥
¯
d
max
/H
max
,
with H
max
is the maximum dimension of the structure.
Therefore, a lower bound of the internal power may be
evaluated.
φ ≥ s
p min
¯
d
max
/H
max
, (C.4)
in which, s
p min
is the smallest among the plastic capacity (N
P
,
M
py
, M
pz
) of all the sections on the structure.
From the Eqs. (C.3) and (C.4), the maximum displacement
is constrained by an upper bound.
¯
d
max
≤ ξµ
∗
H
max
/s
p min
.
Then, any value of d
0
that satisﬁes
d
0
≥ ξµ
∗
H
max
/s
p min
≥
¯
d
max
,
will lead: d
=
¯
d + d
0
is always nonnegative.
With similar argument, the value of d
0
for the shakedown
analysis may be obtained.
References
[1] Hodge PG. Plastic analysis of structures. New York: McGraw Hill; 1959.
[2] Neal BG. The plastic method of structural analysis. London: Chapman &
Hall; 1970.
[3] Massonnet Ch, Save M. Calcul plastique des constructions, vol. 1.
Belgique: Nelissen; 1976.
[4] Cohn MZ, Maier G. Engineering plasticity by mathematical program
ming. Canada: University of Waterloo; 1979.
[5] Smith DL, editor. Mathematical programming method in structural
plasticity. New York: SpringerVerlag; 1990.
[6] Jir´ asek M, Baˇ zant ZP. Inelastic analysis of structure. John Wiley & Sons,
LTD; 2001.
1904 H. Van Long, N. Dang Hung / Engineering Structures 30 (2008) 1895–1904
[7] Parimi SR, Ghosh SK, Cohn MZ. The computer programme DAPS for the
design and analysis of plastic structures. Solid Mechanics Division Report
26. Canada: University of Waterloo; 1973.
[8] Domaszewski M, Stanislawska EMS. Optimal shakedown design of
frames by linear programming. Computers & Structures 1985;21(3):
379–85.
[9] Dang Hung Nguyen. CEPAOan automatic program for rigidplastic and
elastic–plastic, analysis and optimization of frame structure. Engineering
Structures 1984;(6):33–50.
[10] Dang Hung Nguyen. Sur la plasticit´ e et le calcul des ´ etats limites par
´ el´ ement ﬁnis. Th` ese de doctorat sp´ ecial. l’Universit´ e de Li` ege; 1984.
[11] de Saxc´ e G, Ohandja LMA. Une m´ ethode automatique de calcul de l’effet
P  Delta pour l’analyse pas  ` a  pas des ossatures planes. Construction
m´ etallique. 1985 (3).
[12] Liew JYR, Chen H, Shanmugam NE, Chen WF. Improved nonlinear
plastic hinge analysis of space frame structures. Engineering Structures
2000;(22):1324–38.
[13] Kim SE, Park MH, Choi SH. Direct design of threedimensional frames
using advanced analysis. Engineering Structures 2001;(23):1491–502.
[14] Chiorean CG, Barsan GM. Large deﬂection distributed plasticity analysis
of 3D steel frameworks. Computers & Structures 2005;(83):1555–71.
[15] Cuong NH, Kim SE, Oh JR. Nonlinear analysis of space frames using
ﬁber plastic hinge concept. Engineering Structures 2006;(29):649–57.
[16] Liew JYR, Chen H, Shanmugam NE. Inelastic analysis of steel frames
with composite beams. ASCE Journal Engineering Structures 2001;
127(2):194–202.
[17] Jiang XM, Chen H, Liew JYR. Spreadofplasticity analysis of three
dimensional steel frames. Journal of Constructional Steel Research 2002;
(58):193–212.
[18] Orbison JG, McGUIRE W, Abel JF. Yield surface applications in
nonlinear steel frame analysis. Computer Method in Applied Mechanics
and Engineering 1982;(33):557–73.
[19] AISC. Load and resistance factor design speciﬁcation for steel buildings.
Chicago: American Institute of Steel Construction; 1993.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue reading from where you left off, or restart the preview.