You are on page 1of 22

F O U N D A T I O N S O F C I V I L A N D E N V I R O N M E N T A L E N G I N E E R I N G

No. 6 2005
Publishing House of Poznan University of Technology, Poznan 2005
ISSN 1642-9303
Andrius GRIGUSEVICIUS, Stanislovas KALANTA
Vilnius Gediminas Technical University
OPTIMIZATION OF ELASTIC-PLASTIC BEAM
STRUCTURES WITH HARDENING USING FINITE
ELEMENT METHOD
Elastic-plastic beam structures, subjected by distributed loads, optimization
problems are considered in this paper. Mixed method is suggested to form static and
geometrical equations by setting the finite element interpolation functions of internal
forces and displacements. That allows forming optimization problems with restricted
middle cross-sections of beams. General expressions of static and geometrical equations
are presented for a beam subjected by a distributed load. The optimization mathematical
models of elastic-plastic beam structures with linear hardening are formulated as
quadratic programming problems. Results of numerical example are presented and
briefly discussed.
Key words: elastic-plastic beam structures, optimization, linear hardening,
quadratic programming, distributed loading.
1. INTRODUCTION
Systems subjected only by concentrated loads are considered in many
works ([1-12]) intended for beam structures optimization. Optimization
algorithms of beam systems subjected by a distributed load are still not
sufficiently developed. Of course, a distributed load can be changed into a
system of some concentrated loads. However, in this case the size of the problem
increases considerably. Therefore, additional difficulties emerge, since success
of a solution of a non-linear optimization problem also depends on the size of
the problem. Trying to avoid these difficulties second order equilibrium finite
32 Andrius Grigusevicius, Stanislovas Kalanta
elements [13], [14] are used and the structure state equations are created by a
static method. In this case integral displacements of beams subjected by a
distributed load are included in geometrical equations as opposed to middle
section displacements. Meanwhile, the optimization problems of elastic-plastic
structures necessitates to constrain the actual (i.e. not integral) displacements.
For example, the dangerous displacement can be the middle span deflection of
beam, subjected by uniform distributed load. To that end we do not need to
model a beam by two common elements as we can create a beam element, one
degree of freedom of which corresponds to the middle section displacement.
In the papers [9], [15] mathematical models of optimization problems of
beam structures with hardening are formed as non-linear and non-convex
problems of mathematical programming. Such a formulation makes the solution
of the problem more difficult and induces to find alternative solution algorithms.
Therefore a recomposed mathematical model of the problem is suggested in this
article.
The aims of this work are to improve the optimization methods and
algorithms of beam structures subjected by a distributed load, to create a beam
finite element subjected by a distributed load when one of the nodal
displacements of this element corresponds to the middle section deflection and
to improve the formulae methods of structures state of static and geometric
equations.
2. FORMULATION OF STATIC AND GEOMETRIC
EQUATIONS OF THE FINITE ELEMENT
The equations of discrete model of structure are created by assemblage the
ones, corresponding to the separate elements. Therefore, first consider static and
geometric equations of single finite element.
Common element equations are being formed using an equilibrium finite
element method [13], [14]. Interpolation functions of the internal forces
(stresses) are set
( ) ( )
k k k
x x S H S = (2.1)
and first the element static equations are being formed
k k k
S A P = , (2.2)
which describes the relation between element nodal forces
k
P and internal
forces
k
S (Fig. 1). Then by referring to the principal of virtual forces or the
Optimization of elastic-plastic beam structures with hardening using finite element method 33
principal of minimum complementary the energy geometrical equations are
being formed
,
k
T
k k
u A =
(2.3)
or using physical equations,
,
0 k
T
k k k k
u A S D = +
(2.4)
where
k
u - displacement vector corresponded by the element static equations
(2.2),
k
- nodal strains vector. Element flexibility matrix
( ) { } ( )
)
=
k
l
k k
T
k k
dx x x
0
H d H D (2.5)
and distortion (initial or plastic strains) vector
( ) { } ( )
)
=
k
l
k
T
k k
dx x x
0
0 0
H . (2.6)
Here
k
d - flexibility matrix of infinitesimal element;
k
l - finite element length.
b)
a)
Q
k2 k1
Q
N
k
k2
M
k
N
M
k1
1 2
2 1
k4
P
k5
P
k6
P
k1
P
k2
P
P
y
x
k
l
k3
Fig. 1. Finite element under bending and compression or tension with 6 degrees of
freedom: a) generalized forces; b) nodal internal forces
For instance, static equations of a simple element under bending and
compression or tension (Fig. 1), described by linear internal functions are
34 Andrius Grigusevicius, Stanislovas Kalanta
k
k
k
k k
k k
k
k
k
k
k
k
k
k
k
k
k
k
k
N
M
M
l l
l l
M
N
Q
M
N
Q
P
P
P
P
P
P
2
1
2
2
1
1
6
5
4
3
2
1
1
1
1 1
1
1
1 1

= = P (2.7)
Geometrical equations i.e. strain expressions of this element are derived
using equation (2.3):
( )
3 4 1 1
1
k k k
k
k
u u u
l
+ = , ( )
6 4 1 2
1
k k k
k
k
u u u
l
+ = ,
5 2 k k k
u u + = ,
(2.8)
where the directions of displacements
ki
u correspond to the directions of
generalized forces
ki
P .
Such a method of creation of element equations is called a static one.
However, an alternative method of forming element equations is possible, which
we call a mixed method.
We will create an alternative mixed finite element by setting interpolation
function of internal forces (2.1) and displacements interpolation function
( )
( )
( )
( )
( )
( )
k ku
kx
ky
kx
ky
kx
ky
k
x
u
u
x
x
x u
x u
x u H
H
H
u = = = . (2.9)
We are formulating the equations of this element with reference to the principle
of virtual work:
( ) { } ( ) ( ) { } ( )
) )
= =
k k
l
k
T T
k
l
k
T
k k
T
k
dx x x dx x x
0 0
u S S S A ,
where operator
2
2
2
2
dx
d
dx
d
T

= A .
Using functions (2.1) and (2.9) we get such an equation
Optimization of elastic-plastic beam structures with hardening using finite element method 35
( ) { } ( )
k
l
ku
T T
k
T
k k
T
k
k
dx x x u H H S S
)
`

=
)
0
A .
From here the geometrical equations are
( ) { } ( )
k k k
l
ku
T T
k k
k
dx x x u B u H H =
)
`

=
)
0
A , (2.10)
where matrix
( ) { } ( )
)
=
k
l
ku
T T
k k
dx x x
0
H H B A . (2.11)
Then from virtual works equality
{ } { } { }
k
T
k
T
k k
T
k k
T
k
S B u S P u o o o = =
(2.12)
arise static equations
k
T
k k
S B P = .
(2.13)
Thus static and geometrical equations of the element can be formed
twofold making directly static equations and indirectly geometrical equations
or vice versa.
If while forming (2.2) and (2.13) equations the same generalized forces
k
P are considered, then
k
T
k k k
S B S A = .
Furthermore, when internal forces vectors
k
S of equilibrium and mixed element
are equivalent then
T
k k
B A = .
Consequently considering different generalized forces
k
P or vectors
k
S is
possible to get alternative element expressions of static and geometrical
equations.
36 Andrius Grigusevicius, Stanislovas Kalanta
3. STATIC AND GEOMETRICAL EQUATIONS OF THE
ELEMENT SUBJECTED BY DISTRIBUTED LOAD
Often several elements of a beam system are subjected by uniform
distributed loads of intensity
y
p and sometimes
x
p , too. When formulating
optimization problems a second order equilibrium element, one degree of
freedom of which corresponds to integral displacement (Fig. 2), is usually used
to model such beams. Distribution of internal forces in such an element is
described by these interpolation functions [14]:
( ) ,
2
4
2 3
1
3
2
2
2
2
2
1
2
2
k
k k
k
k k
k
k k
k
M
l
x
l
x
M
l
x
l
x
M
l
x
l
x
x M
|
|
.
|

\
|
+ +
|
|
.
|

\
|
+
|
|
.
|

\
|
+ =
( )
3 1
1
k
k
k
k
k
N
l
x
N
l
x
x N +
|
|
.
|

\
|
= .
(3.1)
3 1
k4
P
k5
P
k6
P
k1
P
k2
P
P
k3
y
x
k
l
Q
k3 k1
Q
N
k1
k3
M
k3
N
M
k1
1 3
2 1
k3
u
u
k2
u
k1
u
k6
u
k5
u
k4
=p
kx
2
k7
P
k8
P
u
k7
u
k8
3
2
=p
ky
a)
b)
c)
Fig. 2. Finite element with integral displacements
7 k
u and
8 k
u : a) generalized
forces; b) internal forces of nodes; c) generalized displacements
The stress state of this element is defined by a vector of internal forces
| |
T
k k k k k k
N N M M M
3 1 3 2 1
, , , , = S . Internal forces functions (3.1) are not
compatible with static equations of beam under bending and compression or
tension
( )
y
p
dx
x M d
=
2
2
,
( )
x
p
dx
x dN
=
Optimization of elastic-plastic beam structures with hardening using finite element method 37
a priori. Therefore, to perform these conditions, additional generalized forces
7 k
P and
8 k
P are introduced and two additional equations are formed
( )
( )
3 2 1
2 2
2
7
2
4
k k k
k
k
k
M M M
l dx
x M d
P + = = ,
( )
( )
3 1 8
1
k k
k
k
k
N N
l dx
x dN
P = = ,
where
7 k
P and
8 k
P are the intensities of the forces distributed along the element
length.
Static equations of this element are
3
1
3
2
1
2 2 2
3
3
3
1
1
1
8
7
6
5
4
3
2
1
1 1
4 8 4
1
1
3 4 1
1
1
1 4 3
k
k
k
k
k
k k
k k k
k k k
k k k
kx
ky
k
k
k
k
k
k
k
k
k
k
k
k
k
k
k
N
N
M
M
M
l l
l l l
l l l
l l l
p
p
M
N
Q
M
N
Q
P
P
P
P
P
P
P
P

= = P
(3.2)
These equations correspond to geometrical equations (2.4), which contain the
displacements
( )
)
=
k
l
ky k
dx x u u
0
7
, ( )
)
=
k
l
kx k
dx x u u
0
8
and the element flexibility matrix derived from (2.5) formula is
38 Andrius Grigusevicius, Stanislovas Kalanta
(
(
(
(
(
(
(
(

=
k
k
k
k
k
k
k
k
k
k
k
A
I
A
I
A
I
A
I
EI
l
5 5 , 2
5 , 2 5
2 1 5 , 0
1 8 1
5 , 0 1 2
15
D .
(3.3)
where E - elasticity modulus of material,
k
A and
k
I - element cross-section
area and moment of inertia.
Further aim is to create equations for the element subjected by a distribute
load (Fig. 3) when
7 k
P and
8 k
P are concentrated generalized forces and
7 k
u ,
8 k
u - second node displacements.
3 1
k4
P
k5
P
k6
P
k1
P
k2
P
P
k3
2 1
k3
u
u
k2
u
k1
u
k6
u
k5
u
k4
2
k7
P
k8
P
u
k7
u
k8
3
a)
b)
Fig. 3. Element subjected by a distribute load with linear displacements of middle node:
a) generalized forces, b) nodes displacements
The vertical displacements of this element are described by the fourth degree
polynomial [16]
( )
7
4
4
3
3
2
2
6
3
4
2
3 2
4
4
4
3
3
2
2
3
3
4
2
3 2
1
4
4
3
3
2
2
16 32 16
2 3 8 14 5
2 5 4 8 18 11
1
k
k k k
k
k k k
k
k k k
k
k k k
k
k k k
ky
u
l
x
l
x
l
x
u
l
x
l
x
l
x
u
l
x
l
x
l
x
u
l
x
l
x
l
x
x u
l
x
l
x
l
x
x u
|
|
.
|

\
|
+
+
|
|
.
|

\
|
+ +
|
|
.
|

\
|
+
+
|
|
.
|

\
|
+ +
|
|
.
|

\
|
+ =
,
(3.4)
and displacements ( ) x u
kx
- by second degree polynomial
Optimization of elastic-plastic beam structures with hardening using finite element method 39
( )
8
2
2
5
2
2
2
2
2
4
2 2 3
1
k
k k
k
k k
k
k k
kx
u
l
x
l
x
u
l
x
l
x
u
l
x
l
x
x u
|
|
.
|

\
|
+
|
|
.
|

\
|
+ +
|
|
.
|

\
|
+ = . (3.5)
Expressing (2.11) formula we derive such a matrix of geometrical equations:
10 5 , 12 5 , 2
10 5 , 2 5 , 12
32
14
31
1
1
64
2
32
2
32
32
1
1
14
31
15
1

=
k k k
k k k
k k k
k
l l l
l l l
l l l
B (3.6)
We can see that it is not equal to the transpose matrix of static equations (3.2).
Therefore static and geometrical equations of finite elements shown in Fig. 2 and
Fig. 3 are different. The reason for this are different degrees of freedom of
7 k
u
and
8 k
u . Element shown in Fig. 3 is recommended to apply as it does not
involve integral displacements.
External nodal forces
ki
F of the considered element, equivalent to the
linear distributed load, are derived from formulas
( )
( ) , 8 , 5 , 2 ,
; 7 , 6 , 4 , 3 , 1 ,
0
0
= =
= =
)
)
i dx p x H F
i dx p x H F
k
k
l
kx ki ki
l
ky ki ki
(3.7)
where ( ) x H
ki
- shape functions of the polynomials (3.4) and (3.5). The values
of these forces are shown in Fig. 4.
3 1
k4
F
k5
F
k6
F
k1
F
k2
F
F
k3
2
k7
F
k8
F
=
p l
kx k
6
=
p l
ky k
60
2
=
p l
ky k
30
7
=
2p l
kx k
3
=
p l
ky k
30
16
=
p l
ky k
60
2
-
=
p l
kx k
6
=
p l
ky k
30
7
Fig. 4. External nodal forces of the element
If the beam is subjected only by a linear distributed load ( 0 =
kx
p ), then it
is possible to model it by the element shown in Fig. 2 or Fig. 3. However in this
case
k k k
N N N = =
3 1
and distribution of the displacements ( ) x u
kx
in the
element is linear. Therefore with the view of the decrease of the number of
equations and variables it is preferable to use the element with 7 degrees of
40 Andrius Grigusevicius, Stanislovas Kalanta
freedom for such beams. This element is derived from the element described
above, eliminating the displacement
8 k
u . The stress state of such an element is
described by a vector | |
T
k k k k k
N M M M
, 3 2 1
, , = S and external forces
0
5 2
= =
k k
F F . The matrix of geometrical equations is derived from matrix (3.6)
by only changing the forth row to this one | | 0 , 0 , 15 , 0 , 0 , 15 , 0 and eliminating
the last row and the last column.
4. EQUATIONS OF FREE ORIENTED ELEMENT
In second and third chapters static and geometrical equations of
considered finite elements are presented in local coordinate xy system. However
beam constructions with beam elements which local coordinate xy axis is not
parallel to global all system coordinate ' ' y x axis are often occur in engineering
practice. Static and geometrical equations of all construction are being created
with reference to finite elements equations and are described by global forces P
and displacements u . Therefore the equations of all elements must be expressed
by global forces and displacements.
Relation between generalized forces of free oriented element (Fig. 5) in
local and global coordinate systems is described by equation
'
k k k
P T P = ,
(4.1)
where transformation matrix
o o
o o
o o
o o
o o
o o
cos sin
sin cos
1
cos sin
sin cos
1
cos sin
sin cos

=
k
T
(4.2)
Optimization of elastic-plastic beam structures with hardening using finite element method 41
P
k5
'
P
k5
P
k6
P
'
k4
P
k4
P
'
k8
P
k8
P
'
k7
P
k7
P
'
k1
P
k1
P
k2
P
'
k2
P
k3
'
x
y
'
x
y
o
Fig. 5. Local and global forces of the finite element
Analogically displacements
'
k k k
u T u =
(4.3)
By using formula (4.1) we derive static equation of the element in the
global coordinate system:
k k k k
S A P T =
'
or
k k k
S A P
' '
= ,
(4.4)
where
k
T
k k k k
A T A T A = =
1 '
.
(4.5)
Analogically geometrical equation
' ' '
k k k k k k k k
u B u T B u B = = =
or
0 u B S D = +
' '
0 k k k k k
,
(4.6)
where
k k k
T B B =
'
.
(4.7)
External nodal forces are interdependent by the equation
42 Andrius Grigusevicius, Stanislovas Kalanta
k
T
k k
F T F =
'
.
5. ELEMENT STRENGTH CONDITIONS
Strength conditions of a beam construction with linear hardening material
will be analyzed accepting that axial and shear forces work is negligible in
comparison with bending moments work and can be ignored. These strength
conditions must be secured in all above-mentioned structure design cross-
sections (i.e. finite elements nodes):
, 0
0
> +
i pi i
M M M , 0
0
> + +
i pi i
M M M . ..., , 2 , 1 n i =
(5.1)
here
i
M
0
- limit moment of the i-th cross-section; n - the number of design
cross-sections;
pi
M - plastic moment of the i-th cross-section. Plastic moments
vector of all structure can be expressed through plastic strains vector:
p p
H M = (5.2)
here H - hardening matrix.
M
M
M
pk
0
k
M
i
M
0
M
pi
i
l
i

( h )
M
i
l
M
0
M
pj
j
M
( l )
m1

( t)

m2
m
j
j

( e )
l
j
M
x
( s)
( r )
l l
Fig. 6. Internal forces distribution in the beam
Hardening matrix for a certain type of finite element can be derived
similarly like an elastic stiffness matrix
1
= D K . Plastic deformations can occur
in some parts of the finite element (Fig 6). Additional design points (h, r, s, e, l,
t) need to be introduced to bind these parts. Flexibility growth matrix for any
Optimization of elastic-plastic beam structures with hardening using finite element method 43
plastic zone can be derived using finite element method technique. For instance
flexibility increment matrix of plastic zone ih is:
| |
;
9 5 18 20 9 15 5
12 10 24 40 12 30 20
2 5
30
2 4 2
1 4 3
0 0 1
1
1
2 1
0
4 4
0
2 3
1
3 2 3 2 3 2
3 2 3 2 3 2
3 3 3
2 2 2
0
2
2
2 2
2 2
2 2
1
0
1
i i i i i i i
i i i i i i i
i i i i
l
i i
i i
i i
l
T
p
T
ih
p
ih
I
l
l l l
l l l
dx x x
x
x
l l
l l
l l
I
dx
I
i i



_



_ _

+ +
+ +

=

|
|
|
.
|

\
|

=
|
|
.
|

\
|
=
) )

B k k B d
(5.3)
here I - cross-section inertia moment; _ - hardening ratio; l - length of the
finite element; B - matrix of polynomial multipliers expressed through beam
nodes i, j, k coordinates;
ih
B - matrix of polynomial multipliers expressed
through beam nodes i, r, h coordinates;
p
k , k - polynomial coordinates vectors
for plastic zone ih and finite element k respectively. Second and third rows of
this matrix are irrelevant for finite element flexibility increment matrix since
these rows correspond additional design points r and h. Analogously deriving
matrices for others plastic zones and eliminating additional design points rows
we get such a flexibility increment matrix of k-th finite element:
3 3 3
3
2
3
1
3
2 2
3
1 1
3
2
3
1
3 3 3
5 2
2 5 2 5
2 5
30
j j j j
m m m m m m m m
i i i i
p
k
I
l



_

+ + +

= d . (5.4)
Second row of this matrix is derived adding corresponding members of plastic
zones lm and mt flexibility increment matrices.
Hardening matrix of all construction is derived inversing flexibility
increment matrices:
| |
1
=
p
k
diag d H .
(5.5)
44 Andrius Grigusevicius, Stanislovas Kalanta
6. CONSTRUCTION DISCRETE MODEL EQUATIONS
The equations of a whole construction are being formed from the
corresponding equations of the finite elements s k , ... , 2 , 1 = , conjugated to entire
discrete model.
Strength conditions of the construction from linear hardening material are
described by inequality
( ) 0 S S H S > + +
r e 0
(6.1)
or
0 S H S > +
0
(6.2)
where | |
T
n
S S S , ... , ,
2 1
= S - internal forces vector; - quasi-diagonal matrix
diagonally filled with the blocks of matrices
k
, s k , ... , 2 , 1 = ; H - hardening
matrix created considering to material hardening in the plastic sections;
e
S -
elastic internal forces vector;
r
S - residual internal forces vector. Plastic
multipliers vector is related with plastic strains vector by flow rule

T
p
= , 0 > ,
{ } 0 S H S = +
0
T
.
(6.3)
Construction discrete model, same as a finite element, static and
geometrical equations can be formulated in two ways directly formulating
static equations and indirectly geometrical equations or vice versa.
Lets say that the displacements and forces of all elements are described
by vectors | |
T
s
u u u , ... , ,
2 1
= u , | |
T
s
P P P , ... , ,
2 1
= P , | |
T
s
F F F , ... , ,
2 1
= F . The
displacements of the construction discrete model nodes are described by vector
| |
T
m
u u u , ... , ,
2 1
= u and forces acting in the direction of these displacements are
described by vector | |
T
m
F F F , ... , ,
2 1 0
= F , where m discrete model degree of
freedom.
Direct formulation of geometrical equations. Geometrical equations
of separate unconnected elastic-plastic finite elements s k , ... , 2 , 1 = are described
by system
0 u B DS = +
T
,
(6.4)
Optimization of elastic-plastic beam structures with hardening using finite element method 45
where { }
k
diag D D = , { }
k
diag B B = - quasi-diagonal matrices diagonally filled
with the blocks of matrices
k
D and
k
A . Here is considered, that distortions

T
p
= =
0
.
Separate elements are being connected to the whole discrete model by
displacement compatibility equations
Cu u = , (6.5)
which describe the relation between finite elements nodal displacements u and
nodal displacements of discrete model. Depending on them we derive
geometrical equations of discrete model
0 Bu DS = +
T
,
(6.6)
where
C B B = .
(6.7)
Unknowns of elastic-plastic constructions in analysis and optimization
problems are frequently considered as the residual internal forces
r
S and
displacements
r
u as opposed to the real ones. Since
r e
u u u + = and elastic
strains meet with the equation
0 Bu DS =
e e
, (6.8)
therefore, from the equation (6.6) can be derived geometrical equation, which
relates residual strains and displacements of the construction
0 Bu DS =
r
T
r
.
(6.9)
The principle of virtual displacements will be used for static equation
creation. With reference to equality of virtual works
{ } { } F u S
T T
o o = or { } { } F u S B u
T T T
o o =
(6.10)
we derive
F S B =
T
.
(6.11)
There are static equations of discrete model. Statically possible residual internal
forces
r
S have to meet with the equation
0 S B =
r
T
.
(6.12)
46 Andrius Grigusevicius, Stanislovas Kalanta
Direct formulation of static equations. Static equations of discrete
model can be formulated directly, using finite elements static equations (2.2) and
displacements compatibility equations (6.5).
Static equations of all elements (2.2) compounds such a system of static
equations
S A P = ,
(6.13)
where { }
k
diag A A = - quasi-diagonal static equations matrix, diagonal blocks
of which are matrices of separate blocks
k
A . These equations describe only
forces equilibrium inside every element but do not describe forces equilibrium
between elements. Discrete model static equations of every separate element and
all system we derive with reference to the principal of virtual displacements.
Virtual works of internal and external forces of construction are correlated
with equation
{ } { } { }
0
F u F u P u
T T T
o o o + = or { } { } F u P C u
T T T
o o = .
(6.14)
From here
, F P C =
T
(6.15)
where
.
0
F F C F + =
T
(6.16)
Considering to (6.13) we derive such discrete model static equation:
F AS = , (6.17)
where
A C A
T
= .
(6.18)
From equation (6.17) follows
0 AS =
r
, (6.19)
which meets with residual internal forces of the construction.
Thus static and geometrical equations of a finite elements system can be
formulated using one of above described methods. Equation systems created on
the ground of these methods are equivalent though expressions of these systems
can be different. Applying either method we will derive identical equation
systems only when the internal forces vectors
k
S and displacements vectors
k
u
will be the same. Then we will derive
T
B A = .
Optimization of elastic-plastic beam structures with hardening using finite element method 47
7. MATHEMATICAL MODELS OF OPTIMIZATION
PROBLEMS
Optimization problem of constructions of elastic-plastic linear hardening
material is being formulated as follows: for given construction design schema
and loading find the internal forces, displacements and optimal beam limit
internal forces satisfying given strength and stiffness conditions.
Designing beam constructions usually the same cross-sections are selected
for some beam groups. Lets say that the limit internal forces of such beam
groups are described by the optimized parameters vector
0
M . Relation between
vectors
0
S and
0
M is described by the equation
0 0
GM S = , (7.1)
where G - construction configuration matrix. Stiffness conditions are described
by displacement restrictions
+
s u Lu ,
(7.2)
and optimal criterion is minimal value of function
( )
0 0
M S
T
f = ,
(7.3)
here - weight coefficient vector which members are proportional to the total
length of the same cross-section beams;
+
u - maximal values of restricted
displacements.
Restraints system of the considered optimization problem is made from
three condition groups: equations (static and geometrical) that describe
construction stress-strain state and strength and stiffness restrictions. Thus,
considering to above used notation, construction optimization is expressed by
such a model: find
0
min M
T
,
(7.4)
when
0 S B =
r
T
or 0 AS =
r
,
0 Bu DS = +
r
T
r
,
( ) 0 S S H GM > + +
r e 0
,
( ) { } 0 S S H GM = + +
r e
T
0
, 0 > ,
( )
+
s + u u u L
r e
, 0 M >
0
.
(7.5)
48 Andrius Grigusevicius, Stanislovas Kalanta
Unknowns in this problem are vectors of residual internal forces
r
S and
displacements
r
u as well as limit internal forces and plastic multipliers vectors
0
M and . This is the problem of nonlinear and non-convex programming. As
shown in [14], for the ideal plastic case such a problem can be recomposed into
equivalent quadratic programming problem, which for the positively definite
quadratic matrix has only one solution. Applying the technique proposed in [14]
we derive such a model: find
( ) { }
r e
T T
S S H GM M + + +
0 0
min ,
(7.6)
when
0 S B =
r
T
or 0 AS =
r
,
0 Bu DS = +
r
T
r
, 0 > ,
( ) 0 S S H GM > + +
r e 0
,
( )
+
s + u u u L
r e
, 0 M >
0
.
(7.7)
Using the solution of static and geometrical equations
( ) U D B B D B u = =

T T T
r
1
1
1
,
(7.8)
( ) D D B B D B B D S = |
.
|

\
|
=

T T T T
r
1 1
1
1 1
(7.9)
we can eliminate residual internal forces and displacements from this model.
Then we derive such a problem: find
( ) { } S H GM M + + +
e
T T
0 0
min ,
(7.10)
when
( ) 0 S H GM > + +
e 0
,
( )
+
s + u U u L
e
, 0 M >
0
, 0 > .
(7.11)
8. NUMERICAL EXAMPLE
Described method is realized for three-span beam (Fig. 7). Beam material
elastic modulus kPa 10 205
6
= E , yield limit kPa 240000
0
= o . A list of
displacement restrictions is written in Table 1. The beam cross-section is I-type
(IPE). The dependences among beam cross-section inertia moment I , area A
and limit bending moment
0
M according to [17] are:
Optimization of elastic-plastic beam structures with hardening using finite element method 49
3
3 0 0
b
A a M o = ,
1
1
b
A a I = ,
(8.1)
here
1
a ,
3
a ,
1
b ,
3
b - cross-section shape coefficients (according to [17] for IPE
cross-section 7918716 , 0
1
= a , 8294723 , 0
3
= a , 319975 , 2
1
= b , 660459 , 1
3
= b ).
2m 2m 6m 3m 2m
15m
20 kN
10 kN/m
40 kNm
M
M M
01
02 03
Fig. 7. Beam design schema
Optimization of this beam is performed for two cases:
- beam material is ideal elastic-plastic ( 0 = _ );
- beam material is elastic plastic with linear hardening (
8
10 1 , 6

= _ ).
Iterative solving of the mathematical model (7.6 - 7.7) was used for this
problem. Cross-section inertia moments and relative plastic zones lengths was
freely chosen in the first iteration. Obtained results (Table 1) shows, that the
displacements their limit meanings reached in the first and second spans for both
cases. Objective function (OF) meaning in the second case is smaller by 11,3,
and this shows the expediency of hardening evaluation aiming to use less
material resources.
Table 1. Data and results of numerical example
System Displacement
restrictions, cm
Obtained
displacements, cm
Obtained limit
bending
moments, kNm
Ideal elastic-plastic
0 = _ ;
78 337 OF , =
I span 0 4
1
, u
r
s
II span 0 6
2
, u
r
s
III span 0 5
3
, u
r
s
I span 0 4
1
, u
r
=
II span 0 6
2
, u
r
=
III span 33 1
3
, u
r
=
33 13
01
, M =
56 25
02
, M =
22 26
03
, M =
Elastic-plastic with
hardening
8
10 1 6

= , _ ;
54 299 OF , =
I span 0 4
1
, u
r
s
II span 0 6
2
, u
r
s
III span 0 5
3
, u
r
s
I span 0 4
1
, u
r
=
II span 0 6
2
, u
r
=
III span 46 4
3
, u
r
=
66 10
01
, M =
53 28
02
, M =
14 17
03
, M =

50 Andrius Grigusevicius, Stanislovas Kalanta
9. CONCLUSIONS
1. Optimization algorithms of beam structures subjected by distributed loads
are improved in this paper. Mixed method is suggested to form static and
geometrical equations by setting the finite element interpolation functions of
internal forces and displacements.
2. General expressions of static, geometrical equations and strength conditions
of a beam element subjected by a distributed load are formed using mixed
method. That allows forming optimization problems with restricted middle
cross-sections of the beams. Formulating the equations by static method
there is no such a possibility.
3. Formed general finite element equations allow to unify the equations
formation of all construction. It is shown, that using static and mixed
methods we can form a lot of alternative equations systems for the one
construction. These equations systems are equivalent but their numerical
expressions will be equal only when internal forces vectors
k
S and
displacements vectors
k
u will be the same for the both methods.
4. The optimization mathematical model with non-linear conditions are
modified to quadratic programming problem and applied for optimization of
structures with linear hardening and subjected by distributed loads.
REFERENCES
1. Cyras A.A.: Mathematical Models for the Analysis and Optimization of
Elastoplastic Structures, New York: John Wiley, 1983, 121 p.
2. Borkowski A.: Analysis of Skeletal Structural Systems in the Elastic and
Elastic-plastic Ranges, Warshawa: PWN Elsevier, 1988, 200 p.
3. Atkocinas J.: Design of elastoplastic systems under repeated loading,
Vilnius: Science and Encyclopedia Publishers, 1994, 148 p. (in Russian).
4. Skarauskas V., Merkevicite D., Atkocinas J. Load Optimization of
Elastic-plastic Frames at Shakedown, Civil Engineering (Statyba), Vol 7,
No 6, Vilnius: Technika, 2001, p. 433440 (in Lithuanian).
5. Karkauskas R., Norkus A.: Optimization of Geometrically Nonlinear
Elastic-Plastic Structures Under Stiffness Constraints, Mechanics Research
Communications, Vol 28, No 5, 2001, p. 505512.
6. Yuge K, Kikuchi N.: Optimization of a Frame Structure Subjected to a
Plastic Deformation, Structural optimization, Vol 10, 1995, p. 197208.
7. Kaliszky S., Logo J.: Layout and Shape Optimization of Elastoplastic Discs
with Bounds of Deformation and Displacement, J. Mech. Struct. Mach. 30,
2002, p.171192.
Optimization of elastic-plastic beam structures with hardening using finite element method 51
8. Hayalioglu M.S.: Optimum design of geometrically non-linear elastic-
plastic steel frames via genetic algorithm, Computers & structures, Vol. 77,
2000, p. 527-538.
9. Tin-Loi F.: Optimum Shakedown Design Under Residual Displacement
Constraints, Struct. Multidisc. Optim., 19, 2000, p. 130139.
10. Fedczuk P., Skoworsky W.: Non-linear analysis of plane steel prestressed
truss in fire, Civil engineering and Management, Vol. VIII, No 3, 2002, p.
177-183.
11. Missoum S., Grdal Z., Watson L.T.: A Displacement Based Optimization
Method for Geometrically Nonlinear Frame Structures, Struct. Multidisc.
Optim. 24, 2000, p. 195204.
12. Buhl T., Pedersen C.B.W., Sigmund O.: Stiffness Design of Geometrically
Nonlinear Structures Using Topology Optimization, Struct. Multidisc.
Optim. 19, 2000, p.93104.
13. Kalanta S.: The Equilibrium Finite Elements in Computation of Elastic
Structures, Civil Engineering (Statyba), Vol 1, No 1, Vilnius: Technika,
1995, p. 2547 (in Russian).
14. Kalanta S.: New Formulations of Optimization Problems of Elastoplastic
Bar Structures Under Displacement Constraints, Mechanika, No 5(20),
Kaunas: Technologija, 1999, p. 916 (in Russian).
15. Cyras A.A.: Extremum principles and optimization problems for lineary
strain hardening elastoplastic structures, Applied mechanics (Hpnxnanax
mexannxa), Vol 22, No 4, p. 8996 (in Russian).
16. Kalanta S.: Finite Elements for Modelling Beams Affected by a Distributed
Load, Civil Engineering (Statyba), Vol 5, No 2, Vilnius: Technika, 1999, p.
9199 (in Russian).
17. Saka M.P., Hayalioglu M.S.: Optimum design of geometrically nonlinear
elastic-plastic steel frames, Computers & structures, Vol. 38, No. 3, 1991, p.
329-344.
A. Grigusevicius, S. Kalanta
OPTYMALIZACJA PRJTOWYCH ELASTYCZNO-PLASTYCZNYCH
KONSTRUKCJI WYPRODUKOWANYCH Z WZMACNIAJACEGO SIJ
MATERIALU METODA ELEMENTW SKONCZONYCH
S t r e s z c z e n i e
W rozprawach optymalizacji prtowych elastyczno-plastycznych konstrukcji
rozpatruje si najczsciej systemy obciazane silami skupionymi. Praca jest poswicona
udoskonaleniu algorytmw optymalizacji prtowych konstrukcji obciazonych silami
52 Andrius Grigusevicius, Stanislovas Kalanta
ciaglymi. Rwnania statyczne oraz geometryczne elementu skonczonego i calej
konstrukcji proponuje si tworzyc metoda kombinowana, zadajac funkcje interpolacyjne
przemieszczen oraz sil. To sprzyja sformulowaniu zadania optymalizacji, w ktrym
mozna ograniczyc ugicia srodkowego przekroju w elementach zginanych.
Przedstawiono oglne formy rwnan statycznych oraz geometrycznych dla prta
obciazanego ciaglym obciazeniem. Zadania optymalizacji prtowych elastyczno-
plastycznych konstrukcji wyprodukowanych z liniowo wzmacniajacego si materialu sa
przedstawione jak zadania kwadratowego programowania. Przedstawione sa rezultaty
analizy obliczeniowej.
Received, 26.08.2003.

You might also like