Struct Multidisc Optim 22, 394406 Springer-Verlag 2001

Optimal plastic design of a bar under combined torsion, bending

and shear
W. Egner and M.

Zyczkowski
Abstract Optimal plastic design of a bar or beam under
torsion with superposed bending and shear is obtained
via the boundary perturbation method. Eects of bend-
ing and shear are expressed by a small parameter . The
results of optimization of solid bars and of hollow bars are
shown.
Key words plastic design, combined loadings, solid
bars, hollow bars
1
Introduction
In the present paper we consider optimal plastic design of
a bar under torsion with superposed simultaneous small
bending and shear. This problem will be solved by means
of the boundary perturbation method (BPM) with a bar
of constant circular or annular section assumed as the
basic solution (optimal in the case of pure torsion) and
a small perturbing parameter describing the eect of
bending and shear.
The boundary perturbation method is now widely
used in various branches of solid and uid mechanics.
BPM was applied to determination of shapes of bodies
showing full plastication at the stage of collapse rst
by Kordas and

Zyczkowski (1970). Full plastication, if
possible at all, was then regarded as the rst step to-
wards subsequent optimal plastic design. Ample survey of
the relevant literature was given in the paper by Egner
et al. (1994) and will not be repeated here. We mention
just some more recent related papers by

Zyczkowski and
Egner (1995) and by Egner (1996, 2000a,b), devoted to
optimal plastic design of rotationally symmetric elements
Received May 15, 2000
W. Egner and M.

Zyczkowski
Institute of Mechanics andMachine Design, CracowUniversity
of Technology (Politechnika Krakowska), ul. Warszawska 24,
PL 31-155 Krak ow, Poland
e-mail: [email protected]
and to plane heads of tension members. General approach
to perturbation analysis of optimization problems with
some reference to optimization in plasticity is presented
in a recent monograph by Bonnans and Shapiro (2000).
The problem of combined plastic torsion with bend-
ing (but without shear) was formulated by Handelman
(1944) and Hill (1948, 1950). They used the Prandtl stress
function. The relevant equation in terms of displacements
was derived by Piechnik (1961), Piechnik and

Zyczkowski
(1961). Piechnik applied the perturbation method (but
without boundary perturbation and optimal design) to
obtain eective formulae for a circular cross-section. Ad-
vantages of numerical integration of Piechniks equation
were pointed out by Miller and Malvern (1967). A more
detailed survey is given by

Zyczkowski (1981).
The present paper uses three displacements and mean
stress as the dependent variables. In comparison to Piech-
niks paper two essential generalizations are introduced:
the eect of shear resulting in the loss of longitudinal
homogeneity, and the boundary perturbation leading to
optimal design of the bar.
Plastic optimization of beams under bending with
shear was considered by Heyman (1959) and Rozvany
(1973, 1976) for some relatively simple cost functions.
Rozvany and Hill (1976) proved that even for grillages in
which the beams are permitted to resist torsion, the opti-
mal solution often leads to torsion-free beams. However,
in some structures or structural elements torsion cannot
be eliminated and must be taken into account (

Zyczkow-
ski 1997, 1998). The present paper is devoted to bars or
beams with prevailing torsion.
2
General perturbations for a plastic circular bar under
torsion
2.1
Assumptions, basic solution
As the basic solution (zero-th approximation) we con-
sider a prismatic bar of the length under pure torsion.
Optimal shape of the cross-section is then circular (within
the class of solid, simply connected sections) or annular
395
(within the class of doubly connected sections). Subse-
quent optimization of annular shapes leads to thin-walled
sections, but then a constraint imposed on the loss of sta-
bility of the wall must additionally be introduced. This
problem will not be considered in the present paper: we
simply assume that the ratio of the internal radius b to the
external radius a is prescribed and the wall stability is not
lost. The limit twisting moment M equals
M
t
=
2
3

0
_
a
3
b
3
_
, (1)
where
0
is the yield-point stress in pure shear. For a solid
section we may substitute b = 0. In the following we as-
sume the material to be perfectly plastic, isotropic, in-
compressible, obeying the HuberMisesHencky (HMH)
yield condition. The strains are assumed to be small. For-
mally, the HenckyIlyushin deformation theory is used,
but it may also be considered as the LevyMises theory
of plastic ow if strains are replaced by strain rates. One
end of the bar is assumed to be clamped and the other
free. The cylindrical coordinate system r, , z is intro-
duced with the origin in the centroid of free cross-section
(Fig. 1). The moments are shown as vectors with double
arrows.
In the basic state the bar is under twisting moment
M
z
= M
t
which causes full yielding of the bar (limit
load carrying capacity). The following equations showthe
non-zero stresses, strains and displacements, denoted by
an additional subscript
0
:

z
0
= r ,
z
0
=
0
,
0
=
r
2
0
= Cr ,
C =

2
0
, u

0
=r ( z) , (2)
where denotes unit angle of twist (arbitrary in the limit
state), and is the plastic modulus.
Fig. 1 Scheme of the bar
2.2
Governing equations in the general case
In the general case, under additional perturbing loadings
which cause mainly bending and shear we obtain multi-
axial state of stress and strain. Next we assume change of
the shape of the bar in order to satisfy the yield condition.
So, we can write the plasticity condition in the following
form:
(
r

z
)
2
+(
r

)
2
+(

z
)
2
+
6
_

2
r
+
2
rz
+
2
z
_
= 2
2
0
, where
0
=

3
0
. (3)
The equations of internal equilibrium and the strain-
displacement relations in cylindrical coordinates have the
form

r
r
+
1
r

+

rz
z
+

r

r
= 0 ,

r
r
+
1
r

+

z
z
+2

r
r
= 0 ,

rz
r
+
1
r

+

z
z
+

rz
r
= 0 , (4)

r
=
u
r
r
,

=
1
r
u

+
u
r
r
,
z
=
u
z
z
,

r
=
u

r

u

r
+
1
r
u
r

,
rz
=
u
r
z
+
u
z
r
,

z
=
1
r
u
z

+
u

z
. (5)
The physical equations can be written in the following
form:

k
= (
k

m
) ,
m
=
1
3

jj
, (6)
where
k
denotes Kroneckers symbol and Einsteins
summation convention is introduced. Five of the equa-
tions (6) are independent. Together with the incompress-
ibility condition

r
+

+
z
= 0 , (7)
we obtain a set of 16 equations with 16 unknowns.
According to the boundary perturbation method the
unknowns (including perturbed shape a = a(, z) and
b = b(, z) are expanded into series of certain small pa-
rameters
i
, i = 1, 2, . . . , n,
X =

j=0

k=0
. . .

m=0
X
jk...m

j
1

k
2
. . .
m
n
, (8)
396
where
X =
_

r
,

,
z
,
r
,
rz
,
z
,
r
,

,
z
,
r
,
rz
,
z
,
u
r
, u

, u
z
, a, b

T
, (9)
and individual parameters
i
correspond to various per-
turbing loadings, for example to concentrated moment
M
x
, concentrated force P
1
= P, uniformly distributed
loading q, etc. In what follows, we conne our consid-
erations just to one perturbing loading, namely concen-
trated force. The only small parameter, proportional to
this force, will be simply denoted by . Optimal design
with perturbation due to concentrated bending moment
(much simpler because of absence of shear and unifor-
mity along the axis z) was considered by Bochenek et al.
(1983). Moreover, the exact solution to this plastic op-
timization problem, was derived by

Zyczkowski (1997),
who proved that the solution obtained earlier by Ober-
weis and

Zyczkowski (1997) for a certain family of elliptic
sections satises the optimality condition. On the other
hand, perturbation due to uniformly distributed loading
can go along the lines of the present paper, being slightly
more complicated. Some equations given below will be
general, valid for any perturbing loading.
Making use of the expansions (8) we realise that ten
linear equations (4), (5), and (7) retain their form, just in-
dividual quantities are labelled with relevant additional
subscripts. On the other hand, six remaining non-linear
equations (3) and (6) are subject to changes resulting
in their linearization. From (3) with substituted (2) we
obtain

z
i
= f
i
_

k
(i1)
,
k
(i2)
, . . .
_
, i = 1, 2, . . . , (10)
where
f
1
= 0 ,
f
2
=
1
12
0
_
(
r
1

1
)
2
+(

z
1
)
2
+
(
z
1

r
1
)
2
+6
2
r
1
+6
2
rz
1
_
,
f
3
=
1
6
0
[(
r
1

1
) (
r
2

2
) +
(

z
1
) (

z
2
) +(
z
1

r
1
) (
z
2

r
2
) +
6
r
1

r
2
+6
rz
1

rz
2
] , (11)
and so on, and from (6)

k
i
=
i

j=0

j
_

k
(ij)

m
(ij)
_
. (12)
2.3
Basic equations of the rst correction
On the rst level of correction we obtain from (10) and
(11)
z
1
= 0. The equations of internal equilibrium are
therefore simplied. Making use of (2), (5), and (12) we
express the stresses in terms of displacements and mean
stress
m
1
as follows:

r
1
=
m
1
+

r
1
Cr
=
m
1
+
1
Cr
u
r
1
r
,

1
=
m
1
+

1
Cr
=
m
1
+
1
Cr
2
u

+
1
Cr
u
r
1
r
,

z
1
=
m
1
+

z
1
Cr
=
m
1
+
1
Cr
u
z
1
z
,

rz
1
=
1
2Cr

rz
1
=
1
2Cr
_
u
r
1
z
+
u
z
1
r
_
,

r
1
=
1
2Cr

r
1
=
1
2Cr
_
u

1
r

u

1
r
+
1
r
u
r
1

_
. (13)
Substituting the above relations into internal equilibrium
equations (4) and incompressibility condition (7) we ob-
tain the following set of 4 partial dierential equations
with 4 unknowns:
u
r
1
r
+
1
r
u

+
u
r
1
r
+
u
z
1
z
= 0 ,

m
1
r
+
1
Cr

2
u
r
1
r
2
+
1
2Cr
2

_
u

1
r

u

1
r
+
1
r
u
r
1

_
+
1
2Cr

z
_
u
r
1
z
+
u
z
1
r
_

1
Cr
3
_
u

+u
r
1
_
= 0 ,

m
1

+
1
2C

r
_
u

1
r

u

1
r
+
1
r
u
r
1

_
+
1
2Cr
_
u

1
r

u

1
r
+
1
r
u
r
1

_
+
1
Cr
2

_
u

+u
r
1
_
= 0 ,

m
1
z
+
1
2Cr

r
_
u
r
1
z
+
u
z
1
r
_
+
1
Cr

2
u
z
1
z
2
= 0 . (14)
The solution of the above set of equations depends on
the additional loading of the bar (which is already pre-
loaded by twisting moment). If we assume that loading
acts in the yz plane, we can look for the solution in the
following form:
397
u
r
1
=

n=1
f
rn
(r, z) sin(n) ,
u

1
=

n=1
f
n
(r, z) cos (n) ,
u
z
1
=

n=1
f
zn
(r, z) sin(n) ,

m
1
=

n=1
f
mn
(r, z) sin(n) . (15)
In the case of the horizontal or skew loadings the sine
as well as cosine functions had to be introduced to the
above equations, but for the concentrated force P
1
shown
in Fig. 1, (15) are sucient. Substituting (15) into (14) we
obtain
r
f
rn
r
+f
rn
nf
n
+r
f
zn
z
= 0 ,
2Cr
3
f
mn
r
+2r
2

2
f
rn
r
2
nr
f
n
r
+3nf
n

_
n
2
+2
_
f
rn
+r
2

2
f
rn
z
2
+r
2

2
f
zn
rz
= 0 ,
2nCr
2
f
mn
+r
2

2
f
n
r
2
2n
2
f
n
+2nf
rn
+nr
f
rn
r
= 0 ,
2Cr
3
f
mn
z
+r
2

2
f
rn
rz
+r
2

2
f
zn
r
2
+2r
2

2
f
zn
z
2
= 0 . (16)
These equations may easily be reduced to a set of 2n equa-
tions by eliminating f
n
, and f
mn
. Namely, from the rst
equation (16) we can calculate f
n
f
n
=
r
n
f
rn
r
+
f
rn
n
+
r
n
f
zn
z
, (17)
then from the third equation
f
mn
=
3
2Cn
2

2
f
rn
r
2

r
2Cn
2

3
f
rn
r
3

1
Cn
2

2
f
zn
zr

r
2Cn
2

3
f
zn
zr
2
+
1
2Cr
f
rn
r
+
1
Cr
f
zn
z
. (18)
As the result we obtain the following set of 2n equations
with unknowns f
rn
and f
zn
:
r
4
_

4
f
rn
r
4
+

4
f
zn
r
3
z
_
+r
3
_
4

3
f
rn
r
3
+3

3
f
zn
r
2
z
_

r
2
n
2
_
2

2
f
rn
r
2
+

2
f
rn
z
2
+2

2
f
zn
zr
_
+
n
2
(n
2
1)f
rn
= 0 ,
r
4
_

4
f
rn
r
3
z
+

4
f
zn
r
2
z
2
_
+r
3
_
3

3
f
rn
r
2
z
+2

3
f
zn
rz
2
_

r
2
n
2
_
2

2
f
rn
rz
+4

2
f
zn
z
2
+

2
f
zn
r
2
_
= 0 . (19)
2.4
Boundary conditions
Boundary conditions for a contour in cylindrical coordi-
nates r = a(, z) were derived by Egner et al. (1994). For
a lateral surface free of loadings they may be written in
the form
a
r

r
a
a
z

zr
= 0 ,
a
r

a
a
z

z
= 0 ,
a
zr

z
a
a
z

z
= 0 . (20)
Similar conditions hold for the internal surface r = b(, z)
of a hollow bar. General expansion of (20) into power se-
ries of allowing also for the expansion of a, was also
given in the above-mentioned paper.
3
Eective solution for a solid bar under twisting
moment and concentrated vertical force at the free
end
3.1
The solution of governing equations for the rst
correction
The dierential equations (19) are close to ordinary dif-
ferential equations of Eulers type. Hence, we can look for
the solution in the class of homogenous polynomials of
mth-degree, where m can be any real or complex number.
Here we restrict mto integers. Then the solution takes the
form
f
rn
=

m=0
m

i=0
B
(n)
i,mi
r
i
z
mi
,
f
zn
=

m=0
m

i=0
C
(n)
i,mi
r
i
z
mi
. (21)
For a concentrated force it is sucient to consider n =
1, 3; m = 0, 1, 2, 3, 4. For example, for n = 1 and m = 0, 1,
we look for the solution in the form
f
r1
= B
(1)
00
+B
(1)
10
r +B
(1)
01
z ,
f
z1
= C
(1)
00
+C
(1)
10
r +C
(1)
01
z , (22)
398
and obtain the following general solutions:
u
r1
=
_
B
(1)
00
+B
(1)
10
r +B
(1)
01
z
_
sin ,
u
1
=
_
B
(1)
00
+2B
(1)
10
r +B
(1)
01
z +C
(1)
01
r
_
cos ,
u
z1
=
_
C
(1)
00
+C
(1)
10
r +C
(1)
01
z
_
sin ,

r1
= B
(1)
10
sin ,
1
=
_
B
(1)
10
+C
(1)
01
_
sin ,

z1
= C
(1)
01
sin ,

r1
= B
(1)
10
cos ,
rz1
=
_
B
(1)
01
+C
(1)
10
_
sin ,

z1
=
_
C
(1)
00
r
+C
(1)
10
+C
(1)
01
z
r
+B
(1)
01
_
cos ,

r1
=
1
Cr
_
3
2
B
(1)
10
+C
(1)
10
_
sin ,

1
=
1
2C
B
(1)
10
r
sin ,

z1
=
1
Cr
_
1
2
B
(1)
10
+2C
(1)
01
_
sin ,

rz1
=
1
2Cr
_
B
(1)
01
+C
(1)
10
_
sin ,

r1
=
1
2Cr
B
(1)
10
cos . (23)
Similar solutions can be derived for higher coecients n
and exponents m. They will be given in Sect. 3.2.
3.2
Boundary conditions for the rst correction
For the rst correction the boundary conditions obtained
from the expansions of (20) into power series take the
form:

r
1
|
r=a
0
= 0 ,

r
1
|
r=a
0

0
a
1
z
= 0 ,
a
0

rz
1
|
r=a
0

0
a
1

= 0 . (24)
In order to satisfy the rst equation of (24) we write the
formulae for
r1
using (23) and similar solutions for n =1,
m = 2, 3, 4 and n = 3, m = 0, 1, 2, 3, 4,

r
1
=
1
C
_

_
24B
(1)
40
+5C
(1)
31
_
r
2
+
_
1
2
B
(1)
30
+8C
(1)
03
_
r +
3
_
3C
(1)
40

7
2
B
(1)
13
_
rz +
3
2
B
(1)
10
r
+
B
(1)
01
r
+
_
3
2
B
(1)
11
+2C
(1)
02
_
z
r
+3
_
B
(1)
30
C
(1)
03
_
z
2
r
+
_
C
(1)
40
+
1
4
B
(1)
13
_
z
3
r
_
sin +
1
C
_
40
3
C
(3)
13
r
2
+
29
10
B
(3)
30
r +348C
(3)
04
rz +
40
9
B
(3)
20
+
40
9
C
(3)
12
z +
20
3
C
(3)
13
z
2
+
C
(3)
01
r
+2C
(3)
02
z
r
+
1
10
B
(3)
30
z
2
r
+4C
(3)
04
z
3
r
_
sin3 . (25)
The requirement to satisfy the rst of the above condi-
tions for any and z gives eight relations:
C
(3)
04
= 0 , B
(3)
30
=
200
3
C
(3)
13
a
0
, C
(1)
40
=
1
4
B
(1)
13
,
C
(1)
03
= B
(1)
30
, C
(3)
02
=
20
9
C
(3)
12
,
C
(1)
02
=
51
8
B
(1)
13
a
2
0

3
4
B
(1)
11
,
C
(3)
01
=
40
9
B
(3)
20
a
0
+180C
(3)
13
a
3
0
,
B
(1)
01
=
3
2
B
(1)
10

17
2
B
(1)
30
a
2
0
+24B
(1)
40
a
3
0
+5C
(1)
31
a
3
0
. (26)
In the remaining two conditions we have the same func-
tion a
1
. In order to satisfy both of them the following
condition (Schwarz condition) must hold:

r
1

r=a
0
= a
0

rz
1
z

r=a
0
. (27)
The formulae for
r
1
and
rz
1
are written below:

r
1
=
1
C
__
8B
(1)
40
+
3
2
C
(1)
31
_
r
2
+
_
1
2
B
(1)
30
4C
(1)
03
_
r +
_
9
2
B
(1)
13
3C
(1)
40
_
rz +B
(1)
20

1
2
B
(1)
02
+
_
B
(1)
21

3
2
B
(1)
03
_
z +
_
87
4
C
(1)
31
+120B
(1)
40
_
z
2
+
1
2
B
(1)
10
r
+
1
2
B
(1)
11
z
r
+
_
B
(1)
30
2C
(1)
03
_
z
2
r
+
1
2
B
(1)
13
z
3
r
_
cos +
1
C
_
20C
(3)
13
r
2
+
27
10
B
(3)
30
r +324C
(3)
04
rz +
7
3
B
(3)
20
+
7
3
C
(3)
12
z +
7
2
C
(3)
13
z
2
_
cos 3 , (28)
399

rz
1
=
1
C
__
3C
(1)
40

1
2
B
(1)
13
_
r
2
+
_
21
2
C
(1)
31
+48B
(1)
40
_
rz +
_
2B
(1)
21
+
3
2
B
(1)
03
_
r
_
4C
(1)
02
+
1
2
B
(1)
11
_
+
2
_
B
(1)
30
+4C
(1)
03
_
z +6
_
B
(1)
13
C
(1)
40
_
z
2
+
1
2
_
B
(1)
01
+B
(1)
10
_
1
r
+
_
1
2
B
(1)
02
B
(1)
20
_
z
r
+
1
2
_
3
2
B
(1)
03
B
(1)
21
_
z
2
r

_
29
4
C
(1)
31
+40B
(1)
40
_
z
3
r
_
sin +
1
C
_
20C
(3)
04
r
2

20
9
C
(3)
12
r
20
3
C
(3)
13
rz 4C
(3)
02

2
5
B
(3)
30
z 24C
(3)
04
z
2
+
1
2
C
(3)
10
r
+
B
(3)
20
z
r
+
1
2
C
(3)
12
z
2
r
+
1
2
C
(3)
13
z
3
r
_
sin3 . (29)
Fullment of (27) with (26) taken into account leads to
the following relations:
C
(3)
13
= B
(3)
30
= C
(3)
12
= C
(3)
02
= B
(1)
30
= C
(1)
03
= B
(1)
13
=
C
(1)
40
= B
(3)
20
= C
(3)
01
= B
(1)
11
= C
(1)
02
= 0 ,
B
(1)
10
=112B
(1)
40
a
3
0
24C
(1)
31
a
3
0
. (30)
Then the rst correction of the contour equals
a
1
(, z) = a
11
(z) cos +a
13
(z) cos 3 , (31)
where a
11
(z) and a
13
(z) are polynomials of the third
degree.
3.3
The condition of invariant location of the centroid
Now we demand from the rst correction not to cause
change of location of centroid, otherwise the additional
eccentric twisting would appear. The radius vector de-
scribing the contour of the bar in rst correction is given
by (31) with added a
0
. Then the static moment with re-
spect to the vertical axis y which is required to vanish we
can express in the form
S
y
=
_ _
xdA =
_ _
[a()]
2
cos dr d =
1
3
2
_
0
[a()]
3
cos d . (32)
Substituting (31) into (32) we obtain the following inte-
gral:
2
_
0
_
3 a
2
0
a
1
cos
2
+ a
3
1
cos
4
+6 a
0
a
1
a
3
cos
2
cos 3 +
3 a
3
1
a
3
cos
3
cos 3 +3 a
0
a
2
3
cos cos
2
3 +
3 a
1
a
2
3
cos
2
cos
2
3 + a
3
3
cos cos
3
3
_
d = 0 . (33)
From the above equation we obtain a
11
(z) = 0, and
a
13
(z) may be an arbitrary constant. So we obtain the
additional conditions:
B
(1)
40
=
29
160
C
(1)
31
,
B
(1)
21
=
3
2
B
(1)
03
,
B
(1)
20
=
1
2
B
(1)
02
+
96
5
C
(1)
31
a
2
0
,
B
(1)
01
=B
(1)
10
9B
(1)
03
a
2
0
, (34)
and the function describing the contour in the rst ap-
proximation has the form
a = a
0

C
(3)
10
6C
0
cos 3 . (35)
3.4
The integral conditions of internal equilibrium
In this section we require the sectional forces resulting
from the stress distribution to be equal to those resulting
from external loadings. In the case under consideration
we have three kinds of forces (the condition on normal
force is satised identically): bending moment, twisting
moment and shear force. These conditions can be written
in the following form:
2
_
0
a
_
0

k=0

zk
r
2
sin dr d = tz , (36)
2
_
0
a
_
0

k=0

k
(
rzk
sin +
zk
cos ) r dr d = t , (37)
2
_
0
a
_
0

k=0

zk
r
2
drd = M
z
= M
t
, (38)
where t = P and a = a(z, ). M
t
denotes load-carrying
capacity for twisting of circular cross-section. The upper
limit of integration a is subject to expanding into series:
a(z, ) =

i=0
a
i

i
. In the rst condition of (36) appears
the stress
z
. Taking into account the relations (26), (30),
and (34) we obtain the following formula for
z1
:
400

z
1
=
1
C
_
43
40
C
(1)
31
r
2
3B
(1)
02

384
5
C
(1)
31
a
2
0
9B
(1)
03
z
9
5
C
(1)
31
z
2
+
211
20
C
(1)
31
a
3
0
1
r
_
sin . (39)
Substituting the above formula into the condition (36) we
obtain the following relations:
C
(1)
31
= B
(1)
02
= 0 , B
(1)
03
=
Ct
3a
3
0
. (40)
Taking them into account we obtain from (29)

rz
1
=
3t
2a
3
0
_
a
2
0
r
r
_
sin +
1
2
C
(3)
10
r
sin3 . (41)
Substituting the above result into equations (37) and (38)
we nd them satised identically. Since
z
1
= 0, there is
no additional twisting for the rst correction.
3.5
The DruckerShield condition
At the end we demand to satisfy the necessary optimality
condition of the contour shape. This condition is known
in literature as the DruckerShield condition. According
to it, if the boundary of a structure in the plastic state is
not loaded, then the necessary condition of optimality of
the shape is that the unit power dissipated in the Levy
Mises theory or energy dissipated in the HenckyIlyushin
theory on that boundary is constant. It can be written in
the following form:
D =
ij

ij
|
S
= const . (42)
For the rst correction in view of
z
1
= 0, it has the fol-
lowing shape:
D =
0

z
0
|
r=a
0
+

z
1
|
r=a
0
+
0
a
1

z
0
r

r=a
0
_
+. . . = const . (43)
The above condition results in the following relations:
C
(1)
00
= 8B
(1)
03
a
3
0
, C
(3)
00
=
8
9
C
(3)
10
a
0
. (44)
Finally, we must require to satisfy the boundary con-
dition connected with clamping of the bar: u

(r = a
0
;
=0, ; z =) =0, u
z
(r =a
0
; =/2; z =) =0. These
conditions give certain relations involving the constants
B
(1)
10
, and B
(1)
00
not yet used. The conditions are not shown
here, because these constants will not appear in the fur-
ther analysis. On the other hand, the constant C
(3)
10
in
(35) is not determined by the DruckerShield condition
in the linear approximation (43). However, calculating
the cross-sectional area by integration of (35) squared
and leaving the term with
2
we nd that the minimum
is obtained for C
(3)
10
= 0 and in the following we assume
a
1
= 0.
3.6
List of formulae for the rst correction for solid bar
Finally for the rst correction we have the following for-
mulae:

r
1
=

1
= 0 ,
z
1
=
3t
a
3
0
z sin ,

r
1
=
z
1
= 0 ,
rz
1
=
3
2
t
a
3
0
_
a
2
0
r
r
_
sin ,
a
1
= 0 ,
1
=
Ct
6a
3
0

0
_
a
2
0
r
2
_
cos . (45)
3.7
The second correction for a solid bar
For the second correction the yield condition (3) gives

2
z
1
+3
2
rz
1
+6
0

z
2
= 0 . (46)
Making use of (45) we present
z
2
in the form

z
2
=
3
4
t
2

2
a
6
0

0
_
3
4
_
a
2
0
r
r
_
2
+z
2
_
(cos 2 1) . (47)
Similarly as for the rst correction the number of equa-
tions can be reduced by elimination of stresses and strains
by displacements and mean stress. Making use of (4)(7)
and (45) we eliminate the remaining stress components

r
2
=
m
2
+
1
4
H

0
_
3a
2
0
r

r
3
_
z sin2 +
1
Cr
u
r
2
r
,

2
=
m
2
+
1
4
H

0
_
3a
2
0
r

r
3
_
z sin2 +
1
Cr
2
u

+
1
Cr
2
u
r
2
,

z
2
=
m
2

1
2
H

0
_
3a
2
0
r

r
3
_
z sin2 +
1
Cr
u
z
2
z
, (48)

rz
2
=
1
2Cr
_
u
r
2
z
+
u
z
2
r
_

3
8
H

0
_
r
2
3

2
3
a
2
0
+
a
4
0
3r
2
_
sin2 ,

r
2
=
1
2Cr
_
u

2
r

u

2
r
+
1
r
u
r
2

_
. (49)
In the above equations H is proportional to the concen-
trated force squared, H = t
2
/
2
a
6
0
.
Substituting these stresses to the equilibrium equa-
tions, together with incompressibility condition we ob-
401
tain the following set of four linear nonhomogenous par-
tial dierential equations
u
r
2
r
+
1
r
u

+
u
r
2
r
+
u
z
2
z
= 0 ,

m
2
r
+
1
Cr

2
u
r
2
r
2
+
1
2Cr
2

2
u

2
r

3
2Cr
3
u

+
1
2Cr
3

2
u
r
2

2
+
1
2Cr

2
u
r
2
z
2
+
1
2Cr

2
u
z
2
rz

u
r
2
Cr
3
=
1
4
H

0
_
1
3
+
3a
2
0
r
2
_
z sin2 ,
1
r

m
2

+
1
2Cr

2
u

2
r
2
+
1
Cr
3

2
u

2
+
1
Cr
3
u
r
2

+
1
2Cr
2

2
u
r
2
r
=
3
2
H

0
z
_
1
_
8
9
+
a
2
0
r
2
_
cos 2
_
,

m
2
z
+
1
2Cr

r
_
u
r
2
z
+
u
z
2
r
_
+
1
Cr

2
u
z
2
z
2
=
H

0
_
4
3
r
a
2
0
r
+
3
2
z
2
r
+
a
4
0
r
3
_
sin2 . (50)
The particular solution can be searched within the func-
tions:
u
r
2
= f
r
2
(r, z) sin2 ,
u

2
= f

2
(r, z) cos 2 +
1
2
CH

0
r
3
z ,
u
z
2
= f
z
2
(r, z) sin2 ,
m
2
= f
m
2
(r, z) sin2 . (51)
Hence we obtain the following equations:
f
m
2
r
+
1
Cr

2
f
r2
r
2

1
Cr
2
f

2
r
+
3
Cr
3
_
f

2
f
r
2
_
+
1
2Cr
_
f
r
2
z
2
+

2
f
z
2
rz
_
=
1
4
H

0
z
_
1
3
+
3a
2
0
r
2
_
,
2
r
f
m
2
+
1
2Cr

2
f

2
r
2
+
2
Cr
3
_
f
r
2
2f

2
_
+
1
Cr
2
f
r
2
r
=
3
2
H

0
z
_
8
9
+
a
2
0
r
2
_
,
1
2Cr

2
f
r
2
zr
+
1
2Cr

2
f
z
2
r
2
+
f
m
2
z
+
1
Cr

2
f
z
2
z
2
=
H

0
_
4
3
r
a
2
0
r
+
3
2
z
2
r
+
a
4
0
r
3
_
,
f
r
2
r

2
r
f

2
+
fr
2
r
+
f
z
2
z
= 0 . (52)
Similarly as for the rst correction we eliminate f

2
and
f
m
2
and obtain the following two equations:
r
4
_

4
f
r
2
r
4
+

4
f
z
2
r
3
z
_
+r
3
_
4

3
f
r
2
r
3
+3

3
f
z
2
r
2
z
_

4r
2
_
8

2
f
r
2
r
2
+4

2
f
r
2
z
2
+8

2
f
z
2
zr
_
+12f
r
2
=
6
CH

0
r
3
z ,
r
4
_

4
f
r
2
r
3
z
+

4
f
z
2
r
2
z
2
_
+r
3
_
3

3
f
r
2
r
2
z
+2

3
f
z
2
rz
2
_

r
2
_
8

2
f
r
2
rz
+16

2
f
z
2
z
2
+4

2
f
z
2
r
2
_
=
8
CH

0
_
1
4
a
2
0
r
2
2r
4

3
2
r
2
z
2
a
4
0
_
. (53)
We assume the following form of particular solution for
the functionf
r
2
:
f
r
2
=
1
2
CH

0
r
3
z . (54)
Now, the function f
z
2
must satisfy simultaneously two
equations:

z
_
r
4

3
f
z
2
r
3
+3r
3

2
f
z
2
r
2
8r
2
f
z
2
r
_
= 0 ,

2
z
2
_
r
4

2
f
z
2
r
2
+2r
3
f
z
2
r
16r
2
f
z
2
_
4r
2

2
f
z
2
r
2
=
8
CH

0
_
1
4
a
2
0
r
2
2r
4

3
2
r
2
z
2
a
4
0
_
. (55)
If the particular solution will be written in the following
shape:
f
z
2
= D
40
r
4
+D
20
r
2
+D
02
z
2
+D
04
z
4
+Dln
r
a
0
, (56)
then the rst equation (55) is satised automatically, and
the second one results in
D
40
=
1
3
CH

0
, D
04
=
1
16
CH

0
,
D =2
CH

0
a
4
0
, D
02
=
1
4
D
20

1
16
CH

0
a
2
0
. (57)
Finally we have the particular solution in the following
form:
f
r
2
=
1
2
H
1
r
3
z ,
f
z
2
= D
20
r
2

1
4
_
D
20
+
1
4
H
1
a
2
0
_
z
2
+
1
3
H
1
r
4
+
1
16
H
1
z
4
2H
1
a
4
0
ln
r
a
0
. (58)
402
where the new notation has been introduced: H
1
=
CH/
0
= Ct
2
/
0

2
a
6
0
.
Now, we are looking for the general solution of the ho-
mogeneous part of (53). We restrict general solution to
the following functions:
u
r
2
= f
r
2
(r, z) sin(2) , u

2
= f

2
(r, z) cos (2) ,
u
z
2
= f
z
2
(r, z) sin(2) ,
m
2
= f
m
2
(r, z) sin(2) .
(59)
Using the way shown in (21) we assume
f
r
2
=
4

m=0
4

i=0
E
i(mi)
r
i
z
mi
,
f
z
2
=
4

m=1
4

i=1
F
i(mi)
r
i
z
mi
, (60)
and obtain
f
r
2
=
2
9
E
22
r
4
+E
30
r
3
+E
20
r
2
+E
21
r
2
z +E
22
r
2
z
2
+
E
31
r
3
z +E
00
,
f
z
2
=
1
4
E
31
r
4

1
8
E
21
r
3

1
4
E
22
r
3
z +F
20
r
2

6
5
E
30
r
2
z
3
5
E
31
r
2
z
2
+F
10
r
1
2
E
20
rz
1
4
E
21
rz
2

1
6
E
22
rz
3
+F
01
z
1
4
F
20
z
2
+
1
10
E
30
z
3
+
1
40
E
31
z
4
. (61)
Finally, the second corrections of displacements and
mean stress have the form:
u
r
2
=
_
2
9
E
22
r
4
+E
30
r
3
+E
31
r
3
z +
1
2
H
1
r
3
z +
E
20
r
2
+E
21
r
2
z +E
22
r
2
z
2
+E
00
_
sin2 ,
u

2
=
_
31
72
E
22
r
4
+
7
5
E
30
r
3
+
_
7
5
E
31
+H
1
_
r
3
z +
5
4
r
2
(E
20
+E
21
z +E
22
z
2
)
1
4
_
E
20
+D
20
+
1
4
H
1
a
2
0
_
rz +
3
20
E
30
rz
2
+
1
4
_
1
5
E
31
+
1
2
H
1
_
rz
3
+
1
2
E
00
_
cos 2 +
1
2
H
1
r
3
z ,
u
z
2
=
_
1
4
E
31
r
4
+
1
3
H
1
r
4

1
8
E
21
r
3

1
4
E
22
r
3
z +
E
20
r
2
+D
20
r
2

6
5
E
30
r
2
z
3
5
E
31
r
2
z
2

1
2
E
20
rz
1
4
E
21
rz
2

1
6
E
22
rz
3
+F
01
z
1
4
_
D
20
+
1
4
H
1
a
2
0
_
z
2

1
4
E
20
z
2
+
1
10
E
30
z
3
+
1
16
H
1
z
4
+
1
40
E
31
z
4

2H
1
a
4
0
ln
r
a
0
_
sin2 ,

m
2
=
1
C
_

79
72
E
22
r
2

9
5
E
30
r
_
9
5
E
31
+
17
12
H
1
_
rz
1
8
E
20

1
8
E
21
z
1
8
E
22
z
2

1
2
_
E
20
+D
20
+
7
4
H
1
a
2
0
_
z
r
+
3
10
E
30
z
2
r
+
1
2
_
1
5
E
31
+
1
2
H
1
_
z
3
r
_
sin2 . (62)
3.8
Boundary conditions for the second correction
Boundary conditions for the second correction take the
following form resulting from expansion of (20):

r
2
|
r=a
0
= 0 ,

r
2
|
r=a
0

0
a
2
z
= 0 ,
a
0
a
1

rz
1
r

r=a
0
+a
0

rz
2
|
r=a
0
+a
1

rz
1
|
r=a
0

0
a
2

= 0 . (63)
The rst equation of the set (63) is of the same type as for
the rst correction. In the third equation the terms con-
nected with a
1
are equal to zero and from the Schwarz
condition for a
2
we obtain similar equation as for the rst
correction:

r
2

r=a
0
= a
0

rz
2
z

r=a
0
. (64)
Since the way of satisfying the boundary conditions for
the second corrections is the same as for the rst one, we
will not give here the details and write the nal form of
the second correction for the external contour:
a
2
=
H
1
a
0
C
0
_
17
48
a
2
0

F
10
4H
1
a
0

5
8
z
2
_
cos 2 +
H
1
a
0
C
0
z
2
4
+C
a
. (65)
403
The integral equations of the internal equilibrium for
shear forces and bending moments are satised automat-
ically. The equilibrium condition for the twisting moment
which can be written in the following form:
M
z
=
_
A

z
r dA =
2
_
0
a
_
0
_

0
+
2

z
2
_
r
2
dr d =
M
t
+0 +0
2
+. . . (66)
(where a = a
0
+
2

2
) allows us to calculate constant C
a
:
C
a
=
3
10
H
1
a
3
0
C
1

0
. (67)
Now, the formula for a
2
can be written in the following
form:
a
2
=
H
1
a
0
C
0
__
17
48
a
2
0

F
10
4H
1
a
0

5
8
z
2
_
cos 2 +
1
4
z
2
+
3
10
a
2
0
_
. (68)
We should satisfy the Drucker-Shield condition. For the
second correction it can be written in the form (non-zero
terms only):
(
z
1

z
1
+
0

z
2
+
z
2

z
0
+
rz
1

rz
1
)
r=a
0
+
2C
2
0
a
2
= const . (69)
It turns out that this condition can be satised only ap-
proximately (terms of type z
0
cos 2 only). Using it we
can calculate the constant F
10
.
F
10
=
13
12
H
1
a
3
0
. (70)
After taking into account the above constant, dening the
dimensionless small parameter in the form:
=
t
a
2
0

0
, (71)
and taking into account the results of the rst correction
we can write the function describing lateral surface in the
following dimensionless form:
a
a
0
= 1+
2
_
5
8
_
1
z
2
a
2
0
_
cos 2 +
1
4
z
2
a
2
0
+
3
10
_
. (72)
The conditions of support for the second correction
are of the same form as for the rst correction: u
2
(r =
a
0
; = 0, ; z = ) = 0, u
z2
(r = a
0
; = /2; z = ) = 0.
The condition for displacement u
z2
is satised automat-
ically, thus the condition for u
r2
makes it possible to cal-
culate the constant E
00
.
At the end we compare the obtained shape with the
exact optimal shape of a bar under torsion and bending
in plastic range obtained by

Zyczkowski (1998), who con-
sidered the similar problem but without shear eect and
used this shape as an approximation if the bending mo-
ment is variable along the axis z. As a result the ellipse
with the axes given by the following relation:
a
x
a
0
= 1
2P
2
z
2
9M
2
t
+. . . ,
a
y
a
0
= 1+
4P
2
z
2
9M
2
t
+. . . , (73)
was obtained. Axes a
x
and a
y
are placed in the plane
z=const (Fig. 2), and a
0
is a radius of a bar under pure
torsion at the stage of collapse:
a
0
=
3
_
3M
t
2
0
. (74)
When using boundary perturbation method with shear
taken into account, the axes of a cross-sectionare given by
the following formulae resulting from (72):
a
x
a
0
= 1+
37
90
P
2
a
2
0
M
2
t

3
18
P
2
z
2
M
2
t
+. . . ,
a
y
a
0
= 1
13
90
P
2
a
2
0
M
2
t
+
7
18
P
2
z
2
M
2
t
+. . . . (75)
Comparing the relations (73) and (75) we can see that in
equation (73) all correcting terms depend on z. It is the
result of neglecting the eects of shear, taken into account
in (75). Bending eect, which is described by the terms
proportional to z
2
is for both solutions similar.
3.9
Numerical example
The example was calculated for /a
0
=4 and =0.15. The
following interpretation for the above data can be pre-
sented. The value of small parameter can be understood
as a ratio of bending moment in clamped cross-section
M
b
to the torque:
M
b
M
t
=
t
2
3

0
a
3
0
=
3
2

a
0
. (76)
For the above data this ratio is equal M
b
/M
t
= 0.9.
Figure 2 shows the corresponding optimal shape.
Figure 3 shows the shape of the longitudinal contour
of the bar, obtained as sections in the planes xz and yz.
It is seen from the third of equations (11) that the
third correction governed by
3
must contain the terms
with cos 3 which cannot be avoided as in the rst cor-
rection. The symmetry of the cross-section with respect
to the axis y is then lost. However, the calculations of
this correction are very cumbersome and will not be given
here.
404
Fig. 2 Optimal shape of the solid bar
Fig. 3 Variation of semiaxes along the bar
4
Eective solution for a hollow bar
4.1
Boundary conditions for the rst correction for a bar
with annular cross-section
For bending and torsion the layers which are close to the
axis of symmetry of a bar are not appropriately exploited.
So we may expect further prots for annular cross-section
of bars (doubly connected cross-sections). Such an ap-
proach was applied by Bochenek et al. (1983).
In the case of an annular cross-section of a bar the
form of the boundary conditions given by (20) and their
expansion into power series do not change. However,
we must remember that they must be satised at both
boundaries a(z, ) and b(z, ). It turns out that the exact
fullment of all equations is impossible. Therefore we de-
mand for these equations to be satised identically for the
terms z
0
and z
1
. For the terms with z
2
and z
3
which ap-
pear for example in the equation for
r1
, the fullment is
possible if we take into account higher powers of variables
r and z. On the other hand, taking into account the higher
powers makes impossible to satisfy the boundary condi-
tions for the terms connected with these higher powers.
It is clear that we can nd the solution with any preci-
sion we want, but always the terms with highest powers
would not full some equations. Hence, the obtained solu-
tion can be understood as an approximate solution. This
approximation is better if we consider sections closer to
the free end of a bar. For z 0 the solution leads to the
exact one.
It should be mentioned that in the case of the annu-
lar cross-section under consideration, only the solution
connected with trigonometric functions of a single angle
will be taken into account. The calculations are more
complicated here, since we have two independent surfaces
to design. This justies neglecting the terms with triple
angle. The experiences from the previous problem show
that these terms make only the volume bigger. Moreover
the analysis for a single angle gives chances to nd a non-
trivial correction for a = a(r, z) and b = b(r, z) as early
as for the rst correction. We have two independent sur-
faces to design, so the rigid movement of these surfaces
described by the solution with trigonometric functions of
a single angle is possible. The condition of invariant lo-
cation of centroid will not cause a zero solution. Then
further considerations dier from the previous one only
in the range of neglected terms with triple angle, fulll-
ing of boundary conditions in the terms with z
0
and z
1
and the necessity of satisfying the boundary conditions at
both lateral surfaces.
After satisfying boundary conditions at both surfaces
we obtain the following relations for the rst correction of
the shape:
a
1
=
1
c
0
_
3B
13
a
3
0

_
2B
21
+
3
2
B
03
_
a
2
0

5
2
B
11
a
0

1
2
(B
01
+B
10
)
_
cos ,
b
1
=
1
c
0
_
3B
13
b
3
0

_
2B
21
+
3
2
B
03
_
b
2
0

5
2
B
11
b
0

1
2
(B
01
+B
10
)
_
cos . (77)
The next step is demanding the invariant location of
centroid. This condition can be written in the following
form:
S
y
=
1
3
2
_
0
_
a
3
b
3
_
cos d =
1
3
2
_
0
_
a
3
0
+3a
2
0
a
1
+
. . . b
3
0
3b
2
0
b
1
. . .
_
cos d = 0 , (78)
405
where
a
1
= C
a
cos ,
b
1
= C
b
cos . (79)
The fullment of the above condition results in the follow-
ing relation:
a
2
0
C
a
b
2
0
C
b
= 0 . (80)
Next, we demand of satisfying the condition of the bend-
ing moment equilibrium. In the rst approximation it has
the form shown below:
2
_
0
a
0
_
b
0

z
1
r
2
sin dr d = tz . (81)
The remaining conditions, describing the equilibrium of
the twisting moment and shear force, are satised identi-
cally. The Drucker-Shield conditions are the same as for
the previous example, but they have to be satised at
both surfaces
_

z
1
+ a
1

z
0
r
_

r=a
0
= const ,
_

z
1
+ b
1

z
0
r
_

r=b
0
= const . (82)
After the above considerations we obtain the solution
with two arbitrary constants. However, detailed analysis
has shown that both of the constants in linear way in-
creased the dimensions of the bar. Therefore in further
considerations we assume them equal to zero. Finally we
can write the rst corrections of the shape in the following
form:
a
1
=
3
2
b
2
0
b
3
0
a
3
0
t

cos ,
b
1
=
3
2
a
2
0
b
3
0
a
3
0
t

cos . (83)
We dene the dimensionless small parameter in the same
way as for the rst example (71) introduce the following
notation: a
0
/b
0
= k, and obtain the following formulae:
a
a
0
= 1+
3
2
k
(1k
3
)
cos ,
b
a
0
=
1
k
+
3
2
k
3
(1k
3
)
cos . (84)
The following relation gives the dimensionless area of
a cross-section:
A
A
0
= 1+
2
9
8
_
1k
4
_
_
1
1
k
2
_ _
1
k
k
2
_
2
. (85)
In the rst approximation it is constant along the axis of
the bar. The second correction will not be given here.
4.2
Numerical example
Figure 4 shows the shape of the contour for the small pa-
rameter = 0.15 and ratio of radii k=2.
It is seen, that as a result of design the wall at the right
side is thicker, and at the left side is thinner. It is caused
by the accumulation of shear stresses at the right-hand
side (from torsion and shear).
Figure 5 shows the relation between dimensionless
area (85) and the ratio of radii k.
Fig. 4 Optimal cross-section of a hollow bar
Fig. 5 Dimensionless cross-sectional area in terms of the
ratio of radii k
406
Three curves are shown which correspond to various
values of small parameter , that means to various con-
tributions of bending and shearing in relation to the basic
load (twisting). The above picture shows that the greater
is the perturbation (force P) the greater are the prots in
volume.
5
Conclusions
1. Eective solutions for optimization of a bar under tor-
sion with superposed bending and shear are obtained
via the boundary perturbation method.
2. The rst correction of cross-sectional shape vanishes
for a solid bar, but is dierent from zero for a hollow
bar.
3. The more thin-walled is the doubly connected cross-
section, the greater is the prot in volume. How-
ever, we should remember that for thin-walled sec-
tions the condition of stability should be introduced.
Such a condition was not considered in the present
paper.
References
Bochenek, B.; Kordas, Z.;

Zyczkowski, M. 1983: Optimal plas-
tic design of a cross section under torsion and small bending.
J. Struct. Mech. 11, 383400
Bonnans, J.F.; Shapiro, A. 2000: Perturbation analysis of op-
timization problems. Berlin, Heidelberg, New York: Springer
Egner, W. 1996: Polynomial boundary perturbation for opti-
mal plastic design of heads of plane tension members. Eng.
Trans. 44, 261281
Egner, W. 2000a: Optimal plastic shape design of heads of
plane tension members with skew bearing surfaces. Eng. Opt.
32, 463483
Egner, W. 2000b: Optimal plastic shape design of rotationally
symmetric elements with conical bearing surfaces. J. Theor.
Appl. Mech. 38, 747765
Egner, W.; Kordas, Z.;

Zyczkowski, M. 1994: Optimal plastic
shape design via the boundary perturbation method. Struct.
Optim. 8, 145155
Handelman, G.H. 1944: A variational principle for a state of
combined plastic stress. Quart. Appl. Math. 1, 351353
Heyman, J. 1959: On the absolute minimum weight design of
framed structures. Quart. J. Mech. Appl. Math. 12, 314324
Hill, R. 1948: A variational principle of maximum plastic work
in classical plasticity. Quart. J. Mech. Appl. Math. 1, 1828
Hill, R. 1950: The mathematical theory of plasticity. Oxford:
Clarendon Press
Kordas, Z.;

Zyczkowski, M. 1970: Investigation of the shape
of non-circular thick-walled cylinders subject to full yielding
at the stage of plastic collapse. Bull. Acad. Pol. Sci., Ser. Sci.
Techn. 18, 839847
Miller, P.M.; Malvern, L.E. 1967: Numerical solution of
Piechniks equation for the combined bending and torsion of
rigid-plastic bars. Proc. 10-th Midwestern Conf. Sol. Mech.,
pp. 571580, Colorado State University
Oberweis, P.;

Zyczkowski, M. 1997: On Palas method for
plastic bending with torsion. J. Appl. Mech. 64, 435437
Piechnik, S. 1961: The inuence of bending on the limit
state of a circular bar subject to torsion. Archiwum Mech.
Stosowanej 13, 77106
Piechnik, S.;

Zyczkowski, M. 1961: On the plastic interaction
curve for bending and torsion of a circular bar. Archiwum
Mech. Stosowanej 13, 669692
Rozvany, G.I.N. 1973: Optimumdesign for bending and shear.
J. Eng. Mech. Div. ASCE 99, 11071109
Rozvany, G.I.N. 1976: Optimal design of exural systems. Ox-
ford, New York: Pergamon Press
Rozvany, G.I.N.; Hill, R. 1976: General theory of optimal force
transmission by exure. Advances Appl. Mech. 16, 183308

Zyczkowski, M. 1981: Combined loadings in the theory of plas-

ticity. Warszawa-Alphen aan den Rijn: PWNNijho

Zyczkowski, M. 1997: Optimal plastic design for combined

torsion with bending. In: Gutkowski, W.; Mr oz, Z. (eds.):
Proc. 2nd World Congr. on Structural and Multidisciplinary
Optimization (held in Zakopane), pp. 809814, Warsaw: Pol-
ish Academy of Sciences

Zyczkowski, M. 1998: Optimal plastic design of statically in-

determinate beams under bending with torsion. In: Ober-
weis, P. (ed.): Mechanik, Festschrift Peter Gummert, pp. 251
262. Berlin: Institut f ur Mechanik, TU Berlin

Zyczkowski, M.; Egner, W. 1995: Boundary perturbation

method applied to optimal plastic shape design of rotationally
symmetric elements. In: Olho, N.; Rozvany, G.I.N. (eds.):
Proc. 1st World Congr. on Structural and Multidsicplinary
Optimization WCSMO-1 (held in Goslar), pp. 511516. Ox-
ford: Pergamon Press