Академический Документы
Профессиональный Документы
Культура Документы
buckling,
thermal
loading,
shells,
453
does not occur. The analysis usually also provides, as byproducts, a state of residual postshakedown stresses and/
or a set of kinematic quantities associated to an R or AP
mechanism.
Strictly understood, SD analysis avoids time-stepping
evolutive solutions and represents, conceptually and
computationally, an important generalization of classical
(rigid-plastic) limit analysis.
Traditional SD theory preserves two basic assumptions of limit analysis: firstly, perfectly plastic (no
hardening) constitutive models; and secondly, 'small
displacements', i.e. equilibrium equations referred to the
original, undeformed geometry. In a number of engineering situations, these assumptions turn out to be unrealistic and should be relaxed.
Hardening behaviour generally has a 'stabilizing effect', in the sense that the safety factor s with respect to
lack of SD (i.e. either AP or R) is generally increased over
the one attainable with perfectly-plastic behaviour at
equal original yield surfaces. For instance, linear kinematic and isotropic hardening models rule out the possibility of R. Sometimes materials (e.g. concrete) and
structural components (e.g. compressed bars, reinforced
concrete beams in bending) exhibit negative hardening,
i.e. 'softening', unstable behaviour. Such local instability
generally has an unstabilizing effect on R thresholds.
Stabilizing or unstabilizing effects in the above sense may
be caused by the non-negligible influence of the deformations on the equilibrium relations.
An interplay may arise between constitutive post-yield
behaviour (hardening or softening of materials or structural components) and the effects of configuration
changes on equilibrium, when both these mechanical
features are not negligible.
Many research contributions have been published
with the purpose of extending the classical SD theory
(centred on the Bleich-Melan theorem and on the NealSymonds Koiter theorem) and the relevant computational methods, in order to allow for geometric effects,
alone or in combination with hardening/softening.
Only some of these extensions are dealt with in available surveys of the SD area, such as those by K6nig 1, and
K6nig and Maier 2 or in books on plasticity, such as
those by Cohn et al 3, Zyczkowski4 and Kaliszky 5, where
a broad conspectus of classical SD theory can be found.
Therefore the SD literature explicitly intended to deal
with geometric effects is concisely surveyed below.
To the writers' knowledge, so far the following general
approaches have been considerably developed and have
been applied to at least academic examples.
(1) A 'second-order' theory proposed by Maier and
coworkers 6 9, generalizes most results of the classical SD
theory (static and kinematic theorems, bounds on postshakedown quantities) to geometric and hardening effects, and preserves its computational simplicity by recourse to linearization of the geometric terms in the
equilibrium equations and to piecewise-linearization of
plastic constitutive laws.
(2) An 'inadaptation' methodology, due to K6nig and
SiemaszkolO 13, is centred on the kinematic SD theorem
which is applied iteratively and, hence, may become quite
laborious. It permits a great generality as for the assumed
constitutive laws, which are accounted for between steps
consisting of conventional (small deformations) SD
analysis. In the limit analysis context an interesting
454
Gp
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2R
(~p
(~p
0.
//
outer surface
inner surface
Z=128.3 mm
Figure 1 (a) Geometry of cylindrical shell and 'primary' axial load; (b) temperature distribution on inner and outer surface of shell
455
Finite
3.0
2.5
R*
I
z~~/~
2.0
\\\
1.5
P.S.D.'',,
-
-k--.
',
\B
\
,
0.5
E.S.D.
',, \ \
\
o.o
. . . .
0.0
0.5
. . . .
1
1.0
~
1.5
Figure 2
0,,
456
element
model
The tube has been discretized in space by using twodimensional, axisymmetric, solid finite elements (not
shells). The mesh adopted is shown in Figure 3. The axial
dimension has been subdivided into six fields of length
42.77 mm each. These fields have been discretized using
an increasing number of elements as, starting from the
left, the region of the temperature discontinuity is approached. Hence, the first field (A in Figure 3) has been
discretized by 8 elements while the third field (D in
Figure 3), which is close to the temperature discontinuity,
has been discretized by 50 elements. All elements employed were either triangular or quadrilateral secondorder elements (6 nodes, 3 Gauss points for triangles and
8 nodes, 9 Gauss points for quadrilateral elements).
Overall, the mesh was composed of 589 nodes and 140
elements. Full integration was adopted over all elements,
in order to have Gauss points as close as possible to the
inner and outer surfaces of the tube. In fact, owing to the
temperature discontinuity along the length and to the
presence of the temperature gradient across the thickness, the bending moment in the cylindrical wall plays an
important role in the nonadaptation mechanism. The
flexural behaviour is characterized by maximum (in
modulus) stresses at the extreme fibres of the thickness.
As usual in finite elements, the elastic-plastic constitutive
law is enforced at Gauss points only. Trial tests with
reduced integration have shown a small fictitious increase of strength due to the fact that the onset of
plasticity at the edges of the wall thickness was missed by
the Gauss points.
In order to investigate the sensitivity of the model to
mesh refinement, a few analyses were also carried out
using a finer mesh with 1561 nodes and 436 elements.
B C
A
42.77
()(I)
oufer
D;
42.77~
42.77
surface
I
inner
surface
>
tt
C
4.277
4.277
1.7ll
Figure 3
1.0
I
.........................
'..........................
0.8
--q ..........................
i
cycles
.
E 0.4
O-t = 1
affer 40 cycles
:6"" 0.2
Wi I / - - - c o a r s e
mesh
E i l/ -fine mesh
o.o
........................
J---
.....................
o
n,"
i
mO.2
50
100
fl
150
200
Axial disfonce Z ( m m
Figure 4
meshes
-6
r,,
R3
250
300
et al.
G e o m e t r i c effects due to t e n s i l e a x i a l s t r e s s e s
200 -
O'p = 0.85
180160-
TZ 140
,,J120
/J
/.-
o 100
-i-
o
EL 8 0 -
"~
J
~ "
60-
~"
small displ.
la rge_~displ.
1.2
E
E
40-
1.0
,.g
"~ 0.8
20-
--
--
--
~
0~
'
l t0
' 5'0
' 8'0
Number of cycles
0.6
"'vYYV
,orge d,sp,
~1. o.4
m
0.2
-6
"'VVI ~ ~.:o.8
"U
17.5
0.0
o,i
-0.2 0
E
Z,~ 1 5 . 0
~_ 0.55
E
(.1
5.0
~p=0.80
0.75
~,~
0.0
.
1'0
.
.
.
210
310
410
.
510
6 JO
710
-o 0.40
N u m b e r of cycles
(~p = 0 . 2
-6
~0.35
810
Figure 5
458
E.S.D.
~=1
_0 0.45
~6 2.5 ~
--
large displ.
E 0.50
7.5
0
CL
~
"(2
~=1
~=0.8
small displ.
large displ.
4~5
R .
small displ.
E o.6o
~10.0
,_
0.65
~12.5
E
O
110
0.30 ~
0
Figure 6
10
15
20
25
30
35
Number of cycles
40
45
et al.
800 -
700-
small d
600E
Z500 -
large displ.
o400E]
)..
~
300-
/ /
~P-"'S"
- D_"
2. 5
3 2 0 0 O'p = 0 . 2
100i
1'0
1's
20'2's'3'o
Number of
35
4'O
4's
cycles
E
P.S.D.
~-t= 2.5
E1.2
=o.2
~-1.0
~5
-8
~3
o.a
E]
r~
.....
0.6
5 ' 1=0
small displ.
large displ.
Number of cycles
4~5
1_,, 0"46
0.44
~ 0.43
__.~0.42
llIllll
-8 0.40
1'o
2'0
3b 3's
Number of cycles
4'O
4J5
Figure 8 Tensile axial load. Deformation histories: (a) radial displacement at Z = 150.5 mm; (b) axial displacement at right end of
tube from small and large displacement analyses. Alternating plasticity mechanism
1.219
17"5 q i
& = 1. 1 2 5
1.200
15.0
,~,12.5
"~10.o
__
1.1 O0
7.5
1.050
1.000
"~
5.0
2.5 ~
0.0
Number
10
12
14
of cycles
Figure 9
460
Empirical formulae
The buckling load of a shell structure, such as the
cylinder considered here, is heavily affected by geometric
imperfections 33'36'37. The theoretical buckling load obtained by the so-called 'linear theory' (which only accounts for second-order effects) is substantially reduced
by the presence of imperfections whose effects are accounted for, in a 'nonlinear theory' by higher-order
terms.
The following buckling loads may be useful as a
reference. Classical (Euler's) elastic stability theory applied to cylindrical shells without imperfections provides
the popular formula oct = (3(1 - v 2 ) ) - WZEhR-1 which
for the present situation gives oct = 1.96ay (i.e.
6or = 1.96). It is worth noting that this formula is exact
only for a shell of length equal to integer multiples of the
half wavelength of the buckling mode and that the excess
of trc, over the yield stress o r points out that buckling will
occur in the plastic range at a lower load.
The effect of imperfections on elastic buckling can be
captured by empirical formulae (obtained as the lower
bound envelope of experiments) such as the one reported
in Reference 33 which makes use of a 'knockdown factor'
q~ in order to reduce the classical critical load, namely
a'~r = q5 ac, , ~b act, q5 being defined as
q5 = 1 - 0.9 [1 - exp( - O.0625~/R/h)
for 0.5 < L/R <_ 5
100 < R/h <_ 3000
The present case is included within the applicability
ranges of this formula which yields q~ = 0.358 and hence,
a'c~ = 0.701%.
The combined effects of imperfections and material
plastic behaviour can again be expressed by empirical
formulae such as those suggested by the European Convention of Constructional Steelwork (ECCS) 38, namely
(see also Galletly and Blachut 39)
6'c'r
loaded pressurized cylinder has been recently investigated by extensive numerical analysis in the elastic
range 4. The results of a similar finite element investigation for the present tube thermally and axially loaded in
the elastic-plastic range are reported in Figure 10, where
the nondimensional buckling load, denoted by t~', is
plotted against the size of the thermal imperfection. Here,
the term 'thermal imperfection' on the abscissa denotes
the maximum deflection, due to the combined action of
the axial load and of the thermal discontinuity, at a point
located at Z = 150.0 mm. The plot refers to a monotonic
(not cyclic) thermal loading (and is marked by the label
N = 0, N being the number of thermal cycles). As observed in Figure 6(a), thermal cycles produce an increasing lateral deflection which acts as an imperfection for the
evaluation of the buckling load. Diagrams of the type
shown in Fioure 10, for N > 0 would be slightly different,
although they would reproduce the same behaviour
qualitatively. From Figure 10, it turns out that the buckling load undergoes a dramatic reduction (with an almost
infinite slope) even for very small lateral deflections (or
'imperfections'). These deflections may initially increase
with thermal cycling even if the structure would finally
shakedown in a small displacement analysis. As a consequence, the primary compressive load, which was originally safe, can become critical.
The ECCS formula 38 used above is based on standard
manufacturing imperfections, therefore it is expected that
the asymptotic value of the buckling load resulting from
the present analysis and affected by thermally induced
imperfections will be lower than the ECCS value a~', =
0.479ay, as is apparent in Figure 10.
The occurrence of buckling corresponds, in Figure 9,
to an infinite slope of the dissipation curve. The higher 6~,
the sooner (i.e. the smaller the number of cycles) the
radial deflection reaches a value which makes the com-
pressive load critical. This has important design implications: in the presence of cyclic thermal loading, an
estimate of the buckling load, on the basis of the undeformed structure with fabrication imperfections only,
may lead to seriously unconservative designs.
A summary of the structural response under compressive loading is illustrated in Figure 11. Here, the loading
combinations which lead to buckling, are plotted in a
plane 6, versus #~ (solid lines). The figure also shows the
shakedown boundaries (dashed lines) already discussed
with reference to Figure 2. Each of the solid curves
corresponds to a different number of cycles. More specifically, the rightmost curve of this type corresponds to a
monotonic increase of the amplitude of the temperature
discontinuity (noncyclic thermal load). We recall that the
loading path is such that the primary load is applied first,
together with the uniform thermal expansion. Then the
temperature discontinuity is applied, its amplitude being
characterized by t~,. Points belonging to this rightmost
curve represent loading combinations such that the
structure buckles before even one cycle can be performed.
Likewise, the second curve from the right is the locus of
loading combinations such that the structure buckles
after 20 cycles, and so on. These curves tend to become
closer and closer as the critical number of cycles increases
so that the last curve, corresponding to 70 cycles, can be
considered as a reasonable reference for design purposes.
3.0
/ N
2.5
70
=
40
20
~.N
1.2
2.0
.
1.0
ETe//~y J
Z/V~
0.8
1.5
2/,/~
P.S.D.
&_
0.6
=
1.0
\
\
0.4
E.S.D.
\
\
0.5
0.2
\
\
\
0.0
0.0
'
1.
I0
'
Thermal
2.0
'
3.0
imperfection
'
4.0
'
5.0
(ram)
Figure 10 Buckling toad versus thermal imperfection for monotonically increasing temperature discontinuity (N = 0)
0.0
'
0.0
0.5
1.0
~
1.5
et al.
1.6
buckling
= 1.125
#p= 0.6
E
E
1.4
1.2
~3
~
~J
z E.S.D.
13
0.8
1'0
Number of cycles
1.8
(~t = 1.150
~p=0.6 D
F~
1'2
buckling
-E
B
E 1.4
G)
L)
0
~Q-1.2
"o
~
~
~lorge
,
El
displ.
E.S.D.
disp,.
5-
"lJ 1 . 0
EI
rw
0.8
'
Number of cycles
-pCr
0.9
80m
0.8
Gt =
0.5
1.0
z 70-~ 60-
0.7
0
G)
EL
0.5
1.5
2.0
~-:~21
o
10
20
30
40
50
2.5
60
ffl
0.3
0.2
2.5
O-p= 0.2
40-
~EL30
o0
0.4
~'~ P.S.D.
70
80
Number of c y c l e s
Figure 12 Buckling load versus number of cycles for varying
thermalload
462
L 50-
0.6
0.1
buckling
small displ.
large displ.
~,
&
10
1'2
14
16
1'8
2'o
Number of cycles
Increment of dissipation per cycle versus number of
cycles for compressive axial load 6 . = 0.2 and for thermal load
5 t = 2.5: small and large displacement analyses
Figure 14
4.5
4.0
T. buckling
"-~5.5
,+..
O'p = 0.2
3.0
~lerge
displ.
.)
~ 2.5
o
~_2.o
-~ 1.5
~_~ 1.0
~J
r~ 0.5
0.0
1'0
1'2
l l4
1'6
118 r-~O
Number of cycles
0.4
0.3
P.S.D.
displ.
.,.. 0.2
E
~ = 2.5
E 0.1
O-p = 0.2
G)
0
~D
-0.0
13_
~large displ.
.......................................................
in
,~
-o -o.1
"buckling
-6
"~
< -0.2
-0.3
~,
10
12
14
Number of cycles
16
18
r ~
20
et al.
Acknowledgments
This paper is dedicated to the memory of Professor J. A.
Krnig.
The results presented in this paper have been obtained
in a CEC-sponsored research project conducted jointly
by the Engineering Research Centre of GEC, Leicester,
UK (Contractor) and the Department of Structural
Engineering, Politecnico di Milano (Subcontractor). The
writers wish to acknowledge with thanks: the fruitful and
friendly cooperation with Dr. P. White in this project;
useful information provided by Dr. K. F. Carter and
Professor A. R. S. Ponter, University of Leicester, and by
Dr. A. Siemaszko, Polish Academy of Sciences; the
support from WG 'Codes and Standards', AG2, through
subcontract of Study Contract RA1-0148-UK, in the
context of the CEC Fast Reactor Coordinating Committee; finally, the permission of the above sponsoring
agencies to publish the present paper. The second author
gratefully acknowledges a fellowship from the Italian
Ministry of Foreign Affairs as part of a joint project
between Tsinghua University, People's Republic of
China and Politecnico di Milano, Italy.
References
1 Krnig, J. A. Shakedown of elastic-plastic structures, Elsevier,
Amsterdam, 1987
2 Krnig, J. A. and Maier, G. 'Shakedown analysis of elastoplastic
structures: a review of recent developments', Nuclear Engng Design, 1981, 66, 81 95
3 Cohn, M. Z. and Maier, G. Engineering plasticity by mathematical
programming, Pergamon, New York, 1979
4 Zyczkowski, M. Combined loadings in the theoo' of plasticity,
PWN, Pol. Sci. Publ., Warsaw, 1981
5 Kaliszky, S. Plasticity: theory and engineering applications, Elsevier, Amsterdam, 1989
6 Maier, G. 'A shakedown matrix theory allowing for workhardening and second-order geometric effects', in Foundations of plasticity, Vol I (A. Sawczuk, Ed.), Noordhoff, Leyden, 1972 pp 417 433
7 Maier, G. 'Upper bounds on deformations of elastic work-hardening structures in the presence of dynamic and second order effects',
J. Struct. Mech, 1973, 2 (4), 265 280
8 Krnig, J. A. and Maier, G. 'Adaptation of rigid-work-hardening
discrete structures subjected to load and temperature cycles and
second-order geometric effects', Comput. Meth. Appl. Mech.
Engng, 1976, 8, 37 50
9 Maier, G. and Novati, G. 'A Shakedown and bounding theory
allowing for nonlinear hardening and second-order geometric
effects with reference to discrete structural models', in Inelastic
Solids and Structures, A. Sawczuk Memorial Volume, (M. Kleiber
and J. A. Krnig, Eds) Pineridge, Swansea, 1990 pp 451 471
10 Krnig, J. A. 'Stability of the incremental collapse' in Inelastic
structures under variable loads, (C. Polizzotto and A. Sawczuk
Eds), COGRAS, Palermo, 1984, pp 329 334
464