Article
Article
Applied Mechanics Reviews Copyright © 2008 by ASME NOVEMBER 2008, Vol. 61 / 060801-1
a moment and also to both bending and shear in the case of clamped
R 1共 兲 = , edge. The local nature of the resulting perturbation was pointed
共1 + sin2 兲3/2
out by Reissner. Similar conclusion was reached by Timoshenko
and Woinowsky-Krieger 关39兴 for the case of an internally pressur-
a ized cylindrical shell with clamped ends. The general mathemati-
R 2共 兲 = for 0 ⱕ ⱕ /2 共2兲
共1 + sin2 兲1/2 cal formulation of the theory of boundary disturbance was given
by Chernyh 关40兴. Adaptation of Chernyh’s theory to internally
where  = b / a is a relative depth of the head and = −2 pressurised thin-walled vessels was discussed by Magnucki 关7兴.
− 1 is a dimensionless parameter. Details about the analytical and numerical investigations of stress
concentrations in cylindrical vessels with ellipsoidal heads can be
The principal radii of curvatures of the middle surface of the
found in Refs. 关41–43兴.
circular cylindrical shell are given as follows: 1 / R1 = 0 共R1 → ⬁兲
It follows from Chernyh 关40兴 that the radial displacement w共1兲 0
and R2 = a, where a is the radius of the cylinder. At the junction
between the cylindrical shell and the torispherical or ellipsoidal and the rotation angle 共1兲
0 for the edge 共 = / 2兲 of the ellipsoidal
head there is a discontinuity of curvatures. In addition, for the head, and similarly for the edge 共z = 0兲 of the cylindrical shell
torispherical head, there is an internal discontinuity of curvature 共w共2兲 共2兲
0 , 0 兲, can be defined in the following form 共see also Ref.
between spherical and knuckle parts. These discontinuities of cur- 关43兴兲:
vatures disturb the membrane stress state in the pressure vessel
and they influence, for example, the strength of the torispherical w0共 j兲 = w̃0共 j兲 + ␣11
共 j兲 共 j兲
Q0 + ␣12 M0
domes. Investigations of the strength of pressure vessels with tori-
spherical heads were discussed, for example, by Galletly 关33兴, 0共 j兲 =
˜ 共 j 兲 + ␣ 共 j 兲Q + ␣ 共 j 兲 M
0 12 0 22 0 for j = 1,2 共3兲
Galletly and Błachut 关34兴, Błachut 关35,36兴, Błachut and Ram-
achandra 关37,38兴. where
Consider a circular cylindrical vessel with an ellipsoidal head
a2 p0 a2 p0
loaded by uniform internal pressure, p0. Assume that a and t2 are w̃0共1兲 = 共2 − − −2兲 , w̃0共2兲 = 共2 − 兲 , ˜ 共1兲 =
0
˜ 共2兲 = 0
0
the radius of the cylindrical shell and the thickness of its wall. Let 2Et1 2Et2
b and t1 be the depth of the ellipsoidal head and its thickness,
respectively. At the junction between the ellipsoidal and cylindri-
cal shells one has uniformly distributed longitudinal force Nz
= ap0 / 2 and two unknown forces, i.e., shear force, Q0, and bend-
共 j兲
␣11
2
= 共− 1兲 j−1 c
E
a
tj
冉冊 3/2
, 共 j兲
␣12 =−
2 2a
c
Et j t j
ing moment, M 0, as illustrated in Fig. 7.
Quantities Q0 and M 0 appear as the result of an interaction
between the connected parts of the pressure vessel. The membrane
共 j兲
␣22 = 共− 1兲 j−1
4
Et2j
c3 冉冊
a
tj
1/2
, c = 关3共1 − 2兲兴1/2
Fig. 7 Forces at the junction between ellipsoidal head and cy- i共 j兲 = ˆ i共 j兲 · p0 for j = 1, i = , and for j = 2, i = z,
lindrical shell due to internal pressure 共6兲
The equivalent dimensionless stresses 共Huber–Mises stresses兲 for
the ellipsoidal head or the cylindrical shell can be written in a
ing edges of both shells are given as follows: for deflections customary form as
w共1兲 共2兲 共1兲 共2兲
0 = w0 , and for angles of rotations 0 = 0 . The solution of ˆ 共eqj兲 = 冑关ˆ i共 j兲兴2 − ˆ i共 j兲ˆ 共 j兲 + 关ˆ 共 j兲兴2
these two algebraic equations gives intensities of the transfer
shear force Q0 and the bending moment, M 0. Next, the stress state for j = 1, i = or for j = 2, i=z 共7兲
in the pressure vessel can be determined 共see Ref. 关43兴 for further
details兲. Finally, dimensionless stresses become as follows. The stress concentration factor for the pressure vessel is defined as
follows:
共1兲
• For the ellipsoidal head 共ˆ —meridional and
ˆ eq,max
ˆ 共1兲
—circumferential stresses兲, ␣SC = 共8兲
ˆ 共eq,0
2兲
x 1x 2 1
ˆ 共1兲 = 共1兲
⫾ fm exp共− 1兲 where ˆ eq,max = max兵ˆ 共1兲 ˆ 共2兲
eq,max , eq,max其 is the maximum dimension-
2 冑1 + sin2
less equivalent stress in the vessel and ˆ 共j兲 eq,max is the maximum
dimensionless equivalent stress in the ellipsoidal head for j = 1 or
x1x2 1 − sin2 冑3
ˆ 共1兲 = − f 共c1兲 exp共− 1兲 共4兲 in the circular cylindrical shell for j = 2. ˆ 共2兲
eq,0 = 2 x2 is the equiva-
2 冑1 + sin2 lent Huber–Mises dimensionless stress of the circular cylindrical
共2兲 shell in the membrane state.
• For the cylindrical shell 共ˆ z —longitudinal and The value of the stress concentration factor can be adjusted/
ˆ 共2兲
—circumferential stresses兲, influenced by means of dimensionless parameters x1, x2, and .
The pressure vessel is efficiently designed when the maximum
x2 equivalent stresses in the ellipsoidal head and in the cylindrical
ˆ z共2兲 = 共2兲
⫾ fm exp共− 2兲, ˆ 共2兲 = x2 − f 共c2兲 exp共− 2兲
2 shell are equal, i.e.,
共5兲 共1兲
ˆ eq,max 共2兲
= ˆ eq,max 共9兲
This condition forms the basis for determining the efficient values
where of the dimensionless structural parameters, x1ef and ef, of pres-
冉 冊 冉 冊
ties that can be adopted in order to reduce this disparity. For
c2 c2 example, one can have an ellipsoidal head of greater thickness t1
h共Mj兲 = ⫿ cos j − ⫾ sin j
3 3 共x1 ⬍ 1兲 or bigger relative depth  共0.5⬍ 兲. As an example, let us
keep depth of the head the same, i.e.,  = 0.5, and equate peak
3 stress in the head with the peak stress in the cylindrical shell. The
共 j兲
hQ = cos j ⫿ sin j for j = 1,2 equalization of maximal values of the equivalent stresses in both
c2 parts of the vessel occurs for effective thickness ratio x1,ef
= 0.645. The result of the corresponding numerical investigation is domes having  ⬎ 0.86 the cylinder itself becomes very stiff by
shown in Fig. 9. The values of the maximal dimensionless equiva- comparison with the dome. Hence this can be loosely interpreted
lent stresses for this case are ˆ 共1兲 ˆ 共2兲
eq,max = eq,max = 96.25. This magni-
as the dome being clamped by the overstiffened cylinder.
tude is much smaller than previously seen in the ellipsoidal head. Assuming that the wall thicknesses of the connected shells are
The stress concentration factor given by Eq. 共8兲 for this design of equal, one could easily find the best relative depth of head. It is
pressure vessel becomes ␣SC = 1.112. seen from Fig. 10 that this corresponds to  = 0.6. Should thick-
Appropriate numerical procedures were developed in order to nesses t1 and t2 of both shells differ considerably from each other,
search for cases that exactly satisfy the condition given by Eq. 共9兲, their connection might give rise to technological difficulties and
for x2 within the range 75艋 x2 艋 200. The results of these calcu- therefore such designs would not be suitable. Generally, changes
lations are plotted in Fig. 10. It is seen here that for  smaller than in the value of the stress concentration factor should be analyzed
0.5 values of x1,ef are sensitive to magnitudes of x2. The lower when selecting effective values of  and x1 parameters. As the
corner corresponds to the peak stress being located at the cylinder- effective ratio of wall thicknesses x1,ef is closely related to , it is
dome junction. For all shallower domes, i.e., with  ⬍ 0.5, the enough to interrelate the values of ␣SC to one of these parameters.
cylinder becomes stiffer and stiffer as  becomes smaller and The results of such an analysis for x2 苸 具75, 200典 are shown in Fig.
smaller. It can be said here that horizontal cylinder behaves as if it 11. It is seen here that the values of stress concentration factor are
was clamped at its end by an “overstiffened” shallow ellipsoidal undesirably high for smaller relative depth of ellipsoidal head, i.e.,
dome. Similar situation exists at the upper corner in Fig. 10. For for  ⬍ 0.5. For  苸 共0.5, 0.65兲 the values of ␣SC remain almost
Fig. 9 The effect of equalization of maximal Huber–Mises equivalent stresses in „a… ellipsoid and in „b… cylinder
for the cases of x1,ef = 0.645, x2 = 100, and  = 0.5
The structure is described by ten parameters: vessel radius a, sketched in Fig. 14. The support is made of steel plate, with its
length L, spacing between supports s, distance between vessel axis hoop angle, s = 120 deg 共Fig. 14共b兲兲. Additional steel lining is
and support base H, characteristic dimensions of the support b0, c introduced between the support and the cylinder. This reinforcing
and e, thicknesses t1 of the head, t2 of the cylindrical shell, and t3 patch extends by c = 140 deg in the hoop direction 共see Fig.
of supports. The shape of the saddle support described above is 13共a兲兲. The width of the lining plate is cn. The magnitude of the
width depends on the vessel’s length. In this case it was assumed the stress level in regions D 共the head兲 and C1 共half-length of the
that cn = L / 45. Numerical calculations were carried out for hori- vessel兲. Relatively small influence is in regions A1 and B 共shell-
zontal pressure vessels with volumes of V0 = 100 m3, 200 m3, and support area兲. However, in region E 共support area兲, a distinct
300 m3. Results for each volume have been produced in the form minimum in the stress magnitude is noticed in the range 0.5
of contour maps of equivalent Huber–Mises stresses. Their values ⬍ b0 / a ⬍ 0.6. The height of the end of the saddle support arm, e,
have been monitored in eight characteristic regions of horizontal also influences the stiffness of the support. Changes in the level of
cylindrical vessels of given capacity, and these are shown in Fig. equivalent stress in various parts of the structure are shown in Fig.
13共b兲. Regions A1, A2, and B surround the saddle support of the 16 for the range 0.094艋 e / a 艋 0.28. The most significant effects
vessel including edges of the support’s cover plate. These are the were noticed in regions A1 and E. Near the support, A1, stresses
places where local stress concentration occurs, as a result of the will increase with the increase of e. This suggests that an elastic
support and vessel’s skin interaction. The maximum equivalent support of relatively low stiffness is the most advantageous solu-
stress values in these regions are selected. Regions C1, C2, and C3 tion for this vessel as it gives low stresses in this region of the
have also been chosen for monitoring the stress levels. They are support. At the same time, stress level in the support, E, will
located at half-length cross section of the vessel at the upper/lower increase if the value of e is too small 共note that for e / a ⬇ 0.11
surfaces, and at the half-height of the vessel. Stress values calcu- there is a local minimum of the equivalent stress兲. Hence, a rea-
lated in these points are also taken into account. Region D is sonable recommendation in this case would be to use supports
located in the ellipsoidal head near the joint with the cylindrical with 0.10艋 e / a 艋 0.l2. The equivalent stress values will also
shell, i.e., the place where large equivalent stresses occur. Region change when the 共s / L兲-ratio changes, i.e., when the saddle sup-
E includes the whole saddle support and again the maximum ports moved along the length of the vessel—see Fig. 17. The
stresses are selected from this area. Plots of maximum values of curves plotted in Fig. 17 suggest that the most favorable location
the equivalent stresses in each region are depicted in Figs. 15–17. of supports would be near the ends of the vessel. For increasing
The influence of the width of support bed, b0, on the stress state of s / L the stress level decreases in almost all regions, including the
the vessel is shown in Fig. 15. The length, b0, does not influence middle part of the vessel, C1. In region C3 maximal equivalent
Fig. 15 Influence of the width of the saddle support bed, b0, on Fig. 16 Influence of the height of the saddle support bed, e, on
the stress level „V0 = 300 m3… the stress level „V0 = 300 m3…
Let membrane stresses , z, and z be solutions of the above
equations. The maximum of Huber–Mises equivalent stress,
eq,max, occurs in the lower part of the middle cross section of the
shell, i.e., at = 共region C3, Figs. 13共b兲 and 17兲. The exact value
of eq,max is given by the following expression 共see Ref. 关7兴兲:
冑冉 冊 冉 冊
loaded beam being supported at two points as a substitute for a
cylindrical vessel being supported at two points is, as sometimes ␥m1a2 ␣SC p01 2
1 2 2
eq,max = 48 1 + + Cc −1 共13兲
found in literature, entirely inappropriate. Variation of stresses for 4t2 2 ␥m1a 2
vessels with different slenderness 共L / a兲-ratios is shown in Fig. 17
共for the case of V0 = 300 m3兲. It is seen here that with the increas-
where ␣SC is the stress concentration factor for ellipsoidal head–
ing magnitude of the 共L / a兲-ratio values of equivalent stresses in
cylindrical shell junction, Cc = 6 / 共1 + 2p兲 is a correction coefficient
regions B, C3, and E also increase. Similar calculations were also
taking into account the effect of saddle support on stress state, and
carried out for horizontal vessels with capacities as small as V0
p = p01 / 100␥m1a is the dimensionless parameter associated with
= 15 m3. The main outcome of these calculations is that the main
uniform internal pressure.
proportions of dimensions should be maintained 共even for capaci-
The coefficient ␣SC takes into account the stress concentration
ties down to V0 = 15 m3兲. However, other rules for dimensioning
caused by the attachment of ellipsoidal heads. The coefficient Cc,
vessels and their supports may be required for the case when the
on the other hand, takes into account all regions where the stress
contribution of hydrostatic pressure is more important relative to
is raised by flexible supports, i.e., area A , B , . . . , E, as depicted in
internal pressure.
Fig. 13共b兲.
2.3 Minimal Thickness of Horizontal Cylindrical Shell. Equation 共13兲 can be used for the evaluation of minimum wall
The following monographs discuss the strength of horizontal cy- thickness of cylindrical shell. If the magnitude of allowable
lindrical tanks: Refs. 关2,4,7兴. Discussion of the strength of thick- uniaxial stress is assumed to be all, then the strength condition
walled high pressure vessels can be found in Ref. 关6兴. eq,max 艋 all leads to the following expression for the minimum
Consider a horizontal cylindrical shell simply supported at both wall thickness of a thin-walled cylindrical vessel 关55兴:
Dⵜ8w +
Et2 4w
a z
2
4 0 w
4 + ⵜ Nzz
2
z
冉 0 w
2 + 2Nz
2
az
0 w
+ N
2
2
=0 冊
共17兲
z2
冉 冊册
共0兲 p01a p01Dins
t2,min = ␣SC = 共15兲 2w 2w w
2 all 4 + 2Nz0 0
+ N + =0 共19兲
冑3␣SC all − p01 az a22 a2
The curvatures, in this case, have the form
where Dins = 2a − t2 denotes inside diameter of the shell.
It is worth noting here that the correction coefficient Cc be- 2w 2w 2w w
comes inactive in Eq. 共15兲, as there are no supports for this case. zz = − , z = − , = − 2 + 2 共20兲
z2 az a
2
a
The minimum value of the stress concentration factor for a
classical ellipsoidal head 共 = 0.5兲 is in the range ␣SC Zukowski 关89兴 and Magnucki 关7兴 found magnitudes of critical
苸 具1.110, 1.115典, Fig. 11. As a result, the minimum thickness of a external pressure for cylindrical vessels using Eqs. 共17兲 and 共19兲.
thin-walled cylindrical vessel with ellipsoidal heads takes the fol- The aim here was to establish differences between predictions of
lowing approximate form: buckling pressures given by both approaches. The intensities of
prebuckling membrane forces for a circular cylindrical shell sub-
共0兲 p01Dins jected to uniformly distributed external pressure p02 are as
t2,min = 共16兲
2all − p01 follows:
1
The above formula is recommended by PD 5500:2003 as the
0
Nzz = 2 ap02, Nz0 = 0, 0
N = ap02 共21兲
minimum thickness of a pressure loaded cylindrical vessel.
Graphical comparison of thicknesses given by Eqs. 共14兲 and 共16兲, The equations of stability 共17兲 and 共19兲 can only be solved ap-
for a horizontal cylindrical vessel with ellipsoidal heads, is shown proximately. The deflection function can be assumed in the form
in Fig. 19. mz
w共z, 兲 = wa sin cos共n兲 共22兲
L
3 Stability of Horizontal Cylindrical Vessels
3.1 Critical Pressure of Circular Cylindrical Shell. The where wa is the amplitude of deflection, and m and n are natural
theoretical basis of thin-walled shells is inherently related to numbers.
theory of surfaces. The fundamental contributions to the develop- Substituting intensities given by Eq. 共21兲 and function 共22兲 into
ment of mathematical theory of surfaces are due to Gauss 共1777– Eq. 共17兲 or Eq. 共19兲, one obtains the following critical external
1855兲, Lamé 共1795–1870兲, Codazzi 共1824–1873兲, and Weingarten pressures.
共1836–1910兲. The general bending theory of shells was formu- From Eq. 共17兲,
再 冎冉冊
•
lated by Aron in 1874 and by Love in 1888. Important mono-
graphs dealing with the linear and nonlinear theory of shells and 共1兲 f a共1兲 t2 3
pcr = min E 共23兲
stability of thin-walled structures are due to Volmir 关56兴, Koiter m,n f b共1兲 a
关57兴, Flugge 关58兴, Baker et al. 关59兴, Brush and Almroth 关60兴,
Pielech and Suchorolski 关61兴, Yamaki 关62兴, Bushnell 关63兴, Murray where
冉 冊
ing a complete review of shell buckling analysis methods by
1 2 m L Bushnell 关96,63兴 and the up-to-date and comprehensive mono-
f b共1兲 = ␣ + n 2 共 ␣ 2 + n 2兲 2, ␣= , =
2 a graph on experimental approach to buckling of shells by Singer
et al. 关97兴.
From Eq. 共19兲,
再 冎冉冊
•
Let us assume that an end closure is subjected to incremental
共2兲 f a共2兲 t2 3 external pressure loading. Then the possible critical pressures
pcr = min E 共24兲 would be associated with the following elastic or inelastic modes:
m,n f b共2兲 a
共i兲 asymmetric bifurcation, 共ii兲, axisymmetric collapse/snap-
where through, or 共iii兲 asymmetric collapse. The shell is unable to sup-
冉冊
port any further load beyond pressures corresponding to 共i兲, 共ii兲, or
共 ␣ 2 + n 2兲 2 4 a
2
共iii兲. Hence these modes are frequently termed as failure modes.
f a共1兲 = 2 关共␣ + n 兲 − 共2n − 1 + 2␣ 兲兴 + ␣
2 2 2 2 2
12共1 − 兲 t2 The above failure modes are illustrated in Fig. 20 for the case of
metallic torispherical and hemispherical domes clamped at the
f b共1兲 = 冉 1 2
2
冊
␣ + n 2 − 1 共 ␣ 2 + n 2兲 2
edge. Figure 20共a兲 depicts the deformed shape of a geometrically
perfect torispherical shell just prior to bifurcation buckling, i.e., at
the pressure magnitude corresponding to the eigenvalue. The shell
deforms in an axisymmetric fashion with a visible bending, at the
Differences between values of critical pressures given by Eqs. spherical cap–knuckle junction in addition to the area around the
共23兲 and 共24兲 amount to 5% for practically useful vessels. How- clamped edge. The corresponding eigenshape is depicted in Fig.
ever, for long cylindrical shells 共 → ⬁, ␣ → 0兲, critical external 20共b兲. It is seen here that at the bifurcation buckling there are 17
pressures from expressions 共23兲 and 共24兲 become circumferential waves. In practice, this failure mode is rarely seen
共1兲
lim 关pcr
␣→0
兴 = min
n
再 n2
12共1 − 2兲
E
t2
a
冎冉冊 3 due to imperfections in geometry and/or in wall thickness, but see
later for a specific case where it was possible to capture it experi-
mentally 共Fig. 24共b兲兲. Figure 20共c兲 depicts a collapsed steel tori-
=
1
E
3共1 − 2兲 a
t2
冉冊 3
for n = 2 共25兲
spherical end closure, which had no cylindrical flange and it was
clamped at the edge. The shell deformed axisymmetrically and the
failure was due to excessive plastic straining. Finally, Fig. 20共d兲
再 冎冉冊
shows a hemispherical closure, which failed by developing a
共2兲 n2 − 1 t2 3
single inward dimple. Large plastic straining developed at the col-
lim 关pcr 兴 = min E lapse pressure and they were caused by local thinning of the wall
␣→0 n 12共1 − 2兲 a
冉冊
and by local deviations from perfect shape. Both imperfections
3
1 t2 occurred during manufacturing and they were not intentional.
= E for n = 2 共26兲
4共1 − 2兲 a Internally pressurized end closures on cylindrical vessels, on
the other hand, can 共i兲 buckle by bifurcation, 共ii兲 develop large
The difference between these values of critical pressures is con- plastic deformations, or 共iii兲 fail through burst. A lot of research
siderable, i.e., approximately 33%. It is worth to note that the efforts have gone into all three forms of failure. The existing
critical pressure for an infinitely long circular cylindrical shell, controversy is associated with “large plastic deformations” mode.
given by Eq. 共26兲, was derived by Bryan in 1888. More recently this has been related to plastic loads—see, for ex-
Barrelled shells can be seen as generalizations of circular cy- ample, Ref. 关98兴 and discussion in Sec. 3.2.2. View of a torispheri-
lindrical shells. Recent studies into buckling of barrelled shells cal head at the bifurcation pressure is depicted in Fig. 21共a兲. A
subjected to external pressure can be found in Refs. 关90–95兴. picture of a steel head, which has not buckled but suffered exces-
sive plastic deformations, is showed in Fig. 21共b兲. None of the
3.2 Critical Pressure for Domed End Closures. It has al- previous two modes, 共i兲 or 共ii兲, resulted in the loss of load carrying
ready been mentioned that torispherical or ellipsoidal heads are capacity, i.e., loss of pressure tightness. However, Fig. 21共c兲
frequently used as end closures on cylindrical pressure vessels. shows a close view of aluminum head, which bursts at pressure
Torispherical heads are usually adopted for smaller pressures, that was about 100% above the corresponding plastic load.
while ellipsoidal closures are used for larger pressures. Uniformly Detailed discussion of externally and internally pressurised clo-
distributed pressure causes, first of all, a membrane stress state in sures is provided next.
the vessel. This is true for the region away from the head–
cylindrical shell joint. It is well known that the stress state is 3.2.1 Torispherical and Hemispherical Heads Loaded by Ex-
disturbed by transverse shearing force and bending moment in the ternal Pressure. Early studies into the stability of externally pres-
vicinity of the junction. This effect considerably influences the surised metallic domed ends have revealed the lack of agreement
critical state for heads. Additionally, in torispherical ends, there is between theoretical predictions of failure pressure and values ob-
a disturbance of membrane state away from the torisphere- tained in experiments. This in turn generated a large volume of
cylinder junction. Analytical treatment of the stability problems theoretical and experimental studies. The literature review of early
for multishell structures, e.g., for cylindrical vessel with ellipsoi- work can be found in, for example, Refs. 关63,33,99–109,4兴. Also,
dal or torispherical end closures, is difficult. In such cases it is review papers by Błachut 关74兴, Teng 关110兴, and Błachut 关17兴 re-
nearly imperative to apply numerical methods—mainly the finite port progress on externally pressurized domed end closures.
element methods. Despite a lot of progress in reducing the disparity between the-
The stress distribution and static stability of pressure loaded oretical predictions and experimental test data, the design of ex-
end closures have long history and they have been investigated for ternally pressurized end closures still relies on an empirical ap-
many decades. The following two sections provide the back- proach 共see Ref. 关103兴兲. It has been known for a long time that an
ground information to some of the early studies into stress and accurate prediction of the bifurcation buckling, snap-through, or
static stability analyses of externally or internally pressurized collapse pressure for domed ends is elusive despite considerable
heads. This is followed by the recent developments in the strength effort over many years. The existing discrepancy between com-
and static stability of pressure loaded end closures. Most of these puted buckling loads and observed experimentally buckling loads
references are related to metallic domes. However, references to led to the establishment of an experimental design code based on
for large industrially manufactured heads. cal curve illustrating the sensitivity of the buckling strength of a
One of the concepts for modeling initial geometric imperfec- torispherical shell to the magnitude of local flattening 共␦0兲 is
tions evolved around a local flattening—see Refs. 关102,107兴. It shown in Fig. 24共a兲. A number of perfect and imperfect domes
has emerged from parametric studies carried out that introduction with local flat patches of increased radius were injection molded
of any flattening into hemispheres leads to a reduction of the
buckling load, while in torispheres a certain threshold magnitude
of flattening can be tolerated without any reduction of the buck-
ling load. Once the threshold imperfection is exceeded the buck-
ling load becomes sensitive to the imperfection magnitude. A typi-
while the flow theory model sometimes failed to predict the oc- common approach to evaluate a plastic load, using a graphical
currence of bifurcation buckling. This study has confirmed the evaluation of the relationship between internal pressure and apex
existing paradox between J2 flow and deformation theories of deflection. Gerdeen 关98兴 pointed out that the evaluation of the
plasticity. While the latter remains physically unrealistic theory, it plastic load should employ a more objective and justified physics
does provide useful results in the plastic buckling analysis of in- criterion, i.e., for a pressure test the relevant quantities are pres-
ternally pressurized domes. It is therefore worth reinforcing the sure and change of internal volume. This approach was adopted
suggestion made by Bushnell 关96兴 that both approaches should be by Błachut et al. 关155兴 to obtain plastic loads for torispherical and
used in order to establish the sensitivity of the predictions to the ellipsoidal mild steel domes. Reasonable comparison of the ex-
two models. The above paradox belongs to a wider class of prob- perimental and numerical results was obtained. However, if these
lems where the discrepancy between the two theories of plasticity domes were to operate in the postyield range where cyclic load
in bifurcation calculations has been studied by Giezen et al. 关144兴, could be present, the role of the plastic load is less clear than the
Tugcu 关145兴, Błachut et al. 关83兴, and Durban and Zuckerman shakedown load. In the above study, the concept of a first-cycle
关146兴. shakedown pressure was adopted to compare the relative values of
The finite element methodology aimed at geometrically linear
plastic and shakedown loads. Plastic and shakedown loads for
and nonlinear elastic-plastic stability analyses was given by Soric
domes made from a strain-hardening material were studied by
关147–149兴 and Soric and Zahlten 关150兴. The effect of eigenshape
Błachut et al. 关156兴. While the role of plastic loads for internally
imperfections on the magnitude of buckling loads was also inves-
pressurized heads is still being studied, it is burst pressure that is
tigated in one of these papers.
The results of 190 experimental tests on internally pressurized of greater value from a practical point of view as it gives an
torispherical heads were collated and analyzed by Miller indication of the margin of safety. This is an important quantity,
关151,152兴. Most analyzed torispheres failed by asymmetric, elas- especially, at a design stage or at an emergency situation. A recent
tic, or inelastic bifurcation buckling. In this paper attention was study into this problem is by Błachut and Vu 关157兴. It was postu-
also paid to the plastic collapse and burst pressures. There is a lated to use the true plastic strain, up, corresponding to the ulti-
continuing discussion to identify the best failure criterion for tori- mate tensile strength 共UTS兲 for computing the magnitude of burst
spherical heads, which do not buckle. The use of plastic collapse pressure. This is illustrated in Fig. 26共a兲 for the case of aluminum
loads, as recommended by Gerdeen 关98兴 is a frequently adopted alloy AA6061-T6, where it is seen that up = 0.0815. One needs not
approach—see also Refs. 关35,153,154兴. The above papers adopt a only the magnitude of plastic strain but also a place where this
冉 冊 冉冊 冉 冊冉冊
Tovstik, is
16E  2
1 − 4 2 pcr,Gal = 285 1 − 125 yp 共30兲
共inl兲 E D Rs D
pct,Tov = 1 + Ct 共29兲
冑3共1 − 兲 ␣e
2 ␣ e
where yp is the yield stress, E is Young’s modulus, rt is the radius
冑
where Ct = 193/ 32冑3共1 − 2兲, e.g., for = 0.3 the value of this of the torus, Rs is the radius of the sphere, D = 2a is the diameter
coefficient is Ct = 1.9106. of the cylinder, and t1 is the thickness of the head.
Magnucki and Szyc 关175,176兴 have also investigated the buck- Magnucki and Szyc 关175兴, inspired by the approach used for
ling behavior of internally pressurized ellipsoidal heads in cylin- torispherical domes and given by Eq. 共30兲, have derived the fol-
drical vessels using the FE. Domes with small depths, i.e., 0.15 lowing expression for critical pressure of internally pressurized
艋  艋 0.30, were considered. The geometry of these shells was the ellipsoidal head with small depths 共 艋 0.3兲:
冉 冊冉 冊
same as in the previous section. Critical internal pressure for the
共inl-lin兲 共inl兲 yp  0.84
1
head with relative depth  = 0.25 was found to be pcr,FEM pcr,red = 26.2yp 1 − 125 共31兲
= 5.54 MPa. The corresponding buckling form is shown in Fig. E s − 1 s0.26␣1.53
e
that the initial portions of the load-apex deflection plots were lin-
where s = 1.2共1 +  兲 / 2 is dimensionless parameter.
2 ear. This, as mentioned earlier, was the basis for the derivation of
plastic loads and subsequent derivation of Eq. 共32兲.
Next, they carried out a series of FEM calculations in order to
The values of critical pressure for four descriptions 共two linear
extract plastic loads of internally pressurized ellipsoidal domed
and two nonlinear兲 are compared in Fig. 31. One immediate ob-
ends. Plastic loads, in their studies, corresponded to the end of
servation is that linear elastic solutions are significantly different
linearity on the internal pressure versus apex deflection plots. Nu-
from nonlinear cases. The results based on nonlinear FEM calcu-
merical results were then approximated by a second order poly-
lations and those obtained from formula 共31兲 are very similar in
nomial, with the use of the least square method. The resulting
the range of 0.1艋  艋 0.3.
formula for critical pressure of ellipsoidal heads has the following
form: 3.4 Critical Sizes of Aboveground Based and Under-
共inl兲
pcr,FEM = 1604yp 冉 1 − 125
yp
E
冊冋 共1 − 0.9兲

␣e
册 2
共32兲
ground Horizontal Cylindrical Vessels. Aboveground and un-
derground horizontal cylindrical vessels, Fig. 33, are usually thin-
walled and they are susceptible to loss of stability. Ziółko 关1兴
A typical view of a deformed elliptical head subjected to inter- provided detailed principles for design of such tanks of various
nal pressure is depicted in Fig. 32. In the FE studies it was found geometries, considering their real manufacturing conditions. De-
sign rules for such vessels can be found in European Standards
EN 286-1 关14兴, EN 12285-1 关179兴, EN 12285-2 关180兴, and PD
5500 关15兴. Despite existence of these rules there are still a number
of unanswered issues about the stability of circular cylindrical
storage vessels.
3.4.1 Aboveground Horizontal Cylindrical Vessels. A horizon-
tal cylindrical vessel filled with liquid and supported at both ends
is potentially exposed to destruction by buckling due to a variety
of different loading scenarios. Huang and Redekop 关181兴 have
determined critical load for a circular cylindrical shell subjected to
bending due to three forces distributed on any surface area. This is
both analytical and numerical studies. The theoretical solution for
the elastic buckling is based on Donnell’s stability equation. Guar-
racino and Fraldi 关182兴 analytically determined the deformation of
circular cross section of a pipe subjected to pure bending. Chan
et al. 关183,184兴 and Tooth et al. 关185兴 presented results of a re-
„inl…
Fig. 31 Critical pressure, pcr , versus relative depth, , of in- search related to the stability of horizontal shells pivoted at both
ternally pressurized ellipsoidal heads ends and filled with fluid. They proposed expressions for admis-
sible stresses using known analytical solutions for axial compres- using Galerkin’s method.
sion or for pure bending of cylindrical shells. Experimental results The resulting four homogeneous linear algebraic equations
were compared with the appropriate values obtained from British need to satisfy the following condition for the existence of the
and European Standards. Kacperski 关186兴 carried out an experi- solution:
mental study into the behavior of horizontal cylindrical vessels
detaij = 0 共36兲
filled with water and pivoted at both ends. Buckling tests were
carried out on cylinders of variable length, constant diameter, and where coefficients aij include material constants E, , specific
constant wall thickness. Saal 关187兴 carried out experimental inves- weight of the liquid ␥m1, and dimensions of cylindrical shell.
tigations into buckling process of laboratory scale, horizontal thin- The total capacity of the horizontal vessel is a sum of capacities
walled cylindrical shells filled with water. A proportion between of the cylindrical shell and two ellipsoidal heads, i.e.,
dimensions of cylindrical shell for various levels of water, for
which the structure lost stability, was determined. Local buckling V0 = Vc + 2Vh 共37兲
of a cylindrical shell subjected to bending was investigated by
where Vc = a is the capacity of the cylindrical shell, Vh
3
Axelrad 关188兴 and Mikkelsen and Tvergaard 关189兴. The stability
= 3 a3 is the capacity of the ellipsoidal head, = L / a is the
2
of cylindrical tanks due to bending and filled with liquid was
studied by Danis and Kacperski 关190兴. Huang et al. 关181兴 pro- relative length of the cylindrical shell, and  = b / a is the relative
vided details about the instability of a cylindrical shell subjected depth of the head. From which, the radius of the vessel can be
to bending. Experimental tests were carried out on commercial written as
steel tanks. The mechanisms of plastic collapse of bent tubes were
analyzed by Elchalakani et al. 关191兴.
Stasiewicz 关192兴 and Magnucki and Stasiewicz 关193,194兴 have
a= 冑
3
3V0
共3 + 4兲
共38兲
determined the critical state of a horizontal cylindrical shell filled Substituting this expression into the coefficients aij of the alge-
with liquid and supported at both ends 共Fig. 18兲. braic equation 共36兲, one can calculate the critical values of relative
Consider a ground-based vessel filled with liquid. This gives wall thickness 共t2 / a兲cr as a function of relative length = L / a for
rise to loading by a positive internal hydrostatic pressure. Assume
constant capacity 共V0 = const兲 of the vessel. Numerical calcula-
that there may be a suction action on the cylinder resulting in a
small external pressure, p02. The total pressure will therefore be tions were carried out for a family of circular cylindrical vessels
共see Fig. 18兲 with capacities V0 = 100, 200, 300 m3. Typical material constants
for steel were adopted, i.e., E = 210.0 GPa and = 0.3. Vessels
p共兲 = ␥m1a共1 − cos 兲 − p02 共33兲 were filled with water of specific weight ␥m1 = 9.81 kN/ m3 and
they were additionally loaded by external pressure,p02
Prebuckling forces can be approximately determined from lin-
= 0.01 MPa or p02 = 0.1 MPa. The results of the analysis are
ear equations of equilibrium of cylindrical shell given by Eq. 共11兲,
shown in Figs. 34 and 35.
and they are
Bottom curves are related to the capacity V0 = 100 m3 and the
4 z 1
0
Nzz = t2zz
0
=− ␥m1L2 cos cos − ap02
3 L 2
4 z
Nz0 = t2z0 = − ␥m1aL sin sin
2 L
4 z
0
N = t2
0
= ␥m1a2共1 − cos 兲cos − ap02 共34兲
L
Assume deflection of the shell in the following form:
冋
w共z, 兲 = t2 w2 + 2
5
兺 w cos共i − 2兲
i=3
i 册 cos共n兲cos
z
L
共35兲
冉冊t2
a cr
= k 1 k2 共39兲
as known.
The stability condition of horizontal cylindrical vessels for the
first load case has the following form 共see Refs. 关54,195兴兲:
where k1 and k2 are parameters. 共 j兲
p0,cr
The values of these parameters for aboveground vessels sub- p0 ⱕ 共42兲
jected to additional negative pressure p02 = 0.01 MPa are k1 css
= 0.001142 and k2 = 0.364 共Fig. 34兲. For the case of vessels sub- 共j兲
where p0,cr is the critical external pressure given by Eq. 共23兲 for
jected to larger additional negative pressure, i.e., p02 = 0.1 MPa,
j = 1 or by Eq. 共24兲 for j = 2, and css is the safety coefficient.
the values of parameters k1 and k2 are k1 = 0.002930 and k2
One can choose the wall thickness of cylindrical vessel, t2, or
= 0.401 共Fig. 35兲.
its length, L, or radius, a, in order to increase the magnitude of
3.4.2 Underground Horizontal Cylindrical Vessels. Under- critical pressure, p0 共as it is clear from Eqs. 共23兲 and 共24兲兲. It is
ground horizontal cylindrical vessels are always loaded by exter- also possible to use internal circular rings to support the cylindri-
nal pressure. The external effect of earth on the relatively thin cal shell in order to further increase the magnitude of buckling
cylindrical shells can lead to buckling phenomena. Two load cases pressure, and this would be in addition to shell’s parameters, i.e.,
for these vessels are distinguished: 共1兲 total earth pressure and t2, L, and a. In this case the total length, L, is now divided into a
vehicle weight pressure 共typical geological condition兲, and 共2兲 hy- number of segments of length, say, L0. Each segment of length L0
drostatic lift pressure of earth with water pressure 共extreme geo- is supported by two rings at both ends—as sketched in Fig. 37.
where Jz is the moment of inertia of the area 共second moment兲 of case, i.e., for the case of hydrostatic lift pressure of earth with
the effective cross section of the ring and R1 is the radius of the water pressure, is similar to buckling of aboveground horizontal
effective ring. cylindrical vessel. External hydrostatic pressure in this case is
It is worth noting here that when evaluating Jz for the T-ring,
some neighboring stiffness of cylinder’s wall needs to be in- p共兲 = − ␥m2关a共1 − cos 兲 + h0兴 共45兲
cluded, making T-ring an I-ring. The second flange in the I-ring is
taken as cylinder’s wall of length 30⫻ t2—as illustrated in Fig. 37 where ␥m2 is the specific weight of water.
关109,179兴. Next, the stability condition for the circular ring can Results for empty steel underground circular cylindrical vessels
finally be written as of capacities V0 = 100, 200, 300 m3 placed in soil containing water
共ring兲 and at the depth h0 = 1 m are shown in Fig. 39 关194兴.
p0,cr
p0 ⱕ 共44兲 The values of parameters k1 and k2 used in Eq. 共39兲 for under-
csr ground horizontal vessels located at the depth h0 = 1 m are k1
where csr = 2css is the safety coefficient for the ring. = 0.002613 and k2 = 0.2775 共for V0 = 100 m3兲. Also, k1 = 0.002821
Buckling of a cylindrical shell segment is not equivalent to the and k2 = 0.2715 共for V0 = 200 m3兲, and k1 = 0.003080 and k2
collapse of the entire vessel, if rings remain stable. Therefore, the = 0.2539 共for V0 = 300 m3兲.
safety coefficient for rings is adopted to be twice as large as the Interesting stability problem of an underground vessel was
safety coefficient for a cylindrical shell. With stability conditions studied by Konjuchov et al. 关196兴. It is postulated that buckles are
given by Eq. 共42兲 for cylinder and by Eq. 共44兲 for supporting not allowed to develop in the outer direction but only in the inner
rings, one can choose an appropriate set of parameters in order to direction. It is shown that buckling loads could be increased by
prevent buckling of the vessel. The ideal situation would be when more than 30%. In practice this could be achieved by appropriate
buckling pressure for cylinder has the same magnitude as buck- support media of such vessels.
ling pressure for the rings. The stability behavior of ring-reinforced cylindrical vessels for
This is now to be illustrated by an example of underground underground applications was studied by Walker and McCall
ring-stiffened steel cylindrical vessel. Consider an empty vessel 关197兴 and Tian et al. 关198兴.
with capacity V0 = 100 m3 being loaded by the mean effective ex-
ternal pressure p0, Eq. 共40兲. Assume that the underground steel
vessel is described by the following parameters: E = 2.05
⫻ 105 MPa, = 0.3, h0 = 1 m, b1 = 0.5 m, b2 = 0.76 m, nw = 6, Q
= 50 kN, and gr = 1800 kg/ m3. Let us also assume that the under-
ground vessel is stiffened by circular rings with standard beam,
T80. Based on conditions 共42兲 and 共44兲 the values of critical wall
thickness, t2,cr, of vessels were calculated and they are shown in
Fig. 38. The domain above the curves in Fig. 38, for fixed number
of rings, n, corresponds to the stable state of the shell and rings.
This represents space of feasible solutions. It is seen from Fig. 38
that only wall thickness was used when buckling conditions given
by Eqs. 共42兲 and 共44兲 were evaluated for a given number of rings,
nr, in the vessel. For example, for nr = 5, there are two curves. The
first is obtained from the condition given by Eq. 共42兲 for the shell.
The second is obtained from the condition given by Eq. 共44兲 for
the ring. It is clear that there is a unique 共 ⬅ L / a兲-ratio where
both buckling pressures are the same. This corresponds to
= 5.74 共giving, t2,cr = 5.2 mm兲. Fig. 39 Relative critical thickness of the wall of underground
Buckling problem of underground vessels for the second load cylindrical tank „h0 = 1 m…
tank are sketched in Fig. 41. The joint for the most frequent 2:1
ellipsoidal head is shown in Fig. 41共a兲. The case of the same wall
Fig. 40 Plot of dimensionless mass of ellipsoidal head versus
thickness is depicted in Fig. 41共b兲. It is seen here that the depth of
its dimensionless depth, 
the dome increases to  = 0.6. Figure 41共c兲 depicts the solution of
Eq. 共53兲 while Fig. 41共d兲 refers to the minimum mass configura-
tion of the dome.
␣SC 共0兲 冑3 p0
t1 = t1,ef = t , t2共0兲 = a 4.2 Strength Shaping of a Dished Head. Classical tori-
x1,ef 2 2 all
spherical and ellipsoidal heads disturb the membrane state of
共thickness of the cylindrical shell in membrane state兲 stresses and cause stress concentration at a junction with cylindri-
共51兲 cal part of a pressure vessel. Minimization of the stress concen-
tration was the subject of a number of past studies. Elin and Hari-
and 1 is mass density of the domed end’s material. tonov 关258兴 used the calculus of variations in order to determine
Hence, dimensionless mass of the ellipsoidal head is deter- the optimal variable thickness of domed head. Middleton 关259兴
mined by the following coefficient: calculated the optimal shape of internally pressurized torispherical
head with two and three radii. Mansfield 关260兴 presented the prob-
␣SC lem of optimal shaping of pressure vessel heads in the membrane
m̂1共兲 = cm 共52兲
x1,ef stress state. The range of minimal relative depth was found, i.e.,
0.599艋 h,min 艋 0.620. Wilczyński 关261,262兴 searched for meridi-
Taking into account functions x1,ef共兲, plotted in Fig. 10; onal shape in order to minimize stress concentrations in thin-
␣SC,min共兲, plotted in Fig. 11; and Eq. 共50兲, one can obtain dimen- walled or thick-walled vessels. His approach was based on the
sionless mass of ellipsoidal domed end-closure as a function of its FEM and the use of Bezier functions. Magnucki and Lewiński
depth, . This relationship is plotted in Fig. 40 共see Ref. 关257兴 for 关263,264兴 optimized the shape of dished head on the cylindrical
further details兲. pressure vessel in the membrane stress state under the condition of
Let us now consider the following two optimization criteria: 共i兲 constant wall thickness.
minimum mass of the ellipsoidal head and 共ii兲 minimum mass Consider general-ellipsoidal head of a pressure vessel, the me-
combined with minimum depth, , of the head. The first criterion ridian of which is in the form of a generalized ellipse—Fig. 42.
leads to the minimum mass of the ellipsoidal head at 共1兲 opt = 0.85. The equation of the ellipse is assumed in the following form:
The wall thickness of the head at the optimum is t1,ef = 0.679t共0兲 2 共1 − 兲k + 2 = 1 共54兲
共see Fig. 40兲. This solution was obtained for parameter x2
苸 具75, 200典.
where = z / b and = r / a are dimensionless coordinates, and k is a
Manufacturing costs of an ellipsoidal head increase with the parameter 共2 ⬍ k兲.
increasing depth. Hence the second criterion, which takes into
The geometric relations for surface of revolution are
consideration not only the mass of the head but also its depth,
appears to be more relevant from a practical point of view. One
possible form the takes into account both parameters, i.e., mass
and height/depth of an ellipsoidal closure, can take the following
form:
min兵m̂1, 其 = min冑m̂21 + 2 共53兲
 
This second criterion leads to the optimum at 共1兲opt = 0.65. The wall
thickness of the head at the optimum is t1,ef = 0.845t共0兲
2 . Again, this
solution was obtained for parameter x2 苸 具75, 200典. Both of the
above solutions have been obtained under the assumption of the
same magnitude of peak stresses in domed end and in cylindrical
shells 共as discussed in Sec. 2, Eq. 共9兲兲. Various schemes for at-
tachment of elliptical dome to the cylindrical part of a storage Fig. 42 Meridional shape of a dished dome
dr
d
= R1 cos ,
dz
d
= R1 sin , r = R2 sin 共55兲 or R̂2 冑1−
R̂2
R̂1
+ 冉冊
1 R̂2
3 R̂
1
2
ⱕ1 共59兲
R̂1 =
R1
=
k 关f 1共兲兴3/2
, R̂2 =
R2
=
k
关f 1共兲兴1/2 共56兲 R̂1共0兲 = R̂2共0兲 = R̂0 = 冑3 and k = 2冑3 共60兲
a 2 f 2共兲共1 − 兲共k−2兲 a 2
The point 共b , a兲 共or 共 , 兲 ⬅ 共1 , 1兲兲 in Fig. 42 belongs to the gen-
where eralized ellipse described by Eq. 共54兲 and, at the same time, to the
冉 冊
f 1共兲 = 共1 − 兲2共k−1兲 + 2
 2
k
关1 − 共1 − 兲k兴
straight portion of the cylindrical shell. It is obvious that there is a
continuity of principal curvatures at the junction between the gen-
eralized ellipsoid and the circular cylindrical shell. At the point
共1,1兲 dimensionless principal radii, as given by Eq. 共56兲, are
冉 冊冉 冊
f 2共 兲 = 2 1 −
1
k
2
− 1 − 共1 − 兲k
k
R̂1共1兲 → ⬁ and R̂2共1兲 = 1—see Fig. 43.
The minimization of the relative depth  = b / a of the dished
head under strength constraint given by Eq. 共59兲 involves the
and  = b / a is a relative depth of the head.
criterion of the following form:
Assume that the equivalent stresses 共Huber–Mises stresses兲 in
the head do not exceed the equivalent stress in the cylindrical min关共k兲兴 共61兲
k
shell when both shells have the same wall thickness, i.e., t = t1
= t2. The resulting strength condition becomes
The above problem needs to be solved numerically. Calculations
冑3 a lead to the optimal solution for the closure’s depth: min = 0.71.
共eqhead兲 ⱕ 共eqcyl兲 = p0 共57a兲 The variation of dimensionless equivalent stress in the optimal
2 t
ellipsoidal head is shown in Fig. 44 共note that it is plotted along
or for the case of dimensionless stresses the axial axis, , and not along the closure’s arc length兲. The
optimal shape can found by the substitution of min = 0.71 into the
冑3 a parameter k in Eq. 共60兲, and the obtained value for it into the
ˆ 共eqhead兲 ⱕ ˆ 共eqcyl兲 = 共57b兲 equation defining the ellipse’s shape, i.e., Eq. 共54兲.
2 t
4.3 Optimal Domes Subject to Buckling. Once confidence
where t = t1 = t2, the thickness of the cylindrical shell, is the same has been established in numerical predictions of the buckling
as the thickness of the dished head, p0 is uniformly distributed strength, then there is considerable scope for structural optimiza-
internal pressure, and a is the radius of the cylindrical vessel. tion of buckling prone domed end closures—as pointed out in
The dimensionless, equivalent stress in the dished head is papers by Błachut 关265,266兴 and Błachut et al. 关105,125兴.
ˆ =
1 R2
2 t
, 冉
ˆ = 1 −
1 R2 R2
2 R1 t
冊 carrying capacity was maximized, i.e., popt = max共min共pi兲兲, where
i = 1 corresponds to bifurcation buckling and i = 2 refers to axisym-
metric snap-through. The use of just few segments led to 15–50%
are the usual longitudinal and circumferential stresses. increase in buckling pressure above buckling pressure of equiva-
The strength condition given by Eq. 共57b兲, after substituting Eq. lent torispheres of constant wall thickness.
共58兲, has the following algebraic inequality: Even within a constant wall thickness it is not instantly clear
冑冉 冊 冉 冊
which torispherical geometry would be the best within a given set
R̂2 R̂2 R̂2
2
冑3 of constraints and range of available geometries. The following
3 1− + ⱕ case illustrates this dilemma. Let us seek solution to the following
2 R̂1 R̂1 2
two problems:
popt = max共min共pi/pH兲兲 and Vopt = max共VT/VH兲, i = 1,2 4.3.2 Optimal Domes Under Internal Pressure. References to
early work into structural optimization of internally pressurized
共62a兲
dome closures on cylindrical shells can be found in Refs.
and 关203,273兴. These two complementary reviews of papers devoted
to optimal design of shells quote about 1000 papers that have
popt = min共min共pi/pH兲兲 and Vopt = max共VT/VH兲 i = 1,2 appeared since the beginning of the last century. Structural opti-
共62b兲 mization of internally pressurized dome ends has played a much
larger role in many areas of pressure vessel technology than opti-
where the design vector contains the following geometrical pa- mization of externally pressurized heads. Structural optimization
rameters that define a torispherical end: 兵Rs / D , r / D , D / t其; VH and with respect to bifurcation buckling has very rarely been at-
VT are the volumes of a hemispherical and torispherical heads, tempted. Instead, optimization where buckling was to be avoided
respectively. A given amount of material WH can be utilized in was carried out. Eschenauer 关274兴 reported shape optimization of
many different ways to form an infinite number of torispherical very light satellite tanks where constraints were added for “no
end closures, all having the same weight and diameter but differ- buckling.” More frequently the problem was posed as optimiza-
ent shape and wall thickness. In turn both the buckling strength tion with stresses in mind, e.g., minimization of the maximum
stress in torisperical head—see papers, for example, by Younsh-
and internal volume VT of torispheres will vary. The above prob-
eng 关275兴, Younsheng and Ji 关276兴, and Zhu and Boyle 关277兴.
lem has been solved by Błachut 关267兴 for the following bounds on
Improvements in computer technology that have occurred recently
design variables:
have made the nonlinear analysis more affordable. This in turn
1.25 ⱕ Rs/D ⱕ 0.55 makes it feasible to carry out structural optimization of vessel
heads well into the inelastic region. The paper by Błachut 关278兴
0.06 ⱕ r/D ⱕ 0.5 addressed the problem of finding the minimum mass of domed
closure on internally pressurized cylindrical vessels. Heads were
D/t ⬎ 0 共63a兲 to be made from mild steel and mode of failure was taken to be
plastic load. Subsequently a series of eight mild steel domes was
The side constraints were assumed in the following form 共WH, CNC-machined and used to benchmark optimal solutions. All tori-
WT ⬅ weights of hemisphere and torisphere, respectively兲: spheres were finally burst and there was a substantial strength
reserve beyond the plastic load in all of them. Optimization of
0.5 ⱕ WT/WH ⱕ 1.0 共63b兲
torispherical heads subject to shakedown was reported by Błachut
The above multicriteria problem, given by Eqs. 共62a兲 and 共62b兲
subject to Eqs. 共63a兲 and 共63b兲, was converted into a scalar form
through the bounded objective function, with the buckling load
being the most important component of the objective function. As
a result, the optimum and the pesimum configurations were found
for elastic and elastic-plastic behaviors of heads. Figure 45 depicts
a sample of three contours of optimal solutions—for three initial
weights associated with hemispheres: DH / t = 100, 500, 1000. Sub-
sequently three mild steel domes were machined using a computer
numeric controlled 共CNC兲, lathe in order to verify the optimum
and the pesimum configurations 共with a hemispherical profile be-
ing the reference shape兲. The best torisphere was five times stron-
ger than the weakest one 共under the assumption that both heads
had the same diameter and the same mass兲. Figure 46共a兲 depicts
the stability contour with test points and view of the best 共opti-
mum兲 and worst 共pesimum兲 torispheres together with the refer-
ence hemisphere after the collapse—Fig. 46共b兲. Further details
can be found in Ref. 关105兴.
Other cases of optimally shaped end closures include CFRP
heads in which meridional profile was approximated by parabolic
segments, several circular segments, and generalized ellipse—see
Ref. 关268兴. Maximization of buckling strength of filament wound
dome closures was investigated by Błachut 关269兴 through optimal
shaping of meridian’s profile and optimal distribution of the wall Fig. 45 Contours of optimal solution for elastic analyses.
thickness. Other papers related to optimal shaping of dome clo- Each contour represents torispherical heads of the same
sures include Refs. 关8,74,107,170,270–272兴. weight.
Fig. 46 „a… Contours of optimal elastic-plastic solutions for The capacity of the tank is
D / t = 300. Test points 6 „reference hemisphere…, 7 „optimum…,
and 8 „pesimum… are shown. „b… Photograph of collapsed V0 = 2V1 + V2 共64兲
heads in „a….
where V1 = 3 a3 = 3 a3 is the capacity of the 2:1 ellipsoidal
2 1
4.4 Optimal Sizes of Aboveground Horizontal Cylindrical The mass of the tank is
Vessel. Aboveground horizontal cylindrical tanks resting on two ms = 2m1 + m2 共66兲
supports are primarily used for storage of liquids. Usually the
capacity of such tanks considerably exceeds the capacity of hori- where m1 = sa2t1 · cm is the mass of the ellipsoidal head, cm = 1
+ 共2 / 冑1 − 2兲ln 共共1 + 冑1 − 2兲 / 兲 for a typical ellipsoidal head
zontal tanks designed for land transportation. Every vessel can be
manufactured as a slender or as a compact thickset, for its fixed
capacity—as sketched in Fig. 47. Under the conditions of normal 共 = 0.5兲 and cm ⬵ 1.38017. m2 = 2sa2t2 is the mass of the cy-
lindrical shell, s is the mass density of the vessel’s material, and
t1 , t2 are the thicknesses of the ellipsoidal and cylindrical shells,
respectively.
The effective thickness of the 2:1 ellipsoidal head with respect
to minimum stress concentration in the vessel 共Sec. 2.1兲 can be
written in the following, approximate form:
t1ef = ct · t2 共67兲
where the mass is given by Eq. 共66兲. After substitutions the ob-
jective function, ms, becomes
ms = 2a2t2共cm · ct + 兲 共69兲
where t共1兲 冑 共 2 兲
2min = ␥m1a / 4all 48共1 + 共␣SC / 2兲共p0 / ␥ma兲兲 + Cc 2 − 1
2 2 1 2
Jan Błachut has received his MSc in Physics (1971, Jagiellonian University, Kraków, Poland), Ph.D. in
1980, and DSc (Habilitation) in 1996. The latter two were in Mechanics and both from Kraków University
of Technology. He has recently been awarded the title of Full Professor. For nearly a decade he worked at
the Institute of Physics, Kraków University of Technology. Then he has moved to Department of Mechanical
Engineering, The University of Liverpool, UK. He has published more than 120 papers in reviewed scien-
tific journals and proceedings. Recently he was cochairmen of Euromech Colloquium in Liverpool and
CISM course in Udine—both devoted to various aspects of structural optimization. He is a full member of
the ECCS Technical Working Group 8.4 on Buckling of Shells and a member of the Sub-Committee on
Design Methods PVE/5—British Standards Institution. His research interests include buckling of shells,
integrity of pressure vessels, and structural optimization of components aimed at on land, in the sea, and in
the air applications.