Professional Documents
Culture Documents
Introduction
There are a number of applications in which steel remains the material of
choice for cOQveyance or storage of liquids, particularly liquids under internal
pressure. Steel is often used as a liner in pressure tunnels and penstocks for
hydroelectric plants, water conveyance, and sewage conveyance. Steel liner
is also often used in pipe jacking operations beneath major facilities, and for
the conveyance of gases under pressure (e.g., natural gas).
Of the common structural materials, only steel is considered impervious
and leakproof for liquids and gases under external and internal pressure.
Reinforced concrete is not considered impervious, although it can be made
relatively impervious for low head water systems. High-pressure systems us-
ing concrete [such as Reinforced Concrete Cylinder Pipe (RCCP) and Pre-
stressed Concrete Cylinder Pipe (PCCP)] usually have a thin (1.5 mm) steel
cylinder that makes the pipe impervious. Over the last 30 years, PCCP, which
is wrapped and stressed by high-strength, but brittle, steel strands, has sup-
planted steel pressure pipe in many applications. However, due to a number
of recent failures of PCCP resulting from the stress corrosion of the strands,
many agencies are rethinking their options and considering the use of steel.
Plastic pipe shows promise in some applications, but its long-term use is
unproven by case histories, and may be limited by its relatively low modulus
of elasticity and creep problems.
Structural steel with low to medium strength (yield strength ranging from
200 to 400 MPa) has many advantages for pressure piping systems. Steel is
isotropic, has a high modulus of elasticity and strength, creeps negligibly, is
ductile, and can be made impervious by continuous welding or the use of
gasketed joints. Steel pipes are capable of resisting high internal pressures.
This is particularly true for tunnels in competent rock, where interaction (load
sharing) with the rock may be used to reduce the thickness of steel (or the
stress in the pipe). For large-diameter (greater than 5 m) pressure tunnels that
require an impervious liner, steel is almost always used due to the difficulties
in installing other types of liner, such as very thick and heavy concrete pipe
or composite systems with cast-in-place concrete. Of course, the potential for
corrosion must be investigated when steel is used.
In many cases where a steel liner is used to resist internal pressure by
membrane action, a loading condition assuming net external pressure must
also be considered. This condition occurs most often in pressure tunnels that
are quickly unwatered for inspection, for maintenance, or due to leakage
problems. Net external pressure can also exist when a low-pressure tunnel
passes through an area that has a high ground-water table. In such conditions
JOURNAL OF ENERGY ENGINEERING / DECEMBER 1998/55
geles, the writers formalized an analysis and design procedure for a 3.66-m-
diameter steel-lined tunnel acted on by varying external ground-water heads.
During this process the writers discovered problems and inconsistencies with
existing design methodologies for steel tunnel liners and errors in some of
the buckling equations presented in various sources.
This paper covers the buckling analysis of steel tunnel liners that are acted
on by external pressure and are encased by concrete backfill of sufficient
strength and continuity that the steel liner cannot buckle outward beyond the
surface of the concrete encasement. The result is a comprehensive design
procedure for both plain and stiffened steel tunnel liners subjected to external
hydrostatic pressure. In addition to discussing the concepts behind some of
the buckling equations previously derived by others, the writers will introduce
new concepts of analysis and provide revised equations applicable to both
plain and stiffened liners.
Where needed for resistance against buckling, stiffened steel liner has the
following advantages:
PERPENDICULAR
ROLLED PARALLEL ROLLED OR
ROLLED TEE CHANNEL ROLLED PLATE CUT PLATE
FIG. 1. Typical Stiffener Configurations
The tenns "service load," "working stress," and "allowable stress" are used
as defined in the Manual of Steel Construction-Allowable Stress Design
(1991) and the Boiler Code. The maximum anticipated external design pres-
sure under service conditions is Pd, and the corresponding computed design
tangential stress (also known as the working stress) is <Td' The working stress
must always be no greater than the allowable stress.
Minimum values for the factor of safety (FS) for the various types of
buckling failure, stiffener weld design, and the other potential failure modes
for a steel tunnel liner must be detennined by the designer. The factors of
safety quoted herein are typical; however, appropriate values for different
structures and failure modes will depend on many conditions, including the
accuracy with which the design loading is known, the importance of the
structure, the specified (or realistically attainable) fabrication tolerances, and
the reliability of the theory that describes the failure mode.
Buckling Theories
In the past, the following theories have been used for modeling the buck-
ling of a plain or stiffened steel liner under external pressure:
In addition, recent work by El-Sawy and Moore (1997) in the related field
of pipeline rehabilitation using liners of various types gives an indication that
numerical (finite-element) methods of buckling analysis may be functional in
the future. EI-Sawy and Moore have correlated critical buckling pressures
computed using a numerical model to a theory of restrained thin cylinder
liner buckling developed by Glock (1977). They then extended their results
to cases of liners with local imperfections, oval liners, and loose-fitting liners.
To the writers' knowledge, numerical modeling efforts have been limited to
the cases of plain cylinder liners, and they do not address buckling modes
that consider yielding of the liner material (as do the rotary symmetric and
single-lobe theories discussed in the following sections). For these reasons,
numerical methods serve a limited design purpose at this time.
In the discussion that follows, a summary of the underlying concepts be-
hind these buckling theories and the applicability of these theories to the
design of steel liners is presented.
it can no longer maintain its shape, and it will become unstable and fail by
buckling in a manner analogous to a slender column.
Initially, consider the case of a slender column of length Leo cross-sectional
area A, moment of inertia I, and modulus of elasticity E. Theory states that
a column will buckle in a shape similar to a plucked string of length Leo
forming an integer number of nodes. The number of nodes (n) formed de-
pends upon the columns restraint (i.e., if held only at the ends, n equals I;
if also restrained from translation in the middle, n equals 2; etc.).
The critical buckling load (Tcr ) which the column can resist is computed
by the Euler buckling equation
(I)
This equation, with n equal to 1, is used for the analysis and design of slender
columns in accordance with the Manual of Steel Construction (1991). [Those
interested can refer to Gaylord and Gaylord (1972) for a derivation of this
equation.] When a column is restrained only at its ends, the cases with n
equal to 2 or more are neglected. The critical buckling stress (O"cr) in a column
can be obtained from the foregoing equation br dividing (1) by A and sub-
stituting the square of the radius of gyration (r ) for IIA:
(2)
Using a similar approach, one can derive the equation for the critical ex-
ternal hydrostatic buckling pressure (Per) for an unrestrained thin cylinder
(plane stress condition) of radius R and thickness t:
P = (n
2
-
3
1)£1 = (n
2
_ 1)£ (.!...)3 (3)
cr R 12 R
In (3) the moment of inertia (I) is expressed per unit length of cylinder, and
n is an integer equal to 2 or more. When n is equal to 2 the buckling pressure
is minimized. Fig. 2 illustrates free-tube buckled shapes for different node
values.
Eq. (3) requires modification if the thin cylinder is longitudinally re-
strained. In this case, Poisson's effects resulting from the longitudinal restraint
increase the cylinder's effective tangential stiffness. The critical external
buckling pressure is computed per the following equation:
(n
2- 1)£1 1)£ (n
2- (t)3
Per = (1 - v )R = 12(1 - v 2 )
2 3 R (4)
where v = Poisson's ratio of the pipe material. The term (1 1'2) in this
equation increases the cylinder's tangential stiffness due to Poisson's effects.
JOURNAL OF ENERGY ENGINEERING / DECEMBER 1998/59
(J"
(n2- 1)£' (t)2
= -'------'-- - = (n 2
- 1)£'
(r)2
- (6)
cr 12 R R
where (J"d = service tangential stress; D = pipe diameter; k = factor (less than
1.0) that accounts for ground interaction; and FS = applied factor of safety.
Note that (7) is derived from (6) based on an n value of 2.
Cylinder Held Circular at Ends. If it is assumed that the ends of a
cylinder are held circular, buckling under external hydrostatic pressure will
occur with multiple symmetrical lobes, as indicated in Fig. 2; however, the
buckling mode will actually be three-dimensional (3D), and the shape of each
lobe will be roughly a 3D truncated sphere, as shown in Fig. 3.
Equations for the case where a cylinder of length L is restrained against
buckling at its ends have been derived by numerous authors such as von
Mises (1914), Donnell (1976), Southwell (1913), and Hiigge (1960). In the
resultant free-tube buckling equations the integer value of n that produces
the lowest buckling capacity is unknown, and must be determined by trial
and error (unlike the Euler free-tube case where the ends are not held circular
60 I JOURNAL OF ENERGY ENGINEERING I DECEMBER 1998
and the lowest buckling capacity always results when n is equal to 2). The
free-tube buckling equations developed by von Mises (1914) and Donnell
(1976) are
2n - 1 - v)
E(t/R) ) E(t/R)3
2
(2n-l+---,,---
P- ( 2 +
er - (n - 1) [(nli'rrRi + If 12(1 - v 2 ) (nli'rrRi +I
(8)
• Donnell equation
[F 2E + 6Pcr R rr (K + P E _ R rr + F24PPcrRrr =0
y - y - (11)
Fy - PcrR rr R cr rr
where Pcr = critical buckling pressure (for a factor of safety of 1.0); R =
radius to the neutral axis of the cylinder; Rrr = cylinder radius to thickness
ratio = R/t; F y = yield strength of steel; Yo = initial radial gag due to shrinkage
and temperature; K = gap ratio = yjR; and E' = E/(l - v).
Fig. 4 shows curves of R rr versus Pen computed using (11), for various
gap ratios.
The buckled shape (number of lobes) that results in the minimum critical
buckling pressure is not known a priori; however, it can be back-calculated
from the critical buckling pressure and the Euler free-tube buckling equation
62 I JOURNAL OF ENERGY ENGINEERING I DECEMBER 1998
7000 I
FJ E 261MPa
--
E = 100,000 MPa
Downloaded from ascelibrary.org by New York University on 05/10/15. Copyright ASCE. For personal use only; all rights reserved.
-\ + - v =0.19 f--
liner restrained
IoJllitudinally
5000 -
Gap Ratio (010)
'\
I,
--0-0.00
-0-0.03
f-'
--0.06
---0.10
___ 0.10
3000
1000
R .........
:-.0....
1000
- '"-"-
o !
m rn
RIt=R,.
for an infinitely long cylindrical shell [(5)]. The value of n obtained from
(11) and (5) is typically not an integer, indicating that the computed critical
buckling condition is usually not possible geometrically. For verification, one
can solve the following equation, which is a solution with two unknowns
that omits the buckling equation (5), for Per at integer values of n close to
that computed from (11) and (5):
R"
PaR"
- 1) - 6R;, (K + Pcr R ,,) - n
E' 24
2
(--.&.-
PcrR"
- 1)2 = 0 (12)
The use of (12) indicates that, excluding (5), the value of Per is usually
not sensitive to the value of n assumed [Le., (11) and (12) typically generate
similar values of Per for normal ranges of n between 3 and 15]. This factor,
plus the typical noninteger value of n for the general solution for (11) and
(5) as noted previously, suggest an apparent deficiency in the rotary sym-
metric buckling approach.
Single-Lobe Buckling
As for the rotary symmetric buckling model, in single-lobe buckling theory
it is assumed that the magnitude of the gap between the cylinder and the
JOURNAL OF ENERGY ENGINEERING / DECEMBER 1998/63
other side. When the external pressure reaches a critical magnitude, the cyl-
inder will buckle by forming a single-lobe, as shown in Fig. 5.
An example of single-lobe buckling of a steel tunnel liner is shown in Fig.
6. This photograph shows the 12-mm-thick, 6-m-diameter steel liner that was
installed in the Newhall Tunnel, which is part of the Metropolitan Water
District's Foothill Feeder Project.
Equations for modeling single-lobe buckling of a cylinder have been de-
rived by Amstutz (1970) and Jacobsen (1974). The Amstutz buckling equa-
tions have a simpler fonn because they assume constant values (within a
given range) for several variables used in developing the equations. There-
fore, if the values of these variables are out of their assumed range, the
Amstutz equations are invalidated. In particular, the Amstutz equations are
very sensitive to the variations in the variable E (E is a function of the ratio
of the thrust times the square of the radius to stiffness ratio of the liner). The
Amstutz equations are valid only if E is between 5 and 20. Since E is well
below 5 for a variety of stiffened liner geometries, the Amstutz equations are
not valid for many stiffened liners. In addition, the radius of gyration used
in the Amstutz equations for stiffened liners is computed by taking the cross
section of the stiffener plus the length of the liner, which is assumed to
contribute to the moment of inertia of the stiffened section (i.e., the effective
length; see Fig. 7). Since the determination of the effective length is some-
what arbitrary (as discussed later), the radius of gyration used in the Amstutz
equations may be too large (or too small). Using a larger radius of gyration
than is warranted will produce an erroneously high liner stiffness and there-
fore a buckling capacity that is too large.
For the aforementioned reasons, and because the Amstutz equations yield
less conservative (larger) buckling capacities than the Jacobsen equations, the
writers recommend that the Jacobsen equations be used to analyze single-
lobe buckling of plain and stiffened steel liners under hydrostatic external
pressure.
The following three equations are those derived by Jacobsen for single-
lobe buckling of steel liners. Although these three equations are more com-
plicated than the single rotary symmetric buckling equation, they were de-
rived based on similar principles. Given the liner geometry, steel properties,
and gap size, the following three equations can be solved simultaneously (via
JOURNAL OF ENERGY ENGINEERING / DECEMBER 1998/65
Z-O.78}Rt"
We • EFFECTIVE LENGTH
iteration) to compute the critical buckling pressure (Pcr) and the other two
unknowns (ex and 13):
[(9,r/4J32) - 1][11" - ex + 13 (sin cxlsin J3f]
12(sin cxlsin J3)3(ex - 11"K - J3(sin a1sin (3)[1 + (tan2(ex - (3)/4)]}
R
- =0
\fI2J/F (13a)
3
[(9,r/4J32) - 1]1 Pcr4/R
(sin cxlsin (3)3 - E' =0 (13b)
- :;] = 0 (13c)
where R = radius to the neutral axis of the liner (or portion of liner) under
consideration; Lc/ = length of liner under consideration (1 for a plain liner,
center-to-center spacing of stiffeners for a stiffened liner); 1 = effective mo-
ment of inertia of the liner; F = total area of the liner of length Lc/; h =
distance from the neutral axis of liner to the extreme fiber of the liner; F y =
yield strength of steel; E' = E or E/(l - v 2 ), depending upon assumed lon-
gitudinal restraint; K = gap ratio = (radial gap size)/R; P cr = critical buckling
pressure for a factor of safety of 1.0; ex = one-half of the angle subtended to
the center of the cylindrical shell by the buckled lobe; and 13 = one-half of
the angle subtended by the new mean radius through the half waves of the
buckled lobe.
These equations are most easily solved using computer software capable
of solving the three equations simultaneously. For example, a commercially
available software package like MATHCAD (MathSoft Inc.) can be used for
this purpose.
Global Buckling
In a stiffened liner, the steel cylindrical shell typically is of inadequate
thickness to resist the external design pressure by itself (without the stiffen-
ers). If the stiffened portion of the liner were to buckle due to inadequate
66/ JOURNAL OF ENERGY ENGINEERING I DECEMBER 1998
7000 --
F,-161MPa
E = 100,000 MPa
v=O.29
liDerratrained
5000 10
Gap Ratio (%)
-0-0.00
-0-0.03
__ 0.06
\ l\ , __ 0.10
...... 0.10
3000 -0- Euler Free-Tube
1000
1000 1\ "
i'o--. '-n.
x )l....; "-.
o
30 40 50 60 70 80 90 100 no 110
RIt=R,.,
!
70lt0
l'y.262MPa
, E • 200,000 MPa
v=lt.29
Downloaded from ascelibrary.org by New York University on 05/10/15. Copyright ASCE. For personal use only; all rights reserved.
lbaerrestnbaed
-
1\ loa
sapll aOO (%) - BaddiDc Mod
1\
"
__ It.lt0 - Siagle Lobe
- - 0.20 - Siagle Lobe
3000
-
0 r--..
1\ ""
2000 i 1\
o
'"
I
"--.-
"-
ill
--s:::I
m
Rlt=R.<
and gap ratio. Both of these buckling theories take the restraint provided by
concrete backfill into account; however, rotary buckling theory produces less
conservative results.
Based on the foregoing discussion, the theoretical problems with the rotary
symmetric equation, and that experience has shown that global buckling of
steel tunnel liners usually occurs in a single-lobe, the single-lobe buckling
equations are considered appropriate for the analysis of global buckling of
steel tunnel liners (plain or stiffened). Furthermore, as discussed previously,
the writers recommend that the Jacobsen single-lobe equations be used be-
cause they are more general and conservative than the Amstutz single-lobe
equations.
Local Buckling
If the stiffened portion of the liner has a very large stiffness (thereby pre-
cluding global buckling), but the thickness of the steel cylindrical shell is
inadequate, then the liner could become unstable between the stiffeners. The
buckling of portions of the steel cylinder between the stiffeners (which them-
selves remain circular), is referred to as "local" buckling. Given this context,
local buckling is meaningless for a plain steel liner.
68/ JOURNAL OF ENERGY ENGINEERING / DECEMBER 1998
glects the confining effect of the concrete backfill. Additionally, the free-tube
buckling equations quoted earlier were derived under the assumption of sim-
ple support at the ends of the analyzed cylinder length (i.e., the ends of the
cylinder are free to rotate and to approach each other). Continuity of the
cylinder across the stiffeners and torsional rigidity provided by the stiffeners
increase the local buckling strength to some extent; however, the writers
know of no way to estimate the effects of the backfill concrete and partial
fixity at the supports on the buckling strength for this local buckling mode.
Of course, both effects provide added safety.
(14)
- R" = 0 (l5a)
- 1)
(15b)
(sin alsin
Downloaded from ascelibrary.org by New York University on 05/10/15. Copyright ASCE. For personal use only; all rights reserved.
- = 0 (15c)
In order for the critical buckling pressure computed by these equations to
be valid, the tangential stress level in the liner must not exceed 80% of the
yield strength of the steel being used. When the stress level exceeds this
value, the stress-strain behavior of steel tends to become nonlinear. Hence,
if the tangential stress resulting from the computed critical buckling pressure
(computed as PerRit, where R o is the outside radius of the cylinder) exceeds
80% of the yield strength, the modulus of elasticity used in the foregoing
equations should be reduced to an appropriate tangent modulus of elasticity
(Et ) in order for the computed critical buckling pressure to be valid. For most
plain steel liners, the tangential stress resulting from the critical buckling
pressure will rarely approach 80% of the steel yield strength. In the event
that the tangential stress exceeds 80% of the steel yield strength, the following
equation can be used to compute the tangent modulus for a value of tangential
stress (0) and steel yield strength (Fy ):
The use of this reduction is somewhat complex because the critical buckling
pressure is a function of the tangent modulus, and the tangent modulus is
function of the tangential stress level in the liner at the critical buckling
pressure. Since using this reduction can be tedious, and variations in E sig-
nificantly affect the buckling capacity of the liner, the writers recommend
«T
that liners be designed such that the tangential stress d) resulting from the
service external design pressure (Pd) multiplied by the minimum required
factor of safety against single lobe buckling (FS.,.min) remains below 80% of
the yield strength:
PdRoFSslmln
(J"dFSst,mln = . 0.80Fy (17)
t
The use of (16) for reducing the elastic modulus, and the computation of
the resulting factor of safety against single-lobe buckling, are demonstrated
in design example 2.
The allowable tangential stress «(Tall) per the Boiler Code is computed as
(Tan = min GFy> F.) = min(147 MPa, 133 MPa) = 133 MPa (19.3 ksi)
Based on the assumed thickness of 25.4 mm, the outside radius (Ro ) is
1,537 mm (60.5 in.). The minimum thickness required is computed from (14)
as follows:
PdRo (1,240 kPa) (1,537 mm)
t 2:: - -
(Tan
= (133,000 kPa)
= 14.3 mm (0.564 in.)
Therefore, the assumed thickness of 25.4 mm (1.0 in.) is acceptable. Based
on this thickness, the radius to the centerline (R) is 1,524 mm (60 in.) and
the radius-to-thickness ratio (R n ) is 60.
For this example, the desired factor of safety against single-lobe buckling
(FSs'.min) is 1.5. It is further assumed that the liner is restrained longitudinally.
The effective modulus is therefore
E
E' = 2 = 218,364 MPa (31,663 ksi)
(l - v)
Assuming a radial gap (Yo) of 0.61 mm (0.024 in.) between the steel liner
and the backfill concrete, the gap ratio is equal to
Yo 0.61 mm
K =-R = 1,537 mm = 0.0004
Given this information, the three Jacobsen single-lobe buckling equations
[(15a), (I5h), and (I5c)] are solved simultaneously by iteration, and the fol-
lowing results are produced: P cr = 2,030 kPa (294 psi); at = 53.9°; and 13 =
53.4°. The computed factor of safety is
Pcr 2,030 kPa
FS sl = Pd = 1,240 kPa = 1.64 2:: FsI•mio = 1.5
=:} Actual factor of safety is sufficient (if valid)
For the factor of safety computed using the buckling equations to be valid,
the tangential stress «(Tc,) induced in the liner by the critical buckling pressure
must not exceed 80% of the material yield strength. In this case
(2,030 MPa) (1.537 m)
(Tcr = (0.0254 mm) = 123 MPa (17.8 ksi) < 0.80Fy = 177 MPa (25.6 ksi)
The computed safety factor of 1.64 against buckling is therefore valid, and
the liner is acceptable to resist the design external pressure.
JOURNAL OF ENERGY ENGINEERING / DECEMBER 1998/71
-------i-------
Downloaded from ascelibrary.org by New York University on 05/10/15. Copyright ASCE. For personal use only; all rights reserved.
'""
Astf
E-.--.-.-.-...-.-.-.-.----------- __ tt
-r
J=MOMENT OF INERTIA OF EFFECTIVE SECTION
h=MAX(ho, hi)
Z, = Z2= O.78SJRt
Z, '" .!oz
Z2'" %
EFFECTIVE SECTION
FIG. 10. Notation for Stiffened Lining (Channel Stiffener Shown)
Shell Thickness
It is generally assumed that the maximum tangential working stress in the
shell (computed as PdRolt) is unaffected by the presence of the stiffeners, and
this stress is equal everywhere in the shell. (This assumption is close to reality
in most cases, as it can be shown that the stiffeners only affect the tangential
stress in close proximity to the stiffeners.) As for plain steel liner, the max-
imum allowable stress (<Tall), according to the Boiler Code, is the lesser of
2/3 of the steel yield strength and 1/3 of the steel ultimate strength.
74/ JOURNAL OF ENERGY ENGINEERING / DECEMBER 1998
(18)
Downloaded from ascelibrary.org by New York University on 05/10/15. Copyright ASCE. For personal use only; all rights reserved.
The steel thickness computed in this manner will be the thinnest possible;
however, using the thinnest steel will result in the closest required stiffener
spacing. Ideally, the shell thickness should be between about 6 and 20 mm
due to handling and fabrication limitations.
Stiffener Spacing
Free-tube buckling is assumed critical for local buckling of the shell be-
tween stiffeners. The writers recommend that the maximum clear spacing
between the stiffeners (L) allowed for a specific design pressure (Pd ) be de-
termined by both the Donnell and von Mises equations, and the smaller com-
puted spacing be selected for design. In both equations the optimum value
of stiffener spacing for a specific design pressure is obtained by iteration.
This iterative process is as follows:
Moore (1990) recommends that the minimum factor of safety (FSft.min) used
for this local buckling case be about 10% higher than that used for global
(single-lobe) buckling. The writers agree with this recommendation. For ex-
ample, if a factor of safety against single-lobe buckling failure of 1.5 is
chosen for the design, it would be appropriate to use a factor of safety of
1.65 for design against free-tube buckling failure.
If the computed stiffener spacing is too small for practical purposes, then
the shell thickness should be increased and the process repeated to obtain a
revised stiffener spacing.
Effects of Stress Levels on Shell Thickness Requirement. As for the
plain steel liner, in theory in order for the critical buckling pressure (Pcr.ft ,
computed by the free-tube buckling equations in the previous section) to be
valid, the tangential stress level in the shell should not exceed 80% of the
yield strength of the steel being used:
P R P R
U'cr = 0.80Fy t cr.ft 0 (19)
t O.80Fy
Satisfying this criterion often requires using thicker shell steel than is nec-
essary to meet the Boiler Code allowable stress limits.
JOURNAL OF ENERGY ENGINEERING / DECEMBER 1998/75
problem is given by Bugge (1960). Bugge shows that the tangential stress
in the shell at the point of application of the line load is
rr = @--vt3(l - v 2) + C?S2>.. (200)
t -y t smh 2>" + sm 2>"
where
--vt3(1 - v 2 )
>.. = 4, (20b)
2vRt
Dividing the thrust (wR) by the product of the tangential stress (rr) and the
shell thickness (t), one computes the effective length (z):
VRt sinh 2>" + sin 2>"
z= "¢!3(1 _ v2) cosh2>.. + cos2>.. (21)
One can show that the effective length converges to this maximum value
quickly as the stiffener spacing is increased. In fact, for typical stiffener liner
geometries the stiffeners are sufficiently far apart that they can be assumed
not to interact with one another.
Plugging in a typical value of Poisson's ratio for steel of 0.3, (22) simplifies
to the familiar value of 1.56(Rt)ln.
Alternative approaches to calculating an effective length are available. For
instance, the Manual of Steel Construction (1991) addresses the effective
width (length) of compression members in Section B5, its commentary, and
Appendix B. Per the American Institute of Steel Construction, the effective
length of a stiffened steel liner is a function of the stiffener spacing, the shell
thickness, and the stress in the stiffened liner. Using this approach one will
find that for a stiffened liner the effective length will virtually always be 30
or more times the thickness of the shell. Amstutz (1970) uses 30 times the
shell thickness as the effective length of shell in his computations. The ef-
fective length computed by this method will be larger than 1.56(Rt)ln for a
stiffened liner of typical dimensions; for this reason, the writers recommend
that 1.56(Rt)ln be used as the effective length for the analytical procedure
outlined in the following section.
Stiffener Size
The composite section made up of the stiffener and liner shell is designed
to resist an external radial load at buckling equal to the external pressure
times the factor of safety times the stiffener center-to-center spacing, LeI. The
length of the steel shell used in the "effective section" (i.e., for computing
the effective moment of inertia and the neutral axis of the stiffened section)
is 0.78(Rt)ln, on each side of the stiffener, projected from the stiffener as
JOURNAL OF ENERGY ENGINEERING / DECEMBER 1998/77
1. A stiffener type and size are selected, and the following properties of
the composite section are calculated:
a. Total area (F): The sum of the stiffener area (A,If) and the total
area of the shell tributary to the stiffener (Lei' t). This is not the
area of the "effective section."
b. Effective moment of inertia (J): The moment of inertia of the
"effective section."
e. The distance from the neutral axis of the "effective section" to
the extreme outer fiber of the stiffened section (h).
d. The radius to the neutral axis of the "effective section" (R.).
2. Using the section properties computed above, the general Jacobsen
equations [i.e., (13a), (13b), and (13e)] are solved simultaneously by
iteration to determine Pcr,sl such that Pcr,sIIFSs/.min (FSs/.min is equal to the
minimum factor of safety against single-lobe buckling failure) is greater
than or equal to Pd'
3. If the computed Pcr,sl is such that Pcr,sIIFSs/.min is less than P d , then the
stiffener size needs to be increased and the process repeated from step
1. If the required stiffener size is larger than any commercially available
size, then either a special section must be fabricated or the stiffener
spacing decreased; alternatively, the shell thickness could be increased.
If the computed value of Pcr,sl is such that Pcr,sIIFS.1•min is substantially
greater than Pd , then the stiffener size can be decreased and the process
repeated from step 1. Iterations should continue until an economical
stiffener size is found.
4. As a final check, the average stress in the liner should be computed, as
explained previously. The average tangential stress induced by the crit-
ical buckling pressure over the entire cross-sectional area (ITave•cr) should
be kept below 80% of yield in order for the Jacobsen formulation to
be valid, or the value of E must be modified to determine the actual
factor of safety against single-lobe buckling failure. In other words,
(23)
Stiffener Weld
In general, load at the weld between the stiffeners and the shell comes
from two sources: (I) Horizontal shear forces due to sliding forces at buck-
78/ JOURNAL OF ENERGY ENGINEERING / DECEMBER 1998
('ITa) (1 "J'
'lT
2
(24)
Downloaded from ascelibrary.org by New York University on 05/10/15. Copyright ASCE. For personal use only; all rights reserved.
)
V = EJ p2 -
VQ
T=- (25)
J
(26)
This approach is conservative because more complex analysis shows that the
tangential stress in the stiffener is actually less than average tangential stress
in the stiffened liner under working conditions. More complex methods in
which the actual stress distribution in the stiffened liner is considered can be
used; however, it is rarely necessary to go to this level of detail.
Assuming a line load from a single weld at radius Ro , the service tension
load per unit length (w) is
Asif A'if
W = (I,Ye,d - = PdLcl - (27)
Ro F
The computed shear and tension loads are vectorially combined to produce
a conservative design weld load. The minimum total service (working) force
per unit length (Tweld), for which the weld should be designed is
Tweld = Vw 2
+ (28)
In (28), or is an ultimate force, while w is a working force. Since TWeld is
considered to be a working load, (28) is conservative. If a pair of fillet welds
is used to connect the stiffener to the shell, the size of the fillet welds on
each side of the stiffener shall be such that O.5Tweld is no greater than the
allowable unit load for that size weld.
Some codes do not permit the use of intermittent welds, and most codes
have minimum size fillet welds based upon the thickness of abutting metals.
For example, the Boiler Code requires that the minimum fillet weld leg size
JOURNAL OF ENERGY ENGINEERING / DECEMBER 1998/79
(Tall = min G Fy , F u ) = min (175 MPa, 161 MPa) = 161 MPa (23.3 ksi)
Based on the selected thickness of 25.4 mm, the outside radius (Ro ) is
1,879.4 mm (74.0 in.). The minimum thickness required based on stress con-
siderations is computed from (14) as follows:
t
PdR o
--
(Tall
= (1,034(161,000
kPa) (1,879.4 mm)
kPa)
= 12.1 .
rnm (0.475 10.)
For this example, stiffeners will be used to reduce the thickness of the steel
shell. Based on the foregoing, a thickness of 15.88 mm (5/8 in.) is selected
for the shell thickness. Based on this thickness, the outside radius of the shell
(Ro ) is 1,870 mm (73.62 in.) and the radius to the center line (R) of the shell
is 1,862 mm (73.3 in.). As an initial guess, CIOX20 channel stiffeners in-
stalled on a center-to-center spacing (Lei) of 1,778 mm (70 in.) are selected.
The clear spacing (L) between the stiffeners is therefore 1,524 mm (60 in.).
For stiffened lining notation refer to Fig. 10.
First, the free-tube buckling capacity of the 15.88 mm (5/8 in.) thick shell
between the stiffeners is checked to determine the acceptability of the selected
stiffener spacing. From (10), the approximate number of lobes (napproxima",) that
produces the minimum free-tube buckling pressure is
4 6'11'2[1 - (V)2]1I2 4 611"2[1 - (0.29)2]112
napproximate = (UR)'(tIR) (1,524 mm11,862 rnm)'(15.88 mm11,862 rnm)
=10
Table 2 presents critical free-tube buckling pressures (Per•ft ), computed using
the von Mises equation, (8), for a range of n values and an L of 1,524 mm
(60 in.).
The minimum value of P er•ft , from Table 2, is 1,675 kPa (243 psi) at n
values of 9 and 10. Similarly, using (9) for Donnell free-tube buckling, the
80 I JOURNAL OF ENERGY ENGINEERING I DECEMBER 1998
Per,.
n [kPa (psi»)
(1 ) (2)
7 2,703 (392)
8 1,924 (279)
Downloaded from ascelibrary.org by New York University on 05/10/15. Copyright ASCE. For personal use only; all rights reserved.
9 1,675 (243)
10 1,675 (243)
11 1,800 (261)
12 1,993 (289)
minimum value of P er.ft is 1,752 kPa (254 psi) at an n value of 10. In this
case, the von Mises equation yields a lower free-tube critical buckling pres-
sure; therefore, the factor of safety against free-tube buckling (FSft) is
Per ft 1,675 kPa
FSjl =- ' = = 1.62
Pd 1,034kPa
For this example, a minimum factor of safety against single lobe buckling
(FSs1.min) of 1.4 is desired. From the foregoing, it is seen that the factor of
safety against free-tube buckling (FSft) is 16% larger than the minimum re-
quired factor of safety against single lobe buckling (FSs1.min)' The factor of
safety against free-tube buckling (FSjl) is acceptable since it provides more
than the 10% additional capacity, as recommended previously. Hence, the
stiffener centerline spacing (Lcl ) of 1,778 mm (70 in.) is acceptable for the
C1OX20 channel stiffeners [given a shell thickness of 15.88 mm (5/8 in.)].
For information purposes, per (19), the tangential stress in the shell should
not exceed 80% of the yield strength of the steel being used for the computed
factor of safety against free tube buckling to be valid:
Per,jlR o (1.675 MPa) (1,879 mm)
<J'cr =- t
-= (15.88 mm)
= 198 MPa (28.7 ksl).
< 210 MPa (30.4 ksi)
Therefore, this unessential criterion is met, and the factor of safety of 1.62
against free-tube buckling (FSjl) is valid.
Next, the stiffener size must be checked for single-lobe buckling, using
the Jacobsen equations (13a), (13b), and (13c). For a CI0X20 channel:
The effective lengths (Zl and Z2) of the liner on either side of the channel
stiffener, as shown in Fig. 10, are
h. = .o..(t/_2.:....)('-.-t_·
• (t· W.) + A Sif
(1/2)(15.88 mm)'(522 mm) + (69.6 mm - 15.4 mm + 15.88 mm)(3,794 mm2)
h· = -=---....:....:...----'-'----:-:-':-.......:..
, (15.88 mm)(522 mm) + (3,794 mm2)
= 27.45 mm (1.08 in.)
J= 1 + (t + b -.i _ h.)2A
f • sif
+ W.t
12
3
+ (h. _
'2
:.)2 (t.W)
•
= 49.64 in.2
It is assumed that the liner has gasketed joints; hence, it is not restrained
longitudinally and E' is equal to 200,000 MPa (29,000 ksi). It is further
assumed that the gap ratio (K) is 0.0003. Given these values, the Jacobsen
equations (13a), (13b), and (13c) are solved simultaneously by iteration to
obtain P cr = 1,748 kPa (254 psi); a = 85.235°; and = 84.985°. The computed
factor of safety against single lobe buckling is
P cr 1,748 kPa
FS S
I =-P = 1,034 kPa = 1.69 > F./_' = 1.4 D
d
(-
sin sin(84.985°)
In.)
a = (:;;:L) tan(a - 13) = (1,842'iT nun) tan(85.235° - 84.985°) = 2.56 nun (0.10 10.)
.
Q = Astf{h - x) = (3,794 nun2 )(58.0 nun - 15.4 mm) = 161,624 mm3
2
V
'ITa)
= EJ (--;:- (1p2 - -;;
'lT )
= (200,000 MPa) (1.14 X 10- m
5 4
)
('IT(2.56 mm»)
1.842 mm
2
1 'lT )
( (1,863 m)2 (1.842 m)2
. VQ (-26.09 kN) (1.61624 X 10-4 m3 )
V= -26.09 kN (-5.87 kips) T = -J 10 5 4
(1.14 X m)
= -370 kN/m (-2.11 kips/in.)
The tension load per unit length of weld (w) is computed from (26) and
(27) as follows:
U ave"
PdRoLcf
=- F - =
(1,034 kPa) (1,879.4 mm)(1,778
(32,029 mm2)
nun) = 10.788 MPa (1565'
. ksl)
2
(0.003794 m ) .
W = Uave,d =-Astj
R
= (107.88 MPa) 18
• 794 m
= 218 kN/m (1.24 kips/m)
o
The net load per unit length of weld (Tweld) of the stiffener is therefore com-
puted from (28):
r
TWeld = Vw 2 + = V(218 kN/mi + (370 kN/mi = 429 kN/m (2.45 kips/m)
Hence, the load per unit length for each weld is 214 kN/m (1.22 kips/in.).
Per the Boiler Code, a fillet weld of 3.2 mm (118 in.) size has a capacity of
280 kN/m (1.6 kips/in.); therefore, a 3.2 mm (l/8 in.) weld would be suffi-
cient. However, the minimum weld size permitted on each side of the stiff-
ener, per the Boiler Code, is 6.4 mm (114 in.).
The foregoing computations were all based on a 15.88 mm (5/8 in.) shell
with ClOX20 channel stiffeners spaced at 1,778 mm (70 in.) centerline spac-
ing. However, this may not be the most economical design. In order to de-
termine the most economical design, many factors must be considered. The
net liner weight alone can be assessed by defining an equivalent thickness
(t.q = FILc/) that represents the thickness of a plain steel liner that weighs the
same as the stiffened liner. Using the same methodology, complete analyses
are made of various shell thickness and stiffener combinations for the design
criteria in this example. Table 3 presents a summary.
JOURNAL OF ENERGY ENGINEERING / DECEMBER 1998/83
"-
c...
oC
Jl
o'Tl
m
z
m
Q
m
z
Ci)
TABLE 3. Design Example 3: Suitable Shell ThlcknesslStiffener Combinations
Z t Lei t"'l Per,. Pc".,
m
m [mm (in,)] Stiffener [mm (in.») [mm (in,») [kPa (psi») FS. [kPa (psi») FS.,
Jl
Z (1 ) (2) (3) (4) (5) (6) (7) (8)
Ci)
"-
12,7 (112) ClOX20 1,168 (46) 15.9 (0.626) 1,648 (239) 1.59 1,792 (260) 1.73
o
m
14.3 (9/16) ClOX20 1,448 (57) 16.9 (0.665) 1,662 (241) 1.61 1,778 (258) 1.72
(") 15.9 (5/8) CI0X20 1,778 (70) 18.0 (0.709) 1,676 (243) 1.62 1,748 (253) 1.69
m 19.0 (3/4) C12X25 2,591 (102) 20.8 (0.820) 1,717 (249) 1.66 1,929 (280) 1.87
3: 25.4 (1) None N/A 25.4 (1.000) N/A N/A 1,593 (231) 1.54
OJ
m
...
Jl
<0
<0
(Xl
computed values of p cr•ft for free-tube buckling and Pcr"l for single-lobe buck-
ling are 1,648 kPa (239 psi) and 1,792 kPa (260 psi), respectively. With Pd
of 1,034 kPa (150 psi), the factor of safety against free-tube buckling (FSft)
is 1.59, which is 3% higher than the minimum recommended value of 1.54
(i,e., 1.1 X FS s1•mm = 1.1 X 1.4). The factor safety against single-lobe buckling
(FS s1 ) is 1.73, which is 24% higher than the minimum value of 1.4. The
validity of these safety factors is checked as follows.
For the 12.7 mm (112 in.) shell, R o is equal to 1,867 mm (73.49 in.). The
service stress is therefore
Therefore, the factor of safety of 1.59 computed for free-tube buckling is not
valid. Using the method illustrated in design example 2, it is determined that,
for this example, the actual safety factor against free-tube buckling is between
1.38 and 1.59. By iteration, it can be shown that the actual factor of safety
is about 1.47.
Review of Shell Stresses in Stiffened Liner with 12.7 mm (112 in.) Shell
Plate. From Young (1989) one can derive equations that closely approxi-
mate the actual load transfer between the stiffener and the steel cylinder at
service conditions, as well as the stress in the shell. For channel stiffeners,
the line load transfer (w) at the two welds at spacing d (i.e., depth of channel)
can be computed using (29). This equation neglects axial restraint in the
cylinder and the effect of adjacent stiffeners. Neglecting the effect of adjacent
stiffeners is valid unless stiffeners are very close to one another. The equation
is simplified by assuming that the axial stress is uniform across the stiffener
cross section, and that all forces and pressures act at the centerline radius of
the cylinder, R:
JOURNAL OF ENERGY ENGINEERING 1 DECEMBER 1998/85
where
'1' = 1 +
1/4
_ 3(1 - v 2 )
A- 22 ]
[
Rt
d = distance between the two stiffener connection welds (depth of the chan-
nel); Rsif = radius to the neutral axis of the stiffener; P d = design external
pressure on the shell; t = thickness of the shell; w = line load per stiffener
connection point; and Asif,cp = area of stiffener per connection point (e.g.,
Asif,cp = 0.5A sif for stiffeners with a fillet weld on each side).
For a stiffener connected to the shell via a single connection point (e.g., a
tee stiffener with a groove weld between the stiffener web and the shell), the
value of 'l' should be set to one and Asif,cp should be set to the full stiffener
area (A sif). For channel stiffeners, since there are two connection points per
each channel, Asif,CP should be set to one half of the full stiffener area (A sif).
In addition, for channel stiffeners with depths of 250 mm (10 in.) or more,
the value of'l' is approximately unity. Using v of 0.29, 'l' of 1, and (29),
the line load per each stiffener connection point is approximated as follows:
(30)
w = CtlASt/,cp) +
For the 12.7 mm (112 in.) thick shell in design example 3, d = 254 mm
(10 in.), R = 1,860.4 mm (73.24 in.), P d = 1,034 kPa (150 psi), and A sif =
3,794 mm2 (5.88 in?). Using the more precise equation (29) for w, the fol-
lowing are computed:
1,034 kPa )
w = ( [0.0127 m/(0.5)(0.OO3794 m2)] + [l/1.55v'(1.8604 m)(0.0l27 m)]
The value of 123 kN/m (703 lb/in.) is higher than the actual value computed
previously, and the recommended method of computing w in an earlier sec-
tion is shown to be conservative.
Fig. 11 shows a summary of the stresses along the 12.7 mm (112 in.)
cylinder plate, with the stiffener weld point at x = 0 mm. With d of 254 mm
(10 in.) and L of 914 mm (36 in.), x = -127 mm (-5 in.) represents the
midpoint of the channel stiffener, and x = 457 mm (18 in.) represents the
midpoint of the cylinder between stiffeners. The lower curve in Fig. 11 rep-
resents the computed axial tangential stress (without bending), which is min-
imum (91 MPa = 13.2 ksi) near the welds (at x = 0) and maximum (154
MPa = 22.3 ksi) at x = 366 mm (14.4 in.). In reality, tangential (secondary)
-
160 ,-- ,- - ,-- -----
150 - e-- -- -- -
V/ V
Ii' II
6 130 - -- c- -
.
z
fl
00 120 f-"'- '-"-' . . . ... .. i"'--' '-" '-"' f--'" • ,,"--""
..•.
r= 110 /1\ _.... _.
... f- - -- -- f--
i'-
I
I
100
90
- 1'\
f-- -
V
I
t'- ..... I
I
-_.-
J Pd
f--
1034 kPa
-
1--
.- -
.. -
f--
80
-125 -75 -25 25 75 125 175 225 275 325 375 425
Distance From Stiffener Connection Point (mm)
FIG. 11. Tangential Stress In Stiffened Liner with 12.7 mm (1/2 In.) Shell Plate
From design example 3, if the stiffeners are ignored the axial stress in the
12.7 mm (1/2 in.) shell is 152 MPa (22 ksi), which is shown by the upper
horizontal line in Fig. II. The average tangential service stress across the
entire area of the stiffened liner is 121 MPa (17.6 ksi), which is shown by
the lower horizontal line in Fig. II. Longitudinal stresses from bending are
not shown, but are a maximum of 103 MPa (15 ksi) at the weld location.
ACKNOWLEDGMENTS
The writers would like to thank the Metropolitan Water District of Southern California for
the opportunity to acquire the information presented in this paper and for the two photographs
of a liner buckled in a single-lobe mode.
APPENDIX. REFERENCES
Amstutz, E. (1970). "Buckling of pressure-shaft and tunnel linings." Water Power,
22(Nov.), 391-400.
Boiler and pressure vessel code, section VIII-rules for construction of pressure
vessels. (1995). American Society of Mechanical Engineers, New York.
Bo£Ot. (1957). "Aambage d'in cylindre a paroi mince, place dans une enveloppe
rigide et soumi a une pression exterieure." La Houille Blanche, Grenoble, France,
6,231-245.
Donnell, L. H. (1976). Beams, plates and shells. McGraw-Hill, New York.
EI-Sawy, K., and Moore, I. D. (1997). "Parametric study for buckling of liners: Effect
of liner geometry and imperfections." Trenchless pipeline projects-practical ap-
plications, L. E. Osborn, ed., ASCE, Reston, Va., 416-423.
Aiigge, W. (1960). Stresses in shells. Springer, Berlin.
Gaylord, E. H., and Gaylord, C. N. (1972). Design of steel structures. McGraw-Hill,
New York.
Glock, D. (1977). "Oberkritisches verhalten eines starr umrnantelten kreisrohres bei
wasserdruck vos aussen und temperaturdehnung." Der Stahlbau, Berlin, 7, 212-
217.
Jacobsen, S. (1974). "Buckling of circular rings and cylindrical tubes under external
pressure." Water Power, 26(Dec.), 400-407.
Manual of steel construction-allowable stress design. (1991). American Institute of
Steel Construction, Chicago.
Moore, E. T. (1990). "Designing steel tunnel liners for hydro plants." Hydro Rev.,
9(Oct), 72-89.
Southwell, R. V. (1913). "On the general theory of elastic stability." Philosophical
Trans., 213A, 187-244.
Standard specifications for highway bridges. (1992). 15th Ed., American Association
of State Highway Transportation Officials, Washington, D.C.
"Steel penstocks." (1993). ASCE manuals and Rep. on Engrg. Pract. No. 79, New
York.
Timoshenko, S. (1934). Theory of elasticity. McGraw-Hill, New York.
Vaughan, E. W. (1956). "Steel linings for pressure shafts in rock." J. Power Div.,
ASCE, 82, 949-1-949-40.
von Mises, R. (1914). "Der kritische aussendruck zylindrischer rohre." Zeitscrift
V.D.I., 58, 750-755.
88/ JOURNAL OF ENERGY ENGINEERING / DECEMBER 1998
David Berti
Downloaded from ascelibrary.org by New York University on 05/10/15. Copyright ASCE. For personal use only; all rights reserved.
Principal
Jacobs Associates Engineers and Consultants
San Francisco, CA
Richard Stutzman
Director of Marketing
Jeffco Painting and Coatings
Martinez, CA
Formerly, Senior Civil Engineer
Ensign & Buckley Consulting Engineers
Larkspur, CA
Eric Lindquist
Associate
Jacobs Associates Engineers and Consultants
San Francisco, CA
Maryam Eshghipour
Staff Engineer
Jacobs Associates Engineers and Consultants
San Francisco, CA