ARTICLE IN PRESS

International Journal of Pressure Vessels and Piping 84 (2007) 762–772

www.elsevier.com/locate/ijpvp

Fitness-for-service assessment of spherical pressure vessels

with hot spots
P. Tantichattanont, S.M.R. Adluri, R. Seshadri
Faculty of Engineering and Applied Science, Memorial University, St. John’s, NL, Canada A1B 3X5
Received 23 November 2005; received in revised form 1 December 2006; accepted 20 December 2006

Abstract

Spherical shapes are used in industry as hemispherical vessel heads or high-pressure storage vessels due to the inherent strength
associated with the shape. Structural integrity of such components needs to be evaluated periodically to prevent failure of the vessels
under operating conditions. The paper develops a method for Level 2 (as defined by API 579 [(2000). Fitness-for-service, API 579.
Washington, DC: American Petroleum Institute]) fitness-for-service estimation of spherical shapes subject to local hot spots where the
temperatures are elevated due to local damage. The decay length for spherical shells is determined, and the size of hot spot to be identified
as local is proposed. A lower bound ‘‘remaining strength factor’’ (RSF) for spherical pressure vessels containing hot spots is formulated
by the application of Mura’s variational formulation and the ma-multiplier method. The effectiveness of the proposed Level 2 method is
evaluated and demonstrated through an example.
r 2007 Elsevier Ltd. All rights reserved.

Keywords: Fitness-for-service; Hot spots; Decay length; Spherical shell; Remaining strength factor; Mura’s formulation; Structural integrity

1. Introduction Levels 1–3 are progressively more sophisticated. Each

Structural integrity is of considerable importance in
order to avoid failures of mechanical components and
structures in a number of industrial sectors. The ability to
demonstrate the structural integrity of an in-service
component that sustained some damage or contains a flaw
is termed as integrity assessment or fitness-for-service
(FFS) and is extensively dealt with by assessment
procedures such as R6 [1]. The FFS evaluations are
conducted periodically to determine whether a component
with existing damage is suitable for continued service until
the end of some desired period of operation such as the
next shutdown, a specific future date or the end of its useful
life. The assessments include determination of current
serviceability to ensure safe operation in the present
condition, and remaining service life of the equipment.
For pressurized equipment in operating plants, API 579
[2] prescribes three levels of structural integrity evaluations.
Corresponding author. Tel.: +1 737 3800; fax: +1 709 4042.
E-mail address: adluri@engr.mun.ca (S.M.R. Adluri).
assessment level provides a balance between the degree of
conservatism, the amount of required input, the skill of the
practitioner, and the complexity of the analysis. Level 1
assessments are the most conservative screening criteria
that generally include the use of charts and tables, which
can be implemented by plant technicians with a minimum
quantity of inspection and component information. Level 2
assessments involve detailed calculations intended for use
by plant engineering personnel with the help of a
recommended procedure. Level 3 assessments require a
full rational analysis by specialists where advanced
computational techniques such as nonlinear finite element
analysis are engaged.
The procedures in API 579 are developed to overcome
the shortcomings of the former inspection codes for
pressure vessels and piping which are mainly based on
empirical data and past experience [3]. In developing them,
extensive validation based on both numerical analysis and
physical testing has been applied to various damage modes
such as metal loss and crack-like flaws. In that regard,
further enhancements to Level 2 procedures in damaged

0308-0161/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijpvp.2006.12.003
 ARTICLE IN PRESS
P. Tantichattanont et al. / International Journal of Pressure Vessels and Piping 84 (2007) 762–772
areas such as hot spots are of significant interest. The
current paper focuses on estimating fitness for service of
pressurized components of spherical shape subject to
damage in the form of localized hot spots with elevated
temperatures.
Localized ‘‘hot spots’’ are typical of damage that occurs
in ageing pressure vessels, piping or storage tanks. They are
caused by damage due to loss of refractory lining on the
inside wall of pressure components or due to maldistribu-
tion of flow containing catalyst and reactive fluids. They
are detected through thermography or temperature sensi-
tive paints. An FFS assessment is required to determine the
acceptance of continued service of a component containing
hot spots. The parameter generally used in such assess-
ments is the remaining strength factor (RSF). The RSF is
defined as the ratio of the collapse pressure of the damaged
component to the collapse pressure of the undamaged
component.
For loads in excess of the limit load it is not possible to
have a statically admissible stress field that falls inside the
yield surface. In addition, the stress fields for limit loads
calculated by using kinematically admissible velocity fields
should be on the yield surface. Mura et al. [4,5], eliminated
such strict requirements by introducing the concept of
integral mean of yield criterion to the variational formula-
tion.
Among several limit load multipliers arising from
Mura’s extended variational formulation, the ma-multiplier
proposed by Seshadri and Mangalaramanan [6] has
proven to offer significantly improved lower-bound
estimates. The ma multiplier is determined on the basis of
a linear elastic finite element analysis in conjunction
with the nesting surface theorem and the idea of
leap-frogging to the limit state. Seshadri [7] evaluated
the RSF for thin-walled cylindrical pressure compo-
nents containing hot spots based on the mo, mL and
ma multipliers. The concept of the localized effect of
discontinuities on the cylindrical shell is discussed in detail,
and the concept of reference volume is introduced as the
kinematically active portion that participates in plastic
action.
Indermohan and Seshadri [8] proposed a Level 2 FFS
methodology for evaluating locally thinned areas and local
hot spots in cylindrical shells. For the problem considered
by them, the recommended RSF calculated using ma-
multiplier provides a close lower bound approximation
compared with nonlinear finite element analysis model.
The ideas proposed lead to a simple and yet practical Level
2 assessment of RSF. These ideas include the use of decay
lengths and other intuitive concepts meshed with varia-
tional principles.
As mentioned above, the present work focuses on
developing a Level 2 structural integrity assessment
method for spherical pressure vessels containing local hot
spots. It is based on extended variational formulations in
plasticity similar to those presented by Seshadri [7]. The
concept of decay length is applied to the calculation of
763
reference volume and RSF. The size of hot spot area that
can be labelled as local hot spot is identified.
2. Decay lengths in spherical shells
The estimation of decay length is of substantial
importance in design and integrity assessment because it
identifies the reference volume and is necessary to identify
the interaction of multiple local loads. Reference volume
defines the containment of effects of local stresses and
strains acting on the structure. Decay or characteristic
length can be estimated by evaluating the effect of a local
force on the spherical shape. Decay length is defined as the
distance from the applied force to the point where the effect
of the force is almost completely dissipated or at least
becomes negligible. A larger decay length generally
indicates better energy dissipation of the structure and
leads to higher loading capacity whereas a smaller decay
length suggests severe local effects due to the applied
forces. It must be noted that the decay lengths studied in
the current paper are based on elastic analysis. The elastic
decay lengths are likely to be smaller than those calculated
from plastic analysis. This leads to an overestimation of the
damage severity and thus results in conservative RSFs. It
must also be noted that the entire Level 2 procedure
developed here and the variational formulation associated
with it are based on using elastic analysis to simulate the
limit behaviour. Hence, the use of elastic analysis for
estimating decay lengths (although slightly conservative in
the present case) is justified. A more detailed discussion on
this is given in [14].
2.1. Spherical shell loaded by concentrated normal force
Lukasiewicz [9], among others, has discussed the
problem of spherical shells extensively. The shell differ-
ential equations in terms of the radial displacement w and
stress function F, take the form:
      
2 1þn 1 h2
r2 þ 2 D r2 þ 2 w 1 r2 F
R R R 5ð1  nÞ
 2 
ð2  nÞh 2
¼ 1 r Pz , ð1aÞ
10ð1  nÞ
    
2 1 1n w n 2
r2 þ 2 r2 þ 2 F þ ¼ r Pz , (1b)
R Eh R R 2E
where R is the mean radius, h is the shell thickness, v is
Poisson’s ratio, Pz is the component of body forces normal
to the shell surface, and E is the modulus of elasticity of the
material. The flexural rigidity of the shell is expressed as
D ¼ Eh3/12(1v2), and r2 ¼ (1/l2)(q2/qx2+q2/qy2) is the
Laplacian operator in terms of x ¼ x̄=l and y ¼ ȳ=l; where,
l is a certain characteristic length.
Consider a spherical shell loaded by a concentrated
normal force P in the outwards direction at x ¼ 0 and
y ¼ 0 as shown in Fig. 1. The force can be represented by
 ARTICLE IN PRESS
764 P. Tantichattanont et al. / International Journal of Pressure Vessels and Piping 84 (2007) 762–772

y The decay angle jd for a spherical shell when n ¼ 0.3 can

be expressed as
P   !
rd l 2:1454
rl jd ¼ arcsin ¼ arcsin pffiffiffiffiffiffiffiffiffi . (3)
x R R=h
h
s The interaction of hot spots can be neglected when the
angle between the outer edges of the damaged areas is more
than 2jd.
Seshadri [7] derived the decay length of cylindrical shells
R
in the circumferential direction by applying a line load
along the generator of a cylinder and obtained an
equivalent jd ¼ 6.1/(R/h)1/4. The decay angle for a sphere
is less than 20% of that for a cylinder of the same R/h ratio
 for practical cases. This is because spheres have double
curvature compared to the single curvature of cylinders.
Deformations in a sphere are dissipated much more rapidly
than those in a cylinder. Thus, hot spot effects in spheres
Fig. 1. A spherical shell subjected to normal concentrated force.
are much more localized and hence have an adverse effect
compared to cylinders.
In the above, the decay length has been computed through
the use of a point load on the shell. This simulates the decay
means of a Fourier integral. The closed form expression for
in the meridional direction due to the shear at the rim of the
radial displacement w is then obtained as
hot spot. Similar decay lengths can be obtained due to edge
moments along the rim of the hot spot, as shown below.
Pl 2
w¼ ff þ f 2 þ f 3 þ f 4 g, (2)
2pD 1
2.2. Edge effect bending moments
where the functions f1 ¼ kei r, f 2 ¼ kR ½ð1 þ vÞððp=2Þ
pffiffiffiffiffiffiffiffi
Y o ðr 2kR Þ þ ker rÞ þ ð1=2Þrker0 r; f3 ¼ (Ze) ker r, and Stresses in a spherical shell with constant thickness
pffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi loaded symmetrically with respect to its axis can be
f4 ¼ Z/4r ker0 r. The length l ¼ Rh= 4 12ð1  n2 Þ; r ¼ r̄=l;
described by two differential equations of the second order
and kR( ¼ l2/R2) is the coefficient p characterizing
ffiffiffiffiffiffiffiffi the
of the form [10]
thinness of the shell. The function Y 0 ðr 2kR Þ is the Bessel
function of the second kind. The functions ker r, kei r are d2 Qj dQj
the real and imaginary parts of the Kelvin function, Kn, 2
þ cot j  ðcot2 j  nÞQj ¼ EhV , (4a)
dj dj
stated as evpi/2 Kv(zepi/4) ¼ kern z+kern z and ker0 r is the
first derivative of ker r with respect to r. For an isotropic d2 V dV R2 Qj
2
shell, e ¼ vh2/10(1v)l2, Z ¼ h2/5(1v)l2, where e and Z are þ cot j  ðcot j þ nÞV ¼  , (4b)
dj2 dj D
coefficients expressing the effects of stress normal to the
middle surface and the effect of transverse shear deforma- where Qj is the shearing force, the angle j is as shown in
tion, respectively. Fig. 2, and V is the angle of rotation of a tangent to a
The function f1 in Eq. (2) corresponds to Reissner’s meridian.
solution obtained by means of the theory of shallow shells, The solutions for Eqs. (4a) and (4b) can be estimated by
f2 presents the result of the use of the complete non-shallow the method of asymptotic integration and a satisfactory
shell Eqs. (1a) and (1b), and f3 and f4 are the effects of approximation for the shear force can be obtained as
transverse shear and normal deformations. Qj ¼ Celc sinðlc  gÞ, (5)
The radial displacement w has a peak near the loading
point and decays drastically away from this point in an where c ¼ aj, a is the angle from the crown to the edge
oscillating pattern. The amplitude of w in the second and of the sphere, the constants C and g are to be determined
subsequent half-waves is considerably smaller compared to from the edge conditions at j ¼ a, and l is obtained from
that in the first half-wave. The decay length, rdl can be  2
R
defined as the distance from the loading point to the first l4 ¼ 3ð1  n2 Þ . (6)
h
point where zero radial displacement occurs. This is chosen
as the characteristic decay length for the current study. The The bending moments can be determined as
values of rd can be easily verified (using Eq. (2)) to vary 
D dV R p
from 3.91 to 4.28 for all practical R/h values and the Mj ¼ ¼ pffiffiffi Celc sin lc  g þ
conservative value of 3.91 is applied in the derivation R dj l 2 4
below. and M y ¼ nM j . ð7Þ
 ARTICLE IN PRESS
P. Tantichattanont et al. / International Journal of Pressure Vessels and Piping 84 (2007) 762–772 765

p conditions are (Mj)j ¼ a ¼ 0 and (Nj)j ¼ a ¼ Na and the

constant g is equal to 451. This is similar to the fixed–fixed
case with a ¼ 901.
In each case, the effects of edge bending moments are
considered negligible after the first half-sine wave of
R  Eq. (7). The angle c1 at which zero moments first occur is


 
1 3p
c1 ¼ þg . (10)
l 4
If the conservative value of g equal to zero is applied, the
decay angle cd can be calculated for v ¼ 0.3 as
"  2 #1=4 rffiffiffiffi
3p 3p 1 h h
cd ¼ ¼ ¼ 2:140 . (11)
4l 4 1:47 R R
N N
It must be noted that g and hence cd will have slightly
larger values than above if the hot spot is much more
M M
flexible than the adjacent volume.
H H
2.3. Comparison of the decay angles
Fig. 2. (a) A spherical shell with built-in edges subjected to normal
pressure, (b) membrane stresses, and (c) superposed edge conditions.
The decay angles obtained by using Eqs. (3) and (11)
computed for concentrated loads and edge moments for
In the case of a spherical shell with a built-in edge different R/h ratios, are shown in Table 1.
subjected to a uniform normal pressure p as shown in Fig. 2a, It can be seen that the angle jd is only slightly higher
the membrane forces (Fig. 2b) are Nj ¼ Ny ¼ (pR/2) than cd for most practical cases. This may be due to the
which are superposed by the horizontal forces and the double curvature present in the shell and might prove a
moments uniformly distributed along the edge as in Fig. 2c. useful result for further work. In a real hot spot of
These forces and moments are of such magnitude that the significantly large size, the decay length is influenced by
corresponding horizontal displacement is equal and opposite both normal loads (pressures) and edge moments at the
to the displacement caused by the membrane compression, boundary of hot spots. Therefore, the actual decay angle is
and the corresponding rotation of the edge is equal to zero. likely to be in between jd and cd. Conservatively, cd can
Hence, the edge conditions can be expressed as be used for calculations for Level 2 assessments of decay
length since a smaller decay angle would indicate a more
pR2 ð1  nÞ adverse effect.
ðM j Þj¼a ¼ 2
and
4l 
1 pRð1  nÞ 3. Hot spot evaluation
ðN j Þj¼a ¼ . ð8Þ
cos a 2l sin a
3.1. Variational formulation
The constants C and g for spherical shells with built-in
edges subjected to uniform pressure p can thus be Limit analysis is generally employed to determine the
determined by Eq. (8) and substitution of c ¼ 0 into load carrying capacity of pressure components at which
Eqs. (5) and (7). The constant C is the magnification factor uncontained plastic flow occurs. An alternate procedure for
for functions (5) and (7), whereas g corresponds to the classical lower bound limit load evaluation based on a
phase shift of the sine function and directly affects the variational formulation was first developed by Mura et al.
decay length. The constant g for fixed edges is [5]. Pan and Seshadri [11] applied the concept of integral
  mean of yield with postulated statically admissible stress
1 1 fields that exceed the yield stress in portions of the
g ¼ tan (9)
1 þ cos2 a component. The expression for integral mean of yield
The angle g has the value ranging from 26.51 to 45.01 for a criterion employing the concept of reference volume can be
less than 901.
Table 1
Similarly, for spherical shells with edge moments where
The decay angles jd and cd for different R/h ratios
the rotations are restrained and the displacements are
allowed, the boundary conditions are (Mj)j ¼ a ¼ Ma and R/h 20 30 40 50 75 100 200
(Nj)j ¼ a ¼ 0. For this case, the constant g is equal to zero.
jd (deg) 28.7 23.1 19.8 17.7 14.3 12.4 8.7
If radial displacements and rotations at the edges of
cd (deg) 23.5 19.2 17.7 14.9 12.0 10.5 7.4
spherical shells are allowed to move freely, the boundary
 ARTICLE IN PRESS
766 P. Tantichattanont et al. / International Journal of Pressure Vessels and Piping 84 (2007) 762–772
written as
Z h i actual equivalent stress. Therefore, even if the uniform
m0 f ðs̄0ij Þ þ ðj0 Þ2 dV ¼ 0: (12)
VR
The superscript ‘‘0’’ refers to statically admissible stress
distributions that are in equilibrium and m0 is the plastic
flow parameter from the associated flow rule. The
deviatoric stress s̄0ij corresponds to the impending limit
state, where, s̄0ij ¼ m0d s0ij and m0d is the limit load multiplier
and the deviatoric stress s0ij equilibrates the applied set of
loads. The parameter j0 is a point function that takes on a
value of zero if s̄0ij is at yield and remains positive below
yield.
The Tresca and von-Mises yield criteria can be
expressed, respectively, as
f ðs̄0ij Þ ¼ jm0d se  sy j ¼ 0 and
f ðs̄0ij Þ ¼ ½ðm0d se Þ2  s2y  ¼ 0, ð13Þ
where se is the relevant equivalent stress and sy is the
appropriate temperature dependent yield stress.
The integral mean of yield criterion can be applied to a
pressure vessel with local hot spots subject to the von-
Mises yield criterion assuming the primary stresses to be
uniformly distributed within the adjacent volume, VC, and
the hot spot volume, VH. The flow parameter m0 is taken as
a non-zero constant for all the elements in the model.
Hence it drops out of the equation. This and other similar
effects are discussed in detail by Pan and Seshadri [11]. The
integration of Eq. (12) leads to
ððm0d s0eC Þ2  s2yC ÞV C þ ððm0d s0eH Þ2  s2yH ÞV H ¼ 0, (14)
where the suffix H refers to the hot spot region and suffix C
refers to the active volume away from the hot spot. If the
thickness is constant throughout the reference volume, the
effective stresses in both regions are the same and are equal
to se. The upper bound limit load multiplier m0d for the
von-Mises criterion can then be obtained as
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
s2yC V C þ s2yH V H
m0d ¼ , (15)
s2e V R
where VR ¼ VC+VH is the reference volume.
Similarly, the multiplier m0d for the Tresca criterion can
be written as
syC V C þ syH V H
m0d ¼ . (16)
se V R
Although Indermohan and Seshadri [8] applied the
assumption of uniform stress distribution throughout each
volume in the calculation of the multiplier m0d , this uniform
stress is only used to obtain a practical and effective way to
calculate approximate lower bound limit load multipliers
and does not represent the exact stress distribution in the
component. A similar assumption is employed in the
present work. The assumption is justified since the primary
stress that leads to the limit load is considered as a uniform
membrane stress for the present application. It will be
shown later that the RSF is not directly dependent on the
stress assumption is somewhat influenced by the bending
action, in the limit, the RSF, being a ratio of stresses will
not change significantly. However, bending effects are still
very important and are included in the form of the decay
length calculation which defines the active (or reference)
volume that participates in forming the limit mechanism.
This justification can be verified from the inelastic finite
element analysis results referenced later [14].
3.2. Reference volume for spherical shell
In the current study, irregular shapes of hot spots are
assumed to be represented by an equivalent circular shape.
For a hot spot identified through the included angle ja in a
spherical shell as shown in Fig. 3, the volume of the hot
spot can be calculated as
V H ¼ pR2o hð1  cos ja Þ, (17)
where Ro is the outside radius of the sphere and h is the
shell thickness.
The adjacent volume is defined as the strip volume
surrounding the hot spot where the shell might participate
in plastic action. This is given by
V C ¼ pR2o hðcos ja  cosðja þ cd ÞÞ, (18)
where cd is as defined in Eq. (11).
The reference volume VR obtained as VR ¼ VC+VH is
the total volume of the kinematically active portion of the
shell. It must be pointed out that VR as computed here can
be improved upon by better representation of the active
volume through more accurate analysis.
3.3. Equivalent stress
Temperature changes in a local area affect the material
properties of the component in the proximity of the area.
In order to calculate the elastic thermal stresses and strains
due to the temperature change, we need effective elastic
modulus and thermal coefficient. These are obtained as
weighted averages by the volumes corresponding to each
temperature zone. Thus, the effective elastic modulus and
= 0
Hot Spot Volume (VH)
Reference Volume (VR = VH + VC)
Adjacent Volume (VC)
R. a 
d
= 
2
Fig. 3. Equivalent circular hot spot and reference volume.
 ARTICLE IN PRESS
P. Tantichattanont et al. / International Journal of Pressure Vessels and Piping 84 (2007) 762–772 767

thermal coefficients can be written as where Z ¼ m0d =mL , and mL ¼ syH/se is the classical lower
ECV C þ EHV H aC V C þ aH V H bound multiplier. From Eqs. (15) and (16), we can see that
E eff ¼ and aeff ¼ . (19) Z is independent of the equivalent stress se. RSFa is then
VR VR
expressed as
This is slightly different from the approach used by ma
Seshadri [7] where a simple average was used. RSFa ¼ . (24)
m0u
The thermo-elastic equivalent primary stress of a
spherical shell subject to internal pressure and temperature Note that a finite element analysis is not required to
change is calculated as estimate RSFa since the structure is considered as
Pd Ri E eff aeff DT determinate for obtaining average stresses through the
se ¼  , (20) thickness of the shell.
2h 2
The remaining strength factor RSFL based on the
where Pd is the design pressure, Ri is the inner radius of the classical lower bound multiplier mL is
sphere, and DT ¼ THTC is the temperature change due to
loss of refractory lining leading to the formulation of hot spot. mL
RSFL ¼ . (25)
m0u
3.4. Remaining strength factor (RSF)

In API 579, the concept of a RSF is utilized to define the
acceptability for continued service of a component contain-
ing a flaw. In terms of a plastic collapse load, the RSF is
defined as
PLD
RSF ¼ , (21)
PLU
where PLU and PLD are the plastic collapse loads of
the component before and after damage, respectively. If the
calculated RSF is greater than the allowable RSF, the
component can continue to be in service without any repair
or remediation till the next scheduled major maintenance.
The recommended value for the allowable RSF is 0.90 for
equipment in process services. In the current study,
allowable RSF is estimated as the ratio of required
thickness to design thickness in the undamaged state. The
hot spot is assumed not to interact with any other types of
damage such as corrosion or with other nearby hot spots.
Three types of RSF are computed in this paper. The
upper bound remaining strength factor RSFU can be
obtained from the m0d as
m0d
RSFU ¼ , (22)
m0u
where m0d is a statically admissible multiplier calculated by
Eq. (15) or (16) as discussed earlier and m0u ¼ syC =se is the
multiplier for undamaged vessel.
The second RSF is calculated based on the improved
lower bound multiplier called ma proposed by Seshadri and
Mangalaramanan [6]. The ma is determined on the basis of
two linear elastic finite element analyses and is validated to
be a better approximation than classical lower bound limit
analysis using the theorem of nesting surfaces. The ma
multiplier is given by the expression
2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffi
2 2
2Z þ ZðZ  1Þ ð1 þ 2 
ma ¼ 2m0d 4 pffiffiffi 2
2
ðZ þ 2  5ÞðZ þ 2 þ 5Þ
3.5. ‘Local’ hot spots
Most structural materials such as steel have less
strength at high temperature. When a hot spot occurs in
a spherical shell with internal pressure, the hot region
undergoes higher deformations than the cooler part. At the
junction where hot and cold regions meet, the cold area
acts as the support constraining the hot zone from free
deformation. If the temperature difference between the two
zones is very high, the cold part can be regarded as fixed
end support where the edge displacements due to
temperature change are fully restrained. As the tempera-
ture difference decreases, the cold zone may not be much
stiffer than the hot zone and the edge effects become less
pronounced.
The hot spots are defined as ‘local’ when the resisting
forces due to higher stiffness of the cold region have not
damped out and the pure equilibrium stress does not occur
at any place inside the hot spot area. Therefore, the
maximum ‘local’ hot spot angle size can be defined by the
decay angle c1 [14].
The decay angle c1 obtained by Eq. (10) depends on the
value of l which represents the thinness of the shell and
the angle g. If the hot spot is very flexible compared to the
remaining part, we can assume a sort of fixed support at
the hot spot boundary. For this case, the value of g is given
by Eq. (9). If the hot spot stiffness is the same as that of the
adjacent part, it implies that there is no differential rotation
(indicating the presence of rigid connection). For this case,
g ¼ 0. The actual g will be between the two extremes. If we
assume that the relative stiffness is proportional to the ratio
of the Young’s moduli of the two regions, g can be assumed
to vary linearly from that given by Eq. (9) g ¼ tan1
(1/(1+cos2 a)) for EH/EC ¼ 0 to g ¼ 0 for EH/EC ¼ 1.0. It
pffiffiffi ffi3 can be shown that for practical values of EH and EC, g
ZÞðZ  1 þ 2Þ values are small. Therefore, a conservative value of g ¼ 0
pffiffiffi 5,
is assumed for all cases and the maximum angle size
of local hot spots is equal to cd calculated by using
(23) Eq. (11).
 ARTICLE IN PRESS
768 P. Tantichattanont et al. / International Journal of Pressure Vessels and Piping 84 (2007) 762–772

If the angle ja defining the size of the hot spot is larger
than cd, it means that there is a certain area within the hot
spot that is in pure membrane action defined by
equilibrium stresses alone. This area will accordingly fail
in pure membrane action controlled by syH. Hence for this
case, RSFL is used. This implies that even if the hot spot is
only slightly larger than the ‘‘local’’ hot spot limit, the
failure is ‘‘as if’’ a ‘‘global’’ hot spot has occurred. It must
be noted that the actual structure would in all likelihood
have other reserve strengths. ‘‘Failure’’ here is simply
indicative of the limit criterion being surpassed. This
discussion is in very good agreement with inelastic finite
element analyses [14].
The remaining strength factor RSFa provides a lower
bound approximation for the component with small local
hot spots where decay lengths from either side overlap
significantly. A linear interpolation between RSFa and
RSFL is suggested for intermediate local hot spots.
The transition angle from small to intermediate hot spots
is suggested as
p given pressure. The RSFs, however, are ratios of these
cL ¼ .
4l
Fig. 4 shows the recommended remaining strength factor
RSFr corresponding to the sizes of hot spots defined by
angle ja. For hot spots smaller than cL, RSFa is
recommended. Hot spots that are larger than cd as stated
in Eq. (11) are considered global hot spots. For them,
RSFL is recommended. For hot spots of intermediate sizes,
RSFa at the angle cL is first calculated and RSFr is
obtained by linear interpolation of RSFa at the angle cL
and RSFL. In summary, the recommended remaining
Hot
spot
size
RSF
RSF
Line
a
varia r
tio n

Shell
strength factor RSFr for spherical pressure vessels can be
calculated as below:
For ja pcL ; RSFr ¼ RSFr ¼ RSFa ;
For cL oja pcd ; RSFr ¼ RSFr ¼ ð1:5  0:637lja Þ
RSFa at cL þ ð0:637lja  0:5ÞRSFL ;
For ja 4cd ; RSFr ¼ RSFL :
(27)
It must be noted that RSFU, RSFa, and RSFL as defined
in Eqs. (21)–(24) are non-dimensional and hence only
depend on the ratios of physical parameters and not on
actual values. This gives rise to some useful observations.
The theory above for the calculation of RSF is very likely
common to all shapes and sizes of pressure vessels provided
that we take the shape into account at proper places. The
shape influences the equivalent primary stress. For
ffiffiffi se in Eq. (20) will use pRi/2h for spheres and
example,
p
3pRi 2h for cylinders. The individual load factors m0u ;
mL, etc., are all inversely proportional to se for any
(26) load factors. Thus, the RSFs are not directly dependent on
the shape-induced primary stress se. This can be easily
verified. Therefore, the RSF for any pressure vessel is
only a function of the geometrical and material property
ratios for a given situation and is independent of the
actual applied load. The shape, of course, influences the
decay length as defined by the angle cd and hence VC. But,
for the same angle cd, the values of RSFU, RSFa, and
RSFL are the same for a sphere and a cylinder. It can be
postulated that for other types of geometries, we can use
the same argument and obtain a simplified Level 2
estimation of RSF. In terms of property ratios, we can
express RSFs as below. Let ffi rV ¼ VH/VC, rs ¼ syH/syC, and
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ra ¼ ð1 þ r2s rV Þ=ð1 þ rV Þ. Then
Z ¼ ra =rs , (28a)
RSFU ¼ ra , (28b)
2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffi pffiffiffi ffi3
2Z2 þ ZðZ  1Þ2 ð1 þ 2  ZÞðZ  1 þ 2Þ
RSFL RSFa ¼ 2ra 4 pffiffiffi pffiffiffi 5,
ðZ2 þ 2  5ÞðZ 2 þ 2 þ 5Þ
(28c)
RSFL ¼ rs . (28d)

L
3.6. Distortion due to bulging

d Deformation due to hot spots could result in bulging of

the component. Bulging is the local inward or outward
deviation from the original spherical geometry. Strain due
to bulging (Fig. 5) and the included angle [7] of the bulge
CL can be shown to be
 
Rb jb  Ro ja Ro sin ja
Fig. 4. Recommended remaining strength factor for various sizes of hot b ¼ and jb ¼ arcsin ; (29)
spots. Ro ja Rb
 ARTICLE IN PRESS
P. Tantichattanont et al. / International Journal of Pressure Vessels and Piping 84 (2007) 762–772 769

Bulge
b

Shell b Rb o
Bulge

Ro
Dmax

a
Do

CL

Fig. 5. Hot spot bulging geometry.

where eb is the maximum membrane strain in the bulge. By
limiting eb to the maximum value permitted for the average
strain through the thickness (as defined in the limit criteria
for the component), we can find Rb and jb from Eq. (29).
Radial displacement of the assumed spherical bulge db is
then given by
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
db ¼ ðRb  do Þ  R2b  ðRo sin ja Þ2
where
do ¼ ð1  cos ja ÞRo .
The out-of-roundness ratio is computed as the ratio of the
maximum diameter of the sphere due to bulging
Dmax( ¼ Do+db), and the original diameter of the sphere
Do, as shown in Fig. 5. If the out-of-roundness ratio
exceeds a certain recommended value, RSFr should be
taken as RSF3. This will be an additional limit criterion for
the shell strength based on excessive deformation. It may
be noted that the calculation above ignores the rotation
restraining effect of the surrounding region and assumes a
hinge support for the hot spot zone. Hence, the bulge
displacements are overestimated.
4. Illustrative example and discussion
The following example is given to demonstrate the
proposed Level 2 integrity assessment method in a sphere
under internal pressure (The values given in parentheses
are in the Imperial system.):
ASTM material SA 516 Grade 55
Shell inside radius (Ri) 0.556 m (21.9 in.)
Operating pressure 1.72 MPa (250 psig)
Design pressure (Pd) 2.59 MPa (375 psig)
Operating temperature 32.2 1C (90 1F)
Design temperature 37.8 1C (100 1F)
Corrosion Allowance (CA) 0.0016 m (1/16 in)
Joint Efficiency (Ej) 1.0
4.1. Required thickness calculations
Design thickness, hd, for the spherical shell can be
determined as [12]
Pd Ri
hd ¼ ¼ 0:0076 m ð0:301 in:Þ:
2SE j  0:2Pd
The required shell thickness is h ¼ hd+CA ¼ 0.0092 m
(0.363 in.).
Therefore, a 0.01 m (3/8 in.) thickness is specified.
ð30Þ The allowable RSF can be expressed as RSF ¼ hd =h ¼
0:802: Note that this RSF* is based on design requirements.
Outside radius of the sphere, Ro ¼ 0.566 m (22.275 in.).
In this case, the R/h ratio is 58.9 and for v ¼ 0.3 using
Eq. (6), l ¼ 9.865. The decay angle of spherical shells is
calculated from Eq. (11), cd ¼ 13.71. The decay length
measured outside of the sphere is Sd ¼ cd  Ro ¼ 0.135 m
(5.320 in.)
The decay length along the meridian
pffiffiffiffiffiffi of cylindrical shells
suggested by Seshadri [7] is 2.5 Rh while the circumfer-
ential decay length is 6.10 (R3h)1/4. For a cylinder with the
same diameter and thickness as the example sphere, the
meridian and circumferential decay lengths are 0.183 m
(7.195 in.) and 1.235 m (48.635 in.), which are much larger
than the spherical decay length. This is in line with the
discussion earlier stating that hot spot effects are much
more localized in spheres with very small decay lengths
compared to those in cylinders.
4.2. Evaluation of RSF
The material properties of carbon steel for a temperature
range 37.8–316 1C (100–600 1F) are listed in Table 2. For
the following demonstration, hot spot temperature is
assumed as 204 1C (400 1F).
The dimension of the hot spot is chosen with ja ¼ 10.01.
From Eq. (26), cL ¼ 4.61. From Eq. (11), cd ¼ 13.71.
Since cLojapcd, this is the case of intermediate hot spot
and RSFr is calculated from the second of Eq. (27).
 ARTICLE IN PRESS
770 P. Tantichattanont et al. / International Journal of Pressure Vessels and Piping 84 (2007) 762–772

Table 2
Material properties for SA 516 Grade 55

Temperature (1C) 37.8 (100 1F) 93.3 (200 1F) 149 (300 1F) 204 (400 1F) 260 (500 1F) 316 (600 1F)

E (  109 Pa) 202.0 198.6 195.1 191.0 188.2 184.1

a (  106 mm/mm/1C) 9.95 10.6 11.3 11.9 12.4 12.9
sy (  108 Pa) 2.07 1.88 1.83 1.77 1.69 1.53

The angle for the reference volume is jR ¼ ja+cd ¼
23.71.
From Eq. (17), hot spot volume (VH) ¼ 2.911  104 m3
(17.761 in.3).
From Eq. (18), adjacent volume (VC) ¼ 13.226 
104 m3 (80.711 in.3).
Kinematically active volume (VR) ¼ VH+VC ¼ 16.137 
104 m3 (98.472 in.3).
From Eq. (19), the effective elastic modulus is
Eeff ¼ 199.9  109 Pa and the effective thermal coefficient
is aeff ¼ 10.3  106 mm/mm/1C.
The thermo-elastic primary stress of the damaged shell is
calculated by using Eq. (20), se ¼ 96.3 MPa. Comparing
with syH, se indicates that there is no general yielding
through membrane action. It must be noted that se need
not be computed to estimate RSF values although it serves
a useful purpose in comparing with syH.
Using the above, mu ¼ syc =se ¼ 2.149 and mL ¼
syh =se ¼ 1.841.
From Eq. (25), RSFL ¼ 0.857.
The RSFU and RSFa calculated based on Tresca and
von Mises yield criteria are as follows:
Tresca criterion:
From Eq. (16), md ¼ 2.094.
From Eq. (22), RSFU ¼ 0.974.
Z ¼ md/mL ¼ 1.137.
From Eq. (23), ma ¼ 2.000.
From Eq. (24), RSFa ¼ 0.931.
Von Mises criterion:
From Eq. (15), md ¼ 2.096.
From Eq. (22), RSFU ¼ 0.976.
Z ¼ md/mL ¼ 1.139.
From Eq. (23), ma ¼ 2.002.
From Eq. (24), RSFa ¼ 0.932.
Similarly, RSF2 calculated using the angle ja ¼ cL is
RSF2 at the angle cL ¼ 0.940.
From the second of Eq. (27), RSFr ¼ 0.890.
If the maximum strain in the bulge eb is 1%, from
Eq. (29), Rb ¼ 0.333 m (13.095 in.) and jb ¼ 17.21.
From Eq. (30), db ¼ 6.25 mm (0.246 in.). This corre-
sponding value from inelastic analysis is 2.30 mm
(0.091 in.).
From Eq. (31), rout-of-roundness ¼ 1.006p1.01.
Therefore, the recommended RSFr ¼ 0.890 is used for
the present example.
To verify the above-recommended RSF value, an
inelastic Finite Element Analysis is performed using
ANSYS [13] to determine the inelastic strength. The sphere
is modelled by shell element SHELL93 using the elastic–
plastic material behaviour with plastic tangent modulus
equal to 1.38  108 Pa (20  103 psi). The limit criterion
is chosen to be 1% von-Mises strain. It must be noted
that it is common in the industry to use 1% [e.g., ASME
BPV code clause T-1310] strain limit for primary mem-
brane stress effects, while higher strain limits are used
for secondary or peak stress effects. Since the present
work calculates RSF based on primary stress effects,
a 1% limit is justifiable. The plastic tangent modulus
used in finite element analyses is small compared with
elastic modulus of 202.0  109 Pa (29.3  106 psi) so that the
stress increase beyond yield due to 1% strain is small
compared to yield. Justification for the applicability of the
method for steels with larger tangent modulus is provided
in [14]. The inelastic remaining strength factor RSFi is
calculated from the ratio of the internal pressure that
causes a maximum of 1% average strain (in the hot spot)
and the burst pressure of undamaged sphere (not design
pressure). For the present data RSFi is 0.915. It can be seen
that RSF*oRSFroRSFi. Thus, the recommended RSF
based on the current procedure is acceptable and con-
servative.
The same sphere is further studied for different hot spot
sizes and temperatures for evaluating the effectiveness of
the proposed Level 2 method. The RSFs for the spherical
shell with hot spot temperatures 93.3, 204 and 316 1C using
von-Mises criterion are as shown in Tables 3, 4 and 5,
respectively.
Figs. 6 and 7 compare the recommended RSFr and the
inelastic RSFi obtained from FEA of similar spherical
shells (with inner radius 0.556 m) with R/h ratios equal to
58.9 and 20, respectively. The RSFs are plotted against the
ratio of the hot spot area and the adjacent area. For all
cases of hot spot temperatures, RSFr is shown to offer
good lower bound approximation. For spherical shells
containing hot spots with large included angles, the RSFs
obtained from finite element analyses are shown to be
constant and are equal to RSFL indicating the dominance
of primary membrane effects as discussed earlier. The
‘‘local’’ hotspot limit defined by the angle (cd) as indicated
by the beginning of the horizontal part of the RSF line is
clearly on the conservative side. For very small hotspots,
the failure is less influenced by the bending decay compared
to the stretching action. The stretching decay lengths are
likely to be larger than those of bending action thus
indicating a more diffused effect resulting in higher RSF.
In that sense, for very small hotspots the RSF as computed
above will be conservative. This is also shown by the results
of FEA.
 ARTICLE IN PRESS
P. Tantichattanont et al. / International Journal of Pressure Vessels and Piping 84 (2007) 762–772 771

Table 3 1.000
RSF for a spherical shell with a 93.3 1C (200 1F) hot spot (R/h ¼ 58.9)
0.950
Case no. ja RSFU RSFa RSFL RSFr RSFinelastic RSFi for 93.3°C

1 51 0.994 0.965 0.910 0.963 0.983 0.900

2 71 0.990 0.963 0.910 0.951 0.971 RSFi for 204°C

RSFs
3 101 0.984 0.959 0.910 0.932 0.951 0.850
4 151 0.976 0.954 0.910 0.910 0.927
5 201 0.969 0.949 0.910 0.910 0.915
0.800

RSFi for 316°C

0.750
RSFi for 204°C
RSFi for 316°C
Table 4 RSFi for 93.3°C
RSF for a spherical shell with a 204 1C (400 1F) hot spot (R/h ¼ 58.9) 0.700
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60
Case no. ja RSFU RSFa RSFL RSFr RSFinelastic Ratio of the hotspot area and the adjacent area

1 51 0.990 0.940 0.857 0.936 0.950 Fig. 7. Comparison of RSFr and RSFi for a spherical shell with R/h
2 71 0.985 0.936 0.857 0.918 0.947 ratio ¼ 20.
3 101 0.976 0.932 0.857 0.890 0.915
4 151 0.962 0.924 0.857 0.857 0.878
5 201 0.951 0.918 0.857 0.857 0.863 RSFU, RSFa and RSFL for cylinders will be similar to
those of the sphere (e.g., Figs. 6 and 7) thus indicating their
shape independence properties.

Table 5 5. Conclusions
RSF for a spherical shell with a 316 1C (600 1F) hot spot (R/h ¼ 58.9)

Case no. ja RSFU RSFa RSFL RSFr RSFinelastic

A methodology for Level 2 FFS assessment of spherical
pressure vessels containing hot spots is presented. A decay
1 51 0.984 0.868 0.740 0.863 0.931 angle for spherical shells based on edge effects is proposed.
2 71 0.974 0.865 0.740 0.835 0.886 A reference volume is used in conjunction with a
3 101 0.958 0.859 0.740 0.793 0.828
variational formulation to assess limit strength. The bulge
4 151 0.935 0.850 0.740 0.740 0.764
5 201 0.915 0.841 0.740 0.740 0.747 membrane strain in the post yield zone is limited to 1%. A
RSF depending on the hot spot size is recommended. The
RSF is based on the ma multiplier and integral mean of
yield criterion for highly localized hot spots. For large
1.000 hotspots, it is based on the lower bound multiplier. For
intermediate size hot spots, a linear transition between the
0.950 two is shown to work well. The recommended RSF is
verified to provide good lower bound estimation compared
0.900 RSFi for 93.3°C
to inelastic FEA results. The results are especially close
when the temperature difference between hot and cold
RSFs

0.850 RSFi for 204°C regions is not high. Useful observations about the shape
independence of RSF are presented.
0.800

RSFi for 316°C Acknowledgements

0.750
RSFi for 204°C
RSFi for 316°C
RSFi for 93.3°C
0.700
The financial support provided by the Terra Nova
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 Owner Group (A Petro-Canada operated project), Natural
Ratio of the hot spot area and the adjacent area Sciences and Engineering Research Council of Canada,
and the Canada Research Chair program, which made the
Fig. 6. Comparison of RSFr and RSFi for a spherical shell with R/h
ratio ¼ 58.9.
research possible is gratefully acknowledged.

References
It may be noted that an improved estimate of the
mobilized reference volume might yield a better prediction. [1] R6. Assessing the integrity of structures containing defects. British
However, the aim of the present study is to give the Energy, Gloucester, UK,1995.
[2] API. Fitness-for-service, API 579. Washington, DC: American
practicing engineers a simple yet effective Level 2 method Petroleum Institute; 2000.
of obtaining the RSF without recourse to a detailed study. [3] Sims JR, Hantz BF, Kuehn KE. A basis for the fitness for service
It should also be noted that non-dimensional plots of evaluation of thin areas in pressure vessels and storage tanks.
 ARTICLE IN PRESS
772 P. Tantichattanont et al. / International Journal of Pressure Vessels and Piping 84 (2007) 762–772

Pressure vessel fracture, fatigue and life management. ASME PVP
1992;233:51–8.
[4] Mura T, Lee SL. Application of variational principles to limit
analysis. Q Appl Math 1963;21(3):243–348.
[5] Mura T, Rimawi WH, Lee SL. Extended theorems of limit analysis.
Q Appl Math 1965;23:171–9.
[6] Seshadri R, Mangalaramanan SP. Lower bound limit loads using
variational concepts: the ma-method. Int J Press Vessels Piping 1997;
71:93–106.
[7] Seshadri R. Integrity assessment of pressure components with local hot
spots. ASME Press Vessels Piping Div Publ PVP 2004;480:177–85.
[8] Indermohan H, Seshadri R. Fitness-for-service methodology based
on variational principles in plasticity. ASME Press Vessels Piping Div
Publ PVP 2004;473:83–91.
[9] Lukasiewicz S. Local loads in plates and shells. Warszawa: PWN-
Polish Scientific Publishers; 1979.
[10] Timoshenko S. Theory of plates and shells. 2nd ed. New York:
McGraw-Hill Inc; 1970.
[11] Pan L, Seshadri R. Limit load estimation using plastic flow parameter
in repeated elastic finite element analyses. Trans ASME J Press Vessel
Technol 2002;124:433–9.
[12] Bednar HH. Pressure vessel design handbook. 2nd ed. New York:
Van Nostrand Reinhold Company; 1986.
[13] ANSYS. University research version, 8.1. SASIP, Inc., 2004.
[14] Tantichattanont P. Fitness-for-service assessment for thermal hot
spots and corrosion in pressure vessels. PhD thesis, Memorial
University, St. John’s, NL, Canada, 2006.