External ange loads and `Koves'-method

C.J. Dekker
*
, H.J. Brink
Continental Engineering B.V., Joan Muyskenweg 22, 1096 CJ Amsterdam, The Netherlands
Received 4 June 2001; revised 23 October 2001; accepted 23 October 2001
Abstract
Practical implementation of a much more realistic approach for treating external moments for various major ange assessment methods is
proposed. First, the question of how to deal with torsion in ange rings weakened by bolt holes is addressed. This aspect was not covered by
the original paper of Mr Koves on which our work is based. That our practical implementations result in quite accurate results is shown by an
FEM-analysis of a loose ange ring.
Correctly dealing with external moments, usually the major part of external loads on anges, is important because as soon as design
pressures are known, ange ratings are selected and initial designs for `custom' anges are made. Only in the later stages of a project do all
loads on anges become available and then these external ange loads have to be approved. Usually, the margins between internal pressures
and maximum allowable pressures are not large and hence one can ill-afford the wasteful ways that ange assessment methods presently treat
external loads. q 2002 Elsevier Science Ltd. All rights reserved.
Keywords: Flanges and external loads; Realistic method for external moments implemented in assessment codes; ASME; AD-Merkblatter; Stoomwezen; EN
13445-3 and EN 1591; FEM-analysis proves its accuracy
1. Introduction
1.1. General
Flanges experience two kinds of loads: (internal) pressure
and external loads stemming from the attached piping. At
best, external loads are mentioned in passing but elsewhere
nothing is said about such loads in internationally
recognized codes and/or standards.
The identical ange assessment methods in both divi-
sions of the ASME Boiler and Pressure Vessel Code
Section VIII [1] (in Div. 1: Appendix 2, and in Div. 2:
Appendix 4) do not mention external loads on the ange
connection at all. Note that these are based upon the
method developed by Waters et al. [8].
In the ASME B31.1 Power Piping [2] the existence of
external ange loads is mentioned only in connection
with anges on the safety valve inlet and outlet, refer
to par. 4.2.3-`Analysis of Flange' in the non-mandatory
Appendix II. It amounts to converting the external loads
into an equivalent pressure which is to be added to the
internal pressure. Either this augmented pressure should
not exceed the rated pressure of standard anges or an
ASME VIII Div. 1 assessment shall be carried out
whereby the ange stresses S
H
(though increased with
the longitudinal pressure stress), S
R
and S
T
, respectively,
should not exceed the yield stress at design temperature.
In Germany the AD-Merkblatter [3] (i.e. Blatt B 8 in
conjunction with Blatt B 7) do not mention external
ange loads whereas the `preliminary' DIN 2505 [4]
(dating from April 1990) includes the axial pressure
trust term with the external loads into a force F
R
stem-
ming from the connected pipe. The external moment is
transformed into a force that gives the same (maximum)
stress in the connecting pipe.
F
R
F
Rp
1F
ex
1
4M
ex
d
i
1S
R
_
_
_
_
_
_
_
_
or in ASME-notation
: `total` H
D
H
D
1F
ex
1
4M
ex
B 1g
0
_
_
_
_
_
_
_
_
The Dutch `Rules for Pressure Vessels' [5] requires a
similar approach in sheet `D 0701': in the case of external
ange loads the term representing the pressure thrust load
from the connecting pipe is to be increased by F
ex
1
4M
ex
=D
g
with D
g
the diameter of the effective sealing
diameter (in ASME notation: G).
In the forthcoming EN 13445-3`Unred Pressure
International Journal of Pressure Vessels and Piping 79 (2002) 145155
0308-0161/02/$ - see front matter q 2002 Elsevier Science Ltd. All rights reserved.
PII: S0308-0161(01)00136-3
www.elsevier.com/locate/ijpvp
* Corresponding author. Tel.: 131-020-668-31-41; fax: 131-020-668-
10-98.
E-mail address: [email protected] (C.J. Dekker).
VesselsPart 3: Design' [6] two ange assessment
methods will be included: a main-stream method in
chapter 11 and an alternative assessment method in its
appendices. The main-stream method is based on the
method as developed by Waters et al. [8] (the `ASME'-
method) and just like Appendix 2 of ASME VIII Div. 1
(or Appendix 4 of ASME VIII Div. 2) has no provisions
to include external ange loads. The rather interesting
alternative method, also known as prEN 1591 (pre-
liminary edition of EN 1591), requires that external
ange loads be included (see its clause 4.1.2.2`Addi-
tional external loads') as follows: F
Ri
F
ex
^
4=d
3e
M
ex
with d
3e
being the diameter of the effective
bolt circle.
Remarks: The effective diameter is the mean diameter of
the inscribed circle and the circumscribed circle of the
polygon formed by the C.L.'s of the bolts. For eight bolts
d
3e
< 0:97B
bolt circle
and for 12 bolts d
3e
< 0:99B
bolt circle
:
As EN 1591 is intended for designing `lean' (i.e. just strong
enough) anges, both the maximum as well as the minimum
additional load is to be taken into account. The tensile side
of the moment might determine the maximum load on the
bolts and the compressive side could well be determining
the maximum load on the gasket. That is the reason for the
` 1' sign as well as the ` 2' sign.
Codes and/or standards seem either to ignore the
existence of external loads (though they require account to
be taken of all loads on the pressure equipment considered
or part thereof!) or convert the essential non-uniform
external moment loading on a ange into a uniform axial
load that can be dealt with in a more or less similar way to
the pressure thrust load. In the case of Appendix II of B31.1
this is done in the guise of an equivalent pressure. The
diameter involved in this moment conversion is either the
average diameter of the pipe connected to the ange (as in
Fig. 1) or, as a concession to the conservative nature of this
conversion, the effective diameter of the gasket or even the
(effective) bolt circle.
C.J. Dekker, H.J. Brink / International Journal of Pressure Vessels and Piping 79 (2002) 145155 146
Nomenclature
Remarks on nomenclature
(i) In discussion of various codes we tried to adhere to the considered code's nomenclature with the exception
of F
ex
and M
ex
which we used throughout for the axial force component and bending moment component,
respectively, of the external load.
(ii) A cylindrical coordinate system is used for the ange ring with r, the radial coordinate, u, the tangential
angle coordinate and z, the axial coordinate (z-axis coincides with the ange's C.L.).
(iii) Symbols not mentioned here are either dened in the text or one has to refer to the relevant code.
M
ex
external bending moment on a ange connection
F
ex
external axial force on a ange connection
H
D
hydrostatic end force on area inside of ange, ASME-notation
H
G
gasket load, ASME-notation
A outside diameter of ange, ASME-notation
B inside diameter of ange, ASME-notation
C bolt circle diameter, ASME-notation
G diameter of the location of the gasket load, ASME-notation
g
0
thickness of hub at small end, ASME-notation
h
D
lever arm for H
D
(radial distance bolt circlecircle on which H
D
acts), ASME-notation
D
hd
diameter of the circle on which H
D
acts, ASME inspired notation
d
i
inside ange diameter, DIN 2505 notation
S
R
thickness of the pipe connected to the ange (compare with ASME's g
0
), DIN 2505 notation
F
Rp
hydrostatic end force stemming from connected pipe (ASME's H
D
), DIN 2505 notation
F
R
force from pipe on ange due to pressure and external load, DIN 2505 notation
D
g
diameter of the effective sealing diameter of the gasket (ASME's G), D 0701 notation
F
Ri
pipe force on ange due to external load in assessment condition i, prEN 1591 notation
Fig. 1. Conversion of external moment into a uniform load.
1.2. The `Koves'-method
When one considers a ange ring with an external
moment in isolation (using the superposition principle for
linear-elastic systems), then the ange rotation at the
moment's tensile side must be opposite to the rotation at
the compressive side of the moment, see Fig. 2 with the
cross section through the plane of symmetry of the ange
ring with external moment. Hence, these opposing enforced
rotations in the ange ring are checked somehow by the
torsional stiffness of the ange ring. Ignoring this torsional
stiffness by making uniform the external moment's
maximum tensile load, as is done in some of the above
codes, must therefore result in overestimating the additional
stresses due to external moment loads on anges!
Koves derives in his article `Analysis of Flange Joints
under External Loads' [7] a relation between the internal
radial bending moment in the ange ring due to an
uniformly distributed twisting moment and due to a har-
monically distributed twisting moment having the same
amplitude. His factor is in effect the overestimation in
ange-stresses due to assuming uniforming stresses from
external moments. We include here his derivation though
with some minor alterations, complete with his redrawn
sketches, Figs. 35.
An external moment M
ex
is applied on the ange and is
assumed to create a linear stress distribution in the ange
neck. This moment is reduced to a linear distributed load
H
H
D
which is reacted by a bolt load F
H
B
and a gasket load F
H
G
(both are also linear distributed), see Fig. 3.
The ange system is approximated by a ring model with
an externally applied twisting moment M
u
and an external
load F
u
. The `3 load'-system H
H
D
, F
H
G
and F
H
B
is replaced by
an equivalent system consisting of the distributed moment
M
u
and distributed load F
u
. Both act on a circle with the
same radius as the center of twist of the ring's cross-section,
i.e. r
cot
, see Fig. 4.
F
u
F
0
cos u; F
0
is the amplitude of the distributed load;
M
u
M
0
cos u; M
0
is the amplitude of the
distributed moment:
As the reaction loads F
H
G
and F
H
B
are in equilibrium with
the load H
H
D
(the latter resulted from the external moment),
the equivalent load system must also be in equilibrium. The
resulting moment about the r-axis at u 908 from the
equivalent load system has to be zero:
_p
0
F
u
r
cot
cos ur
cot
du 2
_p
0
M
u
cos ur
cot
du
F
0
r
2
cot
_p
0
cos
2
u du 2M
0
r
cot
_p
0
cos
2
u du 0
The above equation yields the result: M
0
r
cot
F
0
(which
implies M
u
r
cot
F
u
)
The internal forces and moments in the ange ring are
functions of the external equivalent load system F
u
and M
u
.
The only non-zero boundary load component at u 08 (the
plane of symmetry) is the radial moment m
R0
. Due to either
symmetry or the nature of the equivalent external load, all
other boundary load components have to be zero.
The internal moments and forces at angle u must be in
equilibrium with the boundary load at u 08 and the exter-
nal equivalent loads of M
u
and F
u
as exercised over the ring
segment from 08 to u, Fig. 5.
The axial downwards directed internal shear force:
V
u

_u
0
F
0
cos ar
cot
da F
0
r
cot
sin u
The radial (inward) directed internal moment (while using
M
0
F
0
r
cot
):
m
u
m
R0
cos u 1
_u
0
M
0
cos a sinu 2ar
cot
da
2
_u
0
F
0
cos ar
cot
sinu 2ar
cot
da m
R0
cos u
C.J. Dekker, H.J. Brink / International Journal of Pressure Vessels and Piping 79 (2002) 145155 147
Fig. 2. Torsion in ange ring due to opposing (local) ring rotations.
Fig. 3. Flange loading H
H
D
(due to M
ex
) and reaction loads F
H
G
and F
H
B
:
The positive directed internal tangential moment (torsional
moment):
T
u
m
R0
sin u 2
_u
0
M
0
cos a cosu 2ar
cot
da
2
_u
0
F
0
cos ar
cot
{1 2cosu 2a}r
cot
da
m
R0
sin u 2
_u
0
F
0
cos ar
2
cot
da
m
R0
sin u 2F
0
r
2
cot
sin u
Now the variation principle of Castigliano's theorem will be
applied to the ange ring. If the strain energy of an elastic
system is expressed in terms of truly independent external
forces (generalized) then the partial derivative of the strain
energy with respect to any external force is equal to the
actual displacement of that force's point of application in
the direction of that force. The truly independent external
forces are the equivalent external load F
u
(note that M
u
is
not independent as M
u
r
cot
F
u
) and the non-zero boundary
load m
R0
.
The strain energy of the ange ring is:
U
_p
0
m
2
u
2EI
1
T
2
u
2GJ
1k
V
2
u
2GA
_ _
r
cot
du
with I, the moment of inertia about the radial axis of the
ange's cross-section, J, the polar moment of inertia of
the ange's cross-section, A, the cross-sectional area of
the ange ring, k, the shape factor of the ring's cross-section
for shear in the z-direction, E, the modulus of elasticity of
the ange ring and G, the shear modulus, G E=21 1n
The rotations at u 0; the point where the external
generalized force m
R0
acts, and also at u p where an
equally large but opposite signed boundary moment acts,
are due to symmetry constraints equal to zero:
2
2m
R0
U
_p
0
m
u
EI
2m
u
2m
R0
1
T
u
GJ
2T
u
2m
R0
1k
V
u
GA
2V
u
2m
R0
_ _
r
cot
du 0
Substituting the following partial derivatives together with
the expressions found earlier for the internal force and
moments into the above integral expression gives:
the partial derivatives:
2m
u
2m
R0

2
2m
R0
m
R0
cos u cos u
2T
u
2m
R0

2
2m
R0
m
R0
sin u 2F
0
r
2
cot
sin u
_ _
sin u
2V
u
2m
R0

2
2m
R0
F
0
r
cot
sin u 0
and then the integral equation results in:
_p
0
m
R0
cos u
EI
cos u 1
m
R0
sin u 2F
0
r
2
cot
sin u
GJ
sin u
_ _
r
cot
du

_p
0
m
R0
cos
2
u
EI
1
sin
2
u
GJ
_ _
2F
0
r
2
cot
sin
2
u
GJ
_ _
r
cot
du
m
R0
1
EI
1
1
GJ
_ _
2
F
0
r
2
cot
GJ
_ _
pr
cot
2
0
And with F
0
r
cot
M
0
one nds:
m
R0

M
0
r
cot
1 1
GJ
EI
_ _
For harmonically distributed twisting moments the inter-
nal radial directed bending moment is a factor 1 1GJ=EI
smaller than in the case of uniformly distributed twisting
moments.
C.J. Dekker, H.J. Brink / International Journal of Pressure Vessels and Piping 79 (2002) 145155 148
Fig. 4. Equivalent load system on the ange ring.
Fig. 5. Internal forces and moments in the ange ring at angle u.
1.3. Our interpretation of the equivalent load
The remaining thing to do is to nd a relation between the
equivalent load and the load H
D
H
together with the reaction
loads F
G
H
and F
B
H
on the ange. Limiting the occurrence of
an external load to the `operating condition' of the ange
and observing that:
ange assessments in the operating condition are
performed with the minimum required ange loads for
all standards and/or codes as mentioned in the rst
paragraph,
the load in the operating condition consists of the internal
pressure and the super-positioned external load which
consists of the line loads H
D
H
(u), F
G
H
(u) and F
B
H
(u) on
the ange and has no internal pressure component,
we assert that F
H
G
u 0 because the minimum gasket
load in operating condition, required to maintain the
pressure sealing of the anges, is not affected as the internal
pressure remains the same under the external ange load.
Hence, no extra gasket line load in operating condition is
required.
Remark on interpreting F
H
G
u 0 in a physical sense: In
almost all instances, the initial bolt force (i.e. gasket seating
load) is too large and during the rst few operating load
cycles yielding will occur in order to get rid of the excess
bolt force. However, yielding should never go beyond the
(minimum) required operating loads augmented with line
loads H
D
H
(u) and F
B
H
(u) stemming from the external load.
That any yielding will stop is assured by the proposed
adaptations to the operating ange moments in the various
ange assessment methods.
Let us consider now a small segment with the external
line loads H
D
H
(u) and F
B
H
(u) and their equivalent loads F
u
and M
u
. Being equivalent, both load systems must be in
equilibrium with each other for each and every innitesimal
segment of the ring, Fig. 6.
H
H
D
u hd cos u; with `hd' the amplitude
dimension : force per length unit and H
H
D
u
acting on a circle with diameter D
hd
;
F
H
G
u fg cos u; with `fg' its amplitude and acting
on a circle with diameter G;
F
H
B
u fb cos u; with `fb' its amplitude and acting
on a circle with diameter C:
Force equilibrium in the z-direction:
hd cos u
D
hd
2
du 2fb cos u
C
2
du F
0
cos ur
cot
du
Moment equilibrium about the u-axis:
hd cos u
D
hd
2
du r
cot
2
D
hd
2
_ _
2fb cos u
C
2
du
C
2
2r
cot
_ _
M
0
cos ur
cot
du
These two equilibrium relations simplify to:
hd
D
hd
2
2fb
C
2
F
0
r
cot
M
0
and
hd
D
hd
2
_
r
cot
2
D
hd
2
_
1fb
C
2
_
C
2
2r
cot
_
M
0
r
cot
From which the following results can be obtained:
M
0
hd
D
hd
C
C
2
2
D
hd
2
_ _
hd
D
hd
h
D
C

4M
ex
pD
2
hd
D
hd
h
D
C
C.J. Dekker, H.J. Brink / International Journal of Pressure Vessels and Piping 79 (2002) 145155 149
Fig. 6. Segmental equilibrium of equivalent load and H
H
D
u plus F
H
B
u on ange ring.
and
fb hd
D
hd
C
1 2
2h
D
C
_ _
hd
D
hd
C
_ _
2

4M
ex
pD
2
hd
D
hd
C
_ _
2
The amplitude of the harmonic twisting moment M
u
on the
ange ring is now expressed as a function of the external
moment M
ex
. Note that making this harmonic twisting
moment a uniform load would still result in correct
tangential bending stresses in the ange ring if the
previously found Koves factor is applied.
The locally maximum additional stresses in the bolts can
be found by multiplying the amplitude of F
H
B
(u) with the
bolt pitch measured along the bolt circle and dividing this by
the individual bolt stress area:
s
bolt
additional due to M
ex

4M
ex
n
b
A
bolt
C
.
2. Verication of tangential bending stresses
In the ange ring's expression of the strain energy the
contribution due to the torsional moment T
u
is given as the
integral of T
2
u
=GJ along its curved length. Note that
the cross-section of the ange ring is not circular but rectan-
gular discounting any conical hub. For non-circular straight
bars the torsional stiffness should have been used in the
strain energy's expression and not, as is done for the curved
ange ring, the polar moment of inertia.
However, using torsional stiffness implies that torsion
associated warping develops fully in the ange ring. For
realistically sized anges the mean ange ring radius
(more precisely: r
cot
) ranges from about three to six times
the ring's width. The strong curvature of such ange rings,
the `contra'-warping of both symmetrical halves of the
ange ring (at the plane of symmetry warping is zero, i.e.
all tangential displacements at u 0 and at u p are
suppressed) and that over an arc of only 1=2p radians
(length about four to nine times the ring's width) the
torsional moment and associated warping goes from zero
to its extreme value, makes it dubious that warping develops
fully. However, dismissing it altogether is another matter.
To investigate whether warping plays an important role or
is negligible, we performed two FEM-analyses. Our model
consisted of a 1808 segment of a ring with a mean radius of
100 mm and a square cross-section with 20 mm sides.
Equally spaced 20 noded brick elements were used to
build the mesh, four elements along the square sides and
12 elements along the length of the segment: in total 4
2

12 192 quadratic brick elements. The z-displacements of

the half circular shaped central row of nodes on the bottom
surface were suppressed as well as the tangential displace-
ments of all nodes in the r, z planes at u 0 and at u p:
Distributed twisting moments were realized by introducing
equally large traction and pressure on the inside faces of the
innermost top row of elements and the innermost bottom
row of elements, respectively, see Fig. 7.
2.1. Uniform pressure and traction along the circumference
The twisting moment per unit length at the `centre-of-
twist' circumference is:
M
u

1
4
ap
3
4
a
r
inside
r
cot

27
32
a
3
p
r
cot
with a the square side of the cross-section, 20 mm, p, the
pressure as well as traction on the inside faces of the upper
and lower row of innermost elements, p 5 MPa, r
inside
, the
inside radius of the ring, r
inside
90 mm 4
1
2
a
The section modulus of the square section:
W 1=6a
3
The theoretical tangential bending stress:
s
u

M
u
r
cot
W

6 27
32
p 25:3 MPa
Due to the fairly large width (20 mm) with respect to the
mean radius (i.e. centre-of-twist), the bending stresses as
found from the F.E. analysis are neither constant at the
top surface of the ring nor at the bottom surface of the
ring. Averaging the tangential bending stress results in
24.8 MPa which compares well with the theoretical tangen-
tial bending stress of 25.3 MPa.
2.2. `Cosinusoidal' distributed pressure and traction along
the circumference
If the amplitude of the cosinusoidal distributed twisting
moment is equal to the uniformly distributed twisting
moment, then the maximum tangential bending stress
would be smaller with the Koves factor.
K
v
1 1
GJ
EI
1 1
I
z
1I
r
21 1nI
r
1 1
2I
r
21 10:3I
r
1:769
C.J. Dekker, H.J. Brink / International Journal of Pressure Vessels and Piping 79 (2002) 145155 150
Fig. 7. Mesh of the 1808 segment of the ring for verication purposes.
Hence, the theoretical tangential bending stress:
s
u

M
u
r
cot
WK
v

25:3
1:769
14:3 MPa
If we were to replace the polar moment of inertia by the
rotational stiffness K < 0:1406a
4
; then the reduction
factor would be 1.649. And for the tangential bending stress
one would arrive at 15.3 MPa.
The F.E. analysis conrms that the largest tangential
bending stresses in the ring occur at the plane of symmetry,
i.e. at u 0 and at u p: As before, these bending stresses
are not constant but show, or so it appears, an inverse rela-
tion with the r-coordinate. Averaging gives a value of
15.0 MPa for the tangential bending stress which should
be compared with the theoretical value of 14.3 MPa
(assumption: no warping) and with 15.3 MPa if warping
were to develop fully.
Note that the harmonic distributed twisting moment M
u
has a sibling in the form of a harmonic distributed force F
u
(with F
u
M
u
=r
cot
). Rigid z-supports at the bottom row of
nodes having a radius equal to r
cot
, bring about automati-
cally that distributed force.
3. Practical implementation of the Koves-factor
3.1. Accounting for warping
Though the verication by F.E. analyses indicates that
warping is certainly not fully suppressed, not only the
value of the tangential bending stresses testify to that but
also the found locations of the maximum surface shear stres-
ses of the ring at the midsides rather than at the corners of
the ring's cross section at u 1=2p; we will nevertheless
adopt the Koves factor with J (second moment of inertia)
instead of K (torsional stiffness factor). Reasons for doing so
are: (i) most ange rings are rmly connected with hubs to
tubes and these connections oppose locally the angular
deformations in the ange ring associated with warping
and so suppress warping a little more, and (ii) more impor-
tantly, the strength and stiffness reducing inuence of the
bolt holes drilled into the ange rings will be incorporated in
a thoroughly conservative manner.
In the German codes AD-Merkblatt B-7 and DIN 2505, in
the Dutch `Rules for Pressure Vessels' sheet D0701 and also
in the European prEN 1591 the strength and stiffness reduc-
ing effect of holes drilled into the ange ring is approxi-
mated by subtracting an effective bolt hole diameter from
the actual width of the ange. The so obtained `effective
ange width' is used there for the stiffness and for the
section modulus. This same effective width will be used
here too for the second moment of inertia about the z-axis.
Though the bolt holes are close to the vertical principal axis
and hence have a minor effect on I
z
, we will use the effective
ange width as `height' in the I
z
-formulae. In this way, the
effect of the (reduced) bolt holes on I
z
is maximized as if
they were situated at the very rim of the ange, see Fig. 8.
I
r

1
12
effective flange widthflange thickness
3
I
z

1
12
flange thicknesseffective flange width
3
And with n 0:3 one arrives at:
K
v
1 1
effective flange width
2
1flange thickness
2
2:6flange thickness
2
The effective or reduced bolt hole is given in both German
and Dutch codes as:
reduced B
bolt hole
B
bolt hole
max1 2
ID
flange
1000
; 0:5
with ID
flange
in mm
The effective ange width is the ange width minus the
above reduced bolt hole diameter.
The prEN 1591 gives a different formula but in practice
that works out to be quite similar.
3.2. Adaptation of various ange assessment methods
Adapting codes for external loads while using the Koves
factor, requires two things:
1. to make `uniform' the distributed twisting moment M
u
due to the external moment M
ex
for inclusion in the oper-
ating ange moment in such a way that the correct
tangential bending stresses will result from it, i.e. apply-
ing the Koves-factor.
2. to account for the slightly higher demands on the bolts in
the operating condition.
For various codes and standards the required formula
changes to include the Koves-factor for external moments
are discussed below, though we limit ourselves here to
integral anges. These changes/additions are given within
accolades. Note that external axial forces are uniform and
are dealt with like hydrostatic end forces from inside ange
areas. In all cases, the nomenclature of the considered code
is adhered to except that everywhere M
ex
is the external
C.J. Dekker, H.J. Brink / International Journal of Pressure Vessels and Piping 79 (2002) 145155 151
Fig. 8. Cross hatched area upon which the above Koves-factor is assumed to
be based.
bending moment and F
ex
is the external axial force on the
ange.
3.2.1. ASME ange assessment (and methods based on it)
The total ange moment M
o
is the sum of the various
components of the twisting moment integrated all around
the ange ring. The amplitude of the M
ex
-component of the
twisting moment integrated along its circle of application
(radius is r
cot
) is:
operating M
o
due to M
ex

4M
ex
pD
2
hd
D
hd
h
D
C
2pr
cot

4M
ex
D
hd
2r
cot
C
h
D
Hence, assuming r
cot
A 1B=4 the following adaptation
of the H
D
-formula results in the same M
ex
-contribution
towards the total ange moment but for the addition of
Koves:
H
D
0:785B
2
P 1 F
ex
1
4M
ex
D
hd
A 1B=2
CK
v
_ _
To assure enough bolt area A
m
is available in the operating
condition to absorb the additional bolt stress:
W
m1
H 1H
p
1 F
ex
1
4M
ex
C
_ _
3.2.2. DIN 2505
The ange moment in the operating condition is M
B
(in
German: auere Stulpmoment which more or less translates
as external inside out turning moment) and is in effect
identical to ASME's total ange moment M
o
. To include
the benets of the Koves-factor this formula which has
equation number 11 in par. 4.5 of DIN 2505, is to be adapted
for the operating condition in a similar way as ASME's
H
D
-formula:
M
B
F
Rp
a
r
1F
F
a
F
1F
DB
a
D
1F
ex
1
4M
ex
d
i
1s
R

d
a
1d
i
=2
d
t
K
v
_ _
a
R
As DIN 2505 already provides for external loads (though
not for the Koves-factor) no further changes are necessary.
3.2.3. D 0701 from the Dutch rules for pressure vessels
The radial bending moment as dened in par. 3.1.1. of
D 0701 is equal to ASME's total ange moment M
o
divided
by 2p: Such a moment can be interpreted physically as the
radially directed moment that causes the tangential bending
stresses in the ange ring. Including the Koves-factor
requires the following adaptations:
To obtain correct estimates of the additional bolt stresses
due to the external moment loading, the increase for force F
1
is to be calculated with the diameter of the bolt circle D
f1
instead of the effective gasket diameter D
g
, i.e. F
1
is to be
increased with:
F
ex
1
4M
ex
D
f1
_ _
3.2.4. prEN 1591-1 `calculation method'
In this code the force F
Ri
is the resultant from the external
load in the considered assessment condition (with index i)
whereby d
3e
is the diameter of the equivalent bolt circle and
F
Ai
and M
Ai
are the external force and moment, respectively.
F
Ri
F
Ai
^
4M
Ai
d
3e
This formula will indeed lead to the maximum and mini-
mum additional loads on both the bolts as well as on the
gasket. Within the context of prEN1591 this is important for
establishing that the actual realized bolt force neither over-
loads the bolts nor the gasket though it assures tightness of
the gasket for all assessment conditions which specically
can include more than one set of operating conditions.
In deriving the equivalent load (see `our interpretation of
the equivalent load') we have stated that F
H
G
u 0; i.e. the
additional distributed load on the gasket due to the external
load is zero. As in reality, the system of anges, bolts and
gasket is a linear elastic system, or so it is approximated in
prEN 1591, this assertion is strictly spoken not true.
However, if the stress situation results in (some) plastic
ange deformation then such a situation with F
H
G
u 0;
would be approached in the allowable limit case. If one
were to have only a single signicant operating load condi-
tion and hence one need not fear for possible repeated
plastic deformations going from one operating load condi-
tion to another operating load condition, then applying the
Koves factor to the equivalent harmonic load due to the
external moment would be fully justied.
The Koves factor is best introduced in the so-called load
ratio for the ange:
F
F

F
G
h
G
1F
Q
h
H
2h
p
1 F
ex
1
4M
ex
d
E
d
4
1d
0
=2
d
3e
K
v
_ _
h
H
_
_
_
_
_
_
_
_
W
F
# 1
with here in the case of prEN 1591:
K
v
1 1
b
2
F
1e
2
F
2:6e
2
F
:
C.J. Dekker, H.J. Brink / International Journal of Pressure Vessels and Piping 79 (2002) 145155 152
M
p
4
D
2
i
p
d
1F
ex
1
4M
ex
D
i
1d
f1

D
i
1D
f

2D
f1
K
v
_ _
D
f1
2D
i
2d
f1
1F
2
D
f1
2
D
i
1D
g
2
_ _
1F
3
D
f1
2D
g

4p
4. Verication of ange ring with bolt holes
4.1. Model of ange ring
To show the conservative nature of the Koves-factor with
the polar moment of inertia while using the effective ange
width, we performed a nal FEM-analysis. In order not to be
muddled by the unknown restraining inuence of the hub
and connecting tube on any warping, we selected the unsplit
DN 200 (8
HH
nominal diameter) loose ange from the
German standard DIN 2642.
This ange has eight bolts and in our FEM-model the bolt
holes straddle the plane of symmetry. We used parabolic
elements, mostly 20 noded bricks (so-called hex-elements)
and a few 15 noded wedges (cpenta-elements), in total 1576
elements for the 1808-model. Basically, the ange ring
consists of four equally thick layers of elements each with
seven elements across the ring's width. The undersides of
these loose anges press on rather thin lapped anges which
have a gasket in between them. This is simulated by an extra
ring of elements from the chamfer at the ange's inside to
about the outside diameter of the lapped anges. Anticipat-
ing some chamfer on the lapped ange's outside corner too,
this latter diameter is set at 264 mm instead of 268 mm as
specied in DIN 2642. To approximate the reaction of the
gasket and the lapped anges, the material of this ring is
anisotropic with an elasticity modulus of half the normal
steel value in the z-direction and very small E-moduli in
other directions. Bottom nodes of this ring are xed in the
vertical direction although those bottom nodes in traction
have been released as gaskets can only absorb compressive
loads
The circular rings of elements around the bolt holes have
been extended with a fth ring to simulate bolts. The cross-
sectional area of such a ring is slightly larger than that of an
M20 stud bolt but the ring's height is only a quarter of the
ange thickness. Assigning an elasticity modulus to these
rings of just a quarter of the normal elasticity modulus will
result in an overall axial stiffness of these simulated bolts
which is about equal to that of real bolts. Originally we
made the vertical translations of the top nodes of each of
these rings equal to enable the resultant force to shift away
from the bolt's C.L. as a result of the bolt exercising uneven
`nut'-pressure on the ange due to ange rotation. This `bolt
force'-shift is quite pronounced but not having modeled the
countering effect of bolt bending we neglected in our nal
analysis this `bolt'-shift by freeing the vertical translation of
the top nodes so that our FEM-model is as close as possible
to the ange assessment methods (Fig. 9).
4.2. External moment on the loose ange
If n
b
is the number of bolts and A
b
is the cross sectional
area per bolt then the second moment of inertia of equi-
distant spaced bolts on a circle with diameter C is:
I
bolts

C
2
8
n
b
A
b
Distances of the bolts are either 1=2C sin 67:58 or
1=2C sin 22:58 to the neutral plane and hence the
forces per bolt are easily determined as being
10:46194M
ex
=C; 10.19134(M
ex
/C), 20.46194(M
ex
/C)
and 20.19134(M
ex
/C), respectively, for the four bolts in
our 1808-model. As external moment bolt loads are thought
to be superposed on internal pressure bolt loads, negative
bolt loads are physically possible!
The theoretical tangential stress will be determined in
accordance with the adapted `D 0701'-method from the
Dutch `Rules for Pressure Vessels'. The adapted German
DIN 2505 will result in the same value.
D
i
225 mm, inside diameter of size DN 200 loose
ange according to DIN 2642,
D
f
340 mm, outside diameter,
D
f1
295 mm diameter of the bolt circle,
D
f2
22 mm, diameter of the bolt holes,
d
f
20 mm, thickness of the ange,
D
p
f2
D
f2
1 2D
i
=1000 17:05 mm, reduced bolt hole
diameter,
b
f
(D
f
2D
i
)/2 2reduction due to inside chamfer
(340 2225)/2 2(1/2) 5 5/20 56.875 mm, cross
sectional area lost on reduction of ange width is equally
large as that of the chamber,
(b
f
2D
f2
p
) 56.875 2D
f2
p
39.825 mm, effective ange
width,
K
v
2.910Koves-factor with polar moment of inertia.
Within D 0701 all load transfer for loose anges takes
place at the lap ange's outside diameter and this is reected
by setting D
i
1d
f1
D
g
D
i
264 mm in the earlier
given adapted D 0701's radial bending moment formula.
In effect all lever arms for the radial moment calculation
become equal to the distance between the lapped ange's
outside diameter and the loose ange's bolt circle diameter,
i.e. 295 2264=2 15:5 mm. The extra radial bending
C.J. Dekker, H.J. Brink / International Journal of Pressure Vessels and Piping 79 (2002) 145155 153
Fig. 9. Cross-section of mesh at a bolt hole.
moment due to an external bending moment of M
ex
/
C 10,000 N is: M
radial
3:6286M
ex
=C 36; 286 N mm
The section modulus of the ange ring:
W effective flange widthd
2
f
=6 2655 mm
3
This indicates an additional bending stress of
M
radial
=W 13:67 N/mm
2
but for the irregularity factor c
b
.
Bolt loads are not evenly spread out but are rather `lumpy'
and this factor c
b
is thought to cover the consequential stress
effects. We remark that a similar factor is not dened within
DIN 2505. The numerical value of this irregularity factor is
in our case 1.345 and if one were to apply this factor then the
tangential stress in the ange ring would be (locally)
18.39 MPa.
Our FEM-analysis shows tangential ange bending
stresses of 17.4 MPa, below the theoretical value of
18.4 MPa. This value was obtained by averaging the
tangential stresses along the top edge of the loose ange.
One could equally have chosen the bottom edge of the
ange. The most telling graphical representation of the
FEM-analysis showing the tangential bending stresses is
shown in Fig. 10. Note that two effects are not taken into
account within D 0701 (and DIN 2505): load transfer into the
loose ange does not take place at the very edge of the under-
lying lapped ange but over a wider area which, as a conse-
quence, increases the lever arm for the radial moment, and
the earlier mentioned inwards shift of bolt forces may con-
siderably decrease the lever arms. In the FEM-model only the
rst effect is approximated by the partially xed ring of
elements with the E=2 anisotropic gasket material.
We remark that if we had used the effective bolt hole
diameter d
5e
( 9.59 mm) as dened in prEN 1591 then
the effective ange width would be 47.28 mm and the
Koves-factor would be 3.534. Skipping the use of prEN
1591's effective bolt circle but using instead
d
3e
; d
3
295 mm, the theoretical tangential bending
stress would be 11.3 MPa.
Note that we actually used here d
7
d
7max
which gives
the smallest lever arm and hence the smallest theoretical
stresses for the loose ange but at the same time the largest
stub/collar stresses. In reality any overstressing of the stub/
collar will shift the diameter d
7
of reaction between loose
ange and stub/collar away from d
7max
towards d
7min
. This is
recognized within EN 1591 as one may set d
7
at such a value
between d
7min
and d
7max
that gives at the same time the most
favorable result for the loose ange as well as the stub/
collar. Hence, the above stress value of 11.3 MPa should be
regarded as the minimumtheoretical stress, not as a true stress.
5. Conclusions
The observation that external moments on ange connec-
tions also cause torsion in ange rings is obvious enough but
working out a useful relation between the internal torsion
moment and internal bending moment in the ange ring is
the merit of Mr Koves.
The question of whether to use the torsional stiffness or
the polar moment of inertia for this relationship depends on
C.J. Dekker, H.J. Brink / International Journal of Pressure Vessels and Piping 79 (2002) 145155 154
Fig. 10. Tangential bending stresses in the 1808 loose ange model.
how much torsion associated warping will develop in the
ange ring. We think that using the polar moment is justied
if the weakening effect of the bolt holes is taken into account
in the conservative manner proposed. That this leads to
acceptable stress values erring on the conservative side, is
shown in our FEM-analysis of an 8
HH
loose ange which does
not have the benet of a hub or connected tube to resist
warping as integral anges do.
Under the assumption that there is only one signicant
external moment loading in just the operating condition we
give implementations of the Koves-factor for major ange
assessment methods. These proposals make it possible to
enjoy the benets of the Koves-factor in routine engineering
assessments of ange connections with external moment
loadings.
The aim of EN 1591 is to result in anges as lean as
possible and hence its stress calculation is as accurate as
possible. The shift of the gasket reaction away from the
average gasket diameter towards the bolts as a consequence
of ange rotation is taken into account. The resulting `effec-
tive gasket diameter' means smaller lever arms and hence
smaller ange moments. However, the same ange rotation
also results in a shift of the bolt force location away from the
bolt circle towards the gasket and that phenomenon is not
taken into account. From our initial loose ange FEM-
model where bolt-shift was possible, we learned that this
shift can be substantial. If this bolt-shift could be quantied
sufciently accurate then EN 1591 could be sharpened
further though it is somehow reassuring to know that even
this calculation method is conservative.
Note that the authors do not regard the so-called effective
bolt circle diameter d
3e
to reect the above mentioned bolt-
shift as its value depends primarily on the number of bolts
and is independent of the ange rotation. We personally
would always set d
3e
equal to d
3
.
References
[1] ASME Boiler and Pressure Vessel Code, Section VIII, Divisions 1 and
2. American Society of Mechanical Engineers.
[2] ASME Code for Pressure Piping, B31.1. American Society of
Mechanical Engineers.
[3] AD-Merkblatter, Technical rules for pressure vessels. Verband der
Technischen U

berwachungs-Vereine e. V., Essen (Germany). Edition

July 2000.
[4] DIN Standard 2505: Berechnung von Flanschverbindungen. Deutches
Institut fur Normung e.V., Berlin. Proposed edition from April 1990.
[5] Rules for Pressure Vessels, issue 99-12. Published on behalf of
Stoomwezen BV. The Hague: Sdu Publishers.
[6] prEN 13445-3. Unred Pressure Vessel StandardPart 3: Design.
CEN, TC 54 Working Group C-Design, secretariat c/o Sant'Ambrogio
Servizi Industriali s.r.l., Piazza Carlo Donegani 8, I-20133 Milano,
Italy.
[7] Koves WJ. Analysis of ange joints under external loads. Trans ASME
J Press Vess Technol 1996;118:5963.
[8] Waters EO, Rossheim DB, Wesstrom DB, Williams FSG. Develop-
ment of general formulas for bolted anges. Welding Research Coun-
cil, August 1979, copyright 1949 Taylor Forge and Pipe Works.
C.J. Dekker, H.J. Brink / International Journal of Pressure Vessels and Piping 79 (2002) 145155 155