Professional Documents
Culture Documents
Abstract
The effect of load relaxation due to creep has a major impact on the tightness of bolted flanged joints. Several leak incidents have been
reported in the petrochemical and nuclear power plants related to creep. It is clear that the current flange design code does not address
adequately the effect of high temperature on the mechanical and leakage integrities of bolted flange joints used in pairs or with blind
cover plates. It is possible to include the creep analysis in the design methodology.
This paper outlines an analytical approach based on plate theory used to determine the creep effect of blind cover plates and flanges of
a small size on the bolt and subsequently the gasket load relaxation in bolted flanged joints. It details the theoretical procedure to predict
the creep effect of those flanges considered as plates. The results from the theoretical model are verified by comparison with 3D finite-
element results.
r 2008 Elsevier Ltd. All rights reserved.
Keywords: Bolted flange joints; Bolt load relaxation; Blind cover plate; Plate creep theory; Analytical modeling; Finite-element analysis
0308-0161/$ - see front matter r 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijpvp.2008.01.005
ARTICLE IN PRESS
A. Nechache, A.-H. Bouzid / International Journal of Pressure Vessels and Piping 85 (2008) 486–497 487
Recently, successful models based on the elastic interac- [17]. Due to its relative simplicity, accessibility and ease of
tion of the bolted joint elements were developed to study use, the analytical method is more attractive comparatively
non-metallic gasket creep [10–12]. However, due to the lack to FEM.
of gasket creep data, the models were only validated for In a previous paper [14], it was shown that for flanges
short-term periods ranging from 5 to 10 h. A model based having large ratio of outer to inner diameters, ring theory is
on the creep and elastic interaction of all joint elements not adequate to represent accurately the creep behavior of
presented in [13,14] accurately predicts load relaxation. The the annular flange ring. In addition to cases of the creep
model considers the creep of the bolts and the flange and behavior of bolted joints with cover plates, this paper deals
treats the latter as a ring. The creep effect of other attached with those flanges that behave like plates with central hole.
structures such as the hub and the cylinder is added in [15]. Based on the work of [18,19], a model has been developed
As a result, in some cases it was found that the combined to predict the load loss in these two cases.
bolt–gasket–flange creep induced a loss of gasket load of
more than half of its initial preload value. 2. Analytical model
Finite-element analysis of bolted flanged joints subjected
to creep can be retraced as far as the early 70s. In [16] strain 2.1. Analytical interaction model
hardening creep law of the joints members was carried out
to evaluate the load relaxation. The authors ignore the The basic elements that form a bolted flanged joint are
creep of the gasket due to the lack of creep data. Other the flange, the bolts, the gasket and the cover plate when
more recent FE- studies are cited in the survey by Mackerle applicable. The flange is composed of three different
ARTICLE IN PRESS
488 A. Nechache, A.-H. Bouzid / International Journal of Pressure Vessels and Piping 85 (2008) 486–497
elements namely, the shell, the hub and the flange plate. The bolt is represented by a linear elastic spring while
Creep is a behavior that takes place in the three space the gasket non-linear behavior is considered according
coordinates. A one-dimensional creep behavior such as to the approach detailed in [20].
that already covered in [10,11] in the case of the gasket and
the bolts, generates displacement in the axial direction 2.1.1. Radial displacements and rotations
which produces relaxation of the bolt. In addition, creep in The radial displacements and the rotations at the
the axial direction due to plate bending of the flange ring junctions of cylinder–hub and hub–flange are given in
and the cover plate makes the bolt relax more. Finally, the [21]. These deflections are caused by the discontinuity loads
junctions of the cylinder–hub and hub–flange represent and the cylinder, hub and flange creep.
geometrical discontinuities. As with pressure, when creep
takes place, different radial displacements and distortions 2.1.2. Axial displacement
at these junctions are generated for each individual joint The axial displacements of the different components of
element. As a result, this relative movement creates a the joint are caused by the mechanical and creep loads.
barreling effect of the flange which relaxes the bolts. These are given as follows:
The model used is similar to the one developed in [14] to
which plate bending is added. Equilibrium and compat- 2.1.2.1. Gasket. The axial displacement of the gasket is
ibility in both the radial and axial directions are used to given by:
solve the problem. Fig. 1 shows the analytical model with
the different joint elements treated as follows: F fg ðtÞ X c
wfg ¼ þ w_ g Dt, (1)
Kg
The cylindrical shell is treated using the theory of beam where Kg is the equivalent gasket stiffness that depends on
on an elastic foundation. the level of stress reached during bolt-up and is obtained by
The flange is considered as a circular plate with a central linear interpolation of unloading curves as described in
hole or a fully circular plate in the case of a blind cover [20]. It should be noted that the gasket load Fg changes
plate. with time due to creep.
The hub is treated as cylindrical shell with linear varying
thickness.
2.1.2.2. Bolt. The axial displacement of the bolts is given by:
F fb ðtÞ X c
wfb ¼ þ w_ b Dt, (2)
N Kb
where Kb represents the rigidity of the bolts as defined in [21].
p
Shell
2 3
during operation; it is calculated by the sum of all axial 0 0 1 0 0 0 0
displacements of the individual joint elements 60 0 0 1 0 0 07
6 7
Dn ¼ Swi ¼ Swf , (5) B¼6 7. (12)
40 0 0 0 0 0 05
Dn ¼ wig þ wib þ wip þ wif ¼ wfg þ wfb þ wfp þ wff , (6) 0 0 0 0 0 0 0
therefore C is a matrix formed with the hub influence coefficients for
F ig F ib F fg ðtÞ X c F f ðtÞ the radial displacement, rotation, edge load and moment
þ þ C 5 wif þ C 6 ¼ þ w_ g Dt þ b
Kg Kb Kg Kb applied at the junction with the flange ring and the cylinder
X influence coefficient for radial displacement and rotation,
c f
þ w_ b Dt þ C 5 wf þ C 7 þ C 8 .
2
(7) 1=2
z2 ber0 ð2 Þ z2
1=2
bei0 ð2 Þ
6
The bolts and gasket creep axial displacements are not 6 3=2 3=2
6 z2 z2
6 S1 ð2 Þ S 2 ð2 Þ
taken into account in this paper as these were the subject of 6
6 2 2
previous papers [10,11]. 6
6 E f a3 r2 z1=2 E f a3 r2 z2
1=2
6 2
For a pair of identical flanges, 6 24ð1 n2 Þ S 5 ð2 Þ
24ð1 n2f Þ
S 6 ð2 Þ
6 f
6
C5 ¼ 2 and C 6 ¼ C 7 ¼ C 8 ¼ 0. (8) 6
6 E f a z2
3 1=2 1=2
E f a3 z2
C¼6
6 48ð1 n2 ÞS 9 ð2 Þ S 10 ð2 Þ
For a flange with a cover plate 6 f 48ð1 n2f Þ
6
6
F ig F ib F fg ðtÞ F fb ðtÞ 6 0 0
C 5 ¼ 1; C6 ¼ ; C7 ¼ 00 a0 p 6
6
K 0p K 00p K 0p Kp 6 0 0
X 6
6 0 0
and C 8 ¼ w_ cp Dt. (9) 6
6
6 0 0
4
In addition, by considering continuity at the junctions as
0 0
well as equilibrium of forces and moments, a system of 13 3
1=2 1=2
equations with 13 unknowns is obtained as detailed in [21] z2 ker0 ð2 Þ z2 kei0 ð2 Þ 0 0
7
" # 3=2 3=2 7
A B z2 z2 7
S 3 ð2 Þ S4 ð2 Þ 0 0 7 7
C D 2 2 7
½ C1 C2 C3 C4 P1 M1 P2 M2 uh1 yh1 uh2 yh2 F fb ðtÞ 7
E f a3 r2 z2
1=2
E f a3 r2 z2
1=2 7
2
S 7 ð2 Þ S 8 ð2 Þ 0 0 7 7
¼ ½U, 24ð1 nf Þ 24ð1 n2f Þ 7
7
(10) 7
E f a3 z2
1=2
E f a3 z2
1=2 7
S 11 ð2 Þ S 12 ð2 Þ 0 0 7 7
where A is matrix formed with the hub characteristics for 48ð1 n2f Þ 48ð1 n2f Þ 7
7,
the radial displacement, rotation edge load and moment 1 1 7 7
0 0 7
applied at the junction with the cylinder 2b3s Ds 2b2s Ds 7
7
2 7
1=2 1=2 1 1 7
z1 ber0 ð1 Þ z1 bei0 ð1 Þ 0 0 7
6 2b2s Ds bs Ds 77
6 3=2 3=2 7
6 z1 z1 0 0 0 0 7
6 S 1 ð1 Þ S 2 ð1 Þ 7
6 2 2 7
6 0 0 0 0 7 5
6
A ¼ 6 E f a3 r2 z11=2 E f a3 r2 z1
1=2
0 0 0 0
6 S 5 ð 1 Þ S 6 ð1 Þ
6 24ð1 n2 Þ 24ð1 n2f Þ ð13Þ
6 f
6
6 1=2 1=2
4 E f a3 z 1 E f a3 z 1
S ð
9 1 Þ S 10 ð1 Þ 2 3
48ð1 n2f Þ 48ð1 n2f Þ 0 0 0 0 1 0 0
6 0 0 0 0 0 0 7
1=2 1=2 3 6
6
1 7
7
z1 ker0 ð1 Þ z1 kei0 ð1 Þ 0 0 6 1 0 0 0 0 0 0 7
7 6 7
3=2 3=2 7 6 0 1 0 0 0 0 0 7
z1 z1 7 6 7
6 7
S 3 ð1 Þ S 4 ð1 Þ 0 0 7 6 0 0 1 0 0 0 0 7
2 2 7 6 7
7 D¼6
6 0 0 0 1 0 0 0 7,
7
1=2 1=2 7 6 7
E f a3 r2 z1 E f a3 r2 z1 7, 6 ri K 2 þ 1 tf 7
S 7 ð 1 Þ S 8 ð 1 Þ 1 0 7 6 þ n 0 0 0 1 0 7
24ð1 nf Þ 2 24ð1 nf Þ 2 7 6 E f tf K 2 1 f
2 7
7 6
6
7
7
7 6 r i tf ri E f t3f rb rg 7
3 1=2 3 1=2 7 6 0 0 0 7
E f a z1 E f a z1 5 4 2rm rm 2pri Y prm 5
S ð
11 1 Þ S ð
12 1 Þ 0 1
48ð1 n2f Þ 48ð1 n2f Þ 0 0 0 0 0 C5 C9
(11) (14)
ARTICLE IN PRESS
490 A. Nechache, A.-H. Bouzid / International Journal of Pressure Vessels and Piping 85 (2008) 486–497
2 3
r2 ð2 nf Þ including identical flange pairs and flanges with cover
6 h p uch1 7 plates as shown in Fig. 2 were run. Because of symmetry of
6 2E f gx1 7
6 7 both geometry and loading, only an angular portion that
6 r2 ð2 nf Þ 7
6 h p ych1 7 includes half of the bolt is modeled. In the case of
6 2E f gx21 7
6 7 assemblies with identical flange pairs, only one flange
6 gri 7
6 p 7 including half of the gasket thickness is modeled because of
6 4 7
6 7 symmetry with respect to a plane that passes through the
6 r2h ð2 nf Þg2 7
6 p 7 gasket mid-thickness. In the case of a flange with cover
6 7
6 12ð1 n2f Þ 7 plate, a portion of both components is modeled.
6 7
6 2
rh ð2 nf Þ 7 The general-purpose numerical FE program Ansys 9.0
6 p uh2 c 7
6 7 [24] was used to run two identical flange pairs namely a SO
6 2E f gx2 7
6 7 NPS3 class 150 joint and a WN NPS 4 class 150 joint, and
6 r 2
ð2 n Þ 7
6 h f
p yh2 c 7 three flanges with cover plates namely a NPS 16 class 300
6 2E f gx2 2 7
6 7 joint, a 24 in and a 36 in heat exchanger joints. These
U ¼6
6 gri 7.
7 (15)
6 p 7 flanges are made of SA106 forged steel or SA182T304
6 4 7 stainless steel used with A-193 B7 bolts. The loading is
6 2 2 7
6 rh ð2 nf Þg 7 applied in three stages. The initial bolt-up is first achieved
6 p 7
6 12ð1 nf Þ 2 7 by applying an axial displacement to the bolt to produce
6 7
6 7 the initial target bolt stress of 275 MPa to the SO NPS3
6 ð2 ns Þr2i c 7
6 p us 7 class 150 and 172 MPa to the WN NPS 4 class 150 flange as
6 2E t 7
6 s s 7 well as 205 MPa to the NPS 16 class 300, 24 in and 36 in
6 ys T 7
6 7 heat exchanger flanges used with cover plates. Pressure is
6 7
6 ri K 2
þ 1 7 then applied in the radial direction while the hydrostatic
6 c7
6 E ðK 2 1Þ þ n p uf 7 end thrust is simulated by an equivalent longitudinal stress
6 f f 7
6 7 applied to the shell. Finally, in the last stage, creep was
6 2
ðri rg Þðrg þ ri Þ 2 7
6 p 7 applied to the flange materials. As mentioned previously,
6 2rm 7
4 5 creep of the gasket and bolts was purposely not considered
C 9 F ib þ C 5 wif þ C 10 C 8 in the analysis as the prime focus was the creep contribu-
tion of the flange material. It is to be noted that for the case
And for a pair of identical flanges
of the WN NPS 16 class 300 flange, creep was considered
1 1 Ap with no pressure applied.
C9 ¼ þ and C 10 ¼ p. (16)
Kb Kg Kg Two creep material properties were considered. The
creep constants Ac, m and n are 3.85 1015, 5.35 and 0.22
For a flange with a cover plate
! or 7.5 1017, 5.5 and 1, respectively [25,26]. The first
1 1 1 1 Ap Ap material constants were applied to the NPS 16 class 300
C9 ¼ þ þ and C 10 ¼ þ a0 p, flange with cover plate while the second constants were
K b K g K 0p K 00p K g K 0p
applied to the SO NPS3 class150, WN NPS 4 class 150, 24
(17) and 36 in heat exchangers flanges.
where Ap is the pressurized area given by One type of gasket was used; a corrugated metal
sheet, CMS, for all flanges. The mechanical behavior of
pG2
Ap ¼ . (18) the gasket is represented by non-linear loading and
4 unloading curves of gasket contact stress versus axial
The 13 unknowns namely C1, C2, C3, C4, P1, M1, P2, M2, displacement [14].
uh1, yh1, uh2, yh2 and Ffb are solved. When, after a time
interval, the bolt load Ffb is known, the gasket force is 4. Results and discussion
obtained using the axial equilibrium relationship:
F fg ðtÞ ¼ F fb ðtÞ þ Ap p. (19) It is important to note that the long-term behavior of a
bolted flange joints operating at high temperature is very
This procedure is repeated at each interval of time Dt, until difficult to predict because of several phenomena that can
the required total creep time is reached. take place simultaneously besides creep. Thermal degrada-
tion due to exposure, weight loss due to exposure and
3. Finite-element modeling corrosion are just to name a few. Nevertheless, the major
contributing factor remains creep. Depending on the type
To validate the proposed analytical model that estimates of flange and gasket configuration, the relative importance
the load relaxation due to creep, comparison was made of the creep effect in order is in general the gasket, then the
against the more accurate three-dimensional numerical bolt followed by the flange ring treated as a ring or plate
FE method. Several FE models of different assemblies [10,11,14]. The hub, shell and cover plate are comparatively
ARTICLE IN PRESS
A. Nechache, A.-H. Bouzid / International Journal of Pressure Vessels and Piping 85 (2008) 486–497 491
180 310
Plate creep
Bolt up= 205 MPa Flange creep
No Pressure Plate & flange creep
150 280
Bolt stress relaxation, MPa
120
90 220
Cylinder creep
Flange creep
60 Flange & cylinder creep 190
Bolt-up = 172 MPa
Pressure = 0.68 MPa
CAF Gasket
Solid lines : FE model (Ansys)
30 160 Solid lines : FE model (Ansys)
Dotted lines : Analytical model
Filled markers : Ring theory Dotted lines : Analytical model
unfilled markers : Plate theory
0
130
0 2000 4000 6000 8000 10000
0 2000 4000 6000 8000 10000
Time, hours
Time, hours
Fig. 4. Bolt stress relaxation of the WN NPS 4 cl 150.
Fig. 5. Bolt stress relaxation of NPS 16 cl 300 cover plate.
For the case of flanges with blind cover plates, the 200 Plate Creep
performance of the proposed approach is checked through Flange Creep
Bolt relaxation, MPa
250 This is due to the fact that, in general, cover plates are
Bolt up= 205 MPa more rigid and are subjected to much lower stress than
Pressure=0.69 MPa annular rings. Nevertheless, the model can be used to verify
the adequacy of flange geometry and bolting in high-
200 temperature applications or to simply evaluate the impact
of creep on the bolt load loss.
Bolt relaxation, MPa
Plate Creep
Flange Creep
Plate & Flange Creep Appendix A. Theory of plates subjected to creep
150
Creep of plates subjected to bending has been addressed
by several authors [18,27–29]. For convenience, a method
based on Penny’s approach used in this work is presented
100 hereafter.
Considering the flange as a plate subjected to bending
(Fig. A1) and neglecting the radial displacement caused by
Solid lines : FE model (Ansys)
pressure, the radial and tangential strains including creep
Dotted lines : Analytical model
components are given as follows:
50
0 2000 4000 6000 8000 10000 d2 w 1
Time, hours r ¼ z ¼ ðsr np sy Þ þ cr ,
dr2 Ep
Fig. 7. Bolt stress relaxation of 36 in HE cover plate. z dw 1
y ¼ ¼ ðsy np sr Þ þ cy . (A.1)
r dr E p
120.00 R tp =2
NPS 16 cl 300 Multiplying by tp =2 z dz and noting that M r ¼
Heat Exchanger 24'' R tp =2 R tp =2
100.00 Heat Exchanger 36'' tp =2 sr z dz and M y ¼ tp =2 sy z dz the radial and tangen-
tial moments Mr and My can be obtained as
2
Gasket relaxation, MPa
80.00
d w np dw
Mr ¼ D þ
dr2 r dr
Z tp =2 Z tp =2 !
Solid lines : FE model (Ansys) Ep c c
60.00 Dotted lines : Analytical model r z dz þ np y z dz ,
ð1 n2p Þ tp =2 tp =2
1 dw d2 w
40.00 My ¼ D þ np 2
r dr dr
Z tp =2 Z tp =2 !
Ep c c
20.00 y z dz þ np r z dz , (A.2)
ð1 n2p Þ tp =2 tp =2
A.2. Cover plate bolt load, gasket reaction force. The analysis is carried out
using the superposition method as shown in (Fig. A3):
The case of a cover plate is treated as a circular plate The boundary conditions for these cases are given as
simply supported at the outer radius subjected to pressure, follows:
M r ðro Þ ¼ 0,
yðri Þ ¼ 0,
p
wðro Þ ¼ 0. (A.14)
Fb
The displacement w, the rotation y and the moment
tp curvature are given in [22] (Fig. A3).
simplification, the rate of tangential moment is obtained as creep. This depends essentially on the stationarity of the
discontinuity forces which act between the cylinder–hub
_ C3 C4 _
M y ¼ D ð1 þ np Þ þ ð1 np Þ 2 þ I 2 I 3 þ np G and hub–flange, and depends also on the creep material
2 r
! constants. Thus, the axial displacement and the rotation
Z tp =2 Z tp =2
Ep through the radius of the plate due to the creep of a
_ cy z dz þ np _cr z dz . (A.21)
ð1 n2p Þ tp =2 tp =2
material are calculated.
[17] Mackerle J. Finite element analysis of fastening and joining: a [22] Rudolph Szilard. Theory and analysis of plates, classical and
bibliography (1990–2002). Int J Pressure Vessel Piping 2003;80(4): numerical methods. Prentice-Hall; 1974.
253–71. [23] Bouzid A, Chaaban A. Flanged joints analysis: a simplified method
[18] Penny RK, Hayhurst DR. The deformations and stresses in a based on elastic interaction. CSME Trans 1993;17(2):181–96.
stretched thin plate containing a hole during stress redistribution [24] ANSYS. ANSYS inc Standard Manual, Version 9.0, 2005.
caused by creep. Int J Pressure Mech Sci 1969;11:23–39. [25] Kraus H. Creep analysis. New York: Wiley; 1980.
[19] Jahed H, Bidabadi J. An axisymmetric method of creep analysis for [26] Betten J. Creep mechanics. Berlin: Springer; 2002.
primary and secondary creep. Int J Pressure Vessel Piping 2003; [27] Lin TH. Bending of a plate with nonlinear strain hardening creep. in:
80(9):597–606. Colloquium held at Standford University, California, July, 11–15,
[20] Bouzid A, Champliaud H. Contact stress evaluation of non-linear 1960, International Union of Theoretical and Applied Mechanics,
gaskets using dual kriging interpolation. ASME J Pressure Vessel 1962. p. 43–64.
Technol 2004;126:445–50. [28] Patel SA, Cozzarelli FA, Venkatraman B. Creep of compressible
[21] Bouzid A, Nechache A. An analytical solution for evaluating gasket circular plates. Int J Mech Sci 1963;5(1, Jan–Feb):77–85.
stress change in bolted flange connections subjected to high tempera- [29] Venkatraman B, Sharad A Patel. Creep analysis of annular plates.
ture loading. ASME J Pressure Vessel Technol 2005;127:414–22. Franklin Inst—J 1963;275(1):13–23.