Strength and leakage nite element analysis of a GFRP ange joint

H. Estrada
a
, I.D. Parsons
b,
*
a
Department of Civil Engineering, Vanderbilt University, Nashville, TN, USA
b
Department of Civil Engineering, University of Illinois at Urbana-Champaign, 2207 Newmark C.E. Laboratory, MC 250, 205 North Mathews Street,
Urbana, IL 61801-2352, USA
Received 24 January 1999; accepted 14 February 1999
Abstract
The strength and leakage analysis of a glass ber reinforced plastic (GFRP) modied stub anged joint is presented. The analysis is carried
out using three-dimensional and axisymmetric nite elements. To model the GFRP material, we use an orthotropic material model. To model
the loss of contact between the ange and the gasket, we use a contact condition between these mating parts. In the axisymmetric case, we
allow uid to penetrate the area where contact is lost; the three-dimensional contact formulation, however, does not support this uid
penetration option. We use a commercial nite element package (Abaqus) to perform the analysis. The primary emphasis is on assessing the
strength of the ange and the leak tightness of the gasket. Bolt interaction is not considered, therefore we assume a constant bolt load over the
history of the analyses. 1999 Published by Elsevier Science Ltd. All rights reserved.
Keywords: Leakage; Glass ber reinforced plastic pipe joints; Contact nite element analysis
1. Introduction
Glass ber reinforced plastic (GFRP) anged joints are
widely used in the piping and pressure vessel industries. The
anged joint is used because systems in these industries
usually include pumps, valves and other ttings that require
periodic removal for maintenance. GFRP materials are used
as a result of their corrosion resistant properties. Flanged
joints are susceptible to two failure mechanisms regardless
of their material composition: material failure and leakage.
Therefore, a joint must be properly designed to prevent
these two failures. However, current joint design procedures
to address both these failures are primarily empirical due to
the complexity of the stress and leakage analyses. Previous
investigations of GFRP joints have generally focused on
design for strength, whereas leakage has received little
attention. All except two of the previous investigations
regarding strength are primarily concerned with the transfer
of metallic design calculations, without accounting for the
orthotropy of the composite material properties [1,2]. Sun
[3] presents stress calculations for a GFRP ange joint using
lamination theory and nite element analysis; however, he
did not study leakage. Godwin et al. [4] studied GFRP joint
leakage by investigating the relationship between clamping
force and joint sealing for stub, and full-face GFRP anged
joints using experiments and axisymmetric nite element
analysis. The normal stress on the gasket was found to be
the principal factor affecting the leak tightness of these
joints. Finite element analysis was conducted to predict
the internal pressure at which leakage occurs and the pres-
sure distribution on the gasket. Although tensile normal
stresses developed over a part of the gasket face, no
angegasket separation was allowed, because their nite
element analysis code did not support contact formulations.
In this paper, we describe detailed strength and leakage
three-dimensional and axisymmetric nite element
analyses, in which we use a contact formulation to allow
loss of contact between the mating ange and the gasket as
the internal pressure increases. Also, in the axisymmetric
model, we allow uid penetration into the space where this
contact loss occurs; however, this type of loading is not
supported in a three-dimensional analysis. These nite
element analyses are performed using the commercial nite
element package Abaqus.
The analyses were performed on a modied stub anged
joint (Fig. 1). This GFRP joint is a modied version of a
typical GFRP joint, the stub anged joint, and was devel-
oped by Estrada and Parsons [5] to address some problems
particular to GFRP joint geometries currently in use. The
pipe and hub are lament wound as an integral unit. The
International Journal of Pressure Vessels and Piping 76 (1999) 543550
0308-0161/99/$ - see front matter 1999 Published by Elsevier Science Ltd. All rights reserved.
PII: S0308-0161(99)00021-6
* Corresponding author. Tel.: 1-217-333-2690; fax: 1-217-333-
9464.
E-mail address: idp@uiuc.edu (I.D. Parsons)
metallic backing ring is used to connect the joint to other
members. The joint was proportioned using the design
guidelines presented in Ref. [6], and the ASME gasket
design criteria; the dimensions are shown in Fig. 2.
We also attempted to validate the ASME gasket (leakage)
design formulation using detailed nite element analysis for
a particular joint and gasket. The formulation can easily be
extended to other joint congurations and gaskets. The
ASME code leakage design guidelines are based on two
gasket constants, m and y factors; these are determined
empirically. The yield factor y is dened as the minimum
gasket stress required to cause the gasket material to deform
into the ange face irregularities. The product m p is
dened as the minimum gasket stress needed to hold the
joint sealed under the internal pressure p. These factors
are used to determine the bolt load to be applied for an
effective seal. That is, the y factor is used to determine the
initial bolt load to be applied to the joint, and the m factor is
used to determine the gasket pressure needed to prevent
leakage while the joint is in operation. The implementation
H. Estrada, I.D. Parsons / International Journal of Pressure Vessels and Piping 76 (1999) 543550 544
Fig. 1. Schematic of the modied stub ange joint.
Fig. 2. A 5.625 segment of the modied stub ange.
of these factors is through the calculation of the design bolt
load: each factor is used to compute a bolt load, and the
larger of the two is used to design the joint for strength.
The paper is organized as follows. First, we describe the
evolution of leakage over the loading history of the joint and
give a description of the gasket used in the analyses. We
then discuss the detailed nite element analyses: we
describe the three-dimensional analysis, and then discuss
the analysis for the axisymmetric model. In the three-dimen-
sional model, we make use of the cyclic symmetry of the
system and only consider a 1/64 segment of the total
circumference of the joint (see Fig. 2). Since Abaqus
does not support pressure penetration in the three-dimen-
sional contact analysis, we also conducted an axisymmetric
analysis to properly model joint leakage. In the last part of
the paper we present a summary of the results.
2. Leakage development description
A sequence of load steps depicting the loading history of
the joint is shown in Fig. 3. This gure also shows the
leakage development over the loading history as the contact
area is lost. This contact loss between the gasket and the
ange is due to ange rotation. The rotation is caused by the
bolt load, the hydrostatic end load and the uid penetrating
the space where the contact is lost. In the rst step, only the
bolt load is applied. In this case, the gasket pressure (or
contact pressure) is largest near the bolt and decreases
farther away from the bolt. Since the bolts encircle the
gasket, the gasket pressure decreases toward the inside of
the pipe in the radial direction. In the circumferential direc-
tion the gasket pressure is greatest close to the bolt and
decreases toward the point between two bolts. In the second
step, the loss of contact area has allowed uid to penetrate.
This process continues until the contact area recedes toward
the outer edge of the gasket; at the point where contact is
lost, leakage occurs.
Analytical leakage analysis can be performed numeri-
cally using a contact nite element analysis. We can exploit
the cyclic symmetry of the joint to reduce the size of the
problem by only modeling a wedge section of the joint.
Also, in some cases we can use an axisymmetric model
without loss of practical accuracy. The circumferential
variation of the gasket pressure can only be captured in a
three-dimensional wedge analysis. However, as we will
show later, for closely spaced bolts the circumferential pres-
sure distribution on the gasket is nearly uniform. The stub
diffuses the bolt load uniformly onto the hub, and the bolt
load is transferred almost uniformly to the gasket. There-
fore, the gasket pressure distribution in the circumferential
direction depends on the number of bolts (closely spaced
H. Estrada, I.D. Parsons / International Journal of Pressure Vessels and Piping 76 (1999) 543550 545
Fig. 3. Gasket contact at different load increments.
bolts provide a uniform gasket pressure). The bolt spacing
(usually two to three bolt diameters) is basically determined
by practical limitations, such as the size of the wrench used
to tighten the bolts. Since bolt spacing is specied as close
as practically possible to provide a uniform gasket pressure,
an axisymmetric analysis can be used to study the leakage
problem rather than using a more expensive three-dimen-
sional analysis.
3. Gasket description
Rubber gaskets are usually used to seal GFRP joints in
order to limit the bolt force required to conform the gasket to
the mating GFRP surfaces. Also, these soft rubber gaskets
require a low unit compression to keep joints leak tight
because of their resiliency. The gasket material constants
given in Section VIII of the ASME BPV code [7] are not
related to engineering material constants such as Youngs
modulus and Poissons ratio; however, to conduct a detailed
nite element analysis of leakage, we need the material
behavior of the gasket. Furthermore, rubber is a nearly
incompressible material and its constitutive behavior is
non-linear. Therefore, to properly model the material beha-
vior of rubber, we need its complete stressstrain response.
In this study, we use a synthetic rubber (Neoprene)
gasket, 75 shore durometer hardness, m 1 and y
200 psi [7]. The material properties were extracted from
Ref. [8]. The data is reported in a load-deection diagram,
which was obtained from a gasket compression test using an
XY continuous recorder; consequently, we had to digitize
several points along the plot. The gasket is a standard 1/8 in
2 inches-150# size, 2-3/8 inch inside3-3/16 in outside
diameter, which gives a total surface area of 5.54 inches
2
.
From the load and the total surface area we compute the
nominal stress. From the deection and the gasket thickness
we can compute the nominal strain. With the nominal stress
and nominal strain, we can calibrate a hyperelastic material
model, as shown in Fig. 4. Using a second degree Mooney
Rivlin material model and given a set of experimental test
data, the calibration of the hyperelastic material model can
be done using a least squares t. The analysis of hyperelastic
materials requires the use of hybrid elements in the discre-
tization of the hyperelastic material, because of the incom-
pressibility of the material.
4. Joint loading
Initially, only the bolt load is acting on the joint, i.e. the
bolts are pre-loaded. This bolt pre-load causes an initial
gasket pre-stress (the seating condition). In this state, the
gasket deforms lling the irregularities on the ange face,
insuring full contact over its entire surface. We do not
consider this loading condition in the analysis because the
load required to seat a rubber gasket is small and obviously
leakage is not a problem. The internal pressure is then
applied, and the gasket pre-stress decreases and the bolt
load increases (the operating condition). The pressure pene-
tration described above causes an additional decrease in
gasket pre-stress and an increase in bolt load; as the anging
H. Estrada, I.D. Parsons / International Journal of Pressure Vessels and Piping 76 (1999) 543550 546
Fig. 4. Gasket stressstrain diagram.
Fig. 5. Three-dimensional nite element model.
system deforms, some relaxation of the bolt pre-load occurs.
The nal load carried by the bolts depends on this bolt
ange interaction, which in turn depends on the stiffness
of the anging system and the extent of the uid penetration.
Furthermore, the uid penetration is also a function of the
ange stiffness. The bolt load relaxation is due to compat-
ibility of the system deformation, i.e. the deformation in the
bolts and the ange has to be compatible. This is a complex
interaction and we do not consider it here. We will study the
long-term behavior of the joint; therefore, we assume that
the bolt load remains constant over the history of the load-
ing. The justication for this is that, in practice, the
mechanic will usually retighten the bolts to maintain the
required bolt load to prevent leakage. Other issues of load
relaxation will not be considered, such as creep of the resin
in the composite and other long-term relaxation problems.
5. Finite element models
5.1. Three-dimensional nite element model
The three-dimensional nite element mesh is depicted in
Fig. 5. As explained earlier, we make use of cyclic symme-
try to reduce the size of the model to a 5.625 segment of the
ange. 2156 second-order brick elements with reduced inte-
gration are employed throughout the mesh of the anging
system and the gasket.
The contact between the bottom of the hub and the gasket,
and between the hub and the stub are modeled using contact
pairs. Contact is modeled by the interaction of contact
surfaces dened by grouping specic element faces in the
contacting regions. This contact formulation uses a master
slave concept to enforce the contact constraint. Since the
gasket is not rigidly attached to the hub (it can be blown out
by the internal pressure in cases where soft gaskets are used
and ange faces are very smooth), we allow relative sliding
between the gasket and hub contact surfaces. Small relative
sliding is also allowed between the hub and stub contact
surfaces. We use a standard Coulomb friction model in
the contact formulation to control the sliding. We assume
a coefcient of static friction of 0.8, a very rough surface, in
the gaskethub contact interaction and a value of 0.2, a
smoother surface, in the hubstub contact interaction.
These choices are made arbitrarily because of the lack of
data.
The material properties used in the analysis are given in
Table 1 for the anging system and in Fig. 4 for the gasket.
In the composite portion of the system, we have eight layers
through the pipe thickness, [ ^ 54.7]
4
, which we model
with four elements through the thickness, two layers per
element. The material properties are specied via the full
three-dimensional constitutive relations, in this case trans-
versely isotropic.
The loading is also depicted in Fig. 5, it consists of the
hydrostatic end load, internal pressure, and bolt load. The
hydrostatic end load is smeared over the top surface of the
pipe. The bolt load is taken as 1/64 of the total bolt load and
is smeared uniformly over the rst row of elements adjacent
to the bolt hole, which represents the area over which the
bolt head would transfer the load. Notice that in the real
case, when the ange rotates, the bolt head pressure
would not be uniform over this area; it would be greater
on the portion of the area closest to the inside of the pipe.
H. Estrada, I.D. Parsons / International Journal of Pressure Vessels and Piping 76 (1999) 543550 547
Table 1
GFRP and steel material properties [10]
Pipe-hub, glass/polyester
Longitudinal modulus, E
LL
5.14 10
6
psi
Transverse modulus, E
TT
1.52 10
6
psi
In-plane shear modulus, G
LT
0.44 10
6
psi
Major Poissons ratio, n
LT
0.28
Minor Poissons ratio, n
TT
0.35
Longitudinal telsils strength 118.93 10
3
psi
Longitudinal compressive
strength
88.47 10
3
psi
Transverse tensile strength 10.88 10
3
psi
Transverse compressive strength 19.29 10
3
psi
In-plane shear strength 5.00 10
3
psi
Resin compressive strength 19.29 10
3
psi
Stub, steel
Youngs modulus, E 29 10
6
psi
Poissons ratio, n 0.3
Fig. 6. Axisymmetric nite element model.
Pressure penetration is not applied to this model because the
three-dimensional contact formulation does not support this
type of loading.
As we described earlier, we are only modeling a 1/64
segment of the total ange. Therefore, we have to add
boundary conditions to make the model behave as if the
whole component were being modeled. Symmetry boundary
conditions are applied to the symmetry plane of the gasket,
i.e. the axial displacement at the middle of the gasket is zero.
The planes through angles 0 and 5.625 are also symmetric
and are constrained appropriately, i.e. the circumferential
deformation is zero.
5.2. Axisymmetric nite element model
In this section, we study the development of leakage in
the modied stub ange joint using the pressure penetration
option with contact nite elements. The contact stress distri-
bution on the gasket is studied using slide line elements, and
pressure is allowed to penetrate as contact is lost.
One thousand and three hundred second-order reduced
integration axisymmetric elements are used throughout the
mesh of the anging system and gasket (Fig. 6). The contact
between the hub and the stub is again modeled using a
contact pair. Contact between the gasket and the ange
face is modeled using axisymmetric slide line elements on
the gasket and a slide line that is attached to the hub face.
This contact formulation uses a masterslave concept to
enforce the contact constraint. To model the sliding, in
both contact cases, we use the same friction formulation
as in the three-dimensional case.
The material properties for the gasket and composite
system are the same as those used in the three-dimensional
case with a [0/90]
4
lay-up. As none of the axisymmetric
elements support a laminated material model, we used one
element for every layer, eight elements through the thick-
ness of the pipe, and employed a transversely isotropic
constitutive law.
The stub-material is inhomogeneous due to the discrete
nature of the bolt holes. This is handled by smearing the
material properties used in the bolt hole area of the mesh,
the shaded area in Fig. 6, i.e. using material properties corre-
sponding to a weaker material in the bolt hole area. We
follow a procedure similar to that found in the ASME
code Section VIII used to determine the effective material
properties for perforated plates with holes in an equilateral
triangular array. For the model presented here, the effective
material properties are calculated using an elasticity moduli
reduction factor. This factor is 1 AH/AA, where AH is the
total volume of the bolt holes and AA the volume swept by
the bolt diameter along the entire circumference of the
ange, along the bolt circle diameter. Hence, the reduction
factor is 0.88. The effective in-plane moduli of elasticity are
obtained by multiplying the reduction factor by the ange
modulus. The in-plane Poissons ratio is left unchanged.
The modulus in the hoop direction should be very small
and the hoop Poissons ratio should be zero. The effective
shear modulus is computed from its respective modulus of
elasticity and Poissons ratio.
The loading is depicted in Fig. 6. The internal pressure
loading is broken down into three loads: (i) the internal
pressure that acts on the inside surface of the vessel; (ii)
the hydrostatic end load, which is the membrane stress in
the pipe due to the internal pressure; and (iii) the penetrating
pressure as the contact between the ange face and the
gasket is lost. The total bolt load is smeared over the
upper bolt hole surface as a normal pressure, as shown in
Fig. 6. The uid penetration described above can easily be
simulated in the axisymmetric analysis. The uid pressure
can penetrate into the mating surface interface, until some
area on the surfaces is reached where the contact pressure
between the abutting surfaces exceeds the uid pressure,
cutting off further penetration. The pressure penetration
load starts from the inside of the pipe and penetrates
between the surfaces continuously from this side.
In an axisymmetric analysis, the formulation of the
element takes care of the boundary conditions in the circum-
ferential direction. Therefore, the only boundary conditions
we have to specify in this model are symmetry boundary
conditions on the symmetry plane of the gasket, i.e. the axial
displacement in the middle of the gasket is zero (see Fig. 6).
6. Results and discussion
We analyzed both models using non-linear large displa-
cement analysis. The non-linearities in the problem are due
to the contract conditions and the gasket material behavior.
The three-dimensional model, although only about ve
times larger (in terms of degrees of freedom) than the
axisymmetric model, required 26 times more CPU time to
H. Estrada, I.D. Parsons / International Journal of Pressure Vessels and Piping 76 (1999) 543550 548
Fig. 7. Three-dimensional gasket contact pressure distribution.
run; both problems were solved in 24 load increments. A
true comparison, however, cannot be made because in the
axisymmetric model pressure penetration was also included
but omitted in the three-dimensional analysis. In some
cases, this extra expense in computer time (not to mention
the time needed to generate the more complex three-dimen-
sional model) may not be cost effective, since the results of
the axisymmetric model may be sufciently accurate for
practical design purposes.
6.1. Three-dimensional nite element results
Fig. 7 shows the gasket contact pressure p
contact
, computed
using the three-dimensional analysis. This gure clearly
shows that over three-quarters of the contact between the
gasket and the ange is lost. However, the pressure on the
portion of the gasket that remains in contact is greater than
the internal pressure, and the minimum required gasket
pressure to keep the joint leak tight, i.e. the contact pressure,
is greater than m p 50 psi [9]. The contact pressure in
the circumferential direction is not uniform; this pressure is
higher on the side where the bolt load is applied. However,
the difference is small and we can assume this distribution to
be uniform in the circumferential direction. Therefore, an
axisymmetric analysis of the joint leakage behavior can
provide sufciently accurate results.
Fig. 8 depicts the normal contact pressure, p
contact
, on the
hub at the interface between the hub and the stub. Most of the
surface that was initially in contact separates as a result of the
stub rotation. However, the bearing strength of the composite
is not exceeded, i.e. the pressure on the surface that remains
in contact is less than 20 000 psi. In the circumferential
direction, the contact pressure is nearly uniform over most
of the surface, except at the ends; the contact pressure is
higher on the sides of the bolt. This is the reason for the
non-uniformity of the gasket stress shown in Fig. 7.
Stress concentrations localized around the bolt hole area
govern most maximum and minimum values in the stub. In
fact, the allowable stress is exceeded in this localized
portion. This stress concentration is due to the presence of
the bolt hole. However, the yield stress of the stub material,
steel, is not exceeded. To assess the strength capacity of the
composite material, we use a maximum stress failure criter-
ion. None of the maximum or minimum stresses exceeded
the strength capacity of the composite material.
6.2. Axisymmetric nite element results
Fig. 9 shows the gasket contact pressure computed using
axisymmetric analysis. This gure shows that over three-
quarters of the contact between the gasket and the ange is
lost, which is consistent with the three-dimensional analy-
sis. The axisymmetric analysis also predicts the pressure on
the portion of the gasket that remains in contact to be greater
than the internal pressure, and the minimum gasket pressure
required to keep the joint leak tight.
The ASME code minimum gasket pressure to maintain a
leak tight joint, m p is also shown in Fig. 9, and falls below
the average computed pressure. The ASME gasket design
formulation assumes that contact between the gasket and the
mating ange is lost (pressure penetrates) up to the middle
of the gasket [9]; therefore, the ASME gasket pressure, m
p only acts on the outer half of the gasket. However, the
resulting penetration from the nite element results is about
75% into the gasket, which is depicted by the intersection of
the nite element results, and an internal pressure of 50 psi.
However, the maximum and average nite element surface
pressures are greater than that for the ASME code. These
results show that the bolt load computed using the ASME
gasket design formulation is sufcient to keep the joint leak
tight. The maximum gasket contact pressure for the axisym-
metric case is about one-third of that for the three-dimen-
sional case, as we are allowing pressure to penetrate the
space where contact is lost in the axisymmetric case, and
therefore some of the bolt load is balanced by the pressure
acting on the contact face of the hub. Therefore, the leakage
H. Estrada, I.D. Parsons / International Journal of Pressure Vessels and Piping 76 (1999) 543550 549
Fig. 8. Three-dimensional hub contact surface pressure distribution.
Fig. 9. Axisymmetric gasket contact stress.
results from the three-dimensional case should be used with
caution.
7. Conclusions
The nite element method can be used to study leakage
behavior and to validate the ASME code formulation for
other gasket materials and joint congurations. The
ASME method, although not theoretically exact, is suf-
ciently accurate for all practical purposes, and is far simpler
to implement than the nite element formulation. The
results show that the ASME gasket design guidelines
provide sufcient bolt load to keep the joint leak tight.
However, for critical applications, the nite element formu-
lation can provide greater insight into the behavior of the
particular joint conguration. In the nite element analyses,
we demonstrate that an axisymmetric analysis (which is
more cost effective than a full three-dimensional analysis,
in terms of generating the model and running the analysis)
can be used in analyzing critical designs. However, this
axisymmetric analysis does not capture the stress concen-
trations in the bolt hole area. The leakage behavior, though,
is captured more accurately in the axisymmetric case than in
the three-dimensional case because of the option to allow
pressure to penetrate the space where contact is lost; this
however is not supported in the three-dimensional analysis
in Abaqus at present. Also, the limitation on the lay-up ([0/
90]
4
) for the axisymmetric case must be considered when
using the axisymmetric model.
Acknowledgements
This research is funded by the National Science Founda-
tion, grant number SBC CMS 93-15240.
References
[1] Blach AE, Hoa SV. Bolted ange connections for glass ber rein-
forced platic pipes and pressure vessels. In: Proceedings of the 11th
International Conference on Fluid Sealing, Cannes, France, 1987, pp.
64261.
[2] Graham TE. FRP ange for process pipe and tanks: managing corro-
sion with plastics. In: Tenth Biennial Symposium, November 69,
San Antonio, TX, 1989.
[3] Sun L. Bolted anged connections made of ber reinforced plastic
materials. PhD dissertation, Concordia University, Montreal, 1995.
[4] Godwin EW, Matthews FL, Kilty PF. The design of bolted ange
joints in GRP. Plastics and Rubber Processing and Applications
1986;6:161167.
[5] Estrada H, Parsons ID. A GRP pipe joint for lament wound pipes:
strength analysis and design pressure vessels and piping design,
analysis, and severe accidents. ASME 1996;PVP-331:1319.
[6] Estrada H, Parsons ID. A ber reinforced plastic joint for lament
wound pipes: analysis and design. Civil Engineering Studies, Struc-
tural Research Series No. 616, University of Illinois at Urbana-Cham-
paign, Urbana, Illinois, 1997.
[7] ASME 1996. Boiler and Pressure Vessel Code, Section VIII, Division
2, ASME.
[8] Blach AE. Bolted anged connections with full face gaskets. PhD
thesis, Ecole Polytechnique, Montreal, 1993.
[9] ASME 1996. Boiler and Pressure Vessel Code, Section X, ASME.
[10] Hoa SV. Analysis for design of ber reinforced plastic vessels and
piping, Technomic, 1991.
H. Estrada, I.D. Parsons / International Journal of Pressure Vessels and Piping 76 (1999) 543550 550