Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Proceedings of PVP2007

2007 ASME Pressure Vessels and Piping DivisionofConference


Proceedings PVP2007
July 22-26,
2007 ASME Pressure Vessel2007, San Antonio,
and Piping Texas
Conference
July 22-26, 2007, San Antonio, TX, USA

PVP2007-26643

PVP2007-26643

An Update on Improved Flange Design Methods

Warren Brown
The Equity Engineering Group
Shaker Heights, OH, USA
iwbrown@equityeng.com

ABSTRACT design Code that contained leakage-based design and several other
This paper details further progress made in the PVRC project improvements, such as flange component interaction and gasket
“Development of Improved Flange Design Method for the ASME creep. The European community issued the leakage based flange
VIII, Div.2 Rewrite Project” presented during the panel session on design method in 2001/2, CEN [4, 5] and a standard gasket test
flange design at the 2006 PVP conference in Vancouver. The major method, CEN [6] in 2004.
areas of flange design improvement indicated by that project are
examined and the suggested solutions for implementing the improved The project that is outlined in this paper was adopted to try to
methods into the Code are discussed. Further analysis on aspects such
work as much improvement as possible into the ASME VIII, Div.2
as gasket creep and the use of leakage-based design has been
re-write within the extremely short remaining time frame. The initial
conducted. Shortcomings in the proposed ASME flange design
method (ASME BFJ) and current CEN flange design methods (EN- phase of the project was completed, however, due to time limitations
1591) are highlighted and methods for resolution of these issues are for both authoring and reviewing fairly substantial modifications to
suggested. the Code, the second and subsequent phases of the project were not
continued. The findings of the initial phase will be published as a
Welding Research Council (WRC) Bulletin and this paper offers a
INTRODUCTION summary of the findings and some limited further investigation into
Since the early 1940's in North America and much of the rest of two important aspects of flange design; leakage-based design and
the world, pressure vessel and non-standard piping flanges have been gasket creep.
designed using two gasket factors, “m” and “y” and fairly
rudimentary strength design of the joint components based on the
Taylor-Forge [1] approach. The flange designs have, for the most SUMMARY OF THE PROJECT FINDINGS
part, been very successful. Most joint failures in recent years can be The analysis of existing Code and research work in several
attributed to factors such as temperature effects, gasket degradation significant areas of flange design demonstrate that it is possible to
or joint assembly problems, rather than flange design. With a major greatly improve the flange design methodology in the ASME VIII,
re-write of the ASME VIII, Div.2 Code [2] planned for release in Div. 2 Code, without significantly changing the existing design basis.
early 2007, there was a desire to update the ASME Code with the The existing design methodology, although lacking some crucial
best available technology. The original flange design methods to be areas of joint operational considerations, has a proven track history of
introduced into the re-write included both the traditional and a new producing acceptable flange performance and it was concluded that it
leakage-based design approach (ASME BFJ [3]). However, after is a solid basis upon which to develop an improved method. The
extensive review, the leakage-based approach was removed from the major recommendation is a change from the present method of design
document and the remaining method was, therefore, largely the same using a single load to one of designing with both the minimum and
as the original 1940’s Code. During this same time period, the maximum acceptable loads that the flange will see during operation.
European community wrote and published their improved flange

1 Copyright © 2007 by ASME


Downloaded From: https://proceedings.asmedigitalcollection.asme.org on 06/29/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
The intent of the suggested modifications to the Code method This is evident, when the ratio of ASME BFJ thickness versus ASME
was to provide improvement in areas where advances in technology VIII, Div.2 thickness for a spiral wound, asbestos gasket is plotted in
and testing may have enabled an aspect of joint design to be Fig. 3. It can be seen that the leakage-based method indicates a much
considered, whereas previously it may have been neglected. An thicker flange being required (almost twice the present thickness) for
example of this sort of area is in the mechanical and creep behavior the smaller diameter joints and, by comparison, a thinner flange for
of gaskets. One of the initial activities in the project was to examine the larger diameter flanges. This is counter-intuitive to what would be
the existing flange design methods that are either presently published, expected. It should be noted that, for a given diameter, the flange
or under research for publication, in order to determine what possible thickness will generally be proportional to the square root of the bolt
elements of improvement could be addressed in the ASME VIII, Div. load. The example presented in Fig. 4 demonstrates the opposite end
2 re-write. of the spectrum, where the leakage-based method predicts a flange
thickness of around one half the present method. This is particularly
The findings of this initial study are outlined in Table 1, which concerning, given that the gasket in question (a 1/32 in. thick glass
provides a summary of the areas of joint design that are addressed by fiber gasket) is particularly susceptible to uneven loading caused by
each of the available methods. It can be seen that the CEN Codes do flange deflection between bolts and also due to flange rotation. It,
appear to incorporate many more aspects of joint behavior into the therefore, does not seem that the reduction in flange thickness made
flange design process. The items outlined in the table were assessed possible by using leakage-based design would be advisable in this
for inclusion into the ASME VIII, Div. 2 re-write and suggestion for case.
improvement in a total of 19 areas was made. These areas are
outlined following, with a brief explanation of the deficiencies in the An additional concern with leakage based design, where
present methods and recommendations for improvements. The full tightness limits are used, is that it is impractical to imply that a room
background may be obtained in the forthcoming WRC bulletin of the temperature laboratory test using helium and flat platens can have
project. The intent of this paper is to provide only a brief overview of any relevance to field leakage of a flanged joint. There is simply too
the findings. many other variables that will influence actual leakage levels that are
not accounted for in the available methods (the forthcoming WRC
Flange Design Methodology bulletin has a list of 13 such variables). One reviewer of the WRC
The most significant change proposed in the project was to go bulletin put it best when he stated that leakage based design “can not
from the existing method of designing a flange with little be used to predict leakage rates, but could be used by lawyers to
consideration of actual joint operation, to a method that incorporates hang the flange designer”.
assessment of the full joint operational cycle (as per the intent of
EN1591:2002 [4]). This allows a more comprehensive assessment of Due to the numerous concerns with the existing leakage-based
the likelihood of successful joint operation than the present method of flange design methods, it was proposed that a simpler approach of
flange design. The required considerations for flange design using
using leak rate from the standard ROTT tests [7, 8] to determine “m”
this operational limit approach can be seen in Fig. 1. This approach
& “y” factors. Since the bulk of gasket research to date has been
can be applied to flange design in a simplified way for which the
performed using the PVRC ROTT test method (over 300 gasket
final result is a measure of the gap between the bolt load required to
meet the minimum gasket stress and the maximum permissible bolt leakage test results are available) it was proposed that any new
load, as shown in Fig. 2. The intent would be that in order to design a method be verified for suitability using those tests results. However,
flange, the designer must know the assembly efficiency category and once this was attempted, it was found that the results gave a
the minimum allowable percentage of bolt yield between limits. All surprisingly high level of variability, even among similar gaskets. A
other factors would then be calculated from gasket test data. A selection of the results, from ROTT 9 [8] test data are found in
suitable acceptable percentage bolt load between limits for most Figures 5 to 9. These figures show different ROTT test results, for
designs would be selected and might be in the order of 20% of bolt each test a value of “y” is calculated for a given leak rate. In this way
yield, for example. it is possible to demonstrate the effect of basic test data variation.

Minimum Gasket Stress Based on Leakage Figure 5 shows test results from expanded PTFE (ePTFE)
Both ASME BFJ [3] and CEN [4, 5] contain methods of gaskets. It can be seen that the variation in test results is not very
determining the minimum required gasket stress by leakage-based high, however when the value of “y” is calculated for a leak rate of
methods. However, these methods are largely untried and may not 0.01 mg/mm/sec He, using the method outlined in Brown [9], it can
prove to be effective in producing leak-free joint designs. Preliminary be seen that there is a very large variation in obtained value (from
work in the initial phase of the project demonstrated that the ASME 6.0ksi to 10.5ksi). Given the tight grouping of the results, this seems
BFJ method produced results that were highly variable by surprising, but is most likely explained by the fact that in order to get
comparison to existing flange design methods, with the required a “y” value that is close to the present Code value, a rather low leak
flange thickness ranging from half to almost twice the thickness rate must be used. This results in the need to extrapolate the results
determined using the present ASME VIII, Div. 2 design method. into gasket stress levels that are lower than actually tested by the
What was particularly concerning is that the trends of the data did not ROTT test. Since the extrapolation is being performed on a log scale,
follow a logical pattern in terms of additional conservatism appearing it is highly prone to small differences in the test results. In effect, the
for larger diameter flanges, as would be expected based on industry ROTT test is conducted at leak rates that are two to three orders of
experience that larger diameter flanges tend to be more troublesome. magnitude less than the leak rates that are encountered in the field.

2 Copyright © 2007 by ASME


Downloaded From: https://proceedings.asmedigitalcollection.asme.org on 06/29/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
This effect was highlighted by Sawa and Kobyashi [10] during the realistic leakage-based tests (such as the Japanese test) be examined
2006 PVP and is the reason that the Japanese have adopted a new in order to try to obtain a standardized test for the determination of
tested, which is conducted at higher leak rate levels (closer to reality). the Code gasket factors (“m” and “y”). To aspire to the design of
flanged joints to a specified leak rate shows a lack of practical
The same trend can be seen in Fig. 6, which is for class 150 perspective and will result in many difficulties with sealing in the
spiral wound, graphite filled gaskets. The results are reasonably well field, due to the fact that the ROTT results are not a good indicator of
grouped, and yet the obtained value for “y” varies from 7.5ksi to the gaskets field performance.
11ksi. It could be argued that there is one gasket that does appear to
outperform the others (in the room temperature, flat platen test), Gasket Creep/Relaxation
however it must be kept in mind that the object of flange design is to The CEN Code [6] has introduced a factor to account for short-
establish a suitably robust flange thickness that will accept a range of term creep/relaxation of the gasket. However, the test is conducted at
gaskets. It would not be wise to design a flange using the 7.5ksi a single flange stiffness value (although other stiffness values are
gasket, only to find that gasket did not perform well in the field and optional) and so therefore does not fully differentiate between gasket
now a gasket with a “y” of 11ksi was needed. Designing a flange for creep (constant stress) and gasket relaxation (constant deflection)
one particular gasket is not a desirable solution, particularly when the modes of gasket stress loss. If the joint being designed has a
only measure we are using to compare the gaskets is room significantly different rigidity from the test apparatus rigidity then the
temperature helium leakage tests on flat platens. As already pointed results would not be applicable. In addition, the effects of short-term
out, this has almost no correlation to the actual performance of the relaxation after 4 hours will be minimal for gasket types other than
gasket in the field. sheet gaskets. For typical corrugated, kamprofile or spiral wound
gaskets, a 4 hour test would, at best, only pick up around 60% of the
Figure 7 could be used to justify the need for leakage based total gasket stress creep/relaxation that could be expected to occur in
testing, due to the fact that one compressed fiber gasket requires a joint and, at worst, the test would only be measuring around 20% of
almost twice the seating stress as the other. But the question still the final relaxation (Marchand [11]). It is important to differentiate
remains; Is this an appropriate measure for comparing gasket between the effects of relaxation and creep when considering the
performance? In addition, if the test results are compared for the two gasket load lost during operation and have a test that determines the
identical gaskets, it is shown that the obtained “y” value varies from total relaxation that can be expected.
4.5ksi to 8.5ksi. This illustrates very well the errors introduced by
testing at much lower leak rates than reality, as the result is almost a This is evident when gasket relaxation test results of Kamprofile
doubling of the “y” value between identical tests. This is caused by (Fig. 9) and filled PTFE gaskets (Fig. 10) in a joint with one gasket,
the fact that a small variation between results, when extrapolated to a versus relaxation in a tubesheet joint (two gaskets) are compared. If
significantly different stress or leakage range using a log-log pure relaxation occurs then the gasket stress lost would be identical
function, will result in a significant difference in final obtained result. for both configurations. If pure creep occurred, then the gasket stress
loss would be proportional to the joint flexibility. Using the flange
The results in Fig. 8 were included to highlight the point that flexibility equations outlined in Brown [17], it is possible to compute
designing a flange using a leakage based approach should be done in the ratio of gasket stress loss due to pure creep. Therefore, in this
generic terms, without the presentation of individual gasket results or case for the type PG gasket, the tubesheet joint should lose between
the option of using actual gasket test data if it is based solely on room 1.0 to 1.86 times the gasket stress by comparison to a single gasket
temperature leakage. In this case, three corrugated, graphite filled joint (1.0 < ΔSgtbsht/ΔSgsingle < 1.86). It can be seen in Fig. 9 that, prior
gasket test results are presented. Two have similar results (“y” = to hot torque, the stress lost for the single joint was 8% of the initial
8.5ksi and 9.5ksi) and one appears to be much tighter ( “y” = 5.5ksi). load, and the stress lost for the tubesheet joint was 14%. The actual
It is easy to imagine the flange designers selecting (and specifying) ratio of loss was therefore 1.75, which is close to the 1.86 value,
the tighter gasket, as this would result in a 25% saving in flange indicating that creep appears to dominate in this joint for this gasket
thickness. However, if the gasket construction is examined, it can be type. The more flexible PTFE gasket should have a ratio between 1.0
seen that the tighter gasket actually incorporates a section of PTFE and 1.72 (1.0 < ΔSgtbsht/ΔSgsingle < 1.72). It can be seen (Fig. 10) that
into the design, which explains the low leakage test results. If the the actual ratio is 1.38, which indicates that the mode of gasket stress
gasket was placed in normal refining service (450°F to 650°F) loss is fairly evenly distributed between creep and relaxation.
temperature range, the PTFE would not be effective, and most likely
this gasket would perform worse than the other two gaskets that A third interesting case is also included (Fig. 11), where Belleville
would have been discounted due to poorer room temperature leakage washers are used to modify the flexibility of the joint. The expected
performance. It is interesting to note that the gasket style that tested gasket stress relaxation for the joint with Belleville washers versus
better in the ROTT test is no longer available on the market, whereas the joint without Belleville washers is between 0.86 and 1.0 (0.86 <
the other two are still available. If leakage based flange design had of ΔSgBelle/ΔSgsingle < 1.0), depending on which relaxation mode
been already written into the vessel design Codes, then it is most dominates. The actual test results showed almost no difference
between the loss of gasket stress, with a ratio of only 0.97. This
likely that gasket would hold a significant market share today.
indicates that, for this case, the relaxation mode of gasket stress loss
dominates. From these examples, it can be seen that it is important to
In light of these results, it is recommended that the purpose of account for joint flexibility when considering gasket creep/relaxation.
leakage-based design be re-visited and that research into more

3 Copyright © 2007 by ASME


Downloaded From: https://proceedings.asmedigitalcollection.asme.org on 06/29/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
It is, therefore, proposed that the Code incorporate a series of gasket classified as “secondary stresses” per the definition of ASME VIII,
relaxation/creep tests over a range of joint flexibility values and Div.2 Appendix 4 (ASME [2]). This is due to the fact that the stresses
temperature values to enable the flange designer to be able to select a are due the deformation of the flange to maintain local force
single value of percentage of the initial gasket assembly stress that equilibrium and, since the assembly load acts as a moment and has
will be lost due to gasket creep/relaxation for the particular case of limited flange deflection due to the geometry of the arrangement, it
joint flexibility for which they are designing. To attempt to introduce will not result in yield of the entire flange/hub/shell cross section.
a method that separates out the relaxation effects and creep effects
Taking the ASME stress hopper approach one could therefore argue
and then uses equations to calculate the design percentage loss in
that the secondary stress limit of 3Sm (twice yield) applies. This limit
gasket load would only be introducing unnecessary complication.
also corresponds to the elastic stress limit under which shake-down is
assumed to occur (Spence [13]). Since the present maximum
allowable stress is 1.5Sm (for the longitudinal stress) then twice that
Flange & Bolt Stress Basis
limit (3Sm) would seem to imply that elastic shakedown will occur
One of the major concerns with achieving leak free operation of
and therefore flange mechanical failure would not be expected. In
flanges is determining an appropriate bolt assembly stress. Presently,
addition to this justification, it should be noted that the application of
there is a disconnect between the ASME VIII design bolt load
a stress based limit to elastic shake-down is conservative in the case
(approximately 25% of bolt yield) and normal industry practice for
of a flange. This is due to the fact that the flange arrangement is
assembly of joints to approximately 50% of bolt yield (ASME PCC-1
deflection controlled, rather than load controlled. In other words, the
[12]). This disconnect contributes greatly to the incidence of leakage
bolt length is fixed after assembly and so further flange deflection
in the field. Experience has shown that it is necessary to assemble
will result in a diminished bolt load. This concept explains why
joints to around twice the ASME VIII design bolt stress in order to
industry experience is that properly designed, ductile flanges do not
achieve satisfactory operation. This is due to the fact that the present
mechanically fail even at very high assembly bolt loads. Further
ASME VIII flange design loads are too low to account for all aspects
discussion of this approach can be found in Brown [14].
of flange operation, such as gasket creep/relaxation and the effects of
temperature on bolt load. Since subsequent parts of the project
Maximum Allowable Gasket Stress
suggest it would be wise to include these factors in the design
The CEN [4] method of flange design limits the maximum
calculation, then if the allowable stresses are left unchanged, the
gasket stress based on a test that looks for a compressive failure of
effects of those factors will need to be adjusted from actual test
the gasket construction. This could happen in non-homogeneous
values to a value relative to the design bolt load. Essentially, the
gasket construction, but is unlikely to occur in the majority of
changes in gasket stress will need to be divided by two in order that
(homogeneous) gasket materials. In addition, the test rig used has flat
their effect relative to the design bolt load (versus the actual assembly
platens and so therefore the ability of the gasket to tolerate flange
bolt load) is consistent.
rotation is not established during the test. The ASME BFJ [3] method
of testing for an increase in leakage rate may detect failure in both
Alternatively, (and preferably) the design bolt stress could be
homogeneous and non-homogeneous gaskets, but it is only conducted
adjusted until it corresponds with a more reasonable assembly bolt
at room temperature with flat platens, and so therefore has little
load (e.g.: multiplied by two). However, if that approach was used
relevance to gasket operation in the industrial setting. The
then it would be necessary to also adjust the flange allowable
deficiencies in both approaches are shown in Fig. 12 . The gasket
stresses, otherwise the relative geometry of the flange assembly
material is expanded PTFE (ePTFE), which was assembled to a
would be changed (with very much smaller bolt diameters for the
gasket stress (full width basis) of 5ksi and run for several hours at
same thickness of flange). The most appropriate solution appears to
350°F in the RA.S.T test rig (24 in. raised face, heat exchanger style
be to multiply the bolt and flange allowable stresses by two in order
flange). The CEN method (CEN EN13445-3:2002, table G.9.1) lists a
to be performing the design calculations with an assembly bolt load
Qsmax of 22ksi (at 200°C = 392°F) and the ASME BFJ Table 6.1A
closer to that used in the field. It should be noted that this will, in
lists a Sc of 20ksi (15.4ksi on full width for this case), for this gasket
effect, halve the load that can be applied to the gasket in the field and
material. Obviously, neither of the two methods gave acceptable
so some adjustment to the required gasket assembly stress may also
results for application in the field, as the gasket cross-section has
be required (although that is offset somewhat by the inclusion of
failed at 5ksi and leakage would be expected.
additional factors, such as gasket creep).
Therefore, from field experience it is evident that the method of
Designing with twice the flange and bolt stress may not seem
testing for maximum allowable gasket stress should include both the
appropriate at first, due to the fact that the calculated flange stresses
effect of temperature and the effect of flange rotation. The maximum
may be in excess of yield (twice yield in the longitudinal direction).
allowable stress may be obtained by conducting a test at elevated
However, the magnitude of the flange stresses should not be the only
temperature and measuring either an increase in gasket leakage or a
aspect considered, as likelihood that those stresses will cause
decrease in the elastic recovery of the gasket (or both) in order to
structural failure of the flange or shell is a more important aspect to
determine the maximum allowable gasket stress (versus temperature
consider. Since the flange stresses are local to the flange ring/hub or
and flange rotation) that is permissible.
hub/shell junction locations and are primarily caused by bending
stresses due to the bolt load, they tend to diminish rapidly through the
hub or shell thickness. In addition, the flange stresses may be

4 Copyright © 2007 by ASME


Downloaded From: https://proceedings.asmedigitalcollection.asme.org on 06/29/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Effects of Temperature
The CEN [5] Code has introduced a calculation of the effect of Gasket Effective Width
steady-state differential axial expansion of joint components into the The present ASME Code equations for effective gasket width, while
calculation of the minimum and maximum gasket load. The method relatively simple, do not have a sound technical justification.
does not include the calculation of the component temperatures, Therefore it is difficult to know if this approach can be taken to apply
which is critical to the end result. The analysis also does not include to new gasket types. The implementation of a reasonably simple
the effects of differential radial expansion, which can cause flange equation, such as that included into EN13445-3 [4] would improve
rotation and bolt load loss. Essentially, it only serves to highlight the flange design. However, it would be best to first examine if the
CEN equation gives significantly different results to the simplified
potential poor design practices with respect to dissimilar materials
ASME VIII equation, including comparison to FEA. If the results are
selection of the bolts versus the flanges.
not proven to be sensitive to flange rotation or gasket modulus then
there is little to be gained by the added complexity.
WRC Bulletin 510 [15] contains a complete calculation
Effect of Bolt Hole Spacing
method (using a graphical approach for determining the thermal
This effect is primarily important with thin sheet gaskets on
resistance of components) for both the static and transient effects of
low pressure, large diameter flanges, where flange flexibility may
temperature on the joint. Calculation of both the component
result in insufficient gasket stress between the bolt locations. The
temperatures and the resulting gasket stress levels during different
CEN method [5] uses the gasket unloading modulus to limit the
times of operation is possible. The method has been verified against
variation in gasket stress. However, this does not take into account
numerical modeling (FEA), laboratory tests and field measurements
the fact that often the loading portion of the gasket elastic behavior is
on operating heat exchanger and piping joints. However, the very much different from the unloading portion. Since the effect of
calculation method is not simple and experience has shown that in the bolt spacing is relevant to the gasket loading (assembly) then a more
majority of cases such analysis is not required, as the effect of appropriate limit may be to limit the allowable deflection of the
temperature is not highly variable and can be approximated by flange ring between bolt holes based on a percentage of the gasket
assuming a bolt load of approximately 10% of bolt yield is required loading compressive deflection. Further study into an appropriate
to counteract the effect. modulus or limit to allowable deflection is required before these
limits are introduced into the ASME Code.
It is, therefore, suggested that the Code include guidelines
(limits) for cases when analysis of the effects of temperature is Hub/Shell Junction Stress Check
recommended. Such as – dissimilar material joints, dissimilarly During two studies of maximum flange stresses (Brown [14]
heated joints (tubesheet joints) and flanges with thick vessel shells by and Brown [18]), it was found that the present Code stress checks,
comparison to the flange thickness. For those cases, the analysis which do not include a tangential (circumferential) stress check at the
method outlined in WRC 510 could be used to determine the effect of shell/hub junction location, were not sufficient for defining when
temperature. Suggested limits for when the WRC 510 analysis would gross plastic deformation (GPD) of the flange would occur. During a
be required are listed following: subsequent study of piping flanges using Elastic-Plastic finite
1. Joints with material of dissimilar coefficients of expansion element analysis (FEA), there where several flange configurations
2. Joints where the vessel wall thickness exceeds one fifth the where GPD occurred very near the design bolt load. This problem
flange thickness. was further investigated in Brown [19], where equations for the
3. Joints with dissimilar operating temperatures across the tangential stress are the shell to hub intersection are developed and
joint (tubesheet joints) limits for the flange stresses are determined by comparison to elastic-
4. Joints with a (design temperature x flange OD) of greater plastic FEA results. These limits must be verified for non-standard
than 12,000 °F.in. (i.e: a 24in. flange at 500°F). flange configurations, but if they are found to be applicable, then it
5. Joints where thermal shock loading is expected as part of would be recommended to incorporate them into the ASME Code
normal operation flange design, as a check to ensure that the flange was never the
limiting component, and that the full bolt load (until yield) could be
applied onto the flange before GPD occurs. In this way, the flange
Mechanical Interaction design is ensured to be sufficiently robust for field operation.
The aim of incorporating mechanical interaction into the Code is
to accurately calculate the loss in gasket stress due to pressure. The Effect of Bolt Holes on Rigidity
present ASME Code assumption is that the bolt load remains constant Including a reduction in flange rigidity due to the bolt hole
as pressure is applied, which may be non-conservative, particularly would be important in relatively thin flanges where the bolt hole
for larger diameter joints with stiffer gaskets. The original work by represents a significant portion of the flange ring cross-section. Since
Wesstrom [16], can be adapted to relatively easily incorporate the EN13445-3:2002, Appendix G method will, in general, result in
mechanical interaction into the present Code equations. A thinner flanges it may be proven that such a check on ASME VIII
development of this is shown in Brown [17]. Incorporation of these designed flanges is unwarranted, unless the proposed changes result
equations into the Code are required, as failure due to inadequate in thinner ASME VIII flanges for certain cases. The decision whether
design for pressure does occur on larger diameter joints. to include such a check into the ASME Code is best based on a study
of the relative effect of the bolt hole on the flange rotational rigidity.

5 Copyright © 2007 by ASME


Downloaded From: https://proceedings.asmedigitalcollection.asme.org on 06/29/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Once such a study has been conducted, it will be possible to add a paragraph addressing this effect to the gasket creep section of
determine if it is necessary to perform the check for the effect of bolt the Code.
holes on flange rigidity.

Nubbins Prohibited Table of Standard Bolt Root Areas


The CEN Code [4] does not allow the use of nubbins and there It is recommended that the next ASME Code revision include a
is additional poor American industry experience with nubbins (Veiga table of bolt root and tensile areas for standard bolt sizes.
[20]). It is therefore recommended that nubbins be removed from the
flange facing options in the ASME Code.
Seal Welded Joints
It is recommended that the next ASME Code revision include
Effect of Assembly Method seal welded joints into the flange design cases already covered.
Since it is the proposal of this project that the ASME VIII,
Div.2 re-write include a limit check on gasket/bolt stress (minimum
to seal and maximum against joint component damage) then it would Lap Joint Stub End Stresses
seem appropriate that some measure of the assembly efficiency be In general, lap joint flange stub ends designed to ASME VIII
included in this assessment. However, the actual level of assembly have given good service without a history of failures due to stresses
efficiency is dependant not only on the technique used to tighten the at the stub end. However, the premise of only performing a check on
bolts, but also on the procedure (number of tightening passes, how the stub shear stresses is that the stub is not placed in significant
quickly the bolt stress is increased, etc…) itself (Brown [21]). It is, bending by the bolt load. Since there are no dimensional limitations
therefore, inappropriate to relate assembly efficiency only to the in the Code, it is possible that a design may meet the Code, but the
tightening method, without consideration of the procedure being stub end may be placed in bending and cause joint leakage. While it
employed. is relatively easy to analyze the stub end as a flange, it is
recommended that the approach to be taken in the ASME VIII, Div.2
An additional complication to this is that the flange designer re-write is to place limitations on the flange/stub dimensions and
will often have no knowledge of the future joint assembly method or placement of the gasket, such that the design premise (of the stub
procedure at the time of designing the flange. It is proposed, therefore only being placed in shear) is always satisfied. However, include a
that assembly categories (rather than methods) be included in the bearing and bending check on the stub end, similar to EN13445-
ASME VIII, Div. 2 re-write. When designing the flange, a “normal 3:2002, Section 11.0 would not be difficult to implement.
assembly” category should be used, corresponding to torque control
techniques utilizing a good procedure, such as that outlined in ASME
PCC-1. Optional higher and lower assembly procedure levels could Gasket Width Limitations
be specified or allowed by the vessel owner. In the present Code it is possible to design a gasket that is
impractical or difficult to manufacture or handle in the field. Spiral
wound gaskets are particularly problematic at large diameters and
Operational Flange Rotation Limits become instable if the gasket sealing element width is either too wide
It has been found that flange rotation during operation, due to thermal or too narrow. By tapping into gasket manufacturer experience, it
loading or tubesheet deflection caused by differential pressure, can should be possible to provide better guidelines for maximum and
cause joint leakage (WRC 510 [15], Appendix D). A check for the minimum recommended gasket widths. Such a table should be set out
maximum change in flange rotation during operation (rotation in versus gasket type, as the width limitations are often dependant on
excess of flange rotation after assembly) is a relatively easy check to gasket type.
add to the Code and will reduce the incidence of flange leakage,
particularly in tubesheets operating with differential pressure.
CONCLUSIONS
This article has summarized the results of a recent study aimed at
Flange & Bolt Creep
improving the ASME flange design methods. In addition, new
The effect of bolt and flange creep/relaxation is relatively minor,
material on leakage-based design and creep have been included to
by comparison to gasket creep/relaxation, which also occurs at much
assist in clarifying the position taken on those topics. While the list of
lower temperatures. On this basis, these effects are excluded from the
possible additions to the Code is extensive, it is believed that most of
calculation performed in EN13445-3:2002, Appendix G / EN1591. It
the modifications could be quite easily incorporated and all would
is not envisaged that it will be necessary to include a calculation
result in a net reduction in the incidence of flange leakage in the field.
procedure on these effects in the ASME VIII Code. For flanges and
bolts operating in the creep range, a rough creep assessment may be
performed using the bolt load subsequent to gasket relaxation in order
REFERENCES
to determine an expected final gasket load. Since the number of cases
[1] Waters, E.O., Rossheim, D.B., Wesstrom, D.B., Williams, F.S.G.,
where this will be applicable are small, it is considered sufficient to
1949, “Development of General Formulas For Bolted Flanges”,

6 Copyright © 2007 by ASME


Downloaded From: https://proceedings.asmedigitalcollection.asme.org on 06/29/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Taylor-Forge & Pipe Works, Southfield, Michigan, Reprinted by the ICPVT11-93094, Proceedings of the ASME PVPV conference,
PVRC in 1979. Vancouver, Canada

[2] ASME. 2004, ASME VIII, Div 2, Boiler and Pressure Vessel [19] Brown, W., Reeves, D., 2007, “An Update on Selecting the
Code, American Society of Mechanical Engineers, NY, USA Optimum Bolt Assembly Stress for Piping Flanges”, PVP2007-
26649, Proceedings of the ASME PVPV conference, San Antonio,
[3] ASME-BFJ draft 2000 (2000) "Background and Commentary on TX, USA
Appendix BFJ - New Rules for bolted Flanged Connections with Ring
Type Gaskets", Internal correspondence of ASME Special Working [20] Veiga, J., Kavanagh, N., 2005, “Double Jacketed Gaskets for
Group BFJ. Heat Exchanges Sealability Behavior in Flanges With and Without
Nubbin”, PVP2005-71023, Proceedings of the ASME PVP, Denver,
[4] EN13445-3:2002, “Unfired Pressure Vessels – Part 3: Design” Colorado.
CEN
[21] Brown, W., 2003, “Recent North American Research into
[5] CEN EN1591-1:2001 "Design Rules for Gasketed Circular Several Pressure Vessel Bolted Joint Integrity Issues”, Proceedings of
Flange Connections", European Community for Standardisation Task the 10th ICPVT Conference, Vienna, Austria
Group TC47, CEN
[6] EN13555:2004, "Flanges and their joints - Gasket parameters
and test procedures relevant to the design rules for gasketed circular
flange connections", European Community for Standardisation Task
Group TC47, CEN
[7] ASTM Draft 10 Procedure (2001) "Proposed ASTM Method Draft
No 10 of the standard test method for gasket constants for bolted
joint design", Internal correspondence of the ASTM Committee F03
[8] ASTM Draft 9 Procedure (1994) "Proposed ASTM Method Draft
No 9 of the standard test method for gasket constants for bolted joint
design", Internal correspondence of the ASTM Committee F03
[9] Brown, W., 2005, “Improving and Simplifying Leakage Based
Flange Design”, PVP2005-71342, Proceedings of the ASME PVP,
Denver, Colorado, USA
[10] Sawa, T., Kobyashi, T. Presentation during the ASME PVP
2006 in Vancouver during the Panel Session on Flange Design
Methods.
[11] Marchand, L., Derenne, M., Sakr, O., Bouzid, A-H., 2006,
“Long Term Pressurized Graphite Gasketed Joint Tests”, Welding
Research Council Bulletin 507
[12] ASME PCC-1. 2000, “Guidelines for Pressure Boundary Bolted
Flange Joint Assembly”, American Society of Mechanical Engineers,
NY, USA
[13] Spence, J., Tooth, A. 1994, “Pressure Vessel Design, concepts
and principles”, E & FN SPON, London, UK
[14] Brown, W., 2006, “Flange Assembly Bolt Load Selection Based
on Leak before Break Analysis”, PVP2006-ICPVT11-93075,
Proceedings of the ASME PVPV conference, Vancouver, Canada
[15] Brown, W., 2006, “Analysis of the Effects of Temperature on
Bolted Joints”, Welding Research Council Bulletin 510
[16] Wesstrom, D.B., Bergh, S.E., 1951, “Effect of Internal Pressure
on Stresses and Strains in Bolted-Flange Connections”, Transactions
of ASME, 73, n.5, pp 508-568, ASME, NY, USA
[17] Brown, W., 1993, “Design and Behaviour of Bolted Joints”
Proceedings of the 3rd International Conference on Fluid Sealing,
CETIM, Nantes, France, pp. 111-113
[18] Brown, W., Reeves, D., 2006, “Considerations for Selecting the
Optimum Bolt Assembly Stress for Piping Flanges”, PVP2006-

7 Copyright © 2007 by ASME


Downloaded From: https://proceedings.asmedigitalcollection.asme.org on 06/29/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
SS Spiral Wound (Asbestos) - T3, 250psig, Am (BFJ) / Am (ASME), (3, 10ksi)
Gb = 3400, Gs = 93 , a= 0.3, η= 0.85
Sl = 900, Tpmax =500
3.75
3.50-3.75
3.50
3.25-3.50
3.25
Required Bolt Area Ratios (BFJ/ASME)

3.00-3.25 3.00
2.75-3.00 2.75
2.50-2.75 2.50 Sizes 36"
2.25-2.50 2.25 & above
2.00-2.25 2.00 Tpm ax
1.75 Exceeded
1.75-2.00
1.50
1.50-1.75 1.25
1.25-1.50 1.00
1.00-1.25 0.75
0.25

0.75-1.00 0.50
0.5

0.75

0.25
0.50-0.75
1

0.00
1.5

0.25-0.50
2

Gasket Width (in.)


2.5

24

0.00-0.25
3

48
3.25

Gasket ID (in.)
78

Figure 3 – Spiral Wound Asbestos Gasket Seating Stress Comparison

8 Copyright © 2007 by ASME


Downloaded From: https://proceedings.asmedigitalcollection.asme.org on 06/29/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
1/32" thk Glass Fiber Gasket - T2 Tightness, 50psig, Am (BFJ) / Am (ASME)
Gb = 285, Gs = 117 , a= 0.45, η= 0.85
Sl = 900, Tpmax = 26000
1.25

Required Bolt Area Ratios (BFJ/ASME)


1.00

1.00-1.25 0.75
0.75-1.00
0.50-0.75
0.50
0.25-0.50
0.00-0.25
0.25
0.25

0.5

0.75

0.00
1.5

78
Gasket Width (in.)
2.5

48
3

24
3.25
6
Gasket ID (in.)

Figure 4 – 1/32 in. thick Glass Fibre Gasket Seating Stress Comparison

12000
12000 80 (y = 7.5ksi)
10 (y = 6ksi) 89 (y = 8.5ksi)
10000 92 (y = 11ksi)
10000 106 (y = 9.5 ksi)
72 (y = 11ksi)
24 (y = 10.5 ksi)
8000
8000
Gasket Stress
Gasket Stress

6000
6000

4000 4000

2000 2000

0 0
1E-06 1E-05 0.0001 0.001 0.01 0.1 1 10 100 0.0001 0.001 0.01 0.1 1 10
Leak Rate Leak Rate

Figure 5 – 1/16 in. thick expanded PTFE Gasket (Lr=0.01) Figure 6 – cl.150 Spiral Wound Gasket (Lr=0.003)

9 Copyright © 2007 by ASME


Downloaded From: https://proceedings.asmedigitalcollection.asme.org on 06/29/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
16000
16000 83 (y = 9.5ksi)
14000 85 (y = 5.5ksi)
49 (y = 16.5ksi)
14000
190a (y = 8.5ksi) 77 (y = 8.5ksi)
12000
12000 190b (y= 4.5ksi)
10000

Gasket Stress
Gasket Stress

10000
8000
8000
6000
6000
4000
4000

2000 2000

0 0
1E-07 1E-06 0.00001 0.0001 0.001 0.01 0.1 1 10
0.000001 0.0001 0.01 1 100 Leak Rate
Leak Rate

Figure 7 – 1/16 in. thick Comp. Fiber Gasket (Lr=0.01) Figure 8 – Corrugated Graphite Gasket (Lr=0.001)

14 50
Hot Torque
3.5 90
13 45
Start of RA.S.T. Thermal Cycles
80
12 40 3

70
11 35
2.5
% Gasket Stress Lost
Gasket Stress (ksi)

10 30
60
Gasket Stress (ksi)

Relaxation (%)
2
9 25 50

8 20 40
1.5

7 15 30
1
6 10
20

5 Two Gaskets - Gasket Stress One Gasket - Gasket Stress 5 0.5


ksi - One Gasket, Test 9 ksi - Two Gaskets, Test 14 10
Two Gaskets - % Cumulative Loss One Gasket - % Cumulative Loss
4 0 % - One Gasket, Test 9 % - Two Gaskets, Test 14
0 2 4 6 8 10 12 14 16 18 20 0 0
Time (hours) 0 1 2 3 4 5 6
Time (hours)

Figure 9 – Corrugated Kamprofile Gasket Creep Comparison Figure 10 - Filled PTFE Creep Comparison
3.5 70

3 60

2.5 50
Gasket Stress (ksi)

Relaxation (%)

2 40

1.5 30

1 20

0.5 10
ksi - Without ksi - With
% - Without % - With
0 0
0 1 2 3 4 5 6
Time (hours)

Figure 11 – ePTFE with Belleville Creep Comparison Figure 12 – Failed ePTFE Gasket due to Flange Rotation

10 Copyright © 2007 by ASME


Downloaded From: https://proceedings.asmedigitalcollection.asme.org on 06/29/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Table 1 – Comparison of Present Flange Design Codes
ASME ASME ASME EN13445- EN1591:200
VIII, div. 1, VIII, div. 2, Append. BFJ 3:2002, 13
2
App 2 New Rules Sect. 11
Aspect of Flanged Joint Design 1

Flange Design Basis Taylor- Taylor- Taylor- Taylor- TGL


Forge Forge Forge Forge 32903/13
Includes effect of joint mechanical interaction Partial 4 Partial 4 Partial 4 Yes
Flange Stress Check Yes Yes Yes Yes Partial 5
Flange Rotation Check Yes Yes Yes Yes
Check on Lap Joint Stub Shear Stress Yes Yes Yes
Check on Lap Joint Stub Bearing Stress Yes
Check on Lap Joint Stub Bending Stress Yes Yes
Design of Seal Welded Joints Yes
Gasket Loads based on “m” & “y” Yes Yes Yes
Gasket Loads based on leakage Yes Yes
Flange Allowable Stress Basis ST/3.5, ST/2.4, ST/2.4, 6 ST/2.4, Sy/1.5
Sy/1.5 Sy/1.5 Sy/1.5
Austenitic Allowable Stress Increase Allowed No Yes 7 No Yes (?)
Bolt Allowable Stress Basis ST/4, 8 ST/4, Sy/1.5 ST/4, Sy/3 9 ST/2.4, Sy/1.5
Sy/1.5
Gasket Effective Width Basis Simplified Simplified Simplified Simplified Calculate
Includes Gasket Creep/Relaxation Partial 10
Includes Flange & Bolt Creep
Included Effects of Temperature Partial 11
Includes External Moments & Forces Yes Yes Yes
Maximum Allowable Gasket Stress Yes Yes
Maximum Spacing Between Bolts Yes
Effect of bolt holes on flange rigidity Yes
Operational Flange Rotation Limits
Adjusts for Assembly Accuracy Yes Yes
Nubbins Prohibited Yes
Table of standard bolt stress areas Yes
Gasket Width Limitations Partial 12 Partial 12 Partial 12 Partial 12

1
Based on the revision 7 of the document (current as of 1st January, 2006)
2
Based on the draft document dated February 15, 2006
3
This is also the basis of EN13445-3 Appendix G and some listed aspects (flange stress limits for example) are taken from this appendix, as they are
not specified in EN-1591.
4
It can be argued that the factor “m” accounts for the effects of mechanical interaction (ref. Brown [6]).
5
The stress check in EN1591 includes only a check of the circumferential stresses and flange is allowed to have plastic deformation (ref. EN1591,
1.3.4 b). The other methods include radial and tangential stress checks and use an elastic stress check.
6
Note that due to experience with problems at the higher allowable stresses in large diameter joints, the allowable is reduced by a factor of 0.75 for ≥
2000mm (78in.) diameter flanges. For diameters between 1000mm and 2000mm (39in. and 78in.) this reduction factor is taken to linearly vary from
0.75 to 1.0.
7
The basis for allowing higher allowable stresses for austenitic stainless is that the flange rotation limits should eliminate concerns regarding overly
flexible flanges when designed to higher allowable.
8
Actually, the ASME II, Part D tables list allowable stresses for common materials that are closer to ST/5.
9
Note that the yield value for austenitic bolts is taken at an elongation of 1.0%, rather than 0.2%.
10
The effects of short term relaxation only are included in the present revision of EN13555.
11
Includes only axial expansion and does not detail how to determine temperature.
12
Only considers gasket minimum allowable width. No consideration of maximum width limitations.

11 Copyright © 2007 by ASME


Downloaded From: https://proceedings.asmedigitalcollection.asme.org on 06/29/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

You might also like