Você está na página 1de 10

Proceedings of the ASME 2009 Pressure Vessels and Piping Division Conference

PVP2009
Proceedings of the 2009 ASME Pressure Vessels and Piping Conference
July 26-30, 2009, Prague, Czech Republic
PVP 2009
July 26-30, 2009, Prague, Czech Republic

PVP2009-77730
PVP2009-77730

METAL TO METAL FLANGES LEAKAGE ANALYSIS

L. Bertini, M. Beghini, C. Santus ∗ G. Mariotti

DIMNP, Mechanical Department General Electrics, Oil & Gas


University of Pisa Nuovo Pignone
Pisa 56126, Italy Firenze 50127, Italy

ABSTRACT H Flange height.


The use of a gasket made in soft material is not recommended W Flange width.
for large size centrifugal compressor case flange. The two case Z Bolt axis position.
halves are assembled with bolted flanges and leakage is prevented PB Bolt pitch.
by the metal–to–metal contact under pressure. The prediction of L Leakage distance.
the leakage condition is an important engineering issue for this Lo Flange opening distance.
technology. In the paper an original model able to predict the p Internal pressure, inside the vessel.
leakage condition, based on a Fracture Mechanics approach, is pL Internal pressure causing leakage (leakage pressure).
presented. The flange surfaces interface is regarded as a crack pL,FE Leakage pressure predicted with the FE model.
which can be partially open. As the flanges can not transfer pL,FE,max Max FE model leakage pressure among the casings.
tensile traction, the extension of the open zone, i.e. the crack R Correlation coefficient.
length, is obtained by the condition that the Stress Intensity Fac- σn Crack body nominal stress.
tor K is zero. An analytical model, based on the Weight Function K Fracture mechanics Stress Intensity Factor (SIF).
technique, was applied to find the stress intensity K, and then to h(x, a) Weight Function.
predict the leakage condition. a Crack length.
The paper illustrates a validation of the proposed model by the x Local coordinate.
comparison with a nonlinear Finite Element analysis and the σ0 , σ1 Nominal stress parameters.
results of a full scale experimental test series obtained by a re- K0 , K1 Base solution weight function integrations.
search collaboration between industry and academia. The leak- ao Opening crack length.
age pressure predicted by the model is in good agreement with σn,B Bolt pre–load nominal stress distribution.
the numerical prediction and the experimental results. σn,p Internal pressure nominal stress distribution.
σn,B1 Unitary Bolt pre–load nominal stress distribution.
σn,p1 Unitary internal pressure nominal stress distribution.
NOMENCLATURE dH0 Unidimensional model equivalent bolt hole diameter.
DV Vessel diameter. FB Bolt pre–load.
tV Vessel wall thickness. F1 , F2 Bolt pressure distribution compensation forces.
dB Bolt diameter.
dH Bolt hole diameter.

∗ Corr. author: Ciro SANTUS. Ph.: +39 050 836607, Fax: +39 050 836665.

Email: ciro.santus@ing.unipi.it
1 Copyright
c 2009 by ASME
INTRODUCTION tions and approximations of the geometry have been reached
Gasketless flanges are usually employed to connect the halves yet [12, 13];
of large centrifugal compressor bodies, Fig. 1. 6. The onset of leakage is usually associated with the loss of
contact (zero pressure) between the flange mating surfaces
[6–10], or with the obtainment of a critical tensile stress
Centrifugal compressor (usually a few MPa) required to break the sealant film [6];
7. The actual pre–stress of the bolts, after the initial make-
up, is broadly accepted to be the main reason of leakage
flange unreliability, several procedures have been proposed
to improve the makeup procedure for complex flanged joints
[14–19];
8. FE simulations has also been proposed to investigate the
make-up procedure [20, 21].

The aim of the present study is to propose a simple and reliable


model to predict the metal–to–metal flange leakage. The results
are compared with those obtained by nonlinear FE analysis and
Gasketless flange with full scale experiments, showing satisfying agreement. The
proposed analytical model seems capable to avoid cumbersome
FE simulations, and it is a promising tool for designing and opti-
Figure 1. CENTRIFUGAL COMPRESSOR GASKETLESS FLANGE.
mizing the flange geometry.

Studies on gasketless flanges are quite recent and a simple ana-


NONLINEAR FINITE ELEMENT MODEL
lytical physical model for predicting the leakage condition does
A nonlinear parametric FE model of the flanges was devel-
not exist yet. Indications deduced by the literature which are
oped in order to achieve a deeper understanding of the contact
interesting for developing a predicting tool are summarized in the
pressure flange surface distribution and to produce a set of qual-
following:
ified results to validate the analytical model (detailed explained
1. The following geometric parameters produce the main effects below).
on the leakage rate [1–5]:
(a) surface flatness tolerance, indeed an accurate flatness is FE model assumptions
required to avoid local lack of contact pressure between The FE model was aimed at investigating the contact of the
the flange surfaces; cylindrical portion of the compressor shell, at some distance from
(b) surface roughness, that is required to be low; the end regions. Taking into account the flange bolt cyclic sym-
(c) roughness orientation: scratch orientations parallel to metry, the modeled region was limited to half of the bolt pitch
the main leakage flux direction tend to increase the and it is shown in Fig. 2 (in the figure the bolt is not shown to
leakage rate; simplify the scheme).
2. The negative effects produced by irregularities and poor The main geometry parameters are indicated in Fig. 3. A paramet-
surface roughness can be reduced by introducing specific ric FE model was implemented to investigate the effects of several
sealants (e.g. silicone sealants) between the flanges in order geometrical parameters on the flanged joint contact pattern.
to fill the scratches and compensate local irregularities of the The constraints applied to the model are:
flange surfaces [4, 5]; 1. plane symmetry at the mid plane of the case;
3. Most analyses of the recent flanged joints are based on the 2. zero axial displacement on one of the two transverse section
finite element (FE) method and apply nonlinear models to plane;
investigate the contact pressure distribution between the two 3. equal axial displacement condition (coupled degree of free-
mating flange surfaces [6–10]; dom) applied to the second transverse section planes.
4. Modeling the presence of the sealant with effective material 4. vertical constraint applied to one of the nodes laying on the
properties, appears to be relevant for the accuracy of the symmetry plane of the case, to avoid rigid-body motion (this
results, mainly under cyclic loading [11]; node reaction is essentially zero);
5. Plane (axisymmetric) FE models was proposed by a few
authors, but no general agreement on acceptable simplifica- The simulated loads are:

2 Copyright
c 2009 by ASME
and the internal case surface (leakage distance L) Fig.4. The FE
FE model predicted leakage pressure pL,FE is obtained by increasing the
region internal pressure (using small steps) and observing the opening
front up to the leakage condition, Fig.4.

Leakage
PB
Bolt
pitch Flange Increasing
interface the internal
pressure p

Lo
Figure 2. MODELED REGION OF THE COMPRESSOR CASE.

Open contact
front
tV
dB (a) (b)
Vertical
symmetry DV Figure 4. (a) FE MODEL. (b) OPENING DISTANCE AND LEAKAGE
plane CONDITION.

dH H
PB
Bolt pitch
along the ANALYTICAL MODEL
Transverse axial The analytical model proposed to predict the flange leakage
direction
L is based on the similitude between the flange contact interface
section
plane
Z and the edge crack. Relevant results of the Linear Elastic Fracture
W Mechanics (LEFM) theory can be applied to the problem thus
significantly simplifying the analysis.

Figure 3. MAIN MODEL PARAMETERS.


LEFM background
In the fracture mechanics the nominal stress field usually
1. the bolt pre–load: obtained by an initial interference between indicates the stress distribution at the surface crack if the crack
the surface–to–surface elements of the upper flange surface were not present. Considering an edge crack in a plane body,
and the bolt head base; under a nominal stress distribution σn , the Mode I Stress Intensity
2. the internal pressure, applied to the cylindrical case surface; Factor (SIF) K can be calculated by the following relationship
3. the axial load generated by the internal pressure, inside the [22–24] (if the crack is fully open):
closed case, uniformly distributed on the non–fixed transverse
Z a
section plane.
K(a) = σn (x) h(x, a)dx (1)
The main result of an analysis was the distance Lo between the 0

open contact front and the internal case surface, as shown in Fig.4.
The length Lo is termed here ‘flange opening distance’. where h(x, a) is the Weight Function (WF) for the given crack
The fluid inside the flanged vessel can leak only if the flange geometry, and x is the local coordinate variable ranging from zero
opening distance reaches the bolt hole. So, it is assumed that to the crack size a. Eq.1 is valid under the LEFM hypothesis and
leakage occurs when the flange opening distance is larger (or at when the WF is known (which depends only to the geometry).
least equal) than the minimum distance between the bolt hole The SIF can be directly calculated for any given nominal stress

3 Copyright
c 2009 by ASME
distribution σn (x). Under a nominal stress distributions with com-
pressive components, the crack can be partially open [22–24] No leak.:
(for a length ao ) and closed (with crack faces in contact) for the Lo
remaining portion of its length. A possible example of such situa- Lo < L,
tion is shown in Fig. 5, where the nominal stress is tensile at the
surface and becomes compressive before the crack tip.
Leakage:
Lo = L
L
σ n ( x)
Figure 6. LEAKAGE CONDITION.

Open crack
length ao The flange interface nominal stress σn (x) has two components: the
bolt pre–load stress distribution σn,B (x) and the stress distribution
K (ao ) = 0 a generated by the internal pressure σn,p (x). Since the σn,B (x) stress
distribution is proportional to the bolt pre–load FB and σn,p (x) is
Closed
x proportional to the internal pressure, the flange nominal stress can
crack
be written as:
length

σn (x) = FB σn,B1 (x) + p σn,p1 (x) (4)


Figure 5. PARTIALLY OPEN CRACK.
where σn,B1 is the flange stress distribution generated by unitary
bolt pre–load and σn,p1 is the flange stress distribution generated
As the closed crack face regions are required to transmit com- by unitary internal pressure.
pressive traction only, they behave as a continuous (uncracked) The leakage condition can be rewritten as:
material. Therefore, the open part can be regarded as a crack
having length ao , generally unknown. Also for the open crack FB KL,B1 + pL KL,p1 = 0 (5)
the SIF can be calculated by Eq. 1 where the crack length is
ao . As the contact can not transfer tensile traction, the SIF of
the partially closed crack has to be zero, therefore, the following where the unitary bolt pre–load SIF KL,B1 and unitary pressure
integral relationship holds: SIF KL,p1 are the integrations of the unitary bolt pre–load and
unitary pressure nominal stress distributions, respectively:
Z ao
RL
K(ao ) = σn (x) h(x, ao )dx = 0 (2) KL,B1 = 0 σn,B1 (x) h(x, L)dx
0
(6)
RL
KL,p1 = 0 σn,p1 (x) h(x, L)dx
which determines the unknown length ao .
The Eq. 5 can be solved imposing the bolt pre–load FB to find
Flange LEFM model the leakage pressure pL or introducing a value of pL to find the
A partially separated flanged joint was approximated as a leakage safe bolt pre–load FB .
partially open crack, due to the quite similar load transmission The specific weight function h(x, L) should be found taking into
mechanism, with compressive stresses acting through the closed account the effective flange geometry. Since an analytical closed
contact region and no stress acting on the open region. The form weight function does not exist for the complex flange geom-
leakage condition is assumed the same as the nonlinear FE model etry, the following simplifying assumptions were introduced:
previously described, i.e. when the Lo equals the flange geometry 1. the flanged joint was considered a partially closed surface
length L, Fig. 6. crack in a semi–infinite body, the effects of flange thickness,
At the leakage onset the following equation holds: bolt hole, and limited flange extension, on the crack solution
relationships were neglected;
Z L 2. the nominal stress variation along the pitch distance, between
K(L) = σn (x) h(x, L)dx = 0 (3) two adjacent bolts, was neglected.
0

4 Copyright
c 2009 by ASME
These two assumptions allowed to reduce the analysis to a simple L d 'H
plane scheme. An analytical form of the weight function for this
basic crack geometry can be found in the fracture mechanics text-
books [25, 26]. Moreover, as the nominal stress can be reasonably
approximated by a linear distribution, only two integrations are
required for the two stress distributions reported in Fig. 7: uni- dH
form nominal stress, and linearly variable nominal stress (zero PB
stress at the surface edge). As a consequence only two influence
coefficients are enough for solving the problem:
x
Ra √
K0 = 0 σ0 h(x, a)dx = 1.1215 σ0 πa
Ra √ (7)
K1 = 0 σ1 x/a h(x, a)dx = 0.6820 σ1 πa
Figure 8. EQUIVALENT AREA ASSUMPTION, APPROXIMATED UNI-
DIMENSIONAL MODEL.

as linear model imposing the force equivalence, Fig. 9. Consider-


K 0 = 1.1215σ 0 π a K1 = 0.6820 σ 1 π a ing a flange element of length PB , the opening force generated by
unitary internal pressure is: 1 × PB DV and its axis is assumed to
be on the mid plane of the vessel wall thickness.
x
σn = σ0 σ n = σ1
a
σ n,p1 (0)
x σ n,p1 ( L)
a a
x
σ n,p1 ( x)
L
(a) (b)
Figure 7. WEIGHT FUNCTIONS INTEGRATION: (a) UNIFORM
STRESS, (b) LINEAR STRESS.

SV / 2
Indeed, if the nominal stress distribution is linear these two co-
efficients give the solution, without any other weight function
p = 1MPa
integration. 1× PB DV / 2

Nominal stress approximation Figure 9. UNITARY INTERNAL PRESSURE NOMINAL STRESS DIS-
In order to reduce the entire problem to a bi–dimensional TRIBUTION.
scheme, the flange geometry was approximated to a plane con-
figuration reproducing the bolt hole feature as a continuous slot.
For a crack length L under unitary internal pressure the SIF is
The slot dimension dH0 was obtained by imposing the equivalent
related to the stresses σn,p1 (0), σn,p1 (L), as it follows:
area condition, Fig. 8, Eq. 8:

πdH2 KL,p1 = 1.1215 σn,p1 (0) πL +
dH0 = (8) √ (9)
4 PB 0.6820 (σn,p1 (L) − σn,p1 (0)) πL

Internal pressure. The nominal stress distribution at the The fluid pressure penetrated in the open flange gap is not taken
flange interface, generated by the internal pressure, was evaluated into account in Eq.9. To consider this effect it suffices to replace

5 Copyright
c 2009 by ASME
the terms σn,p1 (0) and σn,p1 (L) with σn,p1 (0)+1 and σn,p1 (L)+1 It is worth noting that the σn,B (0), σn,B (L) are negative (pressures)
respectively. However, the terms σn,p1 (0) and σn,p1 (L) are much so they produce negative SIF, to be algebraically summed to the
larger than 1 because the vessel wall is much smaller than the positive SIF produced by the internal pressure stress distribution.
vessel diameter.

Leakage pressure calculation


Bolt pre–load. Following the approach reported in many By solving the Eq. 5 and using the weight function integra-
machine design theory textbooks (for example the book by tions (Eqs.9,10), the leakage pressure pL can be found:
Shigley and Mischke [27]), the flange contact pressure, generated
by the bolt pre–load, is assumed to be distributed over an effective σn,B1 (0) + 1.552 σn,B1 (L)
double conical volume extending from the nut basis. In the present pL = −FB (11)
σn,p1 (0) + 1.552 σn,p1 (L)
study the bolt sequence three–dimensional geometry is reduced to
a plane scheme, Fig. 8. The bolt pre–load conical pressure is re-
duced to a trapezoidal distribution to be consistent with the plane the negative sign is due to the compressive stresses
scheme. However, the bolt compressive region width is larger σn,B1 (0), σn,B1 (L) and the 1.552 coefficient is the result of the
than the flange surface width W , Fig. 10. The pressure distribu- weight function integration coefficients combination.
tion outside the contact (forces F1 , F2 ) has to be redistributed over
the flange surface width, and then superimposed to the trapezoidal
distribution, obtaining the contact pressure σn,B (x), reported in RESULTS
Fig. 10. The studied cases dimensionless geometries are reported in
Tab. 1. The leakage pressure values were predicted by means of
the FE model, pL,FE , and by means of the analytical model Eq.
11, pL . Twelve cases were studied, with different sizes and dimen-
FB = 1kN Bolt pressure sion ratios. A real component full scale pressurization test was
distribution, performed for each condition. The maximum pressure recorded
F1 F2 during each pressurization test is reported in Tab. 1 (column p).
larger than the Only the first test was intended to produce leakage, while the other
flange surface cases were tested up to the service pressure and the no leakage
was observed. The comparison between the leakage pressure
+
FB pL / pL,FE,max
predicted by the FE model and the analytical model is reported in
Press.distr. Fig. 11.
equivalent to:

σ n,B1 ( L) = F1 , F2
1
σ n,B1 (0)
Bolt pressure 0.8 R 2 = 0.997
x actual
pL / pL,FE,max

distribution 0.6
L FB
0.4
Figure 10. BOLT PRE–LOAD PRESSURE DISTRIBUTION.

0.2

The the unitary bolt pre–load SIF can be calculated from the
σn,B (0), σn,B (L) values: 0
0 0.2 0.4 0.6 0.8 1
pL,FE / pL,FE,max

KL,B1 = 1.1215 σn,B1 (0) πL + Figure 11. COMPARISON BETWEEN THE ANALYTICAL AND THE FE
√ (10)
0.6820 (σn,B1 (L) − σn,B1 (0)) πL LEAKAGE PRESSURE PREDICTIONS

6 Copyright
c 2009 by ASME
Table 1. CASE DIMENSIONLESS GEOMETRICAL PARAMETERS AND PRESSURES.

pL,FE pL p
Casing identif. DV /tV W /tV H/tV Z/tV PB /tV dB /tV dH /tV Leakeage
pL,FE,max pL,FE,max pL
1 28.80 3.880 3.120 1.641 2.301 1.132 1.248 0.813 0.802 0.990 Leakage
2 28.29 4.571 3.486 1.614 2.306 1.200 1.286 0.522 0.512 0.840 No Leak.
3 37.50 4.500 3.050 1.663 2.275 1.050 1.125 0.292 0.293 0.333 No Leak.
4 22.61 3.826 3.130 1.496 2.233 1.035 1.139 0.475 0.435 0.774 No Leak.
5 28.89 4.000 2.867 1.467 1.768 0.733 0.944 0.926 0.908 0.986 No Leak.
6 28.80 3.880 3.040 1.560 2.209 1.012 1.128 0.600 0.594 0.920 No Leak.
7 22.22 3.704 3.259 1.444 2.148 1.048 1.156 0.694 0.686 0.801 No Leak.
8 41.11 4.889 3.156 1.689 2.148 0.956 1.156 1.000 0.976 0.743 No Leak.
9 27.80 4.300 3.040 1.670 2.544 1.000 1.200 0.482 0.463 0.895 No Leak.
10 34.00 4.300 2.850 1.580 2.234 0.940 1.060 0.603 0.623 0.989 No Leak.
11 50.00 4.167 2.833 1.667 1.691 0.717 0.867 0.495 0.514 0.375 No Leak.
12 35.17 4.417 3.000 p / pL2.230
1.575 0.967 1.133 0.271 0.260 0.915 No Leak.

All the points, corresponding to the twelve case geometries re- 1


ported in Tab. 1, were very near the dashed line, that is the
PL = PL,FE locus, with very small scatter (correlation coefficient
R = 0.997. It is evident that the approximations introduced in 0.8
the analytical model were acceptable in terms of the final result
accuracy, at least in the range of geometry parameters considered. 0.6
p / pL

Having obtained very similar leakage pressure predictions, the


FE model and the analytical model can be considered equivalent.
The leakage pressure prediction is compared to the experimental 0.4
pressurization tests in Fig. 12.
According to the model: no leakage is predicted if p < pL , while
0.2
leakage is predicted if p > pL . It is worth noting that: the con- LEAKAGE test
dition p/pL < 1 was true for those tests with no leakage; p was NO leakage tests
very near the pL for the single test that showed leakage, though 0
slightly lower. Obviously, near the critical condition some sort of 0 0.2 0.4 0.6 0.8 1
model uncertainty would result. pL / pL,FE,max

Figure 12. PRESSURIZATION TEST RESULTS. THE TEST PRES-


SENSITIVITY PARAMETRICAL ANALYSIS SURE WAS VERY NEAR THE PREDICTED LEAKAGE PRESSURE FOR
THE TEST THAT SHOWED LEAKAGE.
The analytical model Eq. 11 allows for evaluating the pre-
dicted leakage pressure as function of the flange geometry param-
eters. The model can be implemented on an electronic worksheet, were modified (taking other parameters equal) and the leakage
so the parametric sensitivity analysis can be very fast. On the pressure was calculated. The analyzed parameters are:
contrary, FE analysis would require a complete repetition of the
calculation for each new geometric parameter analysis. 1. Bolt pre–load FB ,
Three casing geometries were considered for the leakage pressure 2. Bolt position Z;
sensitivity analysis: Cases A,B,C; identifications 3,5,7 respec- 3. Bolt pitch PB ;
tively, Tab.1. For each casing geometry, the following parameters 4. Flange width W ;

7 Copyright
c 2009 by ASME
pL / pL,FE,max

5. Flange thickness H. 1
Case C
The leakage pressure dependence to the flange parameters re-
0.8
ported in the present section are just the models results and were
not experimentally validated. However, these results can be very

pL / pL,FE,max
0.6
useful to have indications for flange geometry optimization. A
further validation of the Analytical model is also provided, be- Case B
cause the analytical model results are compared to the FE model 0.4
results for different flange geometries.
0.2
Analytical prediction Case A
Effect of the bolt pre–load FE prediction
0
According to the Eq. 11 the leakage pressure and the bolt 1.4 1.6 1.8 2 2.2 2.4
pre–load are linearly related. The linear dependence between the Z / tV
bolt pre–load and the leakage pressure was also confirmed by the
non linear FE analysis. Indeed, the nonlinearity due to the contact Figure 13. LEAKAGE PRESSURE BOLT POSITION SENSITIVITY.
has no effect on the relation between these two variables, because
the contact region is fixed assuming the onset of the leakage.
pL / pL,FE,max
It is well known that the bolt is designed to be pre–loaded almost sensitivity to the bolt pitch.
up to its maximum strength. To take advantage of the linearity
between pre–load and leakage pressure, larger bolt size is required
to increase the bolt pre–load. However, if a larger bolt diame-
ter is introduced a smaller leakage length L = Z − DH /2 results. 1.4
Therefore, the bolt–size has to be optimized on the specific flange Analytical prediction
1.2 FE prediction
geometry. Case C
1
pL / pL,FE,max

Effect of the bolt position 0.8


The distance Z of the bolt axis from the vessel inner surface Case B
0.6
determines the bolt position, Fig. 3. It is understandable that the
lower the Z distance, the higher the leakage pressure, because 0.4
the bolt closure effect is stronger. This trend was confirmed by
0.2 Case A
Eq.11. If the bolt axis is away from the vessel inner surface
(large value of Z) the term σn,B1 (0) is lower because the bolt 0
1 1.5 2 2.5 3 3.5
pressure distribution on the flange interface reduces away from PB / tV
the bolt axis, Fig.10, moreover the term σn,p1 (0) increases due to
the higher bending arm of the internal pressure force 1 × PB DV , Figure 14. LEAKAGE PRESSURE BOLT PITCH SENSITIVITY.
Fig.9. Both analytical and FE model were applied by introducing
different bolt positions for the case A,B and C geometries. The
results are reported in Fig. 13.
The analytical prediction is similar to the FE results for lower This analysis indicates that also the bolt pitch has to be reduced
values of Z, while the two models slightly differ for larger values to a minimum in order to maximize the leakage pressure. The
of Z. However, both models predict lower leakage pressure for minimum bolt pitch is the nut diameter plus the tubular wrench
higher bolt position. This analysis shows that the bolt position thickness.
has to be reduced to its minimum value that is the vessel wall
thickness plus half the nut diameter, Fig. 3.
Effect of the flange width
Effect of the pitch The relation between the flange width and the leakage pres-
The bolt pitch is also a parameter the leakage pressure is sure is shown in Fig. 15. It is evident an asymptotic trend after a
very sensitive to. Similarly to the bolt position, larger bolt pitch certain flange width value. A design with very large flange width
produces lower leakage pressure because the total bolt closure is not useful, because the part of the flange away from the internal
force reduces and then the terms σn,B1 (0), σn,B1 (L), of Eq. 11, vessel side does not produce any benefit to the flange leakage
are lower (absolute value). Fig. 14 shows the leakage pressure resistance.

8 Copyright
c 2009 by ASME
pL / pL,FE,max

1.4 was based on the analogy between a flange and a partially


Analytical prediction
1.2 FE prediction open crack, loaded under variable nominal stress field;
3. The two models were applied to twelve compressor geome-
1 Case C tries, the results of the analytical model were in good agree-
pL / pL,FE,max

ment with those of the FE model;


0.8
4. The proposed model can be considered as a valid analysis
0.6 tool, especially in the preliminary design stage;
Case B 5. The predictions of the two models were also in good agree-
0.4
ment with the results obtained from the pressurization tests
0.2 Case A of the twelve cases;
6. Further experimental results, preferably under laboratory
0
2 3 4 5 6 conditions, would be required for a sound qualification of the
W / tV assumed leakage condition: flange open contact front up to
the hole diameter perimeter;
Figure 15. LEAKAGE PRESSURE FLANGE WIDTH SENSITIVITY. 7. A sensitivity study was conducted for three compressor cases
having different sizes, varying a few interesting geometric
and operating parameters and analyzing their influence on
Effect of the flange thickness the predicted leakage pressure;
The effect of the flange thickness on the leakage pressure 8. The results of the two models were in good agreement also
is shown in Fig. 16. The results show negligible sensitivity to in terms of leakage pressure sensitivity, and then a further
the flange thickness at least in the investigated range. Therefore, verification of the analytical model was provided;
pL / pL,FE,maxthe flange thickness can be considered as a secondary variable in 9. The sensitivity study indicated that:
terms of the pressure leakage design.
(a) the leakage pressure is proportional to the bolt pre–load;
(b) the actual value of the bolt pre–load is usually known
1.2
with large uncertainty, so it is the most important source
Case C
of variation on compressor case performance, being
1 directly related to the leakage pressure;
(c) leakage pressure increases by increasing flange width,
0.8 but an asymptotic trend is obtained, indicating that the
pL / pL,FE,max

flange width away from the bolt has a small effect on


0.6 the flange contact pressure distribution;
Case B
(d) leakage pressure decreases by increasing the distance
0.4
between the bolt axis and the internal case diameter
0.2 (other parameters constant), and also by increasing the
Case A Analytical prediction
bolt pitch (other parameters constant), these results sup-
FE prediction
0 port the practice to keep bolt axis distance and bolt pitch
2.5 3 3.5 4
as low as possible;
H / tV
10. The LEFM approach to find the leakage pressure can be po-
Figure 16. LEAKAGE PRESSURE FLANGE THICKNESS SENSITIV- tentially extended to other similar problems (e.g. the end zone
ITY. of the compressor cases) by means of the Weight Function
method;
11. One of the main advantages of the fracture mechanics simili-
tude is that it is based on the knowledge of nominal stresses
only, therefore FE nonlinear calculations can be avoided.
CONCLUSIONS
1. A three dimensional Finite Element (FE) nonlinear model
was set up to predict the value of the internal compressor case ACKNOWLEDGMENTS
pressure that causes fluid leakage from the central cylindrical The authors are grateful to Claudia Cagnarini and Paolo
region of the vessel; Romanello from GE Oil & Gas Nuovo Pignone (FI) for the strong
2. Following the LEFM approach an innovative analytical contribution given both in the design phase as well as on the test
model was also set up to predict flange leakage, this model campaign.

9 Copyright
c 2009 by ASME
REFERENCES termination of Bolt Preload in Pipe Flange Connections
[1] Nash, D., and Abid, M., 2004. “Surface sensitivity study With Gaskets Under Internal Pressure”. Journal of Pressure
of non–gasketed flange joint”. Proceedings of the Institu- Vessel Technology, 124(4), pp. 385–396.
tion of Mechanical Engineers, Part E: Journal of Process [14] Abid, M., and Nash, D., 2006. “Structural strength: Gas-
Mechanical Engineering, 218(4), pp. 205–212. keted vs non-gasketed flange joint under bolt up and op-
[2] Nakamura, T., and Funabashi, K., 1991. “Effects of di- erating condition”. International Journal of Solids and
rectional properties of roughness and tangential force on Structures, 43(14–15), pp. 4616–4629.
pressure flow between contacting surfaces”. Lubrication [15] Fukuoka, T., and Takaki, T., 2003. “Elastic Plastic Finite El-
Science, 4(1), pp. 13–23. ement Analysis of Bolted Joint During Tightening Process”.
[3] Arghavani, J., Derenne, M., and Marchand, L., 2003. “Effect Journal of Mechanical Design, 125(4), pp. 823–830.
of Surface Characteristics on Compressive Stress and Leak- [16] Mackerle, J., 2005. “Finite elements in the analysis of
age Rate in Gasketed Flanged Joints”. The International pressure vessels and piping, an addendum: A bibliography
Journal of Advanced Manufacturing Technology, 21(10–11), (2001–2004)”. International Journal of Pressure Vessels
pp. 713–732. and Piping, 82(7), pp. 571–592.
[4] Murtagian, G., Fanelli, V., Villasante, J., Johnson, D., and [17] Mackerle, J., 2003. “Finite element analysis of fastening
Ernst, H., 2004. “Sealability of Stationary Metal–to–Metal and joining: A bibliography (1990–2002)”. International
Seals”. Journal of Tribology, 126(3), pp. 591–596. Journal of Pressure Vessels and Piping, 80(4), pp. 253–271.
[5] Kirkemo, F., 2002. “Design of compact flange joints”. In [18] Mackerle, J., 1995. “Fastening and joining: finite element
Proceedings of 2002 ASME Pressure Vessels and Piping and boundary element analyses – A bibliography (1992–
Conference (PVP2002). Vancouver, BC, Canada, Paper 1994)”. Finite Elements in Analysis and Design, 20(3),
1087. pp. 205–215.
[6] Kawamura, H., Sawa, T., and Yoneno, M., 2003. “FEM [19] Tsuji, H., and Nakano, M., 2002. “Bolt preload control for
stress analysis and sealing performance improvement of box- bolted flange joint”. In Proceedings of 2002 ASME Pressure
shaped bolted flanged joints using silicone sealant under in- Vessels and Piping Conference (PVP2002). Vancouver, BC,
ternal pressure and thermal conduction conditions”. Journal Canada, Paper 1094.
of Adhesion Science and Technology, 17(8), pp. 1109–1125. [20] Fukuoka, T., 2005. “Finite Element Analysis of the Thermal
[7] Abid, M., and Nash, D., 2003. “Comparative study of and Mechanical Behaviors of a Bolted Joint”. Journal of
the behaviour of conventional gasketed and compact non– Pressure Vessel Technology, 127(4), pp. 402–407.
gasketed flanged pipe joints under bolt up and operating [21] Fukuoka, T., and Takaki, T., 2003. “Finite Element Simula-
conditions”. International Journal of Pressure Vessels and tion of Bolt–Up Process of Pipe Flange Connections With
Piping, 80(12), pp. 831–841. Spiral Wound Gasket”. Journal of Pressure Vessel Technol-
[8] Abid, M., 2005. “Determination of Gasketed and Non– ogy, 125(4), pp. 371–378.
gasketed Flanged Pipe Joint’s Capacity Subjected to Com- [22] Beghini, M., and Bertini, L., 1996. “Effective stress inten-
bined Loading: An Experimental Approach”. International sity factor and contact stress for a partially closed Griffith
Journal of Mechanics and Materials in Design, 2(1–2), crack in bending”. Engineering Fracture Mechanics, 54(5),
pp. 35–47. pp. 667–678.
[9] Abid, M., 2006. “Determination of safe operating conditions [23] Beghini, M., Bertini, L., and Fontanari, V., 2005. “Para-
for gasketed flange joint under combined internal pressure metric study of oblique edge cracks under cyclic contact
and temperature: A finite element approach”. International loading”. Fatigue & Fracture of Engineering Materials &
Journal of Pressure Vessels and Piping, 83(6), pp. 433–441. Structures, 28(1–2), pp. 31–40.
[10] Abid, M., 2005. “Determination of Safe Operating Condi- [24] Beghini, M., Bertini, L., and Fontanari, V., 2001. “A weight
tions for Non–gasketed Flange Joint under Combined Inter- function technique for partially closed inclined edge cracks
nal Pressure and Temperature ”. International Journal of analysis”. International Journal of Fracture, 112(1), pp. 57–
Mechanics and Materials in Design, 2(1–2), pp. 129–140. 68.
[11] Roos, E., Kockelmann, H., and Hahn, R., 2002. “Gasket [25] Wu, X.-R., and Carlsson, A., 1991. Weight Functions and
characteristics for the design of bolted flange connections Stress Intensity Factor Solutions. Pergamon Press, Oxford.
of metal–to–metal contact type”. International Journal of [26] Fett, T., and Munz, D., 1997. Stress Intensity Factors and
Pressure Vessels and Piping, 79(1), pp. 45–52. Weight Functions. Computational Mechanics, Billerica, MA.
[12] Estrada, H., and Parsons, I., 1999. “Strength and leakage fi- [27] Shigley, J., Mischke, C., and Budynas, R., 2003. Me-
nite element analysis of a GFRP flange joint”. International chanical Engineering Design, 7 ed. McGraw-Hill Sci-
Journal of Pressure Vessels and Piping, 76(8), pp. 543–550. ence/Engineering/Math.
[13] Sawa, T., and Ogata, N., 2002. “Stress Analysis and De-

10 Copyright
c 2009 by ASME

Você também pode gostar