ISSN 10526188, Journal of Machinery Manufacture and Reliability, 2010, Vol. 39, No. 3, pp. 265–271.

© Allerton Press, Inc., 2010.

Original Russian Text © S.V. Shishkin, S.S. Shishkin, 2010, published in Problemy Mashinostroeniya i Nadezhnosti Mashin, 2010, No. 3, pp. 75–82.

RELIABILITY, DURABILITY, AND WEAR RESISTANCE

OF MACHINES AND CONSTRUCTIONS

Calculation of the Tightness of Flanged Joints

S. V. Shishkin and S. S. Shishkin
Moscow, Russia
Received December 2, 2009

Abstract—On the basis of the theory of an equivalent porous layer, we propose two methods for the
calculation of the leakage of a working environment through the seals in flanges depending on the
material of the pad unit and the nature of the contact of the asperities of the mating surfaces. To deter
mine the distribution of the contact pressure along the radius and width of the seal, a numerical solu
tion of the contact problem with an uncertain border of the contact in the finiteelement formulation
is obtained. The solution of the related hermetic problem is considered. A practical example of the cal
culation of leaks in a flange joint with a flat metal gasket is given.
DOI: 10.3103/S1052618810030106

The efficiency and operational safety of a product are often determined by the integrity of its nodes and
joints. These structures include flange joints, whose seals are achieved by crimping a flat gasket. Based on
operating experience, the recommended tightening forces of the protraction of bolts usually guarantee the
desired degree of tightness of the flanges with a rational choice of the material and width of the pads [1].
However, the problem of quantifying the definition of the leakage of the working environment through the
sealing flanges for facilities of the nuclear and chemical industries remains topical.
For the solution of a hermetic problem, it necessary to know the law of the distribution of the contact
pressure in the sealed junction and the length of the contact, because during compression the gasket may
partially withdraw from the mating surfaces. To do this, we consider the numerical solution of the contact
problem for flanges with an uncertain boundary of the contact on the basis of an ICD.
To simplify the calculation, we represent the flanges in the form of a spatial axisymmetric body under
the action of axisymmetric forces qδ uniformly distributed around a ring with a width equal to s, which is
determined from the equality of its area with an area of support surfaces of the nuts (bolt heads) qδ =
nPδ/2πcs, where Pδ, n are the forces of the tightening of the bolts and the number of connecting bolts, and
c is the radius of arrangement of the bolts.
This assumption is more accurate with a smaller step of the bolts, and allows us to transfer from the
solution of a threedimensional problem with discrete forces to the solving of an axisymmetric problem
for which the relationship between the forces and displacements is simpler.
The finiteelement analog of half of the connection with the main sizes is presented in Fig. 1.
By way of the calculated model of the gasket, we will consider a certain linear elastic layer, the axial
displacement δ(r) equivalent to the compression of the gaskets δg(r), and the general convergence of the
contact surfaces of the parts in both sealing joints δ1c(r) + δ2c(r) as a result of the texture deformation:
δ(r) = δg(r) + δ1c(r) + δ2c(r).
Hence, for perfectly flat surfaces of the bodies and the absence of friction, the kinematic condition for
the contact of mutually conjugate points during the tightening of the connection has the following form:
Δ x – V 1 ( r ) + V 2 ( r ) – δ ( r ) = 0, (1)

where Δx is the kinematic flange approximation when tightening the bolts and Vi(r) is the axial displace
ment of the contact surface of the flange (i = 1, 2) at the point r.
With an accuracy sufficient for engineering calculations, we can assume that the displacement of the
layer in the cross section does not depend on the overall deformations of the bodies and is determined only
by the pressure in this section [2].
During the representation of a nonlinear curve of the contact convergence on the pressure in the form
of a dependence characteristic for a rigidplastic body [2], δic(r) = δi0 + λiq(r), (where δi0, λi are the coef

265
266 S.V. SHISHKIN, S.S. SHISHKIN

ficients of linear approximation), we obtain the following expression for determining the displacement of
the layer:
δ ( r ) = δ 0 + λq ( r ) = ( δ 10 + δ 20 ) + ( λ 1 + λ 2 + h p /E p )q ( r ), (2)
where hp, Ep are the thickness of the pad and the module of elasticity of its material and q(r) is the contact
pressure in the sealed junction.
We express the displacement of the flanges in the equation of contact (1) through an unknown contact
and external load using the appropriate influence function [3]
rb re

∫ ∫
V i ( r ) = 2π ηK i ( r, η )q ( η ) dη + 2π ρK i ( r, ρ )q δ ( ρ ) dρ, (3)
ra rc

where Ki(r, η) is the influence function of the contact load q(η), which demonstrates the axial displace
ment of the point r of the sealing surface of the flange i of the singular ring force applied on the same sur
face at the point η in the axial direction; Ki(r, ρ) is the influence function of the pressure of the nut (bolt
head) qδ(ρ), which shows the axial displacement of point r of the contact surface of the flange i of the sin
gular axial force applied at the point ρ of the outer surface of the flange; and (ra, rb), (rc, re) is the radius of
the beginning and the end of the contact area in a sealed junction and under the bolt head (nut), respec
tively.
Substituting expression (3) for the two flanges (i = 1, 2) into the equation of contact (1) with account
ing for relation (2), we obtain
rb re
nP
∫
Δ x – 2π ηJ ( r, η )q ( η ) dη – λq ( r ) = δ 0 + δ δJ ( r, ρ ) dρ
cs ∫ (4)
ra rc

at the marking of the integral kernels J(r, η) = K1(r, η) – K2(r, η); J(r, ρ) = K1(r, ρ) – K2(r, ρ).
Thus, the integral equation of the elastic flange contact (4) supplemented by the static and kinematic
conditions of the contacts and the equation of equilibrium
rb

∫
nP δ = 2π rq ( r ) dr (5)
ra

allows for the solution to the considered problem.

During the contact sampling by the Fredholm method [3], we obtain a finite system of linear equa
tions. In this case, the continuous function of contact stress is replaced by some columnar distribution,
and the condition of contact (1) is satisfied at the finite number of points (at the center of the partition
sections)
L F
nP
Δx – ∑ [ λ jl + 2πη l J jl ( r j, η l )Δη ]q l ( η l ) = δ 0 + δ
cs ∑ ρ J ( r , ρ )Δρ,
f jf j f
j, l = 1 j, f = 1
(6)
L
n
∑ η q ( η )Δη =
l l l  P δ ,
2π
l=1

where j is the index of the section of the contact surface of the flange, in which we determined the axial
displacement ( j = 1, 2, …, L); l, f is the index of the point of the application of the unit force on the sealing
surface (l = 1, 2, …, L) and on the outer surface of the flange (f = 1, 2, …, F), respectively; Δη is the radial
step of the partition Δη = Δρ; and λjl = λ at j = l and λjl = 0 at j ≠ l.
Note that the limits of the integration (ra, rb) in the integrals of displacements (3) and equilibrium
equation (5) are not known in advance; therefore, system (6) is solved by successive approximations and
the search for the unknown contact boundary using certain procedures.
It is obvious that at every point of the contact, the boundary condition qj(rj) ≥ 0 should be satisfied. At
the formation of the lumen, because of the “cohesion” of the joint, qj(rj) < 0. Then, in the current iteration
the corresponding columns of system (6) are nullified, and in place of the diagonal members we put new

JOURNAL OF MACHINERY MANUFACTURE AND RELIABILITY Vol. 39 No. 3 2010

 CALCULATION OF THE TIGHTNESS OF FLANGED JOINTS 267

unknown values of the lumen χj(rj) ≥ 0, which are ∅80

positive according to the solution. Naturally, the ∅38
reverse procedure is also applied.
∅34
The calculation ends with the full satisfaction of
both the static and kinematic contact conditions
[4]. The solution is obtained as a set of values of
contact stress and the lumen; at that, the location of
the lumen in the matrix column defines its coordi
∅62

28
nate.
The required functions of pressure and the

12.5
lumen are smoothed by strengthening the main 2
diagonal of system (6) to the value λ [5] at the uni qδ
form partitioning grid of the contact zone of the
radius (Fig. 1). To assess the stability and conver

8
gence of the numerical solutions, we used the

1
method of doubling the number of contact elements
∅46

8
during the successive refinement of the radial break
down and the corresponding increase in the order of ∅54
system (6). The application of eightpoint axisym Pδ
metric elements gives practically the same curve of
the distribution of the contact load [6].
The topography of the roughness of the mating
surfaces has a random character; therefore, the geo
metrical parameters of the microgaps and capillar
ies are also described by a probabilistic distribution.
The calculations prove that even with a relatively Fig. 1.
small strength of compression, the ratio of the mean
length of the free path of molecules of the environ
ment to the characteristic size of the microchannels corresponds to the viscous regime of the flow at a tem
perature of –60°C to +250° C and the pressure of the working body of no less than 0.1 MPa [7].
In these circumstances, the representation of the rough space between the surfaces by some equivalent
porous layer, whose parameters change according to the convergence of the contacting surfaces, is the
most effective. This approach allows us to find the value of the leaks of the working body from the well
known equation of the continuity of seepage flow in a stationary isothermal laminar flow [7].
For a steady flow of viscous fluid through a ring contact, the volume flow rate in mm3/s is equal to
2πKh ( p 0 – p f )
Q l = 
, (7)
μ s ln r b /r a
where μs is the coefficient of dynamic viscosity of the medium in MPa s; h is the thickness of the equivalent
porous layer under a load in mm; p0, pf is the pressure of the fluid before the seal and at the exit from it in
MPa; and K is the coefficient of the permeability of the porous layer, which characterizes its ability to pass
through a sealable medium under a pressure drop in mm2.
The value of K is determined by the wellknown Kozeni–Karman equation [7]: K = Um3/B2, where B
is the specific surface of the contact of the material with the sealing medium in mm–1, and m, U is the
porosity and Karman constant U = 0.20–0.22.
In the case of penetration through the seal of the gas medium, the volume flow rate is found from the
ratio
2 2
1.36Kh ( p 0 – p f )
Q g = 
, (8)
μ s p a ln r b /r a
where pa is the atmospheric pressure in MPa.
To seal the flanges, soft metal gaskets are widely used. In this case, during the compression of the seal,
the deformation of the asperity of both contacting surfaces occurs.
The most reliable way to describe the real distribution of metal by the height of the roughness is to con
struct a curve of a bearing surface during the processing of profilograms [8]. Its initial part is well approx
imated by a dependence of the type η(ε) = bεν at 0 ≤ ε ≤ 0.35, where ε is the value of the relative contact

JOURNAL OF MACHINERY MANUFACTURE AND RELIABILITY Vol. 39 No. 3 2010

268 S.V. SHISHKIN, S.S. SHISHKIN

convergence, η(ε) is the relative area of the contact on the base length, b and ν are the coefficients of the
initial portion of the Abbot curve. However, at the relatively high level of the contact load, the value ε may
be higher than 0.35. Therefore, the reference curves should be presented in two sections. For the plot
below the point of inflection ε ≥ εi, we can write
ν ν0
η ( ε ) = bε i + b 0 ( ε – ε i ) at ε i ≤ ε ≤ 1, (9)
where b0, ν0 are the coefficients of the lower section of the reference curve. If we closely approximate the
lower part of the microscopic steps of the conical surface ν0 = 1, then from Eq. (9) at the boundary con
ν
ditions ε = 1, η(ε) = 1, we obtain b0 = (1 – bε i )(1 – εi)–1 = 1.54(1 – 0.35νb). Note that this equation is
true only at b < 0.35–ν, and the largest value b0 = 1.54.
For a metal–metal contact with close values of the hardness of the sealing surfaces, for the probabilistic
model of the distribution of microirregularities based on geometric constructions, we obtained the follow
ing expressions for the parameters of the permeability of the porous layer [9]:
ν + 2/3
2  Y ( ε ),
m = 1 –  1
B =  bε
 1 – 

2–ε RzY ( ε i ) sin ( β long /2 ) sin ( β tr /2 )
i 2/3
(ν + 2)
at
b  ε ν + 1 + b ( 1 – ε )ε ν + b 0 ( – ) ν0 + 1 ,
Y ( ε i ) =   1 ε i ε i = 0.35,
ν+1 ν0 + 1
i i i

where Rz is the height of the microirregularities; βlong, βtr are the average values of the angles at the vortices
of the profile of the longitudinal and transverse roughness. These ratios are derived under the assumption
of the equal processing of the contacting surfaces. At various reference curves (b1, ν1) and (b2, ν2), the same
coefficients (b, ν) can be found from the known expressions of the equivalent roughness [10].
In the absence of waviness, in accordance with [10] the contact convergence and the thickness of the
3
porous layer under the load can be approximately determined from the ratio ε = 0.85 q/HB , h =
(Rz1 + Rz2)(1 – ε), where HB is the Brinell hardness of the material with a soft metal gasket.
For the sealing of the flanges, nonmetallic gaskets are used. In this case, the contact convergence in the
sealing junction is realized through the introduction of solid metallic microirregularities of the flanges in
the pliable environment of pads made from rubber, PTFE, etc.
The distribution of the asperities on real surfaces is random and usually corresponds to the normal law
[8]. This model may be represented by a set of cones with different heights and circular bases situated on
the base plane of the flange and the smooth surface of a gasket, in which the microprotrusions of the flange
are introduced.
The average diameter of the cones is determined by the geometric construction D = Rz tan β/2 , where
Rz is the height of microprotrusions and β is the angle of the cone at the top.
In [9], the following expression for the calculation of the value of the permeability coefficient was
obtained:
2 2 3
D Ra ( 12A – π )
K = 2
π
864A D ( 1 – π/4 ) + DF ( ∇ ) + π∇ ⎛ Γ + ⎞ +  ( 1 – ∇ )Z
2 De 2
⎝ 4 sin η/2 ⎠ 4
q 2 144 q 96Ra q
at A = 1 – 2  , F ( ∇ ) = 1 – ∇ 2 , Γ = 2 ,
πHB ( 12 – π ) HB ( 12 – π ) HB (10)
576  
e = 1 – 
q
, Z =
2
D + ( 4RaA ) ,
2
2 HB
π ( 12 – π )
1 Ra 2 1 Ra 4 1 Ra 6
∇ =  –  1 –  ⎛ A⎞ +  ⎛ A⎞ +  ⎛ A⎞ ,
1 ARa
2 σ 2π 6⎝ σ ⎠ 40 ⎝ σ ⎠ 336 ⎝ σ ⎠
where Ra is the arithmetic mean deviation of the profile within the base length and σ is the meansquare
deviation of roughness from the average line.

JOURNAL OF MACHINERY MANUFACTURE AND RELIABILITY Vol. 39 No. 3 2010

 CALCULATION OF THE TIGHTNESS OF FLANGED JOINTS 269

The analysis of expression (10) shows that the permeability of the compacted interface is mainly deter
mined by height of the microprotrusions, and to a lesser degree by their form.
q
For the adopted model of the contact of roughness, ε = 2  , h = Rz(1 – ε).
πHB
The determination of the leakage of a sealable environment from relations (7) and (8) under the con
ditions of the constant permeability of the contact does not cause difficulties in the design of the flanges
when the pressure is constant across the entire width of the gasket.
The thickness of the porous layer h(ε) and the coefficient of permeability K(ε) depend on the contact
convergence of the surfaces ε(q), i.e., on the final calculation of the contact load q. This leads to certain
complexities in calculating the mass transfer of the working environment through a sealed joint in terms
of its variable permeability, i.e., at a nonuniform distribution of pressure over the radius.
Therefore, the necessary condition for solving the hermetic problem in this formulation is to calculate
the distribution of the pressure drop of a sealable medium along the width of the compaction.
We divide the contact zone on the f elementary ring plots with an uneven partitioning step along the
radius in order to ensure a constant value of the logarithm in the denominator of formulas (7) and (8). We
assume that within each of such sites, the contact load and, thus, the parameters of the porous layer, are
constant. Thus, the discretization of the contact leads to the replacement of the continuous functions q(r),
h(r), and K(r) by some piecewise constant distributions.
It is convenient to write the magnitude of the leakage of the liquid for the ith ring contact zone in the
form of the ratio
2π ri
Q l ( i ) =  ( p i – 1 – p i )h ( r )K ( r ), q i = q ( r i ), ω = ln 
, (11)
μs ω ri – 1
where pi – 1, pi is value of the pressure of the sealed environment at the entrance and at the output of a dis
crete circular area of the contact i.
For the viscous flow regime, the flow of the working body, based on the conditions of its continuity, is
constant at any point of the contact: Q1 = Q2 = … = Qf – 1 = Qf = Ql, (i = 1, 2, …, f). Therefore, for the
adjacent sections i and i + 1, the following equality is true: hi(pi – 1 – pi)Ki = hi + 1(pi – pi + 1)Ki + 1.
Writing a similar expression for all elementary zones of the contact, we obtain the following system of
linear equations of the order f – 1:
( h 2 K 2 + h 1 K 1 )p 1 – h 2 K 2 p 2 + 0… = h1 K1 p0 ,
– h 2 K 2 p 1 + ( h 3 K 3 + h 2 K 2 )p 2 – h 3 K 3 p 3 + 0… = 0,
……………………………………………………………
…0 – h f – 2 K f – 2 p f – 3 + ( h f – 1 K f – 1 + h f – 2 K f – 2 )p f – 2 – h f – 1 K f – 1 p f – 1 = 0,
…0 – h f – 1 K f – 1 p f – 2 + ( h f K f + h f – 1 K f – 1 )p f – 1 = hf Kf pf ,
from the solution of which we find the discrete distribution of the pressure of the working environment
along the radius of the seal. At that, the volume flow rate of the liquid is calculated from relationship (11)
during the substitution in it of the obtained the values of pi – 1 and pi .
Similarly, we obtain the solution to the related hermetic problem for the gas using formula (8).
When m = 0, we achieve the total leakage of the area, i.e., of the entire joint. For the second case, m =
0 at q ≥ 0.428 HB. However, this notion is purely conventional, since it is based on the usual mechanical
picture of the elimination in the process of the convergence of the surfaces of crosslabyrinth microchan
nels and virtually ignores the physical–molecular events.
The solution to the hermetic problem is possible only if Ki(ri) > 0 in all of the sections of the sealed joint.
Consequently, at the sites of zero permeability, (for example, at Ki = Ki + 1 = 0, taking into account the
known distribution of the contact pressure in the formation of system (12)) we assume Ki = 0.5Ki – 1,
Ki + 1 = 0.5Ki, etc., if Ki – 1 > 0.
We calculate the leakage of the working fluid from the same compound flanges of 12X18H10T steel
(E1, 2 = 2.2 × 105 MPa) for one day at a pressure in the pipeline p0 = 0.8 MPa and temperature of 20°C.
The main dimensions of the node are given in Fig. 1. The sealable medium, kerosene, has a dynamic vis
cosity μs = 10–3 Pa s.

JOURNAL OF MACHINERY MANUFACTURE AND RELIABILITY Vol. 39 No. 3 2010

270 S.V. SHISHKIN, S.S. SHISHKIN

δ1, 2k μm q, MPa

2
4 1 24
2 1
3
16
λ1, 2
2
8
1 δ10 = δ20

0
0 3 9 15 21 27 17 19 23 27
q, MPa r, mm

Fig. 2. Fig. 3.

In the form of a seal, a flat aluminum pad is placed (Ep = 7.2 × 104 MPa, HB = 700 MPa) with a thick
ness hp = 1.0 mm and an outer dout = 54 mm and inner din = 34 mm. The flanges are tightened by n = 6
bolts with a force Pδ = 3.0 kN and the radius of the arrangement c = 31 mm. The area of the mounting
surfaces of the nuts (bolt heads) M6 is 389.4 mm2, i.e., s = 2 mm, qδ = 46.2 MPa.
The contact surfaces of the flanges are processed by transverse turning and have the following param
eters of roughness: Rz1, 2 = 4 μm, b1, 2 = 1.5, ν1, 2 = 1.5, βlong = 120°, and βtr = 72°. To simplify the calcu
lation, we will assume that the surfaces of the pad have the same roughness.
The determination of the coefficients of the linear approximation of the curve of the contact conver
gence for one junction of pressure as the result of the deformation of roughness is shown in Fig. 2 (1 is the
curve of the contact convergence and 2 is its linear approximation in the form of a model of a rigidplastic
body), the values of which are equal to δi0 = 2.12 μm, λi = 0.0956 × 10–3 mm/MPa. Then, δ0 = 4.24 μm,
λ = 0.205 × 10–3 mm/MPa.
The pattern of the distribution of the contact pressure in the sealed junction flange (curve 1 is the
numerical solution based on the finiteelement method) is presented in Fig. 3 with a small opening set
near the junction of the channel tube. For comparison, here we show the linear dependence 2 (ring flange
model) obtained based on the engineering calculation of the flange connection with its representation in
the form of structurally unconnected rings by the technique in [11].
During the uniform distribution of the load across the width of the seal, the leakage of kerosene was
found by formula (7) and is Ll = 2Qlt = 0.205 l. As a result of the approximate solutions of the related her
metic contact problem, Ll = 0.268 l, i.e., it is 31% higher as compared to the design calculations due to
the opening of the joint and the uneven contact pressure.
Thus, at the given pressure difference and the type of working fluid, the examined sealing flange has a
low integrity. For a qualitative assessment of a leak, it is sufficient to have the calculated sensitivity of the
test method of not higher than 10–1 mm3 MPa/s.

REFERENCES
1. Raschet na prochnost’ detalei mashin: Spravochnik (Strength Calculation for Machines Elements: Handbook),
Bereg, I.A., Shorr, B.F., and Iosilevich, G.B., Moscow: Mashinostroenie, 1979.
2. Iosilevich, G.B., Kontsentratsiya napryazhenii i deformatsii v detalyakh mashin (Tension and Deformation Con
centrations in Machine Elements), Moscow: Mashinostroenie, 1981.
3. Mikhin, S.G., Integral’nye uravneniya (Integral Equations), Moscow–Leningrad: GITTL, 1947.
4. Parton, V.Z. and Perlin, P.I., Metody matematicheskoi teorii uprugosti (Methods of Mathematical Elasticity The
ory), Moscow: Fizmatgiz, 1981.
5. Tikhonov, A.N. and Arsenin, V.Ya., Metody resheniya nekorrektnykh zadach (Methods for Solving the Incorrect
Problems), Moscow: Nauka, 1974.

JOURNAL OF MACHINERY MANUFACTURE AND RELIABILITY Vol. 39 No. 3 2010

 CALCULATION OF THE TIGHTNESS OF FLANGED JOINTS 271

6. Zenkevich, O. and Chang, I., Metod konechnykh elementov v teorii sooruzhenii i v mekhanike sploshnykh sred
(Finite Element Method in Structural Theory and Mechanics of Continuous Mediums), Moscow: Nedra, 1974.
7. Loitstyanskii, L.G., Mekhanika zhidkosti i gaza (Fluid Mechanics), Moscow: Nauka, 1970.
8. Demkin, N.B., Kontaktirovanie sherokhovatykh poverkhnostei (Contact of Rough Surfaces), Moscow: Nauka,
1970.
9. Babkin, V.T., Zaichenko, A.A., and Aleksandrov, V.B., Germetichnost’ nepodvizhnykh soedinenii gidravlicheskikh
sistem (Hermeticity of Fixed Joints of Hydraulic Systems), Moscow: Mashinostroenie, 1977.
10. Trenie, iznashivanie, smazka. Spravochnik (Friction, Wear, Lubrication: Handbook), Kragel’skii, I.V and
Alisin, V.V., Eds., Moscow: Mashinostroenie, 1978, book 1.
11. Shishkin, S.V., On Contact Pressure Distribution in Circumferential Model of Flange Joint, Probl. Mashinostr.
Nadezhn. Mashin, 2004, no. 2, pp. 38–42.

JOURNAL OF MACHINERY MANUFACTURE AND RELIABILITY Vol. 39 No. 3 2010