You are on page 1of 13

Engineering Failure Analysis 14 (2007) 1680–1692

www.elsevier.com/locate/engfailanal

Approximation of contact stress for a compressed and


laterally one side restrained O-ring
a,*
Hyung-Kyu Kim , Sung-Han Park b, Hwan-Gyu Lee b, Dong-Ryun Kim b,
Young-Ho Lee a
a
Korea Atomic Energy Research Institute, 150 Dukjin-dong, Yuseong-ku Daejeon 305-353, Republic of Korea
b
Agency for Defense Development, 462 Jochiwongil, Yuseong-ku Daejeon, Republic of Korea

Received 1 September 2006; accepted 30 November 2006


Available online 22 January 2007

Abstract

This paper revisits the classical O-ring problem by using an experimental, finite element analysis and conventional the-
ories. Especially, the elastic moduli, deformation shape, friction coefficient and extrusion behaviour were investigated in
detail during the experiments. A computed tomography was used to detect the deformed shape of the O-ring. The used
finite element method results were validated by comparing them with the experimental results. For an application to
the present study, an actual case of a compressed and laterally one side restrained condition was analyzed experimentally
and numerically. As a result of the finite element analysis, the friction coefficient affects the contact stress profile and mag-
nitude considerably. Lindley’s formulae for a contact (compressive) force underestimated the measured force but it showed
similar results to the finite element analysis results. Applicability of the Hertz theory to the contact stress field was dis-
cussed by comparing it with the finite element analysis results in the case of a compressed and laterally one side restrained
condition. Consequently, it was found that the normalized results with respect to the peak stress and contact width fol-
lowed the typical Hertzian profile. So an implementation of the appropriate factors to the Hertz equation can yield an
approximate solution of the contact stresses.
 2006 Elsevier Ltd. All rights reserved.

Keywords: O-ring; Contact mechanics; Finite element analysis; Tomography

1. Introduction

An elastomeric O-ring is one of the most widely used sealing components in mechanical systems. These
seals are mainly classified as a static seal and a dynamic seal. One of the application examples is found in a
solid propellant engine which supports a high pressure. The important property which maintains an O-ring
seal in contact with a seal face is an elastic reaction subjected to an imposed deformation. A general concern
of the O-ring is how to establish the relationship between a deformation and its lifetime. If it is possible to

*
Corresponding author. Tel.: +82 42 868 2111; fax: +82 42 863 0565.
E-mail address: hkkim1@kaeri.re.kr (H.-K. Kim).

1350-6307/$ - see front matter  2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engfailanal.2006.11.061
H.-K. Kim et al. / Engineering Failure Analysis 14 (2007) 1680–1692 1681

obtain a correct relationship, it will be very useful from the viewpoint of its economy and safety in the indus-
tries. This goal can be achieved through a solid mechanical analysis to evaluate the stresses and deformations.
The stress field in an O-ring and the contact pressure distribution are good indicators of a seals’ performance
as well as of its sealing characteristics [1]. Tensile stress in the interior region and the shoulder of the O-ring
can cause a failure. Center crack appears along the vertical direction due to expanding (tensile) stresses and
surface crack occurs along the hoop direction due to high hoop stress [2]. These stresses can be analyzed
by an analytical method, a finite element method and an experimental analysis.
The finite element analysis (FEA) for elastomers has been much less developed. The reason is that the elas-
tomer applications incorporate large displacements, large strains, a non-linear elastic behavior and the Poisson
ratio, approaching 0.5. Hence the FEA results cannot be assumed to be reliable without a reasonable level of
verification. Various experimental and analytical methods for analyzing the stresses in a compressed and lat-
erally unrestrained O-ring can be found in the literature [3–5]. Many researches have only focused on analyz-
ing the reaction force of a rubber material and the contact stresses. However most of the O-rings are used in a
pressurized, laterally restrained application. Few researches are found for case of a problem with a pressurized
and laterally restrained O-ring [6]. Because an O-ring problem becomes more complex, an analytical solution
is rarely found.
In this paper, the mechanics of an O-ring is revisited. A fundamental purpose of present work is to find out
an approximate solution for the mechanics of O-ring by using contact mechanics. To this end, an experimental
study and an FEA are carried out at first. Especially, the specially designed tests provide the elastic moduli,
compressive forces and deformed shapes. The method of FEA is validated through comparing its results with
the test results. Lindley’s formulae and the Hertz theory are used to investigate the applicability of the con-
ventional theories to the present O-ring problem. The validated FEA results for the contact width and peak
stress are compared with the theoretical ones in the case of a compressed and laterally one side restrained con-
dition. An approximate solution of the contact stress, which implements two factors to a function form of the
Hertz equation is considered and discussed.

2. Tests

2.1. Tests for the elastic modulus

The material of the O-ring is chloroprene rubber. First, it is necessary to establish the elastic modulus of it
to analyze the deformation behaviour. Although it might be obtained from the material data, tests were con-
ducted to measure an actual value by using a specimen. This test was also necessary to implement a finite ele-
ment analysis where the appropriate elastic moduli were used depending on the deformation behaviour of each
element. For this purpose, uni-axial and equi-biaxial tests were performed.
The results are illustrated in Fig. 1. From Fig. 1a, the coefficients of the Ogden strain energy function were
determined for the FEA. While, Fig. 1b shows the compressive stress-strain curve, obtained from the equi-
biaxial test results. It is concluded that the elastic modulus is linear to a fractional compression of 30%.
The compressive elastic modulus for the theoretical analysis was evaluated as 0.59 kgf/mm2 (5785.92 kPa),
which was obtained by a linear fitting of Fig. 1b up to a compressive strain of 30%.

2.2. Tests for the compressive forces

The sealing force for a full size O-ring was measured by using a screw compression device of a compression
stress relaxation tester developed by the Agency for Defense Development in Korea. An intrinsic purpose of
this test is to compare the measured compressive forces with the present FEA results for its validation. Of
course, the compressive force is a fundamental data for the present analytical solution approach. The dimen-
sions of the full size O-ring were: a chord diameter of 6.98 mm and a mean diameter of 123.19 mm. The O-ring
was compressed up to 35% of its chord diameter (the fractional compression) during this test. Fig. 2 shows the
results during a loading and an unloading phase. The relationship between the fractional compression and
compressive force is nonlinear but it is found that the degree of nonlinearity is not significant in the test range
of 35%.
1682 H.-K. Kim et al. / Engineering Failure Analysis 14 (2007) 1680–1692

Fig. 1. Test results of the elastic moduli for the chloroprene rubber specimen: (a) measured tensile stress vs. strain; (b) compressive stress
vs. strain deduced from the equi-biaxial test result.

Fig. 2. Test result of the compressive force with respect to the fractional compression.

2.3. Tests for the extrusion behaviour

A series of tests was carried out to investigate the behaviour of an extrusion (i.e. so-called ‘‘forcing-out’’)
into a gap of 1 mm formed between an O-ring groove and its cover when a test block was assembled. Internal
H.-K. Kim et al. / Engineering Failure Analysis 14 (2007) 1680–1692 1683

pressure was exerted by using a hydraulic (water) force to the O-ring which was increased slowly. Then, a con-
stant fractional compression of 23% was maintained. Extrusion displacement measurement was achieved by
using an experimental setup presented in Fig. 3, which was specially devised for the present work. The setup
was composed of a test fixture, in which an O-ring was installed, a CCD laser displacement sensor, and a mea-
surement system. The resolution of the sensor was within 1 lm. Four tests with a CCD laser sensor and one
test with tomography were performed. The extrusion depth of the full size O-ring was measured by using the
CCD laser displacement sensor and the computed tomography.
Simultaneously, the deformed shapes of the O-rings were examined by the computed tomography. The
schematic of the tomography equipment (the second generation) is illustrated in Fig. 4. The measurement con-
dition of the computed tomography was that the X-ray energy was 350 kVp, the tube current was 2 mA and
the focal size was 0.8 mm. The resolution of the image was 1024 · 1024 pixels (51 · 51 mm) so the size of a
pixel was 50 · 50 lm. The section dimension of the test block for measuring the O-ring displacement was
designed to be 20.41 mm in width and 34.6 mm in length to optimize the X-ray’s transmission energy. The
edge of the test block was rounded off to minimize the amount of artifacts. The displacements of the O-ring

Fig. 3. The schematic of test setup for friction coefficient and extrusion depth.

Oring

Initial Collimator X, Z Collimator

Water Pump
Detectors

Radiation Source Scanning Unit

PC & Graphical
Display System

Fig. 4. The schematic of experimental setup for deformed shape measurement.


1684 H.-K. Kim et al. / Engineering Failure Analysis 14 (2007) 1680–1692

Fig. 5. Transient behaviour in the extrusion test from which a static friction coefficient was evaluated.

were measured for internal pressures of 0 kPa and 1378.95 kPa, which were applied to the computed tomog-
raphy. The deformed shape including the expanded diameter and the extrusion depth were compared with the
FEA results for a validation of the method in addition to the compressive force.

2.4. Tests for the friction coefficient

Static friction coefficient (l, termed as ‘‘friction coefficient’’, afterwards) measurement was achieved by
using the extrusion displacement measurement device as already shown in Fig. 3. A friction coefficient was
measured from a movement of the O-ring as the internal pressure of the test fixture was increased, which
was exerted after a 23% fractional compression of the O-ring.
During the extrusion tests described above, the pressure at which the O-ring began to move was also mea-
sured. This critical value of the pressure identifies the force required to overcome a static friction caused by the
compression force (FN) of a squeezed O-ring in a joint. The measured results are shown in Fig. 5. The pressure
when the O-ring began to move was 144.79 kPa. The pressure increase rate seems to have little effect on the
friction coefficient. It was evaluated as l = 0.13 from FN and the pressure at which the O-ring began to move.

3. Conventional theories for the O-ring mechanics

When an O-ring is installed in a groove and compressed by a cover, the contact configuration is regarded as
the Hertz contact generally if a plane strain state is assumed. In this case, the contact pressure profile along the
horizontal direction of an O-ring’s cross section becomes a semi-ellipse, which is the well-known Hertz pres-
sure. By adopting the Hertz contact, Lindley derived simple formulae of the relationship between a compres-
sive force and a fractional compression theoretically and empirically, which are given in Eqs. (1) and (2),
respectively [3,7]. From the tests with the O-rings of the gum compound and the black-filled compound whose
elastic moduli were 0.195 and 0.596 kgf/mm2 (1912.30 and 5844.76 kPa), respectively, the formulae have been
verified. During the derivation, the elastic modulus of the fixture material was regarded to be infinite when
compared with that of the O-ring. The material of the O-ring was also regarded as incompressible (i.e., the
Poisson ratio m = 0.5).
F P
¼ ¼ 1:25  C 1:5 ; ð1Þ
pDdE dE
F P
¼ ¼ 1:25  C 1:5 þ 50  C 6 ; ð2Þ
pDdE dE
where F is the compression load, D and d are the mean and cross section diameters of an O-ring, respectively.
E is the elastic modulus of an O-ring and C (= x/d where x is the diametric compression) is the fractional com-
pression. If we consider a plane problem (i.e., plane strain state), the compressive force exerted at the bound-
H.-K. Kim et al. / Engineering Failure Analysis 14 (2007) 1680–1692 1685

ary of the O-ring cross section, i.e. the compressive force to the unit thickness of an O-ring is F/pD, which is
designated as P in Eqs. (1) and (2).
Using Eq. (1) or (2) and the Hertz theory, the contact width, b and the peak contact stress (the highest value
of the Hertz pressure profile), po can be obtained by
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
6
b¼d  gðCÞ; ð3Þ
p
4P
po ¼ ; ð4Þ
pb
where g(C) is the right hand side of Eqs. (1) and (2). It should be noted that b may be obtained from the Hertz
theory such that [8]
rffiffiffiffiffiffiffiffiffiffiffi
16PR
b¼ : ð5Þ
pE
1m21 1m22
In Eq. (5), R and E* are defined as R1 ¼ R11 þ R12 and E1 ¼ E1
þ E2
in which the subscripts 1 and 2 designate
the differences of the contacting bodies such as the fixture and the O-ring. Since the contact region of the fix-
ture is flat, R simply becomes the radius of the O-ring’s cross section (R = d/2). Again, the elastic modulus of
the fixture can be regarded to be infinitely larger than that of the O-ring and an incompressibility (m = 0.5) is
applied for the O-ring material, Eq. (5) is written as
rffiffiffiffiffiffiffiffi
6Pd
b¼ : ð6Þ
pE
It is recognized that equating Eqs. (3) and (6) can yield Eqs. (1) and (2). Consequently, the Hertz theory was
used in Lindley’s derivation. There may be differing views on this since the O-ring problem violates some of the
assumptions of the Hertz theory. It is validated if the contacting bodies are assumed to be semi-infinite when
compared with the contact dimensions; the strains in the contact region must be infinitesimal to apply the the-
ory of linear elasticity; and the contact surface is frictionless. None of these are sufficiently validated for the O-
ring’s contact problem. This has previously been discussed [9], but the results showed that the Hertz theory
was applicable until a fractional compression of up to 25% [3] or 30% [10]. It implies that a linear elasticity
can be applied approximately to the O-ring problem without a significant error up to the above mentioned
fractional compressions. To investigate this, the maximum fractional compression of 35% was applied in
the present tests and the analysis was done up to 32%, slightly larger than 30%.
Besides the compressive force, width and peak stress, the applicability of the Hertz pressure profile is con-
sidered. The well-known Hertz profile is written as
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2
2x
pðxÞ ¼ po 1  : ð7Þ
b
It has been found that the contact width decreased and the peak stress increased from the Hertz solution as the
semi-infiniteness was violated [11]. They found that the semi-infiniteness was validated if the ratio of the con-
tact width to the thickness of the indented body exceeded 5. The overall shape of a parabola was maintained
even though the semi-infiniteness was validated. In the present O-ring problem, the dimensions of the fixture
(cover and grooved device) are sufficiently larger than that for the ratio of 5. However, it may be compared
with the value d/b in the case of the O-ring itself although the O-ring is an indenter rather than an indented
body when an O-ring’s contact configuration is considered. This will be discussed later.
In the previous works, the internal stresses of a compressed O-ring have seldom been analyzed theoretically
when compared with a contact stress evaluation. A few analyses have been carried out by using the FEA and
experiments (e.g., photoelasticity) [12] rather than the analytical method. However, if the contact stress field is
known, Muskhelishvili’s complex potential approach [13,14] or Timoshenko and Goodier’s solution [15] can
be used for the internal stresses. Curro and Salazar presented a precious clue to it [16]. Present paper does not
deal with the internal stresses either since the primary concern was to investigate the validity of the conven-
tional theories and the FEA on the contact behaviour by comparing their results with the test results.
1686 H.-K. Kim et al. / Engineering Failure Analysis 14 (2007) 1680–1692

4. Finite element analysis

4.1. Method

The basis of the stress analysis programs is that when an elastomer is deformed in a given way, the eventual
overall geometric form and stress conditions will be consistent with a minimum strain energy of the whole
body. The definition of the strain energy density function in terms of the strain invariants is known as a con-
stitutive relation. The choice of it is determined through specimen tests such as uni-axial and equi-biaxial tests.
A least-square interpolation permits a definition of the material constants.
It is known that, if the ratio between the toroidal seal mean diameter and the cross section diameter is larger
than 5, the in-plane strain field in the O-ring cross section can be solved in terms of a plain strain model [17].
Presently used finite element program was the MARc [18], which can cope with finite deformation and hyper-
elastic materials. To be consistent with the full size O-ring tests, an axisymmetric model was adopted. Updated
Lagrangian energy minimization approach was used and the contact element was applied at the contact inter-
faces. The strain energy function used in the FEA was the Ogden form and the Ogden coefficients were cal-
culated by a least square interpolation of the uni-axial and equi-biaxial tests as provided in Fig. 1a. The
friction model used in the FEA was a stick-slip model and the friction coefficient was 0.13 which was deter-
mined from the tests. To solve the convergence problem under a high pressure, an automatic remeshing was
adopted. The remeshing criteria were determined iteratively.

4.2. Validation

To validate the present FEA, the compressive force was investigated at first. This was carried out by com-
paring the obtained reaction force from the FEA with the test results (Fig. 2) and Lindley’s formula (Eqs. (1)
and (2)). The results are given in Table 1. It is found that the FEA underestimates somewhat the compressive
force under a fractional compression of 24% and it overestimates the force at 32%. The closest result appears
at 24%. It is also found that the variation of the friction coefficient has little effect on the compressive force. In
Table 1, the compressive force obtained by Lindley’s formulae is also provided for a comparison. It is inter-
esting to note that the differences between the results by the FEA and the Lindley’s are smaller than those
between the test results and the FEA or Lindley’s formulae. The results by Lindley’s formulae are slightly clo-
ser to the test results than the FEA results. It is somewhat difficult to say that the present FEA is validated
from the investigation of the compressive force. So other comparisons were attempted such as the deformed
shape, the expanded diameter and the extrusion depth of the O-ring.
Figs. 6 and 7 show the deformed shapes obtained from the FEA and tests for a fractional compression of
22.3% with an internal pressure of 0 and 1378.95 kPa, respectively, in the case of the compressed and laterally
one side restrained condition. The FEA and tomography results appeared to be almost the same. Fig. 8 shows
a histogram from which the present images of the computed tomography were analyzed. With a consideration
of this, it is regarded that the present FEA technique can simulate an O-ring deformation. Fig. 9 shows the
expanded diameter of the O-ring obtained from the FEA and tests during a fractional compression increase.
It is found that both results agreed well with each other. Fig. 10 shows the extrusion depth calculated from the
FEA and measured by the tomography and the CCD laser sensor during an increase of the internal pressure.

Table 1
Comparison of the contact force (in N) obtained from the tests, Lindley’s formulae and FEA
Fractional compression (%) Tests FEA Lindley’s formulae
l* = 0.1 l = 0.3 l = 1.0 Eq. (1) Eq. (2)
8 622.53 425.61 426.30 426.39 442.08 442.28
16 1417.36 1176.80 1183.07 1183.96 1250.35 1263.49
24 2537.86 2472.26 2537.96 2517.66 2297.11 2446.46
32 4074.96 4831.05 5081.51 5113.87 3536.57 4375.73
l is a static friction coefficient.
H.-K. Kim et al. / Engineering Failure Analysis 14 (2007) 1680–1692 1687

Fig. 6. The deformed shape obtained from (a) FEA and (b) computed tomography at the fractional compression of 22.3%.

Fig. 7. The deformed shape obtained from (a) FEA and (b) computed tomography at the fractional compression of 22.3% and internal
pressure of 1378.95 kPa.

The prediction by the FEA also agrees well with the measured results. From the above investigations, it was
concluded that the present FEA method was validated.

4.3. Contact stress evaluation for the compressed and laterally one side restrained condition

After the completion of the FEA validation, contact widths and stresses were evaluated for the case of com-
pressed (by 22.3%) and laterally one side restrained condition, which has seldom been treated previously. The
two parameters which can describe the deformed shape are the contact width between the O-ring and the top
(or bottom) surface of the fixture, and that between the O-ring and the lateral groove wall. Fig. 11 displays the
results, where the variation of the contact width with respect to the fractional compression is provided. The
1688 H.-K. Kim et al. / Engineering Failure Analysis 14 (2007) 1680–1692

Fig. 8. Histogram of the image obtained from the computed tomography.

Fig. 9. Comparison of the expanded diameters evaluated from the FEA and measured from tests (l is the friction coefficient).

Fig. 10. Comparison of the extrusion depth evaluated from FEA and measured from tests (l is the friction coefficient).

FEA results are the average values of the last nodes in contact and the first nodes not in contact. The FEA
predicts the contact width at the lateral wall very well. However, there exists some discrepancy in the widths
at the top and bottom surfaces. The cause of the discrepancy between the FEA and tomography results may
H.-K. Kim et al. / Engineering Failure Analysis 14 (2007) 1680–1692 1689

Fig. 11. Comparison of contact widths composed between the O-ring and the top (A), bottom (B) surfaces and side wall (C) obtained from
the FEA and tests (l is the friction coefficient).

be attributed to the pixel size and the resolution accuracy due to a small difference in the density between the
water (1.0) and the O-ring (1.1). It was difficult to determine the boundary of the O-ring and water especially
at the contact edges of the top and bottom surfaces. As the fractional compression increases, the difference in
the contact width at the top and the bottom surfaces increases. However, the variation of the friction coeffi-
cient does not really affect the result as seen in Fig. 9.
The contact stresses were also evaluated for the same case (22.3% compression and laterally one side
restrained condition). As is generally accepted, the seal contact stress is an important factor for evaluating
a seals’ performance [19]. Fig. 12 shows the contact pressure profiles calculated on the contacts between
the O-ring and the groove top and bottom surfaces. pT and pB designate the contact stresses on the contacts
between the O-ring and the top surface, and that at the bottom surface, respectively. Contact stress distribu-
tion shows a parabolic shape like the Hertzian profile. The friction coefficient does not influence the profile
shape itself. However, as the friction coefficient increases (say, at l = 1.0), the maximum contact stresses
and the differences between pT and pB increase. Fig. 13 shows the contact pressure profiles evaluated for
the contact at the lateral wall (designated as pL), which was simultaneously found during the FEA calculation
for Fig. 12. In this case, the contact stress distribution seems to be different from the Hertzian profile. It is
thought that the edge radius formed at the top of the lateral wall caused a local irregular deformation at

Fig. 12. FEA result of the contact pressure profiles between the O-ring and the groove top ad bottom surfaces at the fractional
compression of 22.3%.
1690 H.-K. Kim et al. / Engineering Failure Analysis 14 (2007) 1680–1692

Fig. 13. FEA result of the contact pressure profiles between the O-ring and the lateral wall at the fractional compression of 22.3%.

the rounded edge. Besides this, the influence of the friction coefficient is considerable as was also found in the
results at the top and bottom surfaces.

5. Consideration of an approximate solution for the contact stress

The Hertz theory was applied to the above case, i.e. the compressed and laterally one side restrained con-
dition. The FEA solutions in the case of the fractional compression of 22.3% and l = 0.13 were compared
with the Hertz solution for the contact width (b) and peak stress (po). The results are summarized in Table
2. The FEA results were picked up from the data of Figs. 12 and 13, while Eqs. (4) and (6) were used in
the Hertz theory to obtain the peak stress and contact width, respectively.
For the contact width, the FEA always gives larger values than the Hertz theory. On the contact at the top
and bottom surfaces, the contact width by the FEA is very similar to the Hertz theory. However, it is larger by
almost twice on the contact at the lateral wall. On the other hand, the peak stress results appear oppositely
except the contact at the lateral wall. In short, if we disregard the contact at the lateral wall, the FEA results
gave lower peak stress and larger contact width compared with the prediction by the Hertz theory. It is oppo-
site result to the case of the semi-infiniteness violation aforementioned in Section 3 although d/b is much less
than 5 (it can be calculated by using the contact width in Table 2 and the section diameter of 6.98 mm) in the
present contact configuration. So it can be said that the above phenomenon (larger width and lower peak) is
not due to the dimensions. It may be said to be a specific phenomenon of the O-ring’s contact. It is found that
the ratios between the results from the FEA and that from the Hertz theory are nearly the same on the top and
bottom surfaces.
However, those results are very different on the contact at the lateral wall. A considerable discrepancy
appears in the contact width as well as peak stress between the FEA and the Hertz theory. There may be sev-
eral reasons of it, for instance, the rounded edge at the top of the groove wall and an irregular expansion that
may happen during the compression. Nevertheless, it can be said that the deviation of the contact width and

Table 2
Comparison of the contact width and peak stress evaluated from the FEA and the Hertz theory in the case of compression (22.3%) only
Contact location Compressive force (N) Contact width (mm) Peak stress (kPa)
FEA Hertz Ratio FEA Hertz Ratio
Top 2402.43 4.40 3.78 1.16 1930.73 2090.63 0.92
Bottom 2290.54 4.40 3.69 1.19 1924.21 2041.37 0.94
Lateral 208.49 2.21 1.11 1.99 939.24 615.88 1.53
Note: l = 0.13 is used in the FEA.
H.-K. Kim et al. / Engineering Failure Analysis 14 (2007) 1680–1692 1691

Fig. 14. Investigation of the FEA result on the applicability of the Hertz profile in the case of the compressed (22.3%) and laterally one
side restrained condition (for l = 0.13).

peak stress apart from the Hertz theory is resolved by a certain factor if a pressure profile of the FEA results
follow the Hertzian profile. It was attempted and the result is provided in Fig. 14.
Fig. 14 shows the contact stresses on the top and bottom surfaces, and lateral wall evaluated from the FEA
normalized with respect to contact width (b) and peak stress (po). The normalized Hertzian profile is also illus-
trated for comparison. The used function for the normalized Hertzian profile is as follows:
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pðxÞ 4  x x x
¼ x 1 ¼2 1 : ð8Þ
po b b b b
It is apparently found that the FEA results follow the function form of the Hertzian very well. Therefore, it
is concluded that the contact stress profiles composed between the O-ring and fixture (groove as well) surfaces
are a typical Hertzian profile. The discrepancy can be resolved by appropriate factors that change the contact
width and peak stress. If we designate a and b as control factors for the peak stress and contact width, respec-
tively, and introduce normalized variables  pðxÞ ¼ pðxÞ=po and x ¼ x=b, the following equation becomes an
approximate solution of the contact stress field:
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pðxÞ ¼ 2a bxð1  bxÞ:
 ð9Þ
It is easy to recognize that the peak stress and contact width are linearly proportional to a and 1/b, respec-
tively. On the other hand, it can be said that a and b include the uncertainty of the compressive force as well as
the dimension and the elastic modulus of an O-ring as well if Eqs. (4)–(6) are considered. For the above prob-
lem of the compressed and laterally one side restrained condition (Figs. 12 and 13 for l = 0.13), the ratios of
the contact width and the inverse of those of the peak pressure in Table 2 can be applied to a and b for the
approximate solution of the contact stress field.

6. Concluding remarks

For the purpose of understanding the deformation behaviour of an elastomeric O-ring and its sealing per-
formance, experimental, finite element and contact mechanics analyses were carried out in this research.
Although the O-ring problem has been treated many times and the subject is a relatively common one, it is
worth visiting the problem again with modern technology. For instance, the tests were carried out to obtain
more precise values of the elastic modulus for a more accurate finite element analysis and a computed tomog-
raphy was used to investigate the O-ring deformation when it was compressed and then laterally pressed as
well as when it was compressed only. The present finite element method was validated through comparing with
these test results such as the contact force, deformed shape including extrusion behaviour of an O-ring. In the
analysis of the extrusion behaviour, the influence of the friction coefficient was considerable. It implies that the
friction force should be less to maintain the O-ring’s life. Besides, it was found that the compressive forces
1692 H.-K. Kim et al. / Engineering Failure Analysis 14 (2007) 1680–1692

obtained by Lindley’s formulae were similar to the finite element results but they were comparatively different
from the test results. The present finite element analysis dealt with an O-ring’s deformation when it was com-
pressed and laterally restrained at one side, which has seldom been treated even though it often occurs actually
when an O-ring is inserted into a groove. When we focused on the friction coefficient of 0.13 for the case of the
compressed (22.3%) and laterally one side restrained condition, which was actually measured from the tests, it
was found that a typical function form of the Hertzian pressure profile was apparently found on the contacts
between the O-ring and the top and bottom surfaces as well as that and the lateral wall although the contact
widths and the peak stresses appeared to be different from those in the Hertz theory. So an implementation of
appropriate factors (designated as a and b presently) to the function of the Hertz pressure profile can yield an
approximate solution of the contact stress for an O-ring contact problem.

References

[1] Theyse FH. The inverse hydrodynamic theory and its application in the design of controlled leakage seals between moving parts. in:
Proceedings of the third international conference on fluid sealing BHRA, Cranfield 1967. p. 17–32.
[2] Ebisu T, Yamamoto M, Maekawa H, Onodera A. Fundamental studies on the performance of O-ring for cask. PATRAM
1983;83:672–9.
[3] Lindley PB. Compression characteristics of laterally-unrestrained rubber O-rings. J IRI 1967;1:202–13.
[4] George AF, Strozzi A, Rich JI. Stress fields in a compressed unrestrained elastomeric O-ring seal and a comparison of computer
predictions with experimental results. Trib Int 1987;20(5):237–47.
[5] Strozzi A. Experimental stress-strain field in elastomeric O-ring seals. In: Wieringa H, editor. Experimental Stress Analysis. The
Hague: Martinus Nijhoff; 1986.
[6] Salita M. A simple finite element model of O-ring deformation and activation during squeeze and pressurization. AIAA/SAE/ASME/
ASEE 23rd Joint propulsion conference, San Diego, CA; 1987.
[7] Lindley PB. Load-compression relationships of rubber units. J Strain Anal 1966;1(3):190–5.
[8] Johnson KL. Contact mechanics. Cambridge: Cambridge University Press; 1985.
[9] Blok H. Symposium on lubrication and wear, Houston; 1963.
[10] Fessler H, Ollerton E. J Appl Phys 1957:387–93.
[11] Nowell D, Hills DA. Contact problems incorporating elastic layers. Int J Solids Struct 1988;24(1):105–15.
[12] Dragoni E, Storzzi A. Analysis on an unpressurized, laterally restrained, elastomeric O-ring seal. J Trib 1988;110:193–200.
[13] Muskhelishvili NI. Some basic problems of the mathematical theory of elasticity. Leyden Noordhoff: International Publishing; 1977.
[14] Sokolnikoff IS. Mathematical theory of elasticity. 2nd ed. New York: McGraw-Hill Book Company, Inc; 1956.
[15] Timoshenko SP, Goodier JN. Theory of elasticity. 3rd ed. New York: McGraw-Hill Book Company, Inc; 1970.
[16] Curro JG, Salazar EA. Mechanical behavior of O-rings. Rubber Chem Tech 1973;46:530–9.
[17] Gorelik BM, Bukhina MF, Ratner AV. Variation of contact area in the deformation of rubber cylinders and rings. Sobiet Rubber
Tech 1961;1:10–4.
[18] MARc. General purpose finite element programm 2004 vol. A–D.
[19] Medri G, Strozzi A. Mechanical analysis of elastomeric seals by numerical methods. I&EC Product Res Dev 1984;23:596–600.