Você está na página 1de 11

Tribology International 37 (2004) 689699

www.elsevier.com/locate/triboint
Theoretical study of solid back-up rings for elastomeric seals
in hydraulic actuators
George K. Nikas
+
Mechanical Engineering Department, Tribology Section, Imperial College London, Exhibition Road, London SW7 2AZ, UK
Abstract
The use of back-up rings to support elastomeric seals in high-pressure hydraulic systems is well established in the industry.
However, little is known about the operation of such devices and how they aect or interfere in the sealing mechanism of the
seals they accompany. This study is an attempt to model solid, soft back-up rings in terms of elastohydrodynamic lubrication and
establish their exact role in a sealing system in terms of sealing performance (leakage). The model is applied on a typical case of a
linear hydraulic actuator for a wide range of sealed pressures (135 MPa) and operating temperatures (55 to +135
v
C), and
with one ring installed on the low-pressure side of the actuator. A system of a rubber seal and a properly selected and installed
back-up ring is shown to be often signicantly more ecient in terms of uid leakage than the seal without the back-up ring. A
study reveals the eect of various parameters of the ring (e.g. the elastic modulus and the surface roughness) on the sealing mech-
anism and shows the optimum selection to minimize leakage.
Published by Elsevier Ltd.
Keywords: Anti-extrusion ring; Elastomeric seal; Hydraulic actuator
1. Introduction
Back-up rings (BU rings for short) are devices
resembling washers and used in high-pressure sealing
systems to support elastomeric seals and prevent them
from extruding to narrow gaps when subjected to high
sealed pressures. Therefore, BU rings are also known
as anti-extrusion rings. Their role is often very impor-
tant in protecting seals from extrusion damage and,
thus, maximizing their useful service life. Their appli-
cation is extensive in high-pressure linear hydraulic
actuators where piston rods perform reciprocating
motion and are sealed from the external environment
with rectangular, circular or other type of elastomeric
seals.
Although the role of BU rings is well known in the
industry as anti-extrusion devices, the exact mechanism
of their service has not been extensively studied. For
example, what is the eect of such rings on the leakage
of the seals? Do they simply protect the seals from
extrusion without interfering in the normal seal oper-
ation or do they perform some kind of sealing them-
selves? Such questions are answered in the present
study.
A mathematical model for rectangular rubber seals
for reciprocating piston rods was developed by the
author in Ref. [1]. Part of the model was presented in
Ref. [2] for seals obeying the classic Hookean linear
elasticity model. It is that latter basic model which is
used in the present study for a solid BU ring mounted
on the air-side of a seal, as shown in Fig. 1.
The model assumes a soft ring, as is usually the
case, i.e. a ring with elastic modulus in the order of
0.52 GPa. BU rings are typically made of Polytetra-
uoroethylene (PTFE) reinforced with glass bres.
PTFE has tensile modulus of about 0.5 GPa and
Poissons ratio which could be as high as 0.46 or as
(negatively) large as 12 (expanded form of PTFE).
Extensive information about PTFE and its mechanical
properties can be found in Refs. [36], including com-
posite forms with glass bres. Reinforced PTFE and
similar composite materials are, generally, aniso-
tropiceither transversely isotropic (mechanical
properties are the same on one plane in the solid) or
orthotropic (there exist three perpendicular planes of
+
Tel.: +44-207-594-7236
E-mail address: g.nikas@imperial.ac.uk (G.K. Nikas).
0301-679X/$ - see front matter Published by Elsevier Ltd.
doi:10.1016/j.triboint.2004.02.006
symmetry). The stressstrain equations for transversely
isotropic materials contain ve independent elastic con-
stants instead of the usual two constants (elastic modu-
lus and Poissons ratio) of isotropic materials.
Orthotropic materials on the other hand involve nine
independent elastic constants. These constants are not
easily or readily measured and are aected by tempera-
ture. Lack of knowledge of the values of these con-
stants dictates that the model of this study be
compromised to deal with isotropic, linearly elastic
rings.
The author is not aware of any published theoretical
study involving the modelling of BU rings and, even
experimentally, the work of White and Denny [7] at the
end of World War II oers perhaps one of the very few
insights on this subject. Information will be drawn
from the authors work on seal extrusion modelling
Ref. [8], the experimental study in Ref. [9], and a study
of nonlinear rubber elasticity in Ref. [10] (a summary
of which is included in Ref. [11]), which will be used
for a more accurate description of the behaviour of the
rubber seal under high sealed pressure and extreme
temperatures.
2. Mathematical model
The seal assembly is shown in Fig. 1. It involves a
hydraulic actuator with a reciprocating piston rod.
Nomenclature
A auxiliary variable (Eq. (2))
c
a
, c
b
uid constants
d
cyl
, D
cyl
nominal inner and outer diameter of the actuator cylinder
d
rod
, D
rod
nominal inner and outer diameter of the piston rod
E
cyl
, E
ring
, E
rod
elastic moduli of the cylinder, the ring and the rod
h local lm thickness
p elastohydrodynamic pressure
p
cyl
sealed pressure (Fig. 1)
p
gap
uid pressure at the sealring lower gap
p
ring
rodring average (dry) contact pressure (Eq. (9))
p
0
p
ring
(p
cyl
= 0) (Eq. (10))
q uid mass leakage rate (Eq. (15))
q
ring
, q
seal
mass leakage rates of the back-up ring and the seal
q
system
mass leakage rate of the sealring system (Eq. (14))
R
a
average surface roughness
S
rod
, S
ring
local roughness heights of the rod and the ring
u uid velocity component (Fig. 1)
uu
cyl
; uu
rod
radial surface displacements of the cylinder and the rod (Eqs. (6) and (7))
uu
(rod)
z
; uu
(ring)
z
local normal surface displacements of the rod and the ring (Eq. (15) of Ref. [2])
t uid velocity component (Fig. 1)
V stroking velocity (Fig. 1)
a pressureviscosity coecient at temperature h
a
cyl
, a
ring
, a
rod
thermal expansion coecients of the cylinder, the ring and the rod
DR rodring radial interference at the ring installation
Dh hh
0
e
x
, e
y
, e
z
normal strains of the back-up ring
g actuator-uid dynamic viscosity at temperature h
g
0
g(p = 0)
h operating temperature
h
0
room temperature during the seal and ring installation
l boundary friction coecient at the rodring contact
m
cyl
, m
ring
, m
rod
Poissons ratios of the cylinder, the ring and the rod
q actuator-uid mass density at temperature h
q
0
q(p = 0)
r
x
, r
y
, r
z
normal stresses of the back-up ring
u circumferential coordinate (angle)
690 G.K. Nikas / Tribology International 37 (2004) 689699
Sealing is achieved with a rectangular rubber seal, which
is supported by a solid ring of rectangular cross-section.
Initially (during assembly), the ring may or may not be
in contact with the rod. The computational model has
the capability to check for actual contact at any stage of
the operation of the actuator and establish, based on
the thermomechanical expansion or contraction of the
system elements at dierential temperature and sealed
pressure, if there is contact or not, and the level of inter-
ference between the ring and the rod, if any.
As already explained, the ring is assumed soft, iso-
tropic and linearly elastic, with a typical tensile modulus
between 0.5 and 2 GPa. The piston rod is made of steel
and with a typical elastic modulus of 200 GPa, it is
100400 times stier than the ring. The rubber seal on the
other hand has a typical elastic modulus of 510 MPa at
room temperature and about 100400 MPa at low
subzero temperatures (e.g. at 55
v
C, which is very close
to the lower limit where aircraft hydraulic actuators
operate). Therefore, both geometrically and mechani-
cally, the BU ring of this study resembles the typical elas-
tomeric seal modelled in Ref. [2]. Therefore, the largest
part of the model of Ref. [2] will be used in the present
study for the BU ring as well.
The lubrication of the seal is modelled as described
in Section 2 of Ref. [2], except for the thermoelastic
stress analysis (Section 2.5 of Ref. [2]), which is
replaced here by the more suitable nonlinear elasticity
(MooneyRivlin type) model developed in Ref. [1] and
presented in Refs. [10,11]. For a BU ring that is not in
contact with the piston rod, the full model involves the
analysis of seal extrusion, which consists of checking if
the seal extrudes in the gap between the BU ring and
the rod and, if so, computation of the shape of the
extruded part of the seal as well as of the local contact
pressure between the extruded part and the rod. The
extrusion analysis is presented with details in Ref. [8].
Because of the length and complexity of the relevant
equations and since most have been presented with
extensive details in other studies (Refs. [2,8,10,11]),
only important and new or modied equations are pre-
sented here. Starting with the elastohydrodynamic
aspects of the BU ring model, the appropriate form of
the Reynolds equation is similar to that for a journal
bearing, as was presented in Ref. [2]:
4
D
2
rod

@
2
p
@u
2

@
2
p
@y
2

4
D
2
rod

3
h

@h
@u
A
@p
@u
_ _

@p
@u

3
h

@h
@y
A
@p
@y

6 V g
h
2
(a A)
_ _

@p
@y
=
6 g V
h
3

@h
@y
(1)
where p is the local static pressure of the uid lm
between the rod and the ring, u is the circumferential
coordinate (angle of axis Oz and a point on the outer
surface of the rod on the xz-plane in Fig. 1), h is the
local lm thickness, D
rod
is the outer diameter of the
rod, V is the rod velocity (positive for instrokes and
negative for outstrokes), g is the local uid dynamic
viscosity, and
A =
c
a
(1 c
b
p) [1 (c
a
c
b
) p[
a (2)
where c
a
and c
b
are uid constants and a is the uid
pressureviscosity coecient at the operating tempera-
ture. Constants c
a
and c
b
are those used in the density
pressure formula of the actuator uid, assumed to be
the well-known equation for mineral oils q =
q
0
[1 c
a
p=(1 c
b
p)[, q being the local uid mass
density at operating conditions and q
0
= q(p = 0).
Because of the low stroking velocities (less than 1 m/s)
and sealed pressures (in the order of 140 MPa), the
uid at the rodring contact is expected to behave in a
Newtonian manner without signicant viscous shear
heating. The dynamic viscosity is then calculated from
the Barus formula, g = g
0
e
ap
, where g
0
= g(p = 0).
The lubrication Eq. (1) is analysed and non-dimen-
sionalized as presented in Ref. [2]. The following kine-
matic and boundary conditions apply (see Fig. 1):
At z = 0: u = 0 and t = V
At z = h: u = 0 and t = 0
(3)
The boundary shear stresses are computed as in Ref. [2]:
ss
(rod)
zx
=
h
D
rod

@p
@u
;
ss
(ring)
zx
=
h
D
rod

@p
@u
;
ss
(rod)
zy
=
h
2

@p
@y

V g
h
ss
(ring)
zy
=
h
2

@p
@y

V g
h
(4)
Fig. 1. Seal conguration (half section shown, apart from the rod).
G.K. Nikas / Tribology International 37 (2004) 689699 691
in areas where h ,= 0. At asperity contacts where h = 0,
the surface tractions are
ss
(rod)
zx
= ss
(ring)
zx
= 0 and ss
(rod)
zy
= ss
(ring)
zy
= sgn(V) l p (5)
where l is the boundary friction coecient of the rod
ring contact and sgn(V) is the sign function of argument
V(sgn(V) = 1 if V _ 0 and sgn(V) = 1 if V < 0).
Finally, at the edges of the rodring contact, the pressure
is set as follows: p(y = 0, u)=0 and p(y = W;u) = p
gap
for 0 _ u < 2p, where W is the contact width and p
gap
is
the uid pressure at the small gap (if any at all) between
the seal and the ring.
The quantity of uid in the sealring lower-boundary
gap is generally dicult to predict at a given time and
varies continuously. This small gap (possibly) exists
because both the seal and the ring have in reality
slightly curved corners. The size of the gap at a given
time depends on many factors such as the locally dis-
torted geometry of the seal and the ring under press-
ure, the direction of stroking, the frictional forces at
the rodseal and the rodring interfaces, etc. Because
of the normally high sealed pressure, which results in
the seal being squeezed on the ring, it is believed that
the size of that gap during the operation of the actu-
ator is very small indeed, nearly zero. If it is assumed
that a gap exists, even one with a minute volume, then
it will be lled with leaking uid and air, depending on
the leakage rates of the seal and the ring, and on how
these vary in time. In the case of reciprocating double
rectangular rubber seals without intermediate venting,
it has been shown experimentally (see for example Ref.
[12]) that the interseal pressure is positive and can even
exceed the system pressure after some time of oper-
ation. Applying this observation to the present case of
a minute, non-vented gap, it can be argued that the
inter-sealring pressure will quickly rise as soon as
seal leakage occurs. As a result, the uid continuity
between the seal and the ring will not normally be dis-
turbed by cavitation at the seal outlet since the afore-
mentioned gap pressure will be positive. For practical
computation purposes, it is perhaps realistic to assume
that the inlet pressure to the BU ring is equal to the
average contact pressure between the rod and the ring.
This involves the use of the average contact pressure of
the ring as the entry pressure on the left side of the ring
at the border with the seal (Fig. 1) because it is the ring
which is under lighter pressure with the rod rather than
the seal (which has in most cases a signicantly higher
contact pressure, partially attributed to its incompressi-
bility, whereas the ring is made of compressible
material). It is also emphasized that the leakage of the
seal and of the ring are computed separately before
being compared in order to compute the system leak-
age as is explained later (please see Eq. (14) later in this
study). In this way, the sealring uid continuity is
taken into account, in addition to the regulation of the
inter-sealring pressure via parameter p
gap
, which is
used as boundary condition as explained below Eq. (5).
Moving on with the analysis, both the piston rod
and the actuator cylinder are, generally, assumed to be
hollow, and will deform elastically under the sealed
pressure and from normal changes of the operating
temperature. The corresponding radial surface dis-
placements of the actuator bore and the rod outer sur-
face are computed as in Ref. [2]:
uu
rod
=
p
cyl
D
rod
8 E
rod

D
2
rod
d
2
rod
D
2
rod
d
2
rod
m
rod
_ _

D
rod
2
a
rod
Dh (6)
uu
cyl
=
p
cyl
d
cyl
4 E
cyl

D
2
cyl
d
2
cyl
D
2
cyl
d
2
cyl
m
cyl
_ _

d
cyl
2
a
cyl
Dh (7)
where d
rod
is the nominal inner diameter of the rod, d
cyl
and D
cyl
are the nominal inner and outer diameters of
the actuator cylinder, respectively, a
rod
and a
cyl
are the
thermal expansion coecients of the rod and the actu-
ator cylinder, respectively, E stands for elastic modu-
lus, m stands for Poissons ratio, and Dh = h h
0
, h
being the operating temperature and h
0
being the room
temperature during the ring installation at the factory.
Furthermore, at the rodring contact and because of
the local contact pressure p and associated tractions,
the local elastic normal surface displacements of the
rod and the ring are calculated as analysed in Section
2.3 and Eqs. (11)(15) of Ref. [2].
The BU ring is assumed homogeneous, isotropic and
linearly elastic as already explained. Thus, the Hookean
equations of linear elasticity are applicable:
e
x
a
ring
Dh =
1
E
ring
[r
x
m
ring
(r
y
r
z
)[
e
y
a
ring
Dh =
1
E
ring
[r
y
m
ring
(r
z
r
x
)[
e
z
a
ring
Dh =
1
E
ring
[r
z
m
ring
(r
x
r
y
)[
(8)
where E
ring
, m
ring
and a
ring
are the elastic modulus at
room temperature, Poissons ratio, and the thermal
expansion coecient of the ring, respectively, r stands
for normal stress and E stands for normal strain.Follow-
ing the same analysis as in Section 2.5 of Ref. [2], the
pressure p
ring
at the (dry) rodring contact is calculated
from
p
ring
= p
0

m
ring
1 m
ring
p
cyl
(9)
692 G.K. Nikas / Tribology International 37 (2004) 689699
where
p
0
=
E
ring
1 m
2
ring
[(1 m
ring
) a
ring
Dh m
ring
e
x
e
z
[ (10)
is the contact pressure for p
cyl
=0. In Eq. (10),
e
x
=
D
rod
d
ring
1 (11)
e
z
=
uu
cyl
uu
rod
DR
DR
d
cyl
D
rod
2
_ _ (12)
where DR is the rodring radial interference given at the
ring installation. It is assumed that the ring is squeezed
between the rod and the actuator cylinder, and that
interference DR is caused by compressing the ring on its
outer circumference. Establishing the level (or lack) of
contact at the rodring interface is done via Eq. (9): a
negative (computed) contact pressure p
ring
implies lack
of contact.
The lm thickness at the rodring contact is calcu-
lated from
h = S
rod
S
ring
uu
(rod)
z
uu
(ring)
z
uu
(rod;preloading)
z
uu
(ring;preloading)
z
(13)
where S
rod
and S
ring
are the local roughness heights of
the working surfaces of the rod and the ring, respect-
ively, uu
(rod)
z
and uu
(ring)
z
are the normal elastic surface dis-
placements of the rod and the ring, respectively
(computed from Eq. (15) of Ref. [2]), and the last two
terms in Eq. (13) are owed to the rodring pre-loading
from the initial interference plus any temperature dif-
ferential, computed using pressure p
0
(Eq. (10)).
Having analysed the elastohydrodynamics and con-
tact mechanics of the system, it is now time to calculate
leakage. Both the seal and the BU ring have their own
leakage or interface ow. What is of interest here is the
system leakage, which is the leakage of the sealring
pair. A BU ring in contact with the rod is obviously an
obstruction of any uid ow at the air-side of the sys-
tem (Fig. 1). This may be obvious for a ring installed
pressurized on the rod but if the (often large) environ-
mental temperature variations during system operation
are taken into consideration (e.g. temperature varia-
tions up to 200
v
C in hydraulic actuators for aircraft
landing gear), then it is computed that the contact
pressure between the ring and the rod can vary signi-
cantly from nearly zero (loss of contact) to several MPa.
Analysing the combined mass leakage rate of the seal
(q
seal
) and of the BU ring (q
ring
), the system (seal
ring) mass leakage rate, q
system
, is
q
system
=
0; if q
seal
> 0 and q
ring
< 0
minq
seal
;q
ring
; if q
seal
> 0 and q
ring
> 0
min[q
seal
[; [q
ring
[; if q
seal
< 0 and q
ring
< 0
q
ring
; if q
seal
< 0 and q
ring
> 0
_

_
(14)
where mass leakage rate, q, of the seal or the ring is,
generally, calculated from
q =
_
2p
0
q
_
h
0
h z
h
V z
h z
2 g

@p
@y
_ _

D
rod
2
z
_ _
dz
_ _
du (15)
The signs for the ow rates in Eq. (14) are based on
the coordinate system of Fig. 1, i.e. negative ow indi-
cates uid transferred from left to right, whereas the
opposite is true for positive ow. Thus, negative system
leakage is leakage in the common sense (uid loss from
the actuator), whereas positive system leakage is ben-
ecial leakage because uid returns to the actuator
(carried by the rod during an instroke). Under these
terms and for positive sealed pressure (p
cyl
> 0), the
condition q
ring
> 0 can only hold during an instroke.
3. Application and eects of the back-up ring
The numerical solution of the model presented ear-
lier for the BU ring is similar to that for the seal, as
described in Section 3 of Ref. [2]. Specically, the non-
dimensionalized and discretized version of Eq. (1) (as
in Eq. (3) of Ref. [2]) is solved with the successive
overrelaxation (SOR) method. For a typical case, the
complete analysis (elastohydrodynamic solution for the
sealring pair and computation of leakage and friction)
for steady-state conditions takes about 80 s for rough
contacts with 100 100 gridpoints, using a 1.5 GHz
Pentium-4 PC, with typical total memory allocation of
10 MB during the program execution.
This section is devoted to the study of the eects of a
typical solid BU ring in a hydraulic actuator as shown
in Fig. 1, used in aircraft landing gear. The rectangular
rubber seal used for this purpose has, typically,
remarkably dierent behaviour at subzero temperatures
than at room temperature. Fig. 2 shows the stress
strain diagram of such a seal. For the more accurate
description of the mechanics of this seal, a modied
version of the MooneyRivlin nonlinear model of rub-
ber elasticity is used, as presented in Refs. [1,10,11].
The operating data used in the examples of this section
are presented next.
. Operating data: h = 54, +23 and +135
v
C,
h
0
= 23
v
C, sealed pressure is varied between
0.7 MPa (100 lb/in.
2
) and 27.6 MPa (4000 lb/in.
2
),
V = 0:2 m=s, rodactuator clearance at the air-side
of the seal in the absence of a BU ring is 0.2 mm.
Assumption: p
gap
= p
ring
(explained in Section 2).
. Piston rod: steel hollow cylinder, d
rod
= 47 mm,
D
rod
= 50 mm, E
rod
= 207 GPa, m
rod
= 0:3, a
rod
=
12 10
6
K
1
; average roughness (R
a
) = 0:20 lm,
r.m.s. roughness = 0:23 lm, roughness maximum
peak-to-valley height = 0:8 lm. Actuator: steel hol-
G.K. Nikas / Tribology International 37 (2004) 689699 693
low cylinder, d
cyl
= 60 mm, D
cyl
= 63 mm, E
cyl
=
207 GPa, m
cyl
= 0:3, a
cyl
= 12 10
6
K
1
.
. BU ring: solid ring of rectangular cross-section,
d
ring
= 50 mm, D
ring
= 60 mm, width = 2 mm, E
ring
=
1 GPa (except in Fig. 7), m
ring
= 0:3, a
ring
= 12
10
6
K
1
, rodring boundary friction coefficient =
0:10, rodring initial radial interference = 10 lm
(except in Figs. 68); average roughness (R
a
) =
0:20 lm, r.m.s. roughness = 0:23 lm, roughness maxi-
mum peak-to-valley height = 0:8 lm (dierent rough-
ness values are used in Fig. 9).
. Seal: elastomeric ring of rectangular cross-section,
inner diameter = 49:9 mm, width = 3:5 mm, corner
radius = 0:2 mm, stressstrain properties and elastic
modulus given in Fig. 2, Poissons ratio = 0:499,
thermal expansion coefficient = 2 10
4
K
1
, rod
seal boundary friction coefficient = 0:06, rodseal
initial radial interference = 300 lm, glass-transition-
temperature = 47
v
C; average roughness (R
a
) =
1:50 lm, r.m.s. roughness = 1:73 lm, roughness
maximum peak-to-valley height = 6:0 lm.
. Actuator uid: mineral oil, q
0
= 897:0 kg=m
3
at
54
v
C, 842.2 kg/m
3
at +23
v
C, 762.5 kg/m
3
at +135
v
C; g
0
= 0:8073 Pa s at 54
v
C, 0.4795 Pa s
at +23
v
C, 2:7069 10
3
Pa s at +135
v
C; a =
20 GPa
1
, c
a
= 0:6 GPa
1
, c
b
= 1:7 GPa
1
. (The
pressureviscosity coecient is generally tempera-
ture dependent. However, it is assumed constant in
this study due to lack of data. Nevertheless, it is
estimated that its variation with temperature, if
taken into account, would not aect qualitatively
the results presented later because of the relatively
low pressures involved.)
The above data are from a real aviation application.
The roughness heights of the components were created
numerically using a pseudo-random number generator
so that the target average roughness was achieved.
Typical components (rod, seal and BU ring) were mea-
sured with a Talysurf prolometer. Moreover, the seal
and the BU ring were measured with a LaserSurf
optical (non-contacting) instrument. The mathematical
roughness simulation described previously gives pro-
les symmetric to the mean line with zero skewness.
The seal in particular has blunt roughness features with
a negative coecient of kurtosis, equal to 1.2 (thus a
platykurtic distribution, whereas a coecient of zero
corresponds to the Gaussian distribution). More details
of the simulation and the measured roughness proles
are included in Ref. [1].
In the lubricated regions, the uid continuity equa-
tion is satised through Eq. (1), which is derived from
the equation of uid mass conservation. The regions of
solid contact between roughness asperities are found to
be isolated as in similar studies in the literature dealing
with mixed (partial) lubrication. A representative
example of the extent of asperity interactions for the
studied type of seal is given in Fig. 10 of Ref. [2].
According to the results obtained from the current
study and for the range of operating conditions tested,
both the seal and the BU ring operate in the mixed to
boundary lubrication regime. Specically, at the
maximum sealed pressure studied (27.6 MPa), stroking
velocity of 0.2 m/s, and with 50,000 gridpoints in
each of the two contacts (500 points along the rod axis
and 100 points along the circumference), the compu-
tation gives in the worst case scenario in the rodring
contact an average lm thickness of 0.14 lm. With the
given roughness proles, this corresponds to a lambda
ratio (dened as the average lm thickness over the
composite r.m.s. roughness) of 0.44 and a maximum
portion of gridpoints in asperity contacts in the region
of 45%. Therefore, in the highest load case, the BU
ring operates in what is roughly dened as boundary
lubrication regime (lambda below 1, as in p. 41 of
Hamrock [13]). The seal on the other hand, for the
worst case scenario, operates in the mixed or partial
lubrication regime (lambda between 1 and 5Ref. [13])
but close to the boundary lubrication regime, with a
computed lambda ratio of 1.31 and an average lm
thickness of 2.3 lm. These results, although realistic
and in agreement with similar theoretical results in the
literature (as for example with Ruskell [14]), do not
take into account possible adhesion eects, which
Fig. 2. Typical stressstrain data of an industrial elastomeric seal
material.
694 G.K. Nikas / Tribology International 37 (2004) 689699
would be expected to reduce the average lm thickness,
and there is room for improvement in this area.
The most interesting questions to ask are What is
the eect of a BU ring on leakage? and then How
does it vary with the sealed pressure and the operating
temperature? Fig. 3 provides an answer to the previous
questions for the low subzero temperature of 54
v
C.
The sealed pressure is varied from 0.7 MPa (100 lb/in.
2
)
to 27.6 MPa (4000 lb/in.
2
). The mass leakage rate for
instrokes and outstrokes is shown on the left vertical
axis with the sign convention dened earlier. Similarly,
the mass leakage rate per cycle (a cycle is dened as one
outstroke followed by an instroke with the same strok-
ing velocity and equal stroking length) is shown on the
right vertical axis, assuming that the outlet of the system
is always ooded. According to Fig. 3, during an out-
stroke without a BU ring, negative leakage increases
rather linearly with the sealed pressure, in the same way
that positive leakage increases during the following
instroke under the same conditions. However, with the
introduction of the BU ring (with a 10 lm initial inter-
ference as quoted earlier), the leakage, both for out-
strokes and for instrokes, is dramatically reduced and is
insensitive to the sealed pressure variations. What is of
major concern in the industry is of course the leakage-
per-cycle, which is used for seal performance evaluation.
According to Fig. 3 (right axis), the standard leakage-
per-cycle (leakage during instroke leakage during
outstroke) shows an increase when there is no
pre-loaded BU ring installed in the system, whereas the
presence of the pre-loaded BU ring gives an almost con-
stant leakage rate, regardless of the magnitude of the
sealed pressure. Looking at the two relevant leakage
lines (leakage-per-cycle with and without a ring), it is
seen that those two lines intersect at point P, which
corresponds to p
cyl
15 MPa. On the right of point P
(i.e. for p
cyl
> 15 MPa), the benet of the BU ring in
reducing the leakage-per-cycle is obvious; but on the left
of point P (p
cyl
< 15 MPa), the BU ring actually hinders
the seals operation of reducing leakage. Therefore,
under the given conditions, a pre-loaded BU ring is ben-
ecial in reducing the system leakage only if the sealed
pressure is kept higher than a critical value.
Such a critical pressure is also obvious in Fig. 4,
which refers to the same conditions as Fig. 3, except
for the operating temperature, which is now equal to
+23
v
C. The appearance of the various curves is simi-
lar to that of the curves in Fig. 3 but, this time, point P
has moved to the left, signifying a lower critical press-
ure of about 7.5 MPa, nearly half that of Fig. 3 refer-
ring to 54
v
C. In this case, the ring oers a benecial
reduction of the leakage-per-cycle for a wider range of
sealed pressure. Notice that the leakage-per-cycle with-
out a BU ring has increased signicantly compared
with that in Fig. 3, especially at the high end of the
sealed pressure, mainly because the uid dynamic vis-
cosity at +23
v
C is only 59% of the viscosity at 54
v
C.
On the other hand, the leakage-per-cycle in the pres-
ence of the ring has been signicantly decreased at the
increased-temperature case, attributed both to the
changed viscosity and to the changed contact pressure
considering thermal expansion of the system elements.
Moving on to the high end of the temperature scale,
Fig. 5 shows the leakage results at the operating tem-
perature of +135
v
C. Observing the curve referring to
leakage for an outstroke without a BU ring, it is clear
that leakage increases with the sealed pressure. This
could be attributed to the large reduction of the uid
dynamic viscosity at the high temperature of +135
v
C,
Fig. 3. Eect of the sealed pressure on leakage at 54
v
C.
Fig. 4. Eect of the sealed pressure on leakage at +23
v
C.
G.K. Nikas / Tribology International 37 (2004) 689699 695
which also aects adversely the leakage during an
instroke (again without a BU ring). It is interesting to
see that even during an instroke in the absence of a ring,
leakage, despite being positive at low sealed pressure,
becomes negative (uid loss) beyond a sealed pressure
of about 6 MPa. The latter could be explained by the
increased Poiseuille-type ow at the rodseal contact
beyond a certain sealed pressure, which overcomes the
opposing Couette-type ow in the contact from the
retracting rod. However, it must be mentioned at this
point that the theoretical prediction of the model at
high temperature contradicts experimental results that
show decrease of leakage. This discrepancy could poss-
ibly be explained by an overestimation of the thermal
displacements of the piston rod and actuator cylinder as
derived in the model, although, on the other hand,
experiments done at high temperature are often incon-
clusive because of uid vaporization, which prevents the
accurate weighing of the uid that actually leaked.
The high-temperature results make the eect of a BU
ring on leakage seem even more signicant, as Fig. 5
Fig. 5. Eect of the sealed pressure on leakage at +135
v
C.
Fig. 6. Eect of the rodring interference on leakage for various
sealed pressures.
Fig. 7. Eect of the rodring interference on leakage for variable
ring elastic modulus.
Fig. 8. Eect of the rodring interference on the leakage-per-stroke.
696 G.K. Nikas / Tribology International 37 (2004) 689699
demonstrates that the ring reduces leakage substan-
tially, both during an outstroke and during an
instroke, particularly at higher sealed pressures.
Regarding the leakage-per-cycle, Fig. 5 shows that
point P, which corresponds to the position where a BU
ring under the studied conditions becomes eective (or
ineective) in reducing overall leakage, has again
moved to the left. Now, the critical sealed pressure,
over which the ring is an eective seal in itself, is about
0.5 MPa, which is quite low. In fact, there is no region
in Fig. 5 where the BU ring hinders the sealing per-
formance of the system. On the contrary, it improves
the performance by a margin directly related to the
sealed pressure, and, thus, its presence is essential in
minimizing and practically eliminating leakage.
The eect of the initial rodring radial interference is
examined next in Fig. 6 (referring to h = h
0
= 23
v
C
to eliminate any thermal expansion or contraction
interference), having the sealed pressure as a para-
meter. It is seen that the leakage-per-cycle increases
almost linearly with the rodring interference for the
studied range of interference (110 lm), although the
leakage dierences shown are small (about 4%). On the
other hand, increasing the sealed pressure results in a
reduction of the leakage-per-cycle, which is the
opposite of what was observed for the leakage-per-
cycle without a BU ring in Fig. 4. However, it must be
realized that the seal leakage in the absence of a BU
ring is directly aected by any increase of the sealed
pressure in that the left side of the rodseal contact is
directly exposed to the sealed pressure; this is not the
case for the left side of the rodring contact, which, as
explained earlier in this study, is under a more-or-less
constant uid pressure p
gap
(either zero pressure or, in
the other limiting-case scenario, a pressure approxi-
mately equal to the (dry) average contact pressure,
p
ring
). What does change for the ring when the sealed
pressure is varied is the radial normal surface displace-
ment of the rod and the actuator cylinder (see Eqs. (6)
and (7)), which will aect the rodring contact pressure
and the contact lm thickness. Higher sealed pressure
will increase uu
rod
and uu
cyl
and, thus, release some of the
rodring contact pressure and reduce the lm thick-
ness. The conclusion from Fig. 6 is that, in order to
minimize the leakage-per-cycle, a lightly pressurized
BU ring must be used, and this ring will perform
slightly better at higher sealed pressure.
One important parameter in this analysis is the BU
ring mechanical behaviour. Although not much is read-
ily available about the stressstrain relation and the
contact mechanics of typical BU rings made of com-
posite materials such as glass-bre reinforced PTFE, as
has already been explained, the present model assumes
that the ring behaves in the classic Hookean manner
(isotropic and linearly elastic), with a Poissons ratio in
the order of 0.3 (justiable by the fact that non-expan-
ded and non-reinforced PTFE has a Poissons ratio of
0.46) and with elastic modulus comparable to that of
non-reinforced PTFE (0.30.8 GPa) to reinforced
PTFE (a roughly estimated 0.52.0 GPa).
Fig. 7 shows the eect of the rodring interference
on the system leakage-per-cycle, using the ring elastic
modulus as a parameter. The gure refers to room
temperature (+23
v
C) and average sealed pressure
(p
cyl
= 6:9 MPa = 1000 lb=in:
2
). It is seen that by using
a ring with higher elastic modulus, the leakage-
per-cycle is reduced proportionately. A stier ring will
deform less under the applied elastohydrodynamic
pressure and, thus, the lm thickness at the rodring
contact will be thinner. Although this argument can be
used to predict a reduced leakage-per-stroke, it is not
straightforward to explain the reduction of the leakage-
per-cycle. A further insight is provided in Fig. 8, which
shows the eect of the rodring interference on the
leakage-per-stroke, both for an instroke and for an
outstroke, for the same conditions as in Fig. 7 and for
a ring with elastic modulus of 0.5 GPa. The left vertical
axis shows the dierence of the system leakage at rod
ring interference DR as a percentage of the system leak-
age at interference DR = 0 for the outstroke, whereas
the right vertical axis shows the same for the instroke.
Both axes use the same scale absolute limits. It is seen
that the outstroke leakage is aected more than the
instroke leakage (outstroke/solid curve is above
the instroke/dashed curve) for DR > 2 lm, whereas
the opposite is true for DR < 2 lm. The vertical-line
pattern between the two curves highlights the afore-
Fig. 9. Eect of the ring average surface roughness on leakage.
G.K. Nikas / Tribology International 37 (2004) 689699 697
mentioned dierence (which of course depends on
roughness).
Speaking of roughness, it is best to complete this
study with a presentation of the eect of the surface
roughness of the BU ring on the system leakage. Fig. 9
shows the eect of the average roughness R
a
of the ring
on the leakage-per-cycle at three representative sealed
pressureslow, average and high. The average rough-
ness was varied (with the method described at the
beginning of this section) in order to cover all cases
between a perfectly smooth ring and a ring which is
much rougher than the rod. According to Fig. 9, the
leakage-per-cycle is only weakly aected by the ring
roughness, and this is true regardless of the magnitude
of the sealed pressure (within the studied range of pres-
sures). The eect of the sealed pressure is clearer to see:
lower sealed pressure results in higher leakage-per-
cycle, but the dierences are still rather insignicant
and more immediate when increasing the sealed press-
ure from low values (i.e. from p
cyl
= 0:7 MPa and
upwards). The eect of the ring roughness on leakage-
per-stroke (not shown) is linear, with leakage increas-
ing with roughness, which is explained by the increase
of the average lm thickness at the rodring contact
(easily realized from Eq. (13)). The leakage curves in
Fig. 9 reveal an optimum average roughness to
minimize the leakage-per-cycle, which depends on
the sealed pressure. For example, at p
cyl
= 6:9 MPa
(1000 lb=in:
2
), the leakage-per-cycle is minimized for
ring roughness R
a
= 0, something which is not true for
the curve of p
cyl
= 27:6 MPa. However, the leakage dif-
ferences are, by any standards, negligible and subject to
many external variables, which negates the need to be
preoccupied with special ring selection to meet any spe-
cic roughness criteria other than to protect the rods
surface from abrasion.
4. Conclusion
Solid BU rings of relatively low elastic modulus,
used in hydraulic actuators to support rubber seals for
reciprocating motion (Fig. 1), were shown to behave in
an intrusive manner, aecting the sealing performance
of such systems. The rings, originally intended to pre-
vent seal extrusion at high sealed pressure and thus
prevent seal damage, were shown to actually have seal-
ing properties when installed with an initial interference
with the rod on which they operate. Even an initial
interference of a few microns, in combination with an
average elastic modulus of the BU ring in the order
of 1 GPa, generates sucient contact pressure to estab-
lish sealing at high sealed pressures. This sealing mech-
anism is aected by several factors, e.g. the sealed
pressure (transferred to the BU ring via the adjacent
rubber seal), the operating temperature, the viscosity of
the actuator uid, the surface roughness of the ring,
etc. Naturally, the ring intrusion is valid both during
outstrokes and instrokes, and it could very well
increase or decrease leakage, depending on the stroking
direction.
It was generally found that the leakage-per-stroke
could be signicantly decreased by using a rather
lightly pressurized BU ring (see for example Figs. 35).
The leakage reduction is signicant at high operating
temperature, with the actuator uid having a very low
viscosity (Fig. 5). However, and for standard perform-
ance evaluation purposes, the leakage-per-cycle is the
one most relevant in industrial circles. For the leakage-
per-cycle then, it was found that, for given operating
conditions, there exist a critical sealed pressure over
which a BU ring becomes a more eective seal than the
rubber seal it supports. That critical pressure (corre-
sponding to point P in Figs. 35) is inversely pro-
portional to the operating temperature, with eect of
being very low (compared with the sealed pressure) at
very high temperaturesee for example Fig. 5.
The system leakage-per-cycle was found to slightly
increase with the rodring interference and (also
slightly) decrease with the sealed pressure (Fig. 6). The
elastic modulus of the BU ring on the other hand has a
stronger eect on leakage: the leakage-per-cycle exhi-
bits a rather signicant reduction when using a ring
with higher elastic modulus (Fig. 7).
Finally, the leakage-per-stroke of the sealring sys-
tem is found to be linearly aected by the average sur-
face roughness of the BU ring and the eect is
signicant. However, the leakage-per-cycle is only
weakly aected by the magnitude of the average sur-
face roughness of the ring (Fig. 9), irrespectively of the
sealed pressure. An optimum average surface rough-
ness of the ring is found to minimize the leakage-per-
cycle of the system, but this is readily aected by the
operating conditions (e.g. the sealed pressure), and the
benet in terms of leakage reduction is negligible to
pursue this target any further.
For a nal comment, it is emphasized that the results
presented for the leakage per instroke and for the leak-
age-per-cycle are based on the assumption of a ooded
air-side of the actuator (leaked uid resting on the rod,
readily available to be picked up during an instroke).
Otherwise, negative leakage would be signicantly
increased and, based on the presented results, the
advantage of using a BU ring would be even greater.
Acknowledgements
This study is part of a project which was nancially
supported by Smiths Aerospace Actuation Systems
Cheltenham (UK), Smiths Aerospace Actuation
SystemsWolverhampton (UK), Polymer Sealing
Solutions Ltd (UK), and the British Department
698 G.K. Nikas / Tribology International 37 (2004) 689699
of Trade and Industry through the Civil Aircraft
Research and Demonstration Programme.
References
[1] Nikas GK. Determination of polymeric sealing principles for
end user high reliability. Technical Report DOW-08/01 (Dowty
project), Imperial College London, Mechanical Engineering
Department, Tribology Section, London, 2001.
[2] Nikas GK. Elastohydrodynamics and mechanics of rectangular
elastomeric seals for reciprocating piston rods. Trans ASME,
J Tribol 2003;125(1):609.
[3] Li W, Mays S. Analysis of PTFE material in rotary seals. Pro-
ceedings of the 17th International Conference on Fluid Sealing.
UK: The British Hydromechanics Research Association; 2003,
p. 15765.
[4] Lee N-J, Jang J. The eect of bre-content gradient on the
mechanical properties of glass-bre-mat/polypropylene compo-
sites. Compos Sci Technol 2000;60:20917.
[5] Ganghoer J-F, Brillard A, De Borst R. Description of the
mechanical behaviour of micropolar adhesives. Math Comput
Model 1998;27(7):2349.
[6] Caddock BD, Evans KE. Negative Poisson ratios and strain-
dependent mechanical properties in arterial prostheses. Bioma-
terials 1995;16(14):110915.
[7] White CM, Denny DF. The sealing mechanism of exible pack-
ings. Scientic and Technical Memorandum 3/47, UK Ministry
of Supply, 1947.
[8] Nikas GK. Analytical study of the extrusion of rectangular elas-
tomeric seals for linear hydraulic actuators. Proc Inst Mech Eng
Pt J: J Eng Tribol 2003;217(5):36573.
[9] Rana A, Sayles RS, Nikas GK, Jalisi I. An experimental tech-
nique for investigating the sealing principles of reciprocating
elastomeric seals for use in linear hydraulic actuator assemblies.
Proceedings of the Second World Tribology Congress, Vienna,
Austria. 2001 proceedings are on CD.
[10] Nikas GK, Sayles RS. Nonlinear elasticity of rectangular elasto-
meric seals and its eect on elastohydrodynamic numerical
analysis. Tribology International, this issue. doi:10.1016/j.tri-
boint.2004.02.002.
[11] Nikas GK. Transient elastohydrodynamic lubrication of rec-
tangular elastomeric seals for linear hydraulic actuators. Proc
Inst Mech Eng Pt J: J Eng Tribol 2003;217(6):46173.
[12] Field GJ, Nau BS. Interseal pressure between reciprocating rec-
tangular rubber seals. Proceedings of the Fifth International
Conference on Fluid Sealing, UK. The British Hydromechanics
Research Association; 1971, p. 15765.
[13] Hamrock BJ. Fundamentals of uid lm lubrication. McGraw-
Hill; 1994.
[14] Ruskell LEC. A rapidly converging theoretical solution of the
elastohydrodynamic problem for rectangular rubber seals. Proc
Inst Mech Eng J Mech Eng Sci 1980;22:916.
G.K. Nikas / Tribology International 37 (2004) 689699 699

Você também pode gostar