A Thermo-Hyclrodyoamic Analysis

M. D. Pascovici1 of a Mechanical Face Seal

I. Etsion
Department of Mechanical Engineering,
Technion, Haifa 32000, Israel
A thermo-hydrodynamic analysis is performed for a face-to-face double seal con-
figuration. Temperature and viscosity variations both across and along the sealing
gap are considered and realistic boundary conditions are considered. The energy
equation is solved analytically and the radial temperature variation is presented by
an implicit equation. This approach enables analytical parametric investigation and
gives better understanding of the effects of various parameters on the seal's thermal
behavior.

Introduction
Thermo-hydrodynamic (THD) analysis of mechanical face
seals is desired in cases where information on temperature
distribution in the sealing gap is important. These cases include
phase change due to boiling of sealed liquids and thermal
distortions of the seal mating faces which affect the sealing
performance.
A complete THD analysis is very complex mainly due to the
large number of variables involved in the process. In addition
it is very difficult to formulate the problem so as to be suf-
ficiently general. For these reasons the number of published
THD analyses related to mechanical seals is small and in all
cases many simplifying assumptions are introduced that may
invalidate the results for general use. Another limiting factor
in a number of mechanical seal's THD analyses is the use of
cumbersome commercial finite element analyses (FEA), re-
sulting in numerical solutions which do not allow an elegant
analytical parametric investigation.
Li (1976) solved the problem of thermal deformation in a
mechanical seal. For the solution of the energy equation a
constant amount of heat input at the sealing surfaces was
assumed. The fluid properties were considered constant along
the sealing dam, and the temperature field was obtained from
a finite element numerical solution. Isothermal lines were pre-
sented for both the seat and the ring of the seal. Hughes and
Chao (1980) analyzed the problem of phase change in liquid
face seals for the isothermal and adiabatic bounds. In their
seal model the mating rings are assumed to be semi-infinite
solids. This assumption neglects the heat transferred from the
outer diameter surfaces of seal rings into their surrounding
sealed liquid which in practical application may be an impor-
tant source of cooling. Lebeck (1980) used an actual seal ge-
ometry but considered a common heat convection coefficient
'Visiting Scientist-Politechnic Institute of Bucharest, Bucharest Romania.
Contributed by the Tribology Division of THE AMERICAN SOCIETY OF M E -
CHANICAL ENGINEERS and presented at the STLE/ASME Tribology Conference,
St. Louis, Mo., October 13-16, 1991. Manuscript received by the Tribology
Division January 3, 1991; revised manuscript received June 1991. Paper No.
91-Trib-44. Associate Technical Editor: D. Dowson.
for both the rotating and fixed rings at their surfaces in contact
with the high pressure sealed liquid. This thermal boundary
condition overlooks the difference between forced and free
convection at the boundaries of the rotating and fixed rings,
respectively. Lebeck also neglected temperature gradients across
the film thickness. This assumption results in an unrealistic
boundary condition of zero gradient at both surfaces of the
mating rings (see Pinkus, 1984). Dumbrava and Morariu (1987)
presented a THD analysis for a face to face double mechanical
seal arrangement. An analytical solution was attempted for
the face temperature taking into account heat conduction in
the fluid and in the seal rings, and convection by a surrounding
cooling liquid. The analysis is, however, isothermal and gives
unrealistic constant face temperature along the sealing dam.
Doust and Parmar (1987) presented what may be considered
the most realistic model so far for a THD analysis of me-
chanical face seals. However, they used a finite element com-
puter code which does not enable analytical parametric
investigation or dimensionless representation of the results.
More recently, Waidner (1989) and Salant and Hassan (1989)
treated the THD problem, but here again various assumptions
were made like linear radial temperature variation with con-
stant viscosity and density in the first paper, and a uniform
constant temperature in the second. Zeus (1990) presented
results for temperature distribution obtained from thermal
analysis by means of FEA and showed the isotherms in both
the seal ring and seal seat.
The motivation for the present work originated from the
need to know the face temperature distribution and its relation
to the sealing gap size in controlled mechanical seal developed
by one of the authors, see Etsion et al. (1990), and Etsion
(1990). In this seal, shown in Fig. 1, the stator face temperature
is used as an input to a control system which controls the
closing force, and hence, the sealing gap size, in order to
maintain a preselected face temperature.
An attempt is made in the present paper to relax some of
the simplifying assumptions made in previous THD analyses
and at the same time to refrain from using a cumbersome
numerical solution. It is hoped that the present THD analysis
will give a better insight into the complex problem of THD
lubrication in mechanical seals.

Journal of Tribology OCTOBER 1992, Vol. 114/639

Copyright © 1992 by ASME
Downloaded From: http://tribology.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
 CONTROLLED
WATER INLET PRESSURE
THERMOCOUPLE

WATER ROTOR STATOR

OUTLET

Fig. 1 A controlled mechanical seal

Fig. 2 Seal model

Model d_(p]^dp
=0 (1)
A cross section of the controlled seal is shown in Fig. 1. A dr\ ix dr
face to face double seal arrangement is considered, which is The density, p, and viscosity, \i, at each radial location, r,
symmetrical about a midplan of the rotor. Hence, due to this are described as functions of the local average temperature,
symmetry, only one stator has to be analyzed. The following Tm, (Pinkus and Sternlicht, 1961) in the form
assumptions are made:
1. The mating faces are aligned so that axisymmetry pre- p = p*[\-a(.Tm-T*)} (2)
vails. 0{T T )
2. The fluid is a Newtonian liquid of constant viscosity H = H*e- ">~ * (3)
and density across the height of the sealing gap (these where the asterix indicates known reference values.
properties, however, vary in the radial direction). The local average temperature, T,„, at a radius r, is defined
3. The temperature, 7), of the sealed fluid surrounding the
surfaces at the outer diameters of the rotor and stator 1 ["
is known. < =h ~u Tdy (4)
J0
4. The surfaces exposed to the air are insulated.
5. The convective heat transfer between the stator and the and the temperature T is obtained from the energy equation
sealed fluid is much smaller compared to that of the 2
rotor and hence, can be neglected. d2T
k +>i = 0
Assumptions 4 and 5 actually mean that the stator can be ~tf \i (5)

considered as being perfectly insulated.

6. A steady-state condition is assumed. Equation (5) is based on the assumption that most of the
The model to be analyzed is shown in Fig. 2. Surfaces 1 and heat generated in the lubricating film is rejected by conduction
2 are the mating faces of the stator and rotor, respectively. (Pascovici, 1974). This is true especially in mechanical seals
Surface 3 is the rotor surface in contact with the sealed fluid. where it is desired to eliminate leakage.
The sealing gap is of thickness h(r) where h(r) can be any
arbitrary function of r. Method of Solution
The film thickness ht at the inner radius r, is determined by For the aligned seal the velocity u changes linearly (Couette
a force balance between the closing force acting on the back flow) from zero at the stator face, y = 0, to cor at the rotor
of the stator, and the opening force, due to the pressure dis- face, y = h. Hence,
tribution, p, in the lubrication film. This pressure is obtained
from a solution of the Reynolds equation which for the aligned du cor
(6)
seal has the form ~dy = ~h

Nomenclature
Nu = Nusselt number, 2r0H/k /3 = temperature-viscosity coeffi-
a = half the thickness of rotor ring q = heat flow rate cient
a = half the active thickness of ro- r = radial coordinate J3 = dimensionless coefficient, $Tj
tor ring r = dimensionless radial coordi- fluid viscosity
D = dissipation number, fifoi2rl/kTf nate, r/r0 P fluid density
G = geometry parameter, r0//z,sin2(p T = temperature CO rotor angular velocity
H = convection coefficient Tm = mean local temperature, Eq.
h = film thickness (4) Subscripts
h = dimensionless film thickness, T = dimensionless temperature, / sealed fluid
h/hj 777) / inner radius
K = conductivity ratio, k/k* u = tangential velocity of rotor sur- o outer radius
k = fluid thermal conductivity face 1 at stator face
k* = rotor thermal conductivity y = axial coordinate 2 at rotor face
/ = length of thermal path a = temperature-density coefficient 3 at O.D. rotor surface

640 / Vol. 114, OCTOBER 1992 Transactions of the ASME

Downloaded From: http://tribology.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Substituting in Eq. (5) and integrating twice gives
f2=Tm-l-Dr2e-m»'-l) (20)
(7)
and

r c Ti = Tm + ^Df2e"fl{T'" (21)
~£(?)'* ;'^ (8)
In Eqs. (19) through (21) the temperature is normalized by the
The constants of integration Ct and C2 are obtained from the
thermal boundary conditions. Thus, at surface 1 of the stator, sealed fluid temperature 7). The parameter D is the dissipation
y = h. number in the form p,fO?r\/kTj, K is a conductivity ratio
k/k*, and Nu is the Nusselt Number, having in our case (see
Becker, 1963), the form Ir^H/k. The last parameter, G, is a
and, by assumptions 4 and 5 geometric parameter in the form 7-0//2/sin2<p where h-, is the film
thickness at r = r,.
=0 (9)

At surface 2 of the rotor, y = 0. Results and Discussion

T=T2(r) (10) As can be seen from Eq. (19) the average local dimensionless
temperature, Tm, is expressed in an implicit form that requires
From Eqs. (7), (8), (9), and (10) we have an iterative solution. Nevertheless, the effect of the various
2„2 dimensionless parameters on the temperature distribution is
/tcu r (11) clearly apparent.
kh _ The_temperature variation across the film thickness
and Tx - T2 is from Eqs. (20) and (21)
C2=T2(r) (12)
Tl-T2 = -DT2e-lS{T"> (22)
The constant C\ is the temperature gradient at the rotor face, 2
and contrary to the assumption made in Lebeck (1980) this
and, hence, the dimensional temperature difference, T\ — T2,
gradient is not zero.
at ? = 1 is always less than DT/2 or, substituting D, the
From Eqs. (8), (11), and (12) we have
maximum temperature change across the film thickness is (see
2 2
also Eq. (14))
T=T2 + (13)
2h)h
T,-T2< ^/"'v (23)
Hence, the stator face temperature at y = h is 2k
2 2
From Eq. (23) it is clear that the temperature change is in-
-•T2 + HOI r (14)
dependent of the shape and size of the film thickness h. It is
2k also clear that for a given speed o> and seal size r0 the difference
The relation between the temperature T2(r) and the sealed T\ — T2 depends on the ratio of the fluid viscosity to its thermal
fluid temperature 7) can be obtained by considering a heat conductivity. Hence, the axial temperature change in a typical
balance between the conduction in the rotor from surface 2 to oil seal with 7) about 60°C (jx - 80 mPaS, k = 0.14
surface 3, and the convection from surface 3 to the sealed fluid W/m °C) (see Holman, 1976) for fexample is about 700 times
(see Appendix). This results in the following relations higher than in a water seal (/*/ = 0.5 mPaS, k = 0.65
2/xcoV k* (T2-T3)sin 2<p W/m °C). If the seal rotates at 300 rad/S and has a diameter
(15a) of 100 mm the maximum difference Tx - T2 will be over
h ln(r 0 /r)
and 60 °C for the oil case and only 0.09 °C for the water case.
2
From the above example it is evident that while the assumption
k*(T2-T3)cos <p of constant temperature across the film thickness, made in
= H(Ti-Tf)r0 (156)
ln(r0/r) several THD analyses, may be correct in some cases, it is very
inaccurate in others.
where the angle <p (see Appendix) indicates the direction of the
Rearranging Eq. (19) in the form
heat flow path in the rotor.
Eliminating r 3 between Eqs. (15«) and (15£>) we have ( r , „ - l ) / ( T ' » - 1 » = J D/(r) (24)
2/^y cos <p \n(r0/r) where
T2=Tf+ (16)
h sin lip Hrn + k* 2cosV r2
f(r)=2 ••Kln7)G=+, (25)
The viscosity, ix, in Eq. (16) is the local viscosity ju,(r) which Nu / h 3
by Eq. (3) can be written as
clearly shows that for a given dissipation number, D, the level
Ai = / v e - 3 ( 7 ' « - r / ) (17) of Tm depends on /3 a n d / ( r ) . Reducing /3 will increase the
where /*/ is the viscosity of the sealed fluid at its known tem- level of Tm. Hence, all the analyses based on the assumption
perature Tf. The mean temperature, Tm, is obtained from Eqs. of constant viscosity, namely (3 = 0, along the radial width
(4) and (13) in the form of the sealing gap, overestimate the temperature level. The
2 2 error increases with increasing values of T„, corresponding to
Tm=T2 + lioi r (18) higher D values. Another interesting qualitative result that can
3k be obtained from Eq. (24) is the general trend of the radial
Using dimensionless notation we finally have from Eqs. (16) temperature variation. This temperature variation may have a
(17), and (18) significant effect on the seal performance since it affects the
viscosity and pressure distributions and hence the opening force,
" „ / 2 cos2 <p wn \ 73 r 2
Tm=l+D e-W*-1). (19) fluid film stiffness and the leakage.
Nu / h 3 The radial temperature variation is given by the function

Journal of Tribology OCTOBER 1992, Vol. 114 / 641

Downloaded From: http://tribology.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Table 1 Various properties and corresponding dimensionless
parameters for water and oil seals
Property/parameter Water seal Oil seal
20 3 40
p.f Pa Sec 1(T 0.08
k w/m°C 2
0.65 0.14
H w/m °C 18000 3000
0 0.0175 0.038
0.35 1.52
D 0.014 2.6
K 0.0144 0.0031
Nu 2490 1930 0.90 0.92 0.94 0.96 0.98 1.00
Radial Location , r/r0
Fig. 3 The effect of face coning, y, on seal temperature
f(r) of Eq. (25). Assuming parallel faces, namely h 1, and
differentiating Eq. (25) gives
6cos <p
f'(r)=2? Gf-KG(3rlnr + 7)+- (26)
Nu
Since

Inr = - In I 1 +

and in typical face seals 7 = 1 we can use the approximation 1 ^

f In r = r - 1. Hence, for narrow seals, ? = 1, Eq. (26) can
be approximated by
6cos <p 1
f'(?)=2r K)G + Z (27)
Nu
The function / ' (?) in (27) is a measure of the radial tem-
perature gradient. Since the outer surface at r = r0 is exposed
to the cooling sealed fluid the temperature T„, increases mon-
otonically from r0 to r, (f'(r) < 0). This temperature increase
depends on the value o f / ' ( ? ) and therefore is affected by Nu,
K, and G, as it is evident from Eq. (27), and is independent
of the dissipation number D. Increasing the heat conductivity
k* of the rotor reduces the conductivity ratio K and, hence,
reduces the absolute value of /'(?) which means a more uni-
form radial temperature distribution. The same effect may be
obtained by reducing the convection coefficient H, which in
turn reduces Nu.
_ Although the above results were obtained_ for parallel faces
h = 1 the behavior is similar in the general h j± 1 case as well.
Two typical examples were selected in order to demonstrate
the effects of the various parameters on the temperature dis-
tribution. The seal has an outer radius r0 = 45 mm and a
radius ratio r,/r0 = 0.89. The rotation speed is o = 300
rad/s and the rotor is made of stainless steel having a thermal
conductivity k* = 45 W/m °C.
In the first example the sealed fluid is water at ambient
temperature Tj = 20 °C, while in the second example the fluid
is oil at Tf = 40 °C. The various properties and corresponding
dimensionless parameters are summarized in Table 1. A typical
value of <p = 45 was selected for the direction of the heat flow
path, and a typical gap of h-, = 1 ^m was selected as a reference
case. This yields a typical value G = 45000 for the geometry
parameter.
Of all the temperatures Tu T2, and Tm the stator face tem-
perature Ti is of the greatest interest. This is the highest tem-
perature among the three and also the only one that can be
easily measured. Hence, the stator face temperature in its nor-
malized form Tt/Tf was selected to present the effects of the
various parameters.
Figure 3 presents the effect of face coning on the radial
temperature distribution. The film thickness is described by
h = hi + T(r-ri)
where T is the amount of coning. In a dimensionless form the
film thickness is
h=\+y(?-?i)
0.92 0.94 0.96
Radial Location, r/ru
Fig. 4 The effect of dissipation number, D, and viscosity index, /3, on
seal temperature
where 7 is the normalized coning given by:
7
h,
In Fig. 3 the temperature distributions for the cases of 7 =
45 and 7 = 4.5 are compared with that of 7 = 0 (parallel
faces) for both water and oil. For the small coning (7 = 4.5)
there is a slight reduction in the face temperature. The deviation
increases towards the outer radius amounting to about 6 per-
cent in the oil case and about 9 percent in the water case. A
further increase of the coning to 7 = 45 causes a significant
reduction in the stator face temperature due to the larger sealing
gap which reduces the heat generated by shear stress.
The effects of the dissipation number, D, and the temper-
ature-viscosity coefficient, (3, on the face temperature for par-
allel faces are shown in Fig. 4. In the case of water, (3 = 0.35,
reducing the reference dissipation number by a factor of 5
reduces the temperature by about 23 percent at the outer radius
and by 40 percent at the inner radius. When the reference
dissipation number is increased by a factor of 5 the temperature
increases dramatically by as much as 74 percent at the outer
radius and 86 percent at the inner radius. A further increase
in the dissipation number would cause boiling of the water in
the sealing gap. The error caused by assuming constant vis-
cosity (|8 = 0) along the radial width of the sealing dam is
clearly demonstrated in Fig. 4. This error increases with an
increasing dissipation number and the over-estimation of face
temperature at the inner radius, for example, is growing from
3 percent at the lower D value up to 150 percent in the higher
D value. In the reference case the assumption of constant
viscosity results in over-estimating the inner radius temperature
by 28 percent. Obviously, errors of such magnitude are un-
acceptable and THD analyses based on a constant viscosity
assumption should be avoided. For the oil seal, (3 = 1.52, the
error caused by such assumption is much larger (up to two
order of magnitudes) because of the larger dissipation number.
Assuming constant viscosity results in a higher shear stress
compared to the actual one corresponding to a lower viscosity
due to the temperature rise along the sealing dam. Hence, the
error in face temperature will be larger in cases of higher

642 / Vol. 114, OCTOBER 1992 Transactions of the ASME

Downloaded From: http://tribology.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 1.0
0.90 0.92 0.94 0.96
Radial Location , r/r„
Fig. 5 The effect of conductivity ratio, K, on seal temperature
H" 3
0.90 0.92 0.94 0.96
Radial L o c a t i o n , r/r 0
Fig. 6 The effect of Nusselt number, Nu, on seal temperature
0.2E-4 0.4E-4 0.6E-4
Inversed Geometry Parameter , G _ 1
Fig. 7 The effect of film thickness, hh on stator temperature at inner
radius, rh for the case of parallel faces
viscosity and higher viscosity index (3, as was noted by ex-
amining Eq. (24).
The effect of the thermal conductivity of the rotor, k*, on
the stator face temperature distribution is shown in Fig. 5. As
can be seen a reduction in the conductivity ratio K resulting
from an increase in the thermal conductivity k* can affect
dramatically the temperature gradient (see discussion following
Eq. (27)). An increase of k* reduces the resistance to heat flow
through the rotor material and, hence, enables to carry away
more heat from the interface. This, in turn, reduces the tem-
perature rise along the sealing dam and may prevent undesir-
able phase change. As can be seen from Fig. 5 and from Eq.
(25) a change in K does not affect the temperature at the outer
radius f = 1 but has a greater effect on the temperature as T
decreases.
As was noted from Eq. (25) reducing the heat transfer H,
and hence, the Nusselt number Nu, also results in a more
uniform temperature distribution. This is shown in Fig. 6 for
both water and oil. In contrast to increasing k*, the more
uniform temperature in the case of reducing H is obtained by
an increase of the face temperature that is more pronounced
at the outer radius r = 1 than at the inner radius f = ?,-. Here
the more uniform temperature is a result of a poorer heat
transfer and has no advantage.
Figure 7 presents the effect of the sealing gap size hi on the
stator face temperature (7\),- at the inner radius /",•• The results
are shown for both the water and oil seal for the case of parallel
faces (7 = 0). The inverse of the geometry parameter, G~l,
Journal of Tribology
Downloaded From: http://tribology.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
0.98 1.00 0.90 0.92 0.94 0.96 0.98 1.00
Radial L o c a t i o n , r/r„
Fig. 8 The effect of heat flow path direction, p, on seal temperature
is used to represent the film thickness, and for our case a value
of G~l = 10~5 corresponds to a gap of h, = 0.45 /im. As can
be seen from Fig. 7 there is a nonlinear relation between h\
and {T\)j. The temperature is more sensitive to variations in
the film thickness at small gaps and this sensitivity is greater
in the water seal. At film thickness corresponding to G~l >
7 x 10~5 the effect of film thickness on the face temperature
becomes negligible.
Finally the effect of the heat flow path approximated di-
rection, <p, on the results is shown in Fig. 8. Results obtained
0.98 1.00
with (p values of 30 and 60 deg are compared with the reference
case where <p = 45 deg. As shown in Fig. 8 the differences are
less than 8 percent for the oil in both <p = 30 deg and ip =
60 deg cases. For the water example the differences are less
than 12 percent for <j> = 60 deg and 17 percent, at most, for
(p = 30 deg. These differences are at the outer radius. As the
temperature rises toward the inner radius the differences reduce
to less than 6 percent for the oil and about 10 percent for the
water case. It seems that approximating the heat path direction
by assuming a constant direction <p is the weakest point of the
present analysis. More work is required in order to get a better
feeling for a clearer selection of the <p value. This can be
accomplished by using finite element numerical solutions and
0.8E-4 1.0E-4
comparing their results with these obtained from the present
analytical solution. Such effort is beyond the scope of the
present work and is hoped to be performed later. Nevertheless,
inspite of the uncertainty introduced by the <p approximation
the THD analysis presented here sheds some light on the com-
plex problem of temperature distribution in mechanical face
seals and enables a relatively simple evaluation of the effects
of various parameters on the thermal behavior.
Conclusion
A thermo-hydrodynamic analysis was performed for a sym-
metric face seal configuration. Temperature variations both
across and along the sealing gap were considered as well as
realistic boundary conditions. An approximated shape of the
heat flow path was used to facilitate an analytical solution of
the energy equation. This, in turn, enabled us to present the
temperature variation in the seal interface by an implicit an-
alytical expression. This approach brings into light the more
important parameters affecting the temperature distribution
as well as the extent of their effects. The results clearly show
that assuming constant viscosity along the radial width of the
sealing dam (as is commonly done in various THD analyses)
results in unacceptable over-estimation of the temperature and
should be avoided. The present analytical solution may give a
quick check of the temperature level and radial gradient in
cases where phase change is suspected and it can also be used
as a good starting point for more complex finite element anal-
yses.
Acknowledgment
The research reported here was partially supported by the
OCTOBER 1992, Vol. 114 / 643
 Israel Ministry of Industry and Trade Joint Technion Fund.
The assistance of Mr. N. Galperin in processing the data and
presenting the results is gratefully acknowledged.

References
Becker, K. M., 1963, "Measurement of Convective Heat Transfer from a
Horizontal Cylinder Rotating in a Tank of Water, "Int. J. Heat Mass Transfer,
Vol. 6, pp. 1053-1062.
Doust, T. G., and Parmar, A., 1987, "Transient Thermoelastic Effects in a
Mechanical Face Seal," Proc. 11th International Conference on Fluid Sealing,
BHRA, pp. 407-422.
Dumbrava, M. A., and Morariu, Z., 1987, "Thermohydrodynamic Aspects
of the Double Mechanical Seals," Proc. 11th International Conference on Fluid
Sealing, BHRA, pp. 394-406.
Etsion, I., 1990, "Ideas and Tendencies in Future Mechanical Seal Devel-
opment," Lubrication Engineering, Vol. 46, No. 2, pp. 122-125.
Etsion, I., Palmor, Z., and Harari, N., 1990, "Feasibility Study of a Con-
trolled Mechanical Seal," STLE preprint No. 90-TC-2D-1.
Hughes, W. F., and Chao, M. H., 1980, "Phase Change in Liquid Face Seals
II—Isothermal and Adiabatic Bounds with Real Fluids," ASME Journal of
Lubrication Technology, Vol. 102, No. 3, pp. 350-359.
Holman, J. P., 1976, Heat Transfer, McGraw-Hill. Fig. A1 Isotherms and heat flow lines in a symmetric face to face
Lebeck, A. O., 1980, "A Mixed Friction Hydrostatic Face Seal Model With double seal
Phase Change," ASME JOURNAL OF LUBRICATION TECHNOLOGY, Vol. 102, pp.
133-138.
Li, C. H., 1976, "Thermal Deformation in a Mechanical Face Seal," ASLE
Trans., Vol. 19, No. 2, pp. 146-152.
Morariu, Z., and Pascovici, M. D., 1987, "The Thermal Study of the Double
Mechanical Seals," Proc. 5th Conference on Friction Lubrication and Wear
TRIBOTEHNICA 87, Vol. 2, pp. 319-327 (In Romanian).
Pascovici, M. D., 1974, "Temperature Distribution in the Lubricant Film of
Sliding Bearings Under Intensive Lubricant-Wall Heat Transfer Conditions,"
Wear, Vol. 29, pp. 277-286.
Pinkus, O., 1984, "Anisothermal Flow Films," Israel Jour, of Technology,
Vol. 22, pp. 120-138.
Pinkus, O., and Sternlicht, B., 1961, Theory of Hydrodynamic Lubrication,
McGraw Hill.
Salant, R. F., and Hassan, S. F., 1989, "Large Scale Thermoelastic Instability
in Hydrostatic Mechanical Seals," 12th Int. Conference on Fluid Sealing, BHRA,
pp. 75-88.
Waidner, P., 1989, "Evaporation in the Gap of Face Seals—Theoretical Cal-
culations and Results for Hot Water Application," 12th Int. Conference on
Fluid Sealing, BHRA, pp. 127-146.
Zeus, D., 1990, "Viscous Friction in Small Gaps-Calculations for Non-Con-
tacting Liquid or Gas Lubricated End Face Seals," Tribology Transactions, Vol.
33, No. 3, pp. 454-462.

A P P E N D I X

The Thermal Boundary Conditions

The geometry of the symmetric face to face double seal, as
well as the isotherms and heat flow lines in the rotor are shown
schematically in Fig. A l . The approximation suggested by Fig. A2 Approximation of thermal path shape
Morariu and Pascovici (1987), and by Dumbrava and Morariu
(1987) regarding the thermal path is adopted in the following.
According to this approach the thermal path lines that are Integrating Eq. (Al) and using the boundary conditions:
orthogonal to the isotherms are approximated by straight lines
at a fixed angle <p to the rotor face as shown in Fig. A2. Because
r=r 2 at £ = o
of axial-symmetry all the lines originating at a certain radial T=T, at £ = /
location on the rotor, form a conical envelop. If £ denotes a where /, the length of the thermal path is
coordinate measured from the rotor face along the conical
envelop then the heat flow rate, q, through the cone surface ro-r
!-- COS <p
is by Fourier's law:
dT we obtain:
q%= -k*2ir(r+l; cos <p)8 (Al)
di
T,-T,= -- Qi (A2)
where the angle <p is given by 2wk*8 £ cos <p
substituting
tan <p = -
r0-n <5 = dr sin <p
and performing the integration gives
and a is half of the active thickness of the rotor ring (the actual
thickness a may be larger than a meaning that the center portion *k* (T2-T3)sin 2<p
<7f = - dr (A3)
of the rotor contributes little to the heat removal).

644 / Vol. 114, OCTOBER 1992 Transactions of the ASME

Downloaded From: http://tribology.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 From the heat balance between the inlet heat flow at the rotor From Eqs. (7) and (11)
face 2 and the outlet heat flow at the rotor outer surface 3 we
dT /iCO
have (A7)
dy kh
y= 0
, dT
2-irrdr = q^ (A4) Hence, using Eqs. (A3), (A6), and (A7) in Eqs. (A4) and (A5)
y=0 we have
Also from the heat balance between g^ a t t h e rotor surface 3 2^? k* (T2- X3)sin 2<p
and the heat flow convected by the surrounding fluid we have ^ = (A8)
ln(/b/r)
g(=H(T)-T/)2Trr0dz (A5) and

where (see Fig. A2) k*(T2-Ti)cos,2<p

= H(T3-Tf)ro (A9)
cfe =tfr«tan<p (A6) ln(r 0 /r)

DISCUSSION

R. A. Burton1

This paper is a clearly thought-out and clearly presented

contribution to the very difficult problem of the thermotri-
bology of seals. The discusser has been concerned with dif-
ferent aspects of the problem, with emphasis on thermoelastic
interactions. Although the assumption of axial symmetry ex-
cludes many of these interactions, nevertheless some of the
notions of our analyses may shed some light on the findings
presented here. 5,000 IO.COO 20,000 30,000
Figure 9 shows data for cooling of the cylindrical surface
of a rotating disk. This was published in ASME Transactions Reynold's Number, Re

[1], and showed that the earlier work on long cylinders applied
as well to short cylinders [2, 3]. Subsequent measurements by
A. Kistler, of Northwestern University indicate that this effect
is not associated with gross movements of the fluid, but is
quite local. For example, the film is re-established in a short
distance behind a dam. Figure 1 is therefore felt to represent
a good estimate of heat transfer from a turning cylinder.
Nusselt number, Nu, can be shown to correspond to D/h,
where D is the diameter, and 6" is the thickness of stagnant
coolant that would have the same thermal resistance as the
convective cooling effects. From Fig. 1 it is seen that S/D
varies from about 0.01 to 0.025 over the range of Reynolds
number. The ratio of conductivities, r, for stainless steel and
oil is r~ 133, and for stainless steel and water is r~25. Con-
sequently, the thermal resistance can be expressed as a thickness
of metal, 5m, for operation in the two fluids, where:
8m/D = (8/D)r
For the above numbers we find:
In oil, 3.32>5,„/Z»1.33
In water, 0.63>8 m /D>0.25
To get a bound on expected radial temperature drop, take
the case where all of the heat passes radially outward from /•,-
to/o. For /Q —// = 0.1Z), the temperature drop may be expressed
as a fraction of the drop from rQ to the cooling fluid bulk
temperature, Coolant, since:
(t, - t0)/(t0 - /cooiant) = (/•,, - n)/5m = 0AD/5m
Using the above results to evaluate bm/D, one finds:
(f| "~ 'o)/('o — ^coolant)
In oil 0.03 to 0.075
In water 0.16 to 0.4
'Burton Technologies Inc., Raleigh, NC 27636-8809.
Fig. 9
This suggests that in the oil, the radial temperature drop is
not a tremendously important factor, whereas it is much more
important in water, which is in accord with the results reported
by the authors. The drop of viscosity in the hotter region would
bias the heat generation toward the outer edge of the ring,
shortening the thermal path through the metal. Coning acts
to counter this effect and move the radial distribution of heat-
ing inward.
We have recently studied similar interactions in a related
problem, the coupling of heating and waviness, which may aid
in conceptualizing this coupling more fully [4], in a somewhat
different context.
Additional References
1 Heckmann, S. R., and Burton, R. A., " A Theoretical Study of the Effects
of Cooling on Thermoelastic Contact Instabilities," ASME JOURNAL OF LU-
BRICATION TECHNOLOGY, Vol. 99, 1977, pp. 247-253.
2 Wottring, J. M. S., M. S. Project Report, Northwestern University, Ev-
anston, IL, 1975.
3 Etemad, G. A., "Free Convection Heat Transfer From a Rotating Cylinder
to Ambient Air With an Interferometric Study of Flow," Trans. ASME, 1955,
pp. 1283-1289.
4 Burton, R. A., "The Coupling of Waviness and Heating in a Seal," STLE
Trans., in press.
Authors' Closure
The authors would like to thank Dr. Burton for his com-
ments and interest in the paper. Figure 9 of the discussion is
useful for evaluating the Nusselt number. In the paper we used
very similar results that were published in Becker (1963).
Dr. Burton's approach for estimating the radial temperature
drop is indeed an interesting one, and we are pleased to see
that his results are in accord with ours. It should be noted,
however, that although the dimensionless temperature drop,
as presented here, is lower in the oil as compared to the water,
the actual temperature difference T/-T0 may be the same.
This is because T0- ^coolant is higher in the oil.

Journal of Tribology OCTOBER 1992, Vol. 114 / 645

Downloaded From: http://tribology.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use