International Journal of Heat and Mass Transfer 99 (2016) 149–158

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

journal homepage: www.elsevier.com/locate/ijhmt

Extended lubrication theory for generalized Couette flow through

converging gaps
Emil Barić, Helfried Steiner ⇑
Institute of Fluid Mechanics and Heat Transfer, Graz University of Technology, Inffeldgasse 25/F, A-8010 Graz, Austria

a r t i c l e i n f o a b s t r a c t

Article history: The lubrication theory based approximation of the Navier–Stokes equations represents a widely used
Received 31 July 2015 approach for the computation of viscous force dominated flows through narrow channels or gaps. Its
Received in revised form 16 March 2016 inherent neglect of advective transport becomes increasingly questionable when considering strongly
Accepted 20 March 2016
converging geometries and particularly when considering heat transfer as well. The present work devel-
Available online 12 April 2016
ops an extended lubrication theory based model, which accounts for inertia and convective heat transfer
in terms of first-order perturbations. The extended model is applied to generalized Couette flow through
Keywords:
strongly converging annular gaps typically met in wire coating dies, where the moving wall-driven
Generalized Couette flow
Lubrication theory approximation
motion is characterized by very high local shear rates and steep axial pressure gradients. The evaluation
First-order perturbation of the analytical solutions against numerical results from CFD simulations demonstrates the scope and
Coating dies the limits of the original lubrication theory approximation in describing the local flow and thermal con-
ditions inside the gap. Including first-order convective contributions the extended model is shown to pro-
vide a significantly improved description, especially of the redistribution of the generated viscous heat.
The most pronounced improvements are shown by the temperature profiles predicted at axial positions
with rapidly changing cross-section for fluids with higher Prandtl numbers. The observed gain in accu-
racy is also reflected by very accurate predictions of the total transfer rates at the boundaries in the global
balances of heat and momentum. Using the presently proposed extension the lubrication theory based
model represents a reliable and computationally efficient approach for investigating the flow and heat
transfer inside narrow gaps, which is well applicable to a wide range of different gap geometries.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction
Viscous laminar flow through converging gaps has been widely
studied due to its high technical relevance. Typical applications are
found in the processing of polymer melts, liquid foods and drugs,
or in coating technology, like the coating of magnet wires with a
thin layer of insulation. The present work particularly investigates
generalized Couette flow inside cylindrical gaps, which is essen-
tially characterized by a shear-driven motion of the fluid induced
by a moving wall, and high axial pressure gradients due to the axial
variation of the gap height. As such, the presently investigated type
of flow is well representative for a typical wire coating process in
enameling dies, as schematically sketched in Fig. 1. The primary-
coated wire being drawn through the die acts here as a moving
inner wall driving the flow. Due to the converging geometry the
abundant fresh enamel of the primary coating is removed, and
⇑ Corresponding author. Tel.: +43 316 873 7344; fax: +43 316 873 7356.
E-mail addresses: baric@fluidmech.tu-graz.ac.at (E. Barić), steiner@fluidmech.
tu-graz.ac.at (H. Steiner).
the wire finally exits the die coated with a layer of defined thick-
ness. The computational studies performed on such a flow config-
uration thus far mostly assumed the approximation of the
lubrication theory, putting often much emphasis on the effect of
non-Newtonian flow behavior. Assuming, e.g., a power-law fluid
Flumerfelt et al. [1] presented a lubrication theory based analytical
solution for planar gaps, while Lin and Hsu [2] analyzed axisym-
metric flow through concentric annuli. Later, Lin [3] included heat
transfer into the computations. A comprehensive review of the
basic concepts for the computational analysis of Couette-type flow
and heat transfer in converging gaps was presented by Mitsoulis
[4]. Carley et al. [5] computed the flow and temperature field inside
the conical section of wire coating dies solving numerically a sim-
plified set of the two-dimensional conservation equations with a
semi-implicit finite-difference scheme. Dijksman and Savenije [6]
presented a lubrication theory based analytical solution for a con-
verging annular gap flow introducing a special toroidal coordinate
system. Shah et al. [7] carried out an extensive analytical computa-
tion of both flow and heat transfer of power-law fluids inside cylin-
drical gaps with constant cross-section.

http://dx.doi.org/10.1016/j.ijheatmasstransfer.2016.03.078
0017-9310/Ó 2016 Elsevier Ltd. All rights reserved.
150 E. Barić, H. Steiner / International Journal of Heat and Mass Transfer 99 (2016) 149–158

Nomenclature

cp specific heat capacity [J/kg K] eT expansion parameter for temperature [–]

FD drag force [N] g dynamic viscosity [Pa s]
H enthalpy flow rate [W] k thermal conductivity of the fluid [W/m K]
h die exit gap height [m] q density of the fluid [kg/m3]
L axial length of the die [m]
ngeo geometry parameter [–] Dimensionless numbers
Pg production due to the viscous heating [W] Ec Eckert number [–]
p pressure [Pa] Nu Nusselt number [–]
Qw heat flow into the wire [W] Pe Péclet number [–]
r radial coordinate [m] Pr Prandtl number [–]
rw radius of the wire [m] ReL Reynolds number [–]
rd contour of the die [m]
T thermodynamic temperature [K] Superscripts
Uw velocity of the wire [m/s] ⁄
dimensionless quantity
u axial velocity component [m/s]
V_ volumetric flow rate [m3/s]
Subscripts
v radial velocity component [m/s]
z axial coordinate [m] 0 zeroth-order solution
1 first-order solution
d on the inner surface of the die
Greek symbols w on the surface of the wire
d difference in gap height between die inlet and outlet
[m]
e expansion parameter for velocity [–]

The neglect of the inertia as represented by the advective trans-
port terms becomes increasingly questionable in the case of
strongly converging cross-sections. This simplification markedly
limits the scope of the lubrication theory approximation, when
applied to real-file die geometries. Appropriate extensions to the
lubrication theory have thus far been proposed only for other
narrow gap flow configurations, where the liquid is differently
set in motion and/or the gap geometries are different. Extended
analytical solutions were obtained by Collins et al. [8] for flow in
planar journal bearings, and by Tavakol et al. [9] for pressure-
driven flow through narrow planar channels with axially varying
channel height. The latter non-Couette type flow is of special rele-
vance in biomedical applications considering, e.g., human blood
flow through arteries with varying cross-section due to stenosis
as investigated by Nadeem and Ijaz [10], or peristaltic liquid
motion driven by a periodic wave-like contraction of the channel
walls [11]. These very recent computational studies on biomedical
flow commonly assume only mild contractions in cross-section, so
that they can neglect inertia and convective heat transfer to obtain
still fairly accurate exact solutions of the problem. Hall et al. [12]
computed analytical solutions for pressure-driven converging flow
between stationary coaxial cones, where they considered very low
Reynolds number (Stokes flow) conditions as well as high Reynolds
number flow with non-negligible inertia. A special Couette-type
flow evolving inside a narrow gap between a solid disc and a plate,
assuming one wall rotating and the other as stationary, was
analyzed by Shevchuk [13], who computed a self-similar solution
of the full set the two-dimensional conservation equations of mass,
momentum and heat. This concept represents a very powerful
analytical approach, in that it basically imposes no restrictions
on Reynolds or Prandtl numbers as typically met in the lubrication
theory approximation. The actually considered flow configuration
and boundary conditions must permit a self-similarity transform
of all dependent variables though.
The present work attempts to improve the predictions of the
lubrication theory approximation for generalized Couette flow by
including the effect of the advective transport in terms of
corresponding first-order perturbations. The perturbed solutions
shall particularly provide more accuracy when considering
strongly converging geometries associated with a rapid axial
decrease in gap height. Accounting for the convective transport
of heat shall also improve the predictions for the temperature field,
especially for Prandtl numbers higher than unity. The possible
improvement of the predictions of the extended lubrication theory

Fig. 1. Wire coating die principle. Fig. 2. Converging gap domain.

 E. Barić, H. Steiner / International Journal of Heat and Mass Transfer 99 (2016) 149–158 151

based model shall be assessed by a validation against CFD results, z r

z ¼ ; r ¼
which are obtained from a numerical simulation of the full set of L h
equations.
u vL ph
2
u ¼ ; v ¼ ; p ¼ ;
2. Computational model Uw h Uw gU w L

T V_
The generalized Couette flow through the converging annular T ¼ ; V_  ¼ 2
:
gap is described in cylindrical coordinates, where the z-direction Tw 2p h U w
is aligned with the axis of inner cylindrical wall, moving at a con-
Based on this scaling the non-dimensional representations of
stant speed Uw, as shown in Fig. 2. The radial (cross-stream) direc-
the governing conservation Eqs. (1)–(4) read
tion r varies between the radius of the inner moving wall, r = rw,



2 2 
and the radius of the outer quiescent wall, r = rd(z). L denotes the @u @u @p 1 @  @u

h @ u
axial length of the gap. The flow is assumed as axisymmetric and e u þ v  ¼ þ r þ ; ð10Þ
@z @r @z r  @r @r  L @z2
steady. The fluid is assumed as Newtonian with constant dynamic
viscosity g, thermal conductivity k, specific heat capacity cp, and
2

2

h @v  @v  @p h @ @v  v 
constant density q. e u  þ v   ¼   þ þ 
L @z @r @r L @r @r
  r
The governing equations for axial momentum, radial momen-
4 2 
tum, and energy read h @ v
þ ; ð11Þ

 "
 # L @z2
@u @u @p 1 @ @u @2u
q u þv ¼ þg r þ 2 ; ð1Þ


@z @r @z r @r @r @z @T   @T 1 @  @T
ePr u þ v ¼ r
"
# @z @r  r  @r  @r 

 
2 "
 2
 2
 2 #
@v @v @p @ 1 @ðr v Þ @2v @u @v v
q u þv ¼ þg þ 2 ; ð2Þ þ 2PrEc
h
þ þ 
@z @r @r @r r @r @z L @z @r r
"
 " #

2 2



2  2 # @u @ v  h
@T @T 1 @ @T @u
2
@u v þ PrEc þ :
qcp u þ v ¼k r þ 2g þ þ @r @z L
@z @r r @r @r @r @r r

2 ð12Þ
@u @ v
þg þ ð3Þ
@r @z Z r d
V_  ¼

r u dr ¼ const:; ð13Þ
The shown set of equations is closed by the conservation of r w
mass enforcing a constant volumetric flow rate
Z where the parameters
rd
V_ ¼ 2prudr ¼ const:; ð4Þ
2

h qU w L h 2 cp g U 2w
rw e ¼ ReL ¼ ; Pr ¼ ; Ec ¼ ð14Þ
L g L k cp T w
at any axial position z.
No-slip boundary conditions are prescribed at the upper and the denote the reduced Reynolds number, the Prandtl number, and the
lower radial boundaries: Eckert number, respectively. The parameters e and Pr can be further
combined to a reduced Pèclet number
r ¼ rw : u ¼ Uw ; v ¼ 0; ð5Þ

2
h
r ¼ rd : u ¼ 0; v ¼ 0: ð6Þ
eT ¼ ePr ¼ PrReL : ð15Þ
L
A constant wall temperature is prescribed at the lower radial It can be immediately seen from the lhs of the non-
boundary, while the upper radial boundary is assumed as dimensionalized conservation Eqs. (10) and (12) that the advective
adiabatic: transports of axial momentum and thermal energy are of the order
O (e) and O (eT), respectively. In narrow gap flow the axial length of
r ¼ rw : T ¼ Tw ð7Þ the gap typically exceeds the radial height of the gap significantly,
such that (h/L)2  1. It is noted that this disparity holds true for
@T any local gap height throughout the considered domain, and not
r ¼ rd : ¼0 ð8Þ
@r only near the outlet, where the gap height reaches its exit value
For the pressure, a constant ambient pressure is imposed at the h, which is used here as vertical length scale. Thus, considering
die inlet and outlet, respectively: Reynolds numbers ReL markedly higher than unity (typically O
(ReL)  103) and Eckert numbers of order unity of less, it follows
z ¼ 0; z ¼ L : p ¼ p1 ð9Þ from the definitions of the e and eT in (14) and (15), respectively,

2
2
h h
2.1. Zeroth- and first-order solutions  e; Ec Pr  eT : ð16Þ
L L
The lubrication theory approximation and its first-order exten- The solutions for the axial velocity and the temperature are
sion essentially follow from an order-of-magnitude analysis of the assumed as series expansions of the reduced Reynolds and Pèclet
non-dimensionalized balances of momentum and heat. Using the numbers
length L and the exit height h of the gap, the velocity Uw and tem-
u ¼ u0 þ e u1 þ e2 u2 þ    ; ð17Þ
perature Tw of the inner wall as reference scales as axial and radial
length scales the independent and dependent variables can be
T  ¼ T 0 þ eT T 1 þ e2T T 2 . . . ð18Þ
transformed into
152 E. Barić, H. Steiner / International Journal of Heat and Mass Transfer 99 (2016) 149–158

and introduced into the non-dimensionalized conservation Eqs. respectively, with

(10)–(13). Equating the terms of the same orders of e and eT yields Z r Z r Z r
1 @u0 
a constitutive set of equations for the zeroth and first-order solu- Aðz ; r Þ ¼ 2u0 r dr þ v 0 u0 dr : ð32Þ
tions, respectively. The equations for the zeroth-order, i.e., O (e0) r w r r w @z r w

and O (eT0), read The radial component of the velocity occurring in Eq. (32) is


 computed from the zeroth-order continuity equation as
1 @  @u0 @p
r ¼ 0 ; ð19Þ Z r
r @r
  @r  @z 1 @u0  
v 0 ¼  r dr : ð33Þ
r r w @z
@p0
¼ 0: ð20Þ Substituting the obtained zeroth- and first-order expressions
@r
for the axial velocity into the integral mass balances given by
Z r d Eqs. (21) and (25) yields, upon integration from z⁄ = 0 to z⁄ = 1,
V_ 0 ¼

r  u0 dr ¼ const: ð21Þ the axial variation of the pressure and axial pressure gradients,
r w
which are needed to close the Eqs. (30) and (31).


 2 The radial integration of Eqs. (22) and (26) provides the zeroth-
1 @  @T 0 @u0 and first-order solutions for the temperature written as
 r ¼ PrEc : ð22Þ
r @r  @r  @r  Z r
Cðz ; r Þ  r
Equating the terms of the order O (e) and O (eT) and observing T 0 ¼ 1  
dr þ Cðz ; r d Þ ln  ; ð34Þ
r w r rw
the relations given in (16) provides the following equations for
the first-order solution Z r Z r
Dðz ; r  Þ  Eðz ; r Þ 


 T 1 ¼ 2Ec dr þ dr
@u @u @p 1 @ @u r r
u0 0 þ v 0 0 ¼  1 þ   r  1 ; ð23Þ r w r w
@z @r @z r @r @r r
þ ln  ½2Ec Dðz ; r d Þ  Eðz ; rd Þ; ð35Þ
rw
@p1
¼ 0; ð24Þ involving the integrals
@r 
Z r
2
Z @u0 
r d Cðz ; r Þ ¼ r  PrEc dr ; ð36Þ
V_ 1

¼ r  u1 dr ¼ const: ð25Þ r w @r 
r w
Z r
 


   @u0 @u1 
1 @  @T 1 @u0 @u1  @T 0  @T 0 Dðz ; r Þ ¼ r dr ; ð37Þ
r ¼ Ec 2 þ u þ v : ð26Þ rw @r  @r 
r  @r  @r @r  @r 0
@z 0
@r 
Z r Z r
The boundary conditions for the zeroth and first-order read @T 0  @T 0 
Eðz ; r Þ ¼ r u0 dr þ r  v 0 dr : ð38Þ
rw @z r w @r
r  ¼ r w : u0 ¼ 1; T 0 ¼ 1
ð27Þ
u1 ¼ 0; T 1 ¼ 0;
3. Considered gap flow configuration
@T 0
r  ¼ r d : u0 ¼ 0; ¼0 The geometry and the operating parameter setting of the pre-
@r ð28Þ
 sently considered narrow-gap flow configurations were specified
@T 1
u1 ¼ 0; ¼ 0; very close to real-life operating conditions in wire coating dies,
@r 
where the Couette flow typically evolves inside strongly converg-
z ¼ 0; z ¼ 1 : p0 ¼ 0 ing annular gaps with a very small exit height and a very fast
ð29Þ moving thin wire as radial inner boundary. This particularly leads
p1 ¼ 0;
to extremely high local shear rates, high axial peak pressures, and
respectively. locally very intense generation of viscous heat. The shape of the
The derived sets of equations make obvious that the zeroth- converging gap contour is assumed to follow a cosine-type
order represents the analytical solution in the limit of the lubrica- function
tion theory approximation, where the advective transport terms 
 n 
rd rw d z geo
are totally neglected. The first-order perturbation evidently rd ¼ ¼ þ1þ 1 þ cos p : ð39Þ
extends the lubrication theory approximation to account for the
h h 2h L
contributions of these transport terms. As exemplarily shown in Fig. 3, this assumption offers much
Radial integration of Eqs. (19) and (23) yields the zeroth-order flexibility in covering various types of shapes dependent of the set-
and first-order solutions for the axial velocity component, written ting of the parameter ngeo. The shape of possible contours ranges
as from completely convex for low values of the parameter ngeo, to

 r
 almost step-like concave geometries for the high values. In the
r 1 dp0 r 
ln rd 4 dz
r2 ln rd þ r 2 w 2
d ln r  þ r w ln r
r
present work the parameter ngeo was varied as 0.5, 1, and 10, asso-
u0 ¼ þ ; ð30Þ
w d
r r
ln rd ln rd ciated with a convex, a smooth S-shaped, and a steep S-shaped
w w
profile, respectively. The other parameters occurring in Eq. (39),

2 
3 i.e., the length of the gap, the exit gap height, the difference in
ln rr  ln rr
1 dp1 4 2 2 5 gap height between the inlet and exit, and the diameter of the
u1 
¼ Aðz ; r Þ  Aðz  
; rd Þ w
 2 2
ðr d  rw Þ r  ðr  r w Þ ;
w

ln
r
d

4 dz ln d inner moving wall were set to the fixed values L = 25 mm,
rw rw
h = 5 lm, d = 2.5 mm, and dw = 2rw = 0.15 mm, respectively. The
ð31Þ Reynolds number was always assumed as ReL = 2230. The Prandtl
 E. Barić, H. Steiner / International Journal of Heat and Mass Transfer 99 (2016) 149–158 153

comparable with the inlet conditions of the computations with

the analytical model. Uniform ambient pressure p = p1 and
reference temperature T = Tw were prescribed at the inlet and the
outlet of the computational domain, respectively. Even though
the temperature is not constant at the exit, imposing a constant
value boundary condition does not have any effect on the temper-
ature inside the domain, due to the fact that the flow at the die exit
is only directed outwards. The total size of the computational mesh
was roughly 40,000 cells in the case used for the present
comparison.

5. Computational results

5.1. Flow field

The contour plots depicted in Fig. 5 showing the isotachs of the

axial velocity obtained from the extended lubrication theory solu-
Fig. 3. Outer radial contours for different values of the geometrical parameter ngeo.
tion (zeroth- + first-order solutions) gives a first qualitative insight
into the structure of the flow field. For all considered geometries
the velocity contours exhibit the typical features of converging
and Eckert numbers were consistently varied as Pr = 1 and 7,
Couette-type flow, which basically falls into two sub-regions. In
Ec = 0.7, and 0.1, respectively, in order to realize a different relative
the radial inner sub-region the motion is driven by strong shear
weight of the advective terms in the balances of heat and momen-
forces into the direction of the moving wall. The radial outer
tum. Using the present parameter setting in the definitions (14)
sub-region is associated with a negative axial velocity, as the axial
and (15) the expansion parameters for the first-order contributions
decrease in cross-sectional area causes a large part of the liquid
in the series expansions (17) and (18) read e = 8.91  105,
entrained by the moving wall to flow backwards for continuity rea-
eT = 8.91  105 and 6.24  104, respectively.
son. The contour associated with zero axial velocity, u⁄ = 0, separat-
ing the two sub-regions, basically resembles the shape of the radial
outer contour of the gap for all cases. Since continuity is enforced
4. Numerical model for validation
at all orders of accuracy, both the zeroth-order approximation
and the extended first-order solution predict a very similar struc-
Since narrow gap flow is very hard to access experimentally,
ture of the backflow region in the radial outer zone. However, at
there are practically no experimental data available on the details
axial positions, where the cross-sectional area most rapidly
of the velocity and temperature fields inside the gap. Therefore,
changes as indicated by high axial gradients of the outer contour
CFD simulations were carried out to evaluate the solutions
dr⁄d/dz⁄, notable differences occur in the predicted velocity profiles.
obtained under the assumptions of the lubrication theory
This is illustrated in Fig. 6 showing radial profiles of the axial veloc-
(zeroth-order solution), and to assess the improvement brought
ity at four selected cross-sections for all considered geometries.
about by the inclusion of the advective terms in the first-order
The inclusion of the advective transport in with a slower decrease
extension. The CFD simulations used ANSYS-Fluent 14.2 as the flow
in gap height the deviation of the zeroth-order approximation from
solver. Fig. 4 exemplarily shows the computational domain used in
the other solutions is less pronounced or vanishes almost
the simulations for the gap geometry associated with the parame-
completely. The observed improvements for the velocity profiles
ter ngeo = 1, indicating the individual boundaries. The flow was
are also reflected in the axial variation of the pressure displayed
assumed as steady and axisymmetric. Analogously to the analytical
in Fig. 7. The first-order extension improves here again the predic-
model, the same dynamic boundary conditions (no-slip) and the
tions predominantly in the axial sections, where the radial outer
same thermal boundary conditions (constant wall temperature
contour of the gap converges fastest.
and adiabatic) were imposed at the radially inner and outer bound-
aries, respectively. As can be seen from Fig. 4, the computational
domain is somewhat extended upstream of z = 0 by a short cylin- 5.2. Heat transfer
drical section. The attachment of this short cylinder of 10% length
of the die allows for some axial development of the Couette-type Assuming constant density flow with constant material proper-
flow in order to provide flow conditions at z = 0, which are well ties the heat transfer does not couple back on the transport of

Fig. 4. Outline of the computational domain of geometry ngeo = 1 for CFD simulations.
154 E. Barić, H. Steiner / International Journal of Heat and Mass Transfer 99 (2016) 149–158

Fig. 7. Axial variation of non-dimensional pressure for considered geometries.

Fig. 5. Axial velocity isotachs from extended lubrication theory for the considered
geometries.

Fig. 6. Radial profiles of axial velocity at selected cross sections: zeroth-order Fig. 8. Contours of non-dimensional temperature T⁄ = T/Tw for geometry ngeo = 1
(dashed line), zeroth- + first-order (solid line), numerical CFD (dots). with Pr = 7; white contour line denotes isotherm T⁄ = 1.3.
 E. Barić, H. Steiner / International Journal of Heat and Mass Transfer 99 (2016) 149–158 155

momentum, so that the temperature fields obtained for varying

Prandtl numbers rely always on the same velocity fields computed
for a given geometry. The contours in Fig. 8 exemplarily show the
different solutions for the entire temperature field for the case
ngeo = 1 with Pr = 7. The most salient difference appears in the outer
radial region near the upper wall, where the zeroth-order solution
predicts significantly higher temperatures. This discrepancy is not
due to a markedly different production of viscous heat. As it is
already indicated by the velocity profiles in Fig. 6, the highly
sheared region near the moving wall shows no significant differ-
ences for all solutions. Hence, the resulting production of viscous
heat does not notably differ either, as it is exemplarily shown in
Fig. 9. Therein, the depicted contours of the non-dimensionalized
viscous dissipation obtained from the zeroth-order solution as

 2
@u0
/g;0 ¼ Ec Pr ; ð40Þ
@r 

Fig. 9. Contours of viscous dissipation term in the region near the exit for ngeo = 1 almost coincide with the CFD results. The contours for the first-
with Pr = 7. order extension would agree even closer and have been dropped
for better visibility. Dealing with almost the same local viscous

Fig. 10. Radial profiles of the temperature at selected cross sections for geometry Fig. 11. Radial profiles of the temperature at selected cross sections for geometry
ngeo = 0.5 with Pr = 1 (left column) and Pr = 7 (right column): zeroth-order (dashed ngeo = 1.0 with Pr = 1 (left column) and Pr = 7 (right column): zeroth-order (dashed
line), zeroth- + first-order (solid line), numerical CFD (dots). line), zeroth- + first-order (solid line), numerical CFD (dots).
156 E. Barić, H. Steiner / International Journal of Heat and Mass Transfer 99 (2016) 149–158
heating the observed discrepancy in temperature has therefore to
be rather attributed to the limitation to radial heat conduction as
the only available transfer mechanism in the zeroth-order solution.
This limitation evidently leads to markedly increased temperature
levels near the adiabatic outer wall for the zeroth-order solution.
On the other hand, including the convective transport significantly
reduces the size of the high-temperature region as indicated in
Fig. 8 by the highlighted contour line for T⁄ = 1.3. In the extended,
zeroth- + first-order, solution as well as the CFD results, the region
associated with temperatures T⁄ > 1.3 occupies only a limited zone
near the exit, while it extends across the entire radially upper part
of the flow domain in the zeroth-order solution.
The quantitative gain in accuracy provided by the first-order
extension is illustrated in Figs. 10–12 showing the predicted tem-
perature profiles for all considered geometries at the same selected
axial positions as used in the discussion above for the velocity. In
the case of unity Prandtl number the solutions in general agree
very well for all geometries. The first-order extension still some-
what improves the predictions to a similar extent as observed for
the predicted velocities. For the higher Prandtl number Pr = 7
significant differences appear near the outer radial wall. These
Fig. 12. Radial profiles of the temperature at selected cross sections for geometry
ngeo = 10 with Pr = 1 (left column) and Pr = 7 (right column): zeroth-order (dashed
line), zeroth- + first-order (solid line), numerical CFD (dots).
discrepancies, which have already been addressed in the
discussion of the temperature contours for ngeo = 1 shown in
Fig. 8, become evident for all geometries. The zeroth-order solution
in general strongly overpredicts the temperature in the layer
adjacent to the outer adiabatic wall of the gap. The apparently
stronger deviations for the cases with Pr = 7 can be attributed to
the fact that the advective transport in the non-dimensional heat
balance scales with the product ReLPr, such that its total neglect
becomes more influential for increasing Prandtl numbers. In turn,
accounting for this transfer mechanism in terms of first-order per-
turbations particularly improves the predictions for the higher
Prandtl number, as seen from the respective profiles of the
extended solution (0th + 1st-order) coming much closer to the
CFD results for all geometries. Despite the observed improvements,
the scope of the first-order perturbation of the lubrication theory is
still limited to Prandtl numbers which should not significantly
exceed unity. Using values markedly higher than unity, say,
Pr > 20, as input into the expansion parameter defined in Eq.
(15), would require extending the linear perturbation including
second-order terms of even higher. Due to this inherent
expansion-order dependent accuracy the present perturbation
approach may appear as less attractive than alternative analytical
concepts like solutions based on self-similarity. So the self-
similar solution obtained by Shevchuk [13] for the momentum
and heat transfer of a Couette-type flow inside a narrow gap
between a solid cone and a plate, with one wall rotating, is
basically equally accurate for any Prandtl number. However,
although the presently considered generalized Couette flow also
evolves inside a narrow gap between a moving lower and a station-
ary upper wall, the radial variation of the axial velocity, and, hence
the convective transports, cannot described by self-similar profile
functions. The sign reversal in the axial pressure gradient, as seen
in Fig. 7, leads to a change in curvature of the velocity profile from
Fig. 13. Radial variations of advective transport terms in the horizontal direction,
u0 ð@T 0 =@z Þ (solid line), and vertical direction, v 0 ð@T 0 =@r  Þ (dashed line), and of the
viscous heat production (dots) for ngeo = 1 with Pr = 7 at selected cross-sections.
 E. Barić, H. Steiner / International Journal of Heat and Mass Transfer 99 (2016) 149–158 157

Table 1
Global dynamic and thermal quantities for considered geometries, shown percentages are based on total production of viscous heat P⁄g.

ngeo = 0.5 ngeo = 1 ngeo = 10

0th 0th + 1st CFD 0th 0th + 1st CFD 0th 0th + 1st CFD
Q⁄ 0.0452 0.0451 0.0451 0.0456 0.0455 0.0454 0.0556 0.0552 0.0549
F D 2.2652 2.2733 2.2766 1.3913 1.3950 1.4057 0.3382 0.3441 0.3471

P g ð¼ 100%Þ 1.6451 1.6494 1.6502 1.0525 1.0546 1.0556 0.3674 0.3714 0.3725
 Q w 1.6451 1.5112 1.5111 1.0525 0.9243 0.9228 0.3674 0.2432 0.2424
(100%) (91.6%) (91.6%) (100%) (87.64%) (87.42%) (100%) (65.5%) (65.1%)
Hz ¼ 0 0.1355 0.1356 0.1280 0.1298 0.1243 0.1254
(8.2%) (8.2%) (12.14%) (12.3%) (33.5%) (33.7%)
Hz ¼ 1 0.0027 0.0035 0.0023 0.003 0.0039 0.0047
(0.2%) (0.2%) (0.22%) (0.28%) (1%) (1.2%)

concave to convex near the exit of the die, which conflicts with the
required self-similarity of the profiles.
A more detailed insight into the redistribution of the locally
generated viscous heat is given by Fig. 13, illustrating exemplarily
again the results of the case ngeo = 1 with Pr = 7. The radial varia-
tions of the first-order convective contributions are depicted here
together with the locally generated viscous heat at selected axial
positions. Both the radial and the axial convective term exhibit
positive peaks near the moving inner wall. This effectively reduces
the increase in temperature caused by viscous heating, whose
maximum is always located at the inner wall, from where it rapidly
decreases to almost zero in the rest of the domain as indicated by
the steep radial decrease of the dotted curves.
The observed improvements in the description of the local flow
and temperature fields also translate into more accurate predic-
tions of important global dynamic and thermal quantities. Table 1
lists the predicted volumetric flow rates computed from Eq. (13),
the total drag forces, the total production of viscous heat, the
enthalpy flow rates at the inlet and outlet, and total heat flow into
the wire, obtained for the different geometries and Pr = 7 as
Z  Z 1 Z r
 2
@u   
1 d @u  
F D ¼ 
r dz ; P 
g ¼ Ec Pr r  dr dz ;
0 @r r w 0 
rw @r

2 Z r Z 1 
h d
 @T   
Hz ¼0;1 ¼ Re Pr u ðT   1Þr  dr ; Q w ¼ 
L 
rw 0 @r r w
respectively. Accounting for the advective transport generally leads
to slightly increased total drag forces. A comparison of the predicted
values across the different geometries makes evident that changing
the die shape towards a steeply, almost step-like, converging con-
tour associated with increasingly high ngeo reduces most effectively
the total drag force. A most rapid change from an almost constant
gap height determined by the inlet cross-section to the exit height
very next to the outlet provides obviously a shortest possible region
with very high local shear stress, where a significant part of the
total drag force is generated. The somewhat higher total drag forces
predicted by the first-order extended model are consistently
reflected by a slightly increased total production of viscous heat.
More important differences appear in the heat transport across
the boundaries of the domain. While the convective heat loss at
the exit is always insignificantly small due to the small exit gap
height, a considerable amount of heat is convected backwards
towards the inlet. As a result, between roughly 8 up to more than
30% of the total production of viscous heat are lost through the
convective heat flow at the inlet via Hz ¼0 . This effectively reduces
the heat flow into the moving wall by the same amount, as the adi-
abatic outer radial boundary does not dissipate any heat. On the
other hand, neglecting convective heat transfer the zeroth-order
solution cannot account for any convective losses across the open
boundaries, so that the total production of viscous heat has to be
dissipated completely by the moving wall, which finally leads to
considerably overpredicted values for Q w . The first-order extension
eliminates this deficit, which becomes more significant for the
faster converging gap heights associated with higher values of ngeo,
to a large extent as indicated by the good agreement of with the CFD
results used for validation. A reliably accurate description of the
global heat balance of strongly converging geometries requires
evidently the inclusion of convective heat transfer as provided by
the present first-order perturbations.
6. Conclusions
The generalized Couette flow evolving inside strongly converg-
ing annular gaps is computationally investigated in the framework
of the lubrication theory, considering viscous generation and trans-
fer of heat as well. The high aspect ratio between the inlet and out-
let cross sections of the considered strongly converging geometries
poses a big challenge to this approximation, which inherently
neglect all advective transports. The present work attempts to
extend the scope of the lubrication theory based analysis account-
ing for inertia and convective heat transfer in terms of appropriate
linear first-order perturbations in the balances of momentum and
energy. The assessment of the results of the zeroth-order and the

r w dz ; extended analytical solutions using numerical results computed
from CFD simulations can be summarized as follows:
ð41Þ
(1) Although the total neglect of inertia becomes highly
questionable due to the rapidly decreasing gap heights, the
zeroth-order solution predicts the velocity field fairly well.
Its most salient feature, the formation of a big reverse flow
region near the radially outer quiescent wall is essentially
enforced by continuity and not by inertia. Including first-
order perturbations for the latter still somewhat improves
the predictions of the local velocities in those sections,
where the gap geometry converges fastest.
(2) The local generation of viscous heat mainly occurring close
to the moving inner wall is basically the same for all
solutions.
(3) In contrast to the neglect of inertia, the neglect of the
convective heat transfer strongly limits the scope of the
zeroth-order solution. Especially for the case with the higher
Prandtl number, it strongly overpredicts the temperatures
near the radially outer wall. The first-order extension turns
out to capture the convective redistribution of heat inside
the domain fairly well, which significantly improves the pre-
dictions of the temperature field.
(4) The gain in accuracy provided by the proposed first-order
extension is also reflected in the global heat balance. Depen-
dent of the geometry of the gap a considerable fraction of the
total viscous heat is convected outwards across the inlet
158 E. Barić, H. Steiner / International Journal of Heat and Mass Transfer 99 (2016) 149–158
boundary. Due to the total neglect of convective transfer the
zeroth-order solution completely misses this relevant cool-
ing effect, which finally leads to considerably over-
predicted total wall heat transfer rates into the moving wall.
The first-order extension is shown to predict the convective
loss at the inlet, and, hence, the total wall heat transfer very
well. The good agreement with numerical validation data
from CFD suggests the extended model as a very reliable
analytical approach for describing the balancing of global
heat production and losses in strongly converging gap flow.
Acknowledgements
The authors gratefully acknowledge the financial support from
the Austrian Research Promotion Agency FFG (project no. 833754)
and the industrial partner MAG Maschinen und Apparatebau AG.
References
[1] R.W. Flumerfelt, M.W. Pierick, S.L. Cooper, R.B. Bird, Generalized plane Couette
flow of a non-Newtonian fluid, Ind. Eng. Chem. Fundam. 8 (2) (1969) 354–357.
[2] S.H. Lin, C.C. Hsu, Generalized Couette flow of a non-Newtonian fluid in annuli,
Ind. Eng. Chem. Fundam. (1980) 421–424.
[3] S.H. Lin, Heat transfer to generalized Couette flow on non-Newtonian fluid in
annuli with moving outer cylinder, Int. J. Heat Mass Transfer 35 (1992) 3069–
3075.
[4] E. Mitsoulis, Fluid flow and heat transfer in wire coating: a review, Adv. Polym.
Technol. 6 (1986) 487–497.
[5] J.F. Carley, T. Endo, W.B. Krantz, Realistic analysis of flow in wire-coating dies,
Polym. Eng. Sci. 19 (1979) 1178–1187.
[6] J.F. Dijksman, E.P.W. Savenije, The flow of Newtonian and non-Newtonian
liquids through annular converging regions, RheologicaActa 24 (1985) 105–
118.
[7] R.A. Shah, S. Islam, A.M. Siddiqui, T. Haroon, Exact solutions of a power law
fluid model in post treatment analysis of wire coating with linearly varying
boundary temperature, Appl. Math. 4 (2013) 330–337.
[8] D. Collins, M.D. Savage, C.M. Taylor, The influence of fluid inertia on the
stability of a plain journal bearing incorporating a complete oil film, J. Fluid
Mech. 168 (1986) 415–430.
[9] B. Tavakol, D.P. Holmes, G. Froehlicher, H.A. Stone, Extended Lubrication
Theory: Estimation of Fluid Flow in Channels with Variable Geometry, 2014.
arXiv:1403.2343v1 [physics.flu-dyn].
[10] S. Nadeem, S. Ijaz, Theoretical analysis of metallic nanoparticles on blood flow
through stenosed artery with permeable walls, Phys. Lett. A 379 (2015) 542–
554.
[11] Noreen Sher. Akbar, A.W. Butt, Physiological transportation of casson fluid in a
plumb duct, Commun. Theor. Phys. 63 (2015) 347–352.
[12] O. Hall, A.D. Gilbert, C.P. Hills, Converging flow between coaxial cones, Fluid
Dyn. Res. 41 (2009) 25.
[13] I.V. Shevchuk, Laminar heat and mass transfer in rotating cone-and-plate
devices, ASME J. Heat Transfer 133 (2) (2011). 024502-1.