Heat transfer with mechanically driven thermal

contact resistance at the polymermold interface in

injection molding of polymers
H. Massee
a
,

EE. Arquis
a
, D. Delaunay
b,
*
, S. Quilliet
c
, P.H. Le Bot
d
a
Laboratoire MASTER-ENSCPB, 16 av. Pey Berland, 33607 Pessac, France
b
Laboratoire de Thermocineetique de leecole Polytechnique de luniversitee de Nantes, CP 3023, 44087 Nantes cedex 03, France
c
CERAP, 27 bd. Du 11 novembre, 69100 Villeurbanne, France
d
Inergy, 310 rue de Ransbeek, 1120 Brussels, Belgium
Received 3 December 2001; received in revised form 4 April 2002
Abstract
Plastics are materials which show important volume variations when they undergo temperature changes. During the
injection molding process, polymer parts may shrink and so become unstuck from the mold walls. It drives then to the
apparition of air gaps which may modify both the shape of the part and its cooling time. The cooling of such a polymer
part is studied, taking into account key parameters such as thermal contact resistances, residual stresses, or PvT dia-
gram. The results of the numerical simulation, using a coupled thermomechanical approach, are compared to exper-
iments that are led in parallel.
2003 Elsevier Ltd. All rights reserved.
1. Introduction
During the cooling of an injected polymer, a very
strong coupling exists between the thermal and
mechanical phenomena. In order to ll the mold and to
compensate the eects of reduction of the volume bound
on the thermal shrinkage and possibly on the solidi-
cation shrinkage for the semi-crystalline polymers, the
pressure in the molding cavity is maintained to a very
high level. This is possible if the injection gate that
separates the molding cavity of the runners full of
polymer is not solidied. As the gate is generally thin
to allow the comfortable separation of the piece with
the sprue (matter frozen in the feeding channels of the
piece), this solidication generally occurs early in the
cycle. The decrease of the temperature provokes a
shrinkage that leads to the decrease of the pressure in the
molding cavity. It provokes a strong modication of the
contact conditions between polymer and mold, and even
the apparition of an air gap. The pressure therefore
conditions the cooling. When the piece is no more in
contact with the mold, its shape, so the thickness of the
air gap, is determined by the eld of displacement and
therefore of stresses in the piece. One conceives then
the very strong triangular coupling existing between the
temperature, the heat transfer at the interface and the
pressure or the stresses in the solid after detachment.
The knowledge of the contact resistance is very impor-
tant for the simulation of the injection process since
the temperature of the polymer can be very dierent
from the one of the mold [1], notably at the beginning
of cooling when the heat uxes are very large. The
thermal barrier can be important enough, so that in
some cases, a local heating of the piece happens, that, if
it is not taken into account forbids the calculation of
the shrinkage and of the warpage (stress relaxation
for example). Several works concern thermal contact
resistances between a polymer and a metal. The mea-
surements by classical stationary methods give for
*
Corresponding author. Fax: +33-2-40-68-31-41.
E-mail address: didier.delaunay@polytech.univ-nantes.fr
(D. Delaunay).
0017-9310/$ - see front matter 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijheatmasstransfer.2002.04.001
International Journal of Heat and Mass Transfer 47 (2004) 20152027
www.elsevier.com/locate/ijhmt
thermoplastics [2,3] values in the range [3.5 10
5
to
8.5 10
5
m
2
KW
1
] and for elastomers [4] values of a
superior order of magnitude, around 3 10
4
m
2
KW
1
.
Yu et al. [5] measured, in the molding cavity, the
resistances during a complete cycle. They showed that
the resistance varies strongly during the cycle, with a
mean value of 5 10
4
m
2
KW
1
during the phase where
the pressure in the molding cavity is superior to the air
pressure, to reach large values close to 2 10
3
m
2
KW
1
after the detachment of the piece. When this one occurs,
it may be expected that the shrinkage in the thickness
allows to calculate the thickness d of the air gap and
therefore its thermal resistance R d=k
air
. It has been
veried and is shown in [6]. For a polypropylene, the
evolution of the sum of the resistances measured on the
two faces of a piece after detachment is compared with
the evolution of the thermal resistance equivalent to the
dierence between the calculated thickness of the piece
and the one of the molding cavity. One gets a remark-
able agreement. If the two faces of the piece have a
perfectly symmetrical behavior, the air gap shares then
by half on every face, and for the time where the piece is
not in contact with the mold, the resistance can be
modeled. It is rarely the case, because of the adherence
phenomena, of the corners eects and of the dierent
heat transfer on the faces of a piece. The air gap thick-
ness varies with time and space, as shown by Sridhar [7],
with even instabilities (beating of the piece) observed in
[6]. The thickness of the air gap is in fact conditioned by
the position of the piece in the molding cavity, conse-
quence of the instantaneous stress eld. Only the reso-
lution of the coupled thermal and mechanical problems
permits to foresee the evolution of the resistance
equivalent to the air gap.
Before the detachment, after the lling, the pressure
into the molding cavity may vary from 100 Mpa before
decreasing to reach the atmospheric pressure. One must
wait therefore for important variations of the contact
resistance since the contact pressure is, with the rough-
ness of the surfaces, one of the essential parameters of
which the contact resistance depends on, see [4] for
example. One of the objectives of this paper is to analyze
the evolution of the contact resistance in the post-lling
stage, before the detachment, in order to link it to the
molding conditions and to the nature of the materials in
contact. One will be able thus to calculate the temper-
ature, stresses and pressure elds in cavity in order to
describe heat transfers during the cooling of an injected
piece.
This work is structured in two main parts. The rst
one concerns the analysis of the contact conditions at
the interface between the polymer and the mold. The
methods and the results presented were obtained in the
framework of [8,9]. The second one concerns the simu-
lation of the heat transfers during the solidication of
the piece and the validation attempt. It is the object of
the works of Massee [10]. These works have been led in
cooperation in the AMETH program.
2. Transfers at the polymermold interface
The notion of interface thermal resistance is generally
based on the existence of several ways for the passage of
Nomenclature
C
p
specic heat (J kg
1
K
1
)
E Youngs modulus (Pa)
k thermal conductivity (Wm
1
K
1
)
K bulk modulus (Pa)
p pressure (Pa)
R thermal resistance (m
2
KW
1
)
Ra roughness (m)
TCR thermal contact resistance (m
2
KW
1
)
T temperature (K)
t time (s)
U displacement vector (m)
v specic volume (m
3
kg
1
)
Greek symbols
a linear thermal expansion coecient (K
1
)
b volumetric thermal expansion coecient
(K
1
)
d air gap thickness (m)

ee ee strain tensor ()
U thermal ux (Wm
2
)
k 1st Lamee coecient (Pa)
l 2nd Lamee coecient (Pa)
m Poissons coecient ()
q density (kg m
3
)

rr rr stress tensor (Pa)

s temperature variation (K)
Subscripts and superscripts
air relative to air
ms mold surface
ps polymer surface
0 initial
* relative to the interface
g relative to the glass transition
2016 H. Massee et al. / International Journal of Heat and Mass Transfer 47 (2004) 20152027
the heat ux through the interface, this one including
zones of contact and interstices in which gas is trapped.
The main part of the resistance is due to the constriction
of the heat ux lines in the zone at the neighborhood of
the interface. The statistical analysis of the surface per-
mits to estimate the area and the number of points of
contact and acts as basis to models that give satisfactory
enough results for metals [11]. This approach is with
diculty transposable to the case of the injection where
the polymer surface behavior is dicult to specify and
can evolve for example if some slips occur. A diculty is
also linked to the transient character of the process. In
the beginning of the lling, at the time of the abrupt
contact, one must consider the setting time of the con-
striction resistance. It is possible to estimate it by
s 100r
2
=a, where r is the surface of contact radius and
a the weighted diusivity of the two contacting bodies.
The order of magnitude is the millisecond. One can
consider therefore that the resistances are established.
Their expression leads to determine the surface tem-
peratures of the mold T
ms
t and of the polymer T
ps
t, as
well as the ux Ut which crosses the interface:
Rt T
ms
t T
ps
t=Ut 1
The measurement of the surface temperature of the
mold and of the heat ux that crosses the interface is
relatively comfortable, notably by inverse method to go
back to the surface. This method is impossible to use for
the polymer, in which one cannot introduce any su-
ciently thin sensors so that they do not modify the
temperature eld. The method that we used as well as its
validation is described in the Ref. [1]. We chose an
indirect method that consists in calculating the temper-
ature eld in the thickness of the injected plate, while
supposing that the transfers are 1D in this one, and
while using the experimental heat ux evolutions versus
the time as boundaries conditions. From the moment the
initial temperature eld is known, this calculation is
possible if the thermal properties, and notably their
variation according to the temperature is known. In the
case of an amorphous polymer, it does not pose any
particular problem. In the case of a semi-crystalline
polymer, the source released at the crystallization is not
known with precision in fast cooling, which is the case at
the beginning of the solidication during injection. The
kinetic eects prevent to determine the contact resistance
precisely then, at least during the rst instants. It is the
reason why we used the ABS which is an amorphous
polymer. The specic heat has been determined by using
a dierential scanning calorimeter, in cooling mode, at 8
K/min. The PvT diagrams have been measured by using
a SWO PVT100 device. The thermal conductivity has
been measured using the device and the method de-
scribed in Ref. [12]. The values of these parameters and
their variations with the temperature are indicated in
Fig. 1ac.
As we already underlined it, the initial temperature
eld is necessary to compute the thermal contact resis-
tance. Two dierent methods have been used, that gave
very close results. The rst one consisted in using a
software to simulate the lling of the mold. The second
one, described in the Ref. [1], consisted to use an inverse
method to go back to the initial state. A multiple criteria
method based on the variation of the heat uxes on the
faces of the sample as well as on the integral of this one
has been used. Indeed, the initial temperature distribu-
tion cannot be gotten in an unique manner if one uses
only the heat uxes. On the other hand, if one knows the
nal enthalpy, by adding a criteria on the integral of
these uxes, the solution becomes unique. Fig. 2 shows
an example of temperature eld identied in the thick-
ness of the molding cavity, at the end of the lling. We
also show the eld calculated at the same instant by the
software Moldow

. The proles are very close.

A mold made of steel (40 CMD8) molding plates of
0.012 m 0.014 m0.002 m has been used. On each of
the two faces, in the center of the plate, two heat ux
and surface temperature sensors are placed one in front
of the other. A pressure sensor is placed close to these
thermal sensors. We have veried by simulation that the
surfaces of the molding cavity had an uniform temper-
ature. The maximal temperature gap at the end of lling
is of the order of 1 K. The transfers are therefore well
1D in the thickness.
The heat ux and mold surface temperature sensors
whose schematic views are shown in Fig. 3a and b are
composed of a cylindrical measurement cell, Fig. 3b,
lodged in a body, Fig. 3a. The cell is composed of two
half cylinders whose plan of cut is perpendicular to the
molding cavity surface, in order not to modify the ux.
For the same reason, the metal of the sensor and its
surface roughness are identical to the one of the mold.
The simulation of the thermal behavior of the sensor in
the mold has been done, to verify that the ux that
crosses it is well identical to the one that exists in its
absence. One shows that up to a distance of 5 mm of the
surface, the temperature eld is not disrupted and re-
mains one-dimensional well. Three thermocouples of 12
lm diameter are placed along the isotherms, that is to
say parallel to the surface (Fig. 4). They are intrinsic
thermocouples since the extremities of the chromel and
alumel wires are welded on the body of the sensor and
separated by a portion of isothermal metal. The part of
the junction in the metallic body of the sensor does not
intervene therefore in the measurement but participates
to improve the response of the sensor. Their position in
relation to the surface is xed according to the rules of
the inverse method used to exploit the measurements.
The rst one consists in placing the thermocouple that
will act as boundary condition as far as possible of the
surface, in the zone where the transfers remain 1D. It is
placed therefore at 5 mm. The thermocouple closer to
H. Massee et al. / International Journal of Heat and Mass Transfer 47 (2004) 20152027 2017
the molding cavity is placed as close as the surface as
possible in regard of the technical constraints. It is so at
a depth of 200 lm. An intermediate thermocouple is
placed at a distance of 2 mm, in order to calculate the
residues. Its sensitivity in term of heat ux determination
is weaker than for the one placed closer to the surface.
The wires of the thermocouples are connected to a
reference box in which a platinum probe is placed.
Amplied tension is recorded then by a rapid data
acquisition system, at a frequency of 1 ms. A classical
sequential method [13] is used to get ux and surface
temperature. The good work of the sensors is veried on
a special apparatus described in detail in the Ref. [9]. A
characteristic example of ux and surface temperature is
shown in Fig. 5a and b. At the time of the contact of the
hot polymer, the temperature increases suddenly. If
there were no contact resistance, the contact tempera-
ture would be reached instantaneously. In presence of
this one, the increase rate is limited. The temperature
decreases then, with the progression of the cooling of the
piece. At the time of contact heat ux is theoretically
innite, it reached 10
5
W/m
2
s at the end of the lling,
decreases then to become equal to zero when the piece
is cooled.
Fig. 6 shows the evolution of the thermal contact
resistance, for a cycle corresponding to the evolution of
pressure indicated on this gure, using the right hand
scale. The initial time t 0 s corresponds to the end of
the lling. The pressure decreases in the molding cavity,
because of the frost of the gate that is almost immediate.
Fig. 2. Initial temperature eld in the piece identied by inverse
method and calculated by Moldow.
Fig. 1. Thermal properties of ABS: (a) specic heat, (b) PvT diagram, (c) thermal conductivity.
2018 H. Massee et al. / International Journal of Heat and Mass Transfer 47 (2004) 20152027
The contact resistances that are equal on the two faces
of the piece increase as the pressure decreases. One sees
that they double between the beginning and the end of
the cooling phase where the piece is in contact with the
mold. Their mean values are of 5 10
4
m
2
KW
1
,
which is in conformity with the values observed by Yu
et al. [5]. When the pressure becomes equal to the
atmospheric pressure, the resistances increase brutally
while remaining equal on the two opposed faces of the
cavity. In this case, the air gap distributes itself in an
identical manner on the two faces of the sample.
We show in Fig. 7 the results that we get for dierent
molding cavity surfaces. The results for three dierent
roughness (Ra 0:05 lm, Ra 1 lm, Ra 5 lm) are
presented as well as for a surface of roughness Ra 1 lm
covered with chromium or Teon. The drawn curves are
obtained showing in abscissa the instantaneous value of
the pressure. For every surface, a logarithmic evolution
of the conductance according to the pressure is noticed.
The chromium-plating does not modify the contact
resistance whereas the coating of Teon decreases the
conductance, which was foreseeable. Having some re-
sults for three dierent roughness, it was interesting to
examine the evolution of the conductance according to
the roughness, for dierent values of the instantaneous
pressure. One notes that the conductance decreases when
the roughness increases. The quantity of air captured
between the peaks of the surface explains this evolution.
Although only three values are available, one gets nev-
ertheless a linear evolution of the conductance according
Fig. 3. Technical scheme of the heat ux and mold surface
temperature sensor: (a) a cut of the body of the sensor. Inside
this one, a cylindrical measurement cell is lodged, (b) a view of
one of the two half-cylinders which constitute this cell, for
which a detail is shown in Fig. 4.
Fig. 4. View of the sensor showing the thermocouples wires
along the isotherm.
Fig. 5. Example of measure of heat ux and surface temperature of the mold. (a) heat ux, (b) mold temperature.
H. Massee et al. / International Journal of Heat and Mass Transfer 47 (2004) 20152027 2019
to the roughness. One proposes the empiric law (2) to
calculate the contact resistance between the surface of the
mold and of the polymer
Rt R
0
expPt=P
0
2
R
0
and P
0
are constants that depend on the roughness of
the mold. For condentiality reasons, their exact values
cannot be given here. This result is only valid for a
couple of moldpolymer material, the resistance also
depending, but of secondary manner of the conductivity
of the materials in contact. Indeed, the role of the
interstitial gas is not negligible in relation to the passage
by the points of contact. Its decompression during the
cooling and therefore the variation of its volume
strongly modies the conditions of contact and forbid-
den the use of classical models. A thermomechanical
analysis at a microscopic scale would be necessary to go
further in the modeling.
3. Modeling of the cooling phase
3.1. General presentation
The simulation uses the nite volumes method. The
equations are written in a conservative form, which
makes nite volumes a natural method to solve the
equations of the uids mechanics of the and the heat
transfers [14].
Traditionally, the equations of the solid mechanics
are solved by nite elements methods. Some authors
begin however to use the nite volumes to solve these
equations in particular I. Demirdzic [15,16]. Hattel [17]
uses also a technique of nite dierences.
One applies the small perturbation hypothesis (small
displacements, small strains, small temperature varia-
tions), that supposes that the displacements between two
time steps in the solid material are suciently small. The
position of the nodes of the mesh remains unaltered
during the calculation, but a displacement of these nodes
is calculated during the time.
3.2. Formulation of the equations
At the end of the lling phase, the piece cools and
solidies in the mold. The physical phenomena are
piloted then by the state equation of the polymer, the
displacements equation and the energy equation.
3.2.1. State equation
The specic volume joins the variations of tempera-
ture and pressure to the volume of the material through
the PvT diagram. This relation designed as state equa-
tion may be written in the general form
f P; v; T 0 3
Although theoretical approaches of this state equation
exist, it is generally represented by an empirical model
based on the measurements. We chose the Tait relation
[18], currently used in the literature [19], that gives a
good approximation of the PvT diagram and whose
coecients have been tabulated for many polymers. This
relation is
t0;

TT tp;

TT
t0;

TT
C ln 1
_

p
B

TT
_
4
Where the variation of T
g
f p is linear, its evolution
being deduced from experimental results
T
g
p T
g
0
0:55
3
p 5
One notes that the use of this relation requires the
knowledge of three parameters:
Fig. 6. Evolution of the thermal resistances at the polymer
mold interface.
Fig. 7. Evolution of the conductance according to the pressure.
Inuence of the roughness of the mold.
2020 H. Massee et al. / International Journal of Heat and Mass Transfer 47 (2004) 20152027
the specic volume at atmospheric pressure v0; T,
the adimentional constant C,
the function BT, that has the dimension of a pres-
sure.
Simha et al. [20] showed that C is nearly constant
(better approximation: C 0:0894) and that BT can
be written
BT b
1
e
b
2
T
6
The constant b
2
has a mean value of 4.5 10
3
while b
1
depends on the nature of the polymer, but is propor-
tional to the elastic module of compression K
E
312m
.
We use b
1
3:3 10
2
K, which permits to approach
the measurements satisfactorily (Fig. 1b).
The knowledge of the PvT diagram of the material
permits us, if we consider that the material has a quasi
elastic behavior, to deduce the values of the density and
of thermal expansion coecient, which depend at a time
on the temperature and the pressure
q
1
m
; a
1
m
om
oT
_ _
p
; b
aE
1 2m
7
The use of this last equation includes some limita-
tions however. Indeed, even though it is possible to give
a realistic enough evolution of the Poissons coecient
(for example m 0:2 for T < T
g
and m 0:4 for T > T
g
),
the detachment of the piece makes it the bulk modulus
unrealistic when the pressure at the surface of the piece
becomes equal to the atmospheric one. It is then nec-
essary to use the measured values of the Young modu-
lus, in the relation giving the variation of the bulk
modulus.
3.2.2. Displacements equation
The unknown of this equation are the components of
the displacements vector. In the formulation that we
propose, it is the equilibrium equation that is written
and that allows to obtain the discrete displacements
equation (on an irregular Cartesian mesh) following the
way presented in Fig. 8.
The static equation of equilibrium is
1
V
_ _ _
D
r

rr rrdv 0 8
While using the theorem of the divergence, we obtain
1
V
_ _
S

rr rr ~nnds 0 9
The elastic behavior leads to write the stress tensor

rr rr ktr

ee ee

II

II 2l

ee ee bs

II

II 10
The equation that links the deformation to the dis-
placements is then used

ee ee
1
2
rU r
T
U 11
We get the displacements (or Navier) equation
rkr U r blrU rU
T
c rbs 0 12
The discrete form on the Cartesian volume presented in
Fig. 9 gives
1
dx dy
_ _
S
r
xx
e r
xx
w dy

1
dx dy
_ _
S
r
xy
n r
xy
S dx 0
1
dx dy
_ _
S
r
xy
e r
xy
w dy

1
dx dy
_ _
S
r
yy
n r
yy
S dx 0
_

_
13
where e, w, n and s are the respective indications of the
east, west, south and north faces of a control volume.
The derivatives of the displacements vector U com-
ponents are approached by a nite dierences scheme.
While noticing that the main term derivatives (u or v for
the equation projected respectively according to x or y)
are in the direction normal to the surfaces on which they
are dened, we use a ve points discrete scheme. By
example
ob
n
ox
e
ob
s
ox
e
b
i1;j
b
i;j
Dx
e
b u; v 14
stresses
(equilibrium equation)
strains displacements
(Navier equation)
stress-strain relation Strain-displacements relation
Fig. 8. Displacement equation from equilibrium equation.
n
s
e w

yy

xy

yx

xx
Fig. 9. Control volume for the discrete equations.
H. Massee et al. / International Journal of Heat and Mass Transfer 47 (2004) 20152027 2021
The terms of coupling u (respectively v) for the
equation projected according to x (respectively y) have
their derivatives in a tangent direction to the surface on
which they are dened. We use then a nine points dis-
crete scheme. By example
ob
w
oy
s
b
i;j
b
i1;j
b
i;j1
b
i1;j1

2Dy
s
b v; u
15
We get a matrix system then where the main blocks A
and D contain ve diagonals, while the coupling blocks
B and C contain nine diagonals
A B
C D
_ _
U
V
_ _

/
x
/
y
_ _
16
3.2.3. Energy equation
In the packing phase and after freezing of the gate,
the shear rate is very small and viscous dissipation is
negligible. So, the equation of the energy may be written
in the form of
1
v
C
p
T
oT
ot
aT; pT
op
ot
r kTrT 17
We have already noted the strong inuence of the
resistances on the evolution of the temperature eld.
Besides the contact resistance that has been quantied in
the previous paragraph, a thermal resistance exists due
to the air gap that can come in between the piece and the
mold during the cooling. It is therefore necessary to
know the displacements and the deformation of the
piece to calculate this thermal resistance correctly.
At the time of the detachment, that is to say when the
stress at the polymermold interface becomes equal to
the atmospheric pressure, an initial thickness of the air
gap equal to the value of the contact resistance at this
instant (t 7 s in Fig. 6) is considered. The interface
conductance expresses then
ht
1
d
R
air
Rpt
18
Thermal conductivity and specic heat have been
measured at the laboratory and have been used in the
simulation. The density comes from the PvT diagram
and depends therefore on the pressure and the temper-
ature.
3.3. Discrete equations
The state, displacements and energy equations are
written in discrete form in space and time, by using the
control volume represented in Fig. 9.
For the energy equation a classic Quick scheme is
used. The algorithm consists in solving in a rst step the
energy equation, then the state equation gives vt
n
; p
n1
.
One can calculate then at
n
; p
n1
as well as bt
n
; p
n1

what permits to solve the displacements equation. The

deformations are then calculated and the resolution of
Eq. (10) gives the stresses and therefore the new pres-
sure.
In theory iterations on the pressure are necessary but
we noticed that they do not bring a considerable modi-
cation to the result for very small time steps (10
3
s).
4. Presentation of the results and comparison to the
experiments
We present here two results. The rst one corre-
sponds to a packing pressure of 40 MPa, the tempera-
ture mold regulation being programmed in order to
ensure a good thermal symmetry all along the experi-
ence. It corresponds to a mold dierent of the one used
to determine the relations giving the thermal contact
resistance, but it is constituted of the same material and
has the same surface roughness. It permits to achieve
square pieces of 0.062 m0.062 m and of 3 mm thick-
ness. The ux and surface temperature sensors are ar-
ranged on the two large faces of the piece, as indicated
in Fig. 10.
We present the result of surface temperature recorded
by one of the thermal sensors in Fig. 11. On this graph
are presented the data obtained by the three thermo-
couples constituting the sensor and the surface temper-
ature identied, which is nearly confounded with the
data coming from the thermocouple closer to the sur-
face.
The acquisition time includes three cycles (lling,
packing, cooling, ejection). A very good reproducibility
of the temperatures of a cycle on the other is noticed,
what shows that the press reached its thermal balance
when the measurements have been done.
Fig. 10. Position of the temperature and pressure sensors on
the surface of the molding cavity.
2022 H. Massee et al. / International Journal of Heat and Mass Transfer 47 (2004) 20152027
The surface temperatures being nearly confounded
for the two sensors situated in front one of the other,
only one is shown. One also notes the instant of the
ejection, 16 s after the injection, the evolution of the
surface temperature showing an angular point.
We used the plans of the mold to reproduce the po-
sition as well as the size of the cooling channels. The
heat uxes extracted by these channels are estimated
according to the temperature of the cooling water, that
can be considered constant with a 0.1 K precision. The
heat transfer coecient between the cooling water and
the channel is computed by a classical correlation [21].
The other conditions at the boundaries of the calcula-
tion domain have been calculated taking into account
the mold thermal resistance.
The coupled calculation of heat transfer and
mechanics starts at the end of the lling. The initial
temperature prole is simulated by the cooling of a
polymer initially to uniform temperature equals to 235
64
66
68
70
72
74
76
0 20 40 60 80 100
time (s)
t
e
m
p
e
r
a
t
u
r
e

(

C
)
temperature
sensor 1
sensor 2
sensor 3
Fig. 11. Heat ux sensor measurements (three thermocouples
and the identied temperature).
67
68
69
70
71
72
73
74
75
0 5 10 15 20
time (s)
m
o
l
d

t
e
m
p
e
r
a
t
u
r
e

(

C
)
numerical
experimental
Fig. 12. Comparison between measure and simulation: tem-
perature of the mold.
0
120000
100000
80000
60000
40000
20000
0 5 10 15 20
time (s)
f
l
u
x
e
s

(
W
/
m

)
numerical
experimental
Fig. 13. Comparison between measure and simulation: heat
ux at the interface.
0
50
100
150
200
250
300
350
400
450
0 5 10 15 20
time (s)
p
r
e
s
s
u
r
e

(
b
a
r
)
numerical
experimental
Fig. 14. Comparison between measure and simulation: pressure
at the interface.
-25
-20
-15
-10
-5
0
5
10
15
0 0.5 1 1.5 2 2.5 3
part thickness (mm)
s
t
r
e
s
s

(
M
P
a
)
Fig. 15. Residual stresses trapped in the thickness of the piece
after cooling.
H. Massee et al. / International Journal of Heat and Mass Transfer 47 (2004) 20152027 2023
C put in contact with a mold to 60 C. The beginning of
the calculation corresponds to 0.45 s after the contact
time.
The calculations are done on a Cray J90 with 200
Mops by processor power. The modeling of a 20 s.
cycle requires about 72 h of CPU time.
Figs. 1214 show the comparison between the mea-
surements and the calculation of the mold temperatures,
of the parietal heat uxes and of the pressure.
The calculated and measured temperatures at the
surface of the mold are in very good agreement. It is also
the case for the pressure. This one is imposed at 40 Mpa
during 3 s then decreases freely, the gate being supposed
frozen at this instant. The moment when the value of the
pressure reaches the atmospheric pressure is well pre-
dicted, what shows that the evaluation of the variation
of the Young modulus is correct. The calculated heat
ux is lower than the measured one. We interpret this
dierence by the fact that we did not take in account the
entry of hot polymer during the packing phase. Indeed,
if the calculation makes really vary the mass during the
cooling under constant pressure, it aects to the entering
matter the local temperature during cooling. This creates
an internal energy decit that explains the ux dier-
ences.
Fig. 15 shows the residual stresses distribution at the
time of the ejection of the piece. This result is quite
similar to the results obtained by Struik [22] for an
elastic material. The shrinkage being small, one does not
note detachment of the piece.
The second example corresponds to a non symmet-
rical case where a 10 K dierence is imposed on the two
faces of a 2 mm thickness piece. The packing has been
voluntarily limited to 2.6 Mpa in order to provoke
detachment. Fig. 16 shows the evolution of the surface
temperature of the polymer calculated by the technique
described in the Section 2, in one interval of time where
the detachment of the piece occurs. When it occurs, at
dierent times on the two faces of the piece, the surface
temperature of this one increases (up to 12 K in this
example). This heating is due to the brutal increase of
the polymermold interface resistance. One observes the
eect of a beating of the piece that, at t 4:5 s nds
a new position in the axis of the cavity, to come closer
of the cold side.
The simulation of this case is shown in Fig. 17. The
detachment is observed, as well as the temperature in-
60
65
70
75
80
85
90
95
0 2 4 6 8 10
time (s)
p
o
l
y
m
e
r

s
u
r
f
a
c
e

t
e
m
p
e
r
a
t
u
r
e

(

C
)
hot side
cold side
Fig. 16. Experimental polymer surface temperature deduced from inverse analysis. One notes the eect on these temperatures of the air
gap occurring in the process.
60
65
70
75
80
85
90
95
100
0 10 20 30 40
time (s)
t
e
m
p
e
r
a
t
u
r
e

(

C
)
mold surface (cold)
mold surface (hot)
polymer surface (cold)
polymer surface (hot)
Fig. 17. Calculated evolution of the mold and polymer surface
temperatures.
2024 H. Massee et al. / International Journal of Heat and Mass Transfer 47 (2004) 20152027
crease of the polymer. The amplitude of this increase is
correctly predicted. The two faces become unstuck in a
simultaneous manner, but for surface temperatures close
to those observed experimentally. The beating is not put
in evidence but the sum of the resistances is compliant to
the experimental results. Their evolution is also com-
pliant to the results of Yu et al. [5]. The curve of the
calculated evolution of the thermal contact resistances
on the two faces is shown in Fig. 18. The sum of the
resistances corresponds well to the shrinkage calculated
thanks to the PvT diagram. The phenomenon of beating
does not seem due in this case to the non uniform stress
eld in the thickness of the piece, but to the adherence of
the piece on the side opposed to the ejectors, by suction
eect, caused by the decompression of air trapped in the
roughness of the surface of the mold. This suction eect
could be simulated by a normal negative additive stress
applied to the surface of the piece.
The air gap distribution on the two faces of the
piece and its evolution during time is shown in Fig. 19.
One observes that shape and amplitude of the air gap
on the two faces are not symmetrical. Before 10 s
the distribution is identical on each of the two faces.
Thereafter, the piece takes a prole of S, while coming
closer of the cold wall. At the same instant on one face
of the sample the resistances defer 30%, what corre-
sponds to 7 K temperature dierence. An interesting
result must be noted: the bottom of the piece, solidi-
ed just at the end of the lling, does not contract
itself in thickness but displaces up of 2 mm, drawn by
the shrinkage of the rest of the piece. One can there-
fore expect important variations of specic volume. In
the upper part of the cavity, the gate blocks the
shrinkage.
0
5
10
15
20
25
30
35
40
45
50
0 20 40 60
air gap width (um) air gap width (um)
h
e
i
g
h
t

(
m
m
)
h
e
i
g
h
t

(
m
m
)
t=5s
t=10s
t=20s
t=30s
t=40s
0
5
10
15
20
25
30
35
40
45
50
-60 -40 -20 0
t=5s
t=10s
t=20s
t=30s
t=40s
Fig. 19. Distribution of the air gap on the faces of the piece at dierent times.
0.0E+00
5.0E-04
1.0E-03
1.5E-03
2.0E-03
2.5E-03
3.0E-03
0 10 20 30 40
time (s)
T
R

(
m

.
K
.
W
-
1
)
cold side
hot side
Fig. 18. Evolution of the thermal contact resistances on the
faces of the piece.
H. Massee et al. / International Journal of Heat and Mass Transfer 47 (2004) 20152027 2025
5. Conclusion
In a rst part, we presented a method to measure the
evolution of the thermal contact resistances between an
injected polymer and the mold. we propose a relation
that permits to predict the time evolution of this resis-
tance for a couple of mold material and polymer and a
given roughness. This relation is valid when the piece is
in contact with the mold, before the apparition of the air
gap due to shrinkage. The classical approaches based on
the mechanics of the contact do not allow to get orders
of magnitude and evolutions of the resistances in
agreement with this empirical relation. Theoretical
investigation to interpret this empirical relation consti-
tutes an open eld that would permit to better predict
these resistances. In the present state, only heavy cam-
paigns of measurements permit to evaluate the R
0
and P
0
coecients of the model.
In a second part, we present the simulation of the very
strong thermal and mechanical coupling during the
cooling of a piece in the molding cavity. The calculation
is based on simple hypotheses from a mechanical point of
view since it supposes an elastic material whose Young
modulus and Poissons coecient are temperature
dependent. The contact resistance depends on the normal
stress at the polymermold interface according to the
previously established empirical relation. After the
detachment, the thickness of the air gap is determined by
the position of the piece in the mold resulting of the stress
distribution due to the temperature eld. This model
compared to the measurements on a mold dierent of the
one used in the rst part permits to obtain a good
agreement with the experimental evolutions of mold
temperature and of pressure in the molding cavity, and
gives a correct behavior for the residual stresses.
In a case of strong shrinkage (no packing) with
temperature dierences between the surfaces of the
molding cavity, the detachment of the piece and
the temperature increase on its surface are observed. The
time evolutions of resistances are compliant to the
experiments both in the evolutions and in the reached
values. The eects of gas trapped in the interstices of the
mold roughness are important, notably on the appari-
tion of beatings. [9] showed also that the decompression
of this interstitial gas permits to interpret the time var-
iation of the contact resistance. The surface tempera-
tures show heterogeneity superior to 5 K. The shape of
the piece can be very complex and our results conrm
the importance of a thermal and mechanical coupled
simulation to foresee the cooling, the shrinkage, and the
residual stresses correctly in an injected piece.
The results presented here concern an amorphous
material and they are not transposable to a semi-crys-
talline material. The shrinkages are very important and
the relation between pressure and resistance does not
seem to be as simple as in the studied case.
Acknowledgements
This work has taken place in the frame of the re-
search network AmETh (Ameelioration des Echanges
ThermiquesImprovement of Thermal Exchange)
supported by both the French Ministry of Education
and Technological Research, MENRT, and the Na-
tional Center for Scientic Research, CNRS, SPI
Department).
References
[1] D. Delaunay, Ph. Le Bot, R. Fulchiron, J.F. Luyee, G.
Regnier, Nature of contact between polymer and mold in
injection molding. Part 1: Inuence of a non-perfect
thermal contact, Polym. Eng. Sci. 40 (7) (2000).
[2] O. Rhee, Thermal contact measurement, Technical report
80, Cornell Injection Molding Report, 1994.
[3] J.A. Hall, W.H. Ceckler, E.V. Thomson, Thermal proper-
ties of rigid polymers, I. Measurement of thermal conduc-
tivity and question concerning contact resistance, J. Appl.
Polym. Sci. 33 (1987) 20292039.
[4] S.K. Parihar, N.T. Wright, Thermal contact resistances at
elastomer to metal interfaces, Int. Commun. Heat Mass
Transfer 24 (8) (1997) 10831092.
[5] C.J. Yu, J.E. Sunderland, C. Poli, Thermal contact resis-
tance in injection molding, Polym. Eng. Sci. 30 (1990) 1599
1605.
[6] S. Quilliet, Ph. Le Bot, D. Delaunay, Y. Jarny, Heat
transfer at the polymermetal interface: a method of
analysis and its application to injection molding, in: ASME
Proceedings of the 32nd National Heat Transfer Confer-
ence, HTD-340, vol. 2, 1997, pp. 916.
[7] L. Shridar, W. Yin, K.A. Narh, The eect of shrinkage
induced interface gap on the thermal contact resistance
between the mold and plastic in injection molding, J. Inject.
Mold. Tech. 4 (1) (2000).
[8] Ph. Le Bot, Comportement thermique des semis-cristallins
injectees, Application aa la preediction des retraits, Theese de
Doctorat, Universitee de Nantes, 1998.
[9] S. Quilliet, Transferts thermiques aa linterface polymeere-
meetal dans le proceedee dinjection des thermoplastiques,
Theese de Doctorat, Universitee de Nantes, 1998.
[10] H. Massee, Couplage thermomeecanique lors de la solidi-
cation de mateeriaux polymeeres, Theese de Doctorat, Uni-
versiteede Bordeaux 1, 2000.
[11] J.J. Salgon, F. Robe-Valloire, J. Blouet, J. Bransier, A
mechanical and geometrical approach to thermal contact
resistance, Int. J. Heat Mass Transfer 40 (1997) 1121.
[12] T. Jurkowski, Y. Jarny, D. Delaunay, Estimation of
thermal conductivity of thermoplastics under moulding
conditions: an apparatus and an inverse algorithm, Int.
J. Heat Mass Transfer 40 (17) (1997) 41694181.
[13] J.V. Beck, B. Blackwell, C. St. Clair, Inverse Heat
Conduction, Wiley-Interscience, 1985.
[14] S.V. Patankar, Numerical Heat Transfer and Fluid Flow,
Hemisphere, 1990.
[15] I. Demirdzzicc, D. Martinovicc, Finite volume method for
thermo-elasto-plastic stress analysis, Comput. Methods
Appl. Mech. Eng. 109 (1993) 331349.
2026 H. Massee et al. / International Journal of Heat and Mass Transfer 47 (2004) 20152027
[16] I. Demirdzzicc, S. Muzaferija, Finite volume method for
stress analysis in complex domains, Comput. Methods
Appl. Mech. Eng. 37 (1994) 37513766.
[17] J.H. Hattel, A control volume-based nite dierence
method for solving the equilibrium equations in terms
of displacements, Appl. Math. Modell. 19 (1995) 210
243.
[18] P.G. Tait, Phys. Chem. 2 (1) (1888).
[19] D.W. Van Krevelen, Properties of Polymers, Elsevier,
1990.
[20] R. Simha, P.S.S. Wilson, O. Olabisi, Kolloid-Z 2 (1973) 402.
[21] W.M. Kays, M.E. Crawford, Convective Heat and Mass
Transfer, McGraw-Hill, 1993.
[22] L.C.E. Struik, Orientation eects and cooling stresses in
amorphous polymers, Polym. Eng. Sci. 18 (10) (1978) 811
817.
H. Massee et al. / International Journal of Heat and Mass Transfer 47 (2004) 20152027 2027