Você está na página 1de 15

An analysis of the three-dimensional resin-transfer

mold lling process


Seong Taek Lim, Woo Il Lee *
Department of Mechanical Engineering, Seoul National University, Seoul 151-742, South Korea
Received 13 April 1999; received in revised form 26 July 1999; accepted 21 October 1999
Abstract
Resin-transfer molding (RTM) is a process in which thermosetting resin is injected into a mold cavity pre-loaded with a porous
brous preform. For parts with small thickness, the mold-lling process can be modeled as two-dimensional by neglecting the
resin ow in the thickness direction. However thicker parts require extensive resin ow through the thickness and thus the three-
dimensional eect in the resin ow must be considered. In this study, numerical simulations of three-dimensional mold lling during
RTM were performed. The governing dierential equations were discretized by using the control volume nite element method
(CVFEM). The CVFEM technique was employed bacause of its simplicity in handling the moving boundary problems. The tem-
perature and the degree of cure were also calculated. In order to evaluate the validity of the numerical results, they were compared
with exact solutions for simple geometries, and close agreement was observed. Experiments were also performed. To check the
three-dimensional resin front location as a function of time inside the preform, an optical ber was used as a sensing element. The
agreements between the experimental data and the numerical results were found to be satisfactory. Numerical calculations for
complicated geometries were also performed to illustrate the eectiveness of the computer code developed in this study. # 2000
Elsevier Science Ltd. All rights reserved.
Keywords: Resin-transfer molding (RTM); Finite-element analysis (FEA); Permeability
1. Introduction
Resin-transfer molding (RTM) is a process in which
thermosetting resin is injected into a preheated mold.
After the ber preform is completely wetted out by the
resin, the curing process follows (Fig. 1).
In RTM, the injection pressure, temperature of the
mold, permeability of the ber mat and resin viscosity
are the major processing variables. In general, higher
injection pressure and mold temperature and lower resin
viscosity shortens the manufacturing cycle time. How-
ever, excessive injection pressure may cause deformation
of the mold and wash-out of the ber preform. An
excessively high mold temperature may induce pre-
mature resin gelation and cause short shot. All of the
process variables are interrelated and have eects on the
mechanical properties of the products. It is therefore
essential to predict the eect of the process variables in
order to optimize the conditions.
In simulating the RTM process, the prediction of the
moving boundary is a major concern. Coulter and
Gu c eri [1] have performed numerical simulations using
the nite-dierence method and boundary tted coor-
dinates, while Chan and Hwang [2] used the nite-ele-
ment method. Um and Lee [3] and Yoo et al. [4]
adopted the boundary-element method for mold-lling
simulation. In solving moving boundary problems using
the FDM or FEM, as the resin-ow front advances, the
calculation domain should be redened and the numer-
ical mesh regenerated. The regeneration of the mesh
requires large calculation time as the calculation domain
becomes complicated. Among numerical techniques, the
control volume nite element method (CVFEM) is that
a set of equations is formed for nodal control volumes
and solved as if they are nite elements. As a xed grid
system is employed in CVFEM, there is no need for the
regeneration of the mesh and the simulation for com-
plex geometry can be done rapidly and eectively.
Bruschke and Advani [5] obtained reasonable results for
isothermal resin ow by FE/CV method to predict ll-
ing patterns for complex shell-like molds. Lin et al. [6]
0266-3538/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved.
PI I : S0266- 3538( 99) 00160- 8
Composites Science and Technology 60 (2000) 961975
* Corresponding author. Tel.: +82-2-880-7122; fax: 82-2-883-0179.
E-mail address: wilee@snu.ac.kr (W.I. Lee).
used CVFEM in considering heat transfer and chemical
reaction problems. Kang [7] simulated the mold-lling
process including the temperature and the curing eect.
Most of the previous works were done for thin parts,
where the resin ow is two-dimensional. However,
thicker products with more complex geometry require
extensive ow in the thickness direction. Temperature
and degree of cure also vary signicantly through the
thickness. Young [8] simulated RTM process three-
dimensionally using CVFEM to predict the ow-front
locations during the lling process and Phelan [9]
adopted FEM to solve the governing equation set and
simulated RTM mold-lling process including runner
distribution systems, shell geometries and fully three-
dimensional ows. Also Varma et al. [10] and Lam et al.
[11] suggested the governing equations and the numer-
ical methods for RTM analysis which were applicable to
full three-dimensional analysis.
In this study, an attempt was made to develop a com-
puter code which can simulate the three-dimensional
RTM mold-lling process, including the temperature
and the curing eects. In order to validate the developed
computer code experimentally, the ow-front location
inside the thick preform should be monitored during the
lling process. The resin-front locations were observed
using optical-ber sensors. In order to perform numer-
ical simulation, three-dimensional permeability must be
known. The permeability of the ber preform was mea-
sured also using optical-ber sensors.
2. Problem statements
Consider an RTM mold with an arbitrary three-
dimensional geometry. The mold is loaded with ber
preform with anisotropic permeability. The mold is
thick so that the resin ow through the thickness must
be considered. Resin is injected at specied locations on the
mold surface. Either the injection pressure or the injection
ow rate is given as a function of time. The mold may
be preheated to facilitate the resin ow. The pre-loaded
ber mat, therefore, can be assumed to be heated at the
same temperature. The temperature and degree of cure
of the resin are considered to be maintained constant at
the injection gates. Air vents are ventilated to the
atmosphere. They are assumed to be closed when the
resin front reaches. The resin-front location and the
pressure eld during the lling process are to be pre-
dicted. Also, the temperature distribution and the
degree of resin curing are to be estimated.
3. Mathematical modeling and governing equations
In the RTM mold-lling process, thermosetting resin
ows through the porous ber preform. Therefore, the
ow can be assumed to follow Darcy's law [12]
V

=
1
j
K [ ]VP (I)
where V

is the velocity of the resin ow, j is the visc-


osity of the resin, K [ [ is the permeability tensor of the
ber preform and P is the pressure.
As the resin is incompressible, mass conservation can
be stated as
V V

= 0 (P)
The energy equation for the resin can be written as
follows [6]:
,
r
c
pr
oT
r
ot
,
r
c
pr
V

VT
r
= Vk
r
VT
r
h
v
T
f
T
r
_ _
G
.
(Q)
where ,. c
p
. k are the density, the specic heat and the
thermal conductivity of the resin, respectively. h
v
is the
convection heat transfer coecient between the ber
and the resin, T is the temperature, and is the porosity
of the ber preform. The subscripts r and f represent
resin and ber, respectively. is the viscous dissipation
rates which comes from the heat generated by the resin
ow through the pore. G
.
is the heat generated by the
resin curing reaction and is expressed as [13]
G
.
= H k
1
k
2
o
m
1
( ) 1 o ( )
m
2
(R)
The energy equation for the ber preform can be
written as follows [6]:
1 ( ),
f
c
pf
oT
f
ot
= 1 ( )Vk
f
VT
f
h
v
T
r
T
f
_ _
(S)
Mass conservation of chemical species represents the
conversion between monomer and polymer [6]
oo
ot

V
.

Vo = m
.
(T)
Fig. 1. Resin transfer modeling process.
962 S.T. Lim, W.I. Lee / Composites Science and Technology 60 (2000) 961975
where o is the degree of cure and m
.
is the chemical
reaction rate of the resin, which can be related to the
degree of cure as [13]
m
.
= k
1
k
2
o
m
1
( ) 1 o ( )
m
2
(U)
Constitutive equations can be obtained in the forms
to be used in the CVFEM technique, if the control
volume method is applied to the governing equations
[14]. First, the mass conservation equation [Eq. (2)] is
integrated for control volume and the divergence theo-
rem is applied,
_
g..
V V

du =
_
g..
V

dS = 0 (V)
Eqs. (1) and (8) are combined to yield the following
expression:
_
g..

1
j
K [ ]VP n

dS = 0 (W)
The energy equations of resin and ber [Eqs. (3)(5)]
can be joined into a single equation [Eq. (10)] by the
assumption that the temperature of the resin and the
ber become identical as the resin impregnates the dry
ber [6].
,c
p
oT
ot
,
r
c
pr
V

VT = VkVT G
.
(IH)
where ,c
p
and k are expressed as
,c
p
= ,
r
c
pr
1 ( ),
f
c
pf
k = k
r
1 ( )k
f
(II)
Integration of the combined energy equation [Eq.
(10)] for control volume with the aid of Green's theorem
yields
o
ot
_
g..
,c
p
Tdu
_
g..
,
r
c
pr
TV

dS =
_
g..
kVT n

dS

_
g..
H k
1
k
2
o
m
1
( ) 1 o ( )
m
2
du
(IP)
Likewise, the chemical species mass conservation
[Eqs. (6) and (7)] can be expressed as
o
ot
_
g..
odu
_
g..
o

dS
=
_
g..
k
1
k
2
o
m
2
( ) 1 o ( )
m
2
du
(IQ)
The boundary conditions can be stated as follows.
At the injection gate:
P
inlet
= P
0
or u
inlet
= u
0
T
inlet
= T
0
o
inlet
= o
0
(IR)
At the mold wall:
oP
on
wll
= 0
T
mold
= T
m
(IS)
On the ow front region:
P
front
= 0
k
oT
on
front
= 1 ( ),
f
c
pf
u
n
T
f0
T
_ _
(IT)
Solutions to Eqs. (12) and (13) along with the boundary
conditions [Eqs. (14)(16)] yield the resin front location
as well as pressure, temperature and degree of cure dis-
tributions as functions of time.
4. Numerical simulation
The RTM mold lling process is a moving boundary
problem, which means that the computation domain
changes continuously with time. Therefore, those con-
ventional methods such as FEM and FDM require
mesh regeneration at every time step to describe the
change of the calculation domain. The control volume
method saves much trouble in mesh regeneration when
combined with the volume of uid (VOF) [15] which is
suggested by Tadmor and Broyer [1618]. In order to
solve the governing equations, the CVFEM was used.
Discretization of the three-dimensional domain was
done with a tetrahedron nite element (see Fig. 3). A
tetrahedron element consists of four sub-volumes divi-
ded by four control surfaces. Each control volume is
composed of sub-volumes surrounding a node. For
example, as shown in Fig. 3, the control volume i is
made up of sub-volumes of every element which sur-
rounds the node i. In the CVFEM, the net ux of mass,
momentum, energy and chemical species through the
control surfaces is conserved rigorously within the cor-
responding control volume. In this method, the whole
domain is divided into a xed grid system and a scalar
parameter f is introduced for each cell to represent the
ratio of occupied volume to the total volume. As the
ow front advances, all of the control volumes can be
classied into three categories as (see Fig. 2):
f = 1: main ow region
0 - f - 1: ow front region
f = 0: empty region
S.T. Lim, W.I. Lee / Composites Science and Technology 60 (2000) 961975 963
The ow front lies over the control volumes of which
the lled fraction is between 0 and 1. If f reaches a cer-
tain value, say 0.5, the ll time at the moment is recor-
ded as the ll time of the control volume. After the
entire mold is lled, the lling patterns of the ow front
can be obtained by interpolating the recorded ll times
of the respective control volume [19]. This method
enables us ecient advancement of the ow front which
satises the mass conservation more strictly.
In the energy equation [Eq. (10)] and the chemical
species equation [Eq. (6)], the transient terms are inclu-
ded. For time integration of these equations, an explicit
method was used [20]. The necessary condition for sta-
bility can be expressed as Eq. (17) based upon the Von
Neuman stability analysis [20].
t4
2o
x
x
2

2o
y
y
2

2o
z
z
2

z
_ _
1
(IU)
where a
i
, represents the diusion coecient, u

. v

. w

represent the velocities of the resin ows, t is the time


increment and x. y. z are the characteristic nite-
element lengths.
On the ow-front region, the lled fraction f is upda-
ted at each time step as the ow front advances. There-
fore, the lled fraction f is included in the dierential
term of time. On that region, Eq. (12) can be written in
matrix form as
C
d fT t ( ) ( )
dt
= DT t ( ) f E
S
t ( )
df
dt
E
F
(IV)
where T(t) represents the temperature, DT t ( ) is the
energy convection by resin ow, and f E
S
t ( ) is the heat
released by resin curing.
df
dt
E
F
is the energy absorbed
from the pre-heated ber mat. The energy convection
DT t ( ) into the newly added calculation domain is to
be calculated using the information from the previous
time step (explicit scheme) and the upstream (upwind
scheme). Therefore, the energy convection by resin ow,
the exothermal heat by resin curing and the energy of
the pre-heated ber mat are to be added to the energy of
the newly dened ow-front domain one time step later.
This can be written as
C
n
f
n1
T
n1
= f
n
C
n
tD
n
( )T
n
tf
n
E
n
S
t f
n1
f
n
_ _
E
F
(IW)
where the superscript n represents the nth time step.
The numerical simulation procedure is illustrated in
the ow chart (Fig. 4). Eqs. (12) and (13) include the
generation term which is a function of the temperature
and degree of cure. Therefore, Eqs. (9), (l2) and (13) are
all coupled and must be satised simultaneously. First,
the viscosity calculated based upon the temperature and
the degree of cure at the previous time step is used to
solve Eq. (9). Then the volumetric mass ux through
each control surface can be obtained. Eqs. (12) and (13)
are solved using this volumetric mass ux already
obtained and the resin viscosity is updated for the next
Fig. 3. The shape of the element used in numerical simulation. An
element is divided into four sub-volumes by the control surfaces.
Control volume i consists of sub-volumes from every element which
surrounds node i.
Fig. 2. Illustration of the ow front advancing technique. According
to the lled fraction f, the entire calculation domain can be divided
into three categories. In this gure, solid and dashed lines represent
element and control volume boundaries, respectively (see Fig. 3).
Fig. 4. Flow chart of the numerical simulation procedure.
964 S.T. Lim, W.I. Lee / Composites Science and Technology 60 (2000) 961975
time step. Then, the ow front advances until any con-
trol volume on the ow-front region is completely lled
using the volumetric mass ux and the calculation
domain is redened. In solving Eqs. (12) and (13), a
large truncation error might appear because the con-
vection term is dominant compared to the diusion
term. Also, because the Eqs. (12) and (13) are coupled,
some iteration is required to solve these two equations.
In order to settle this situation, the time step used for
the integration of Eqs. (12) and (13) should be smaller
than that for the ow-front advancement (see Fig. 4)
and hence Eqs. (12) and (13) could be linearized by an
explicit scheme. After the temperature and the degree of
cure at the current time step is calculated, the resin
viscosity is updated. This procedure is repeated until the
entire cavity is completely lled.
Resin viscosity depends on the temperature and the
degree of cure as [21]:
j = j
o
exp
E
RT
ko
_ _
(PH)
The cure model adopted in this study is as follows
[22]:
d[
dt
= k
1
k
2
[
m
1
( ) 1 [ ( )
m
2
(PI)
where [ represents the isothermal degree of cure and k
1
,
k
2
are expressed as
k
1
= A
1
exp
E
1
RT
_ _
. k
2
= A
2
exp
E
2
RT
_ _
(PP)
From the isothermal degree of cure [, the degree of
cure o can be obtained.
o =
H

[
H

= C
1
T C
2
T - T
g
( ). H

= H

T5T
g
( ) (PQ)
where H

is the isothermal heat of reaction which is


dened as the total amount of heat generated from time
t = 0 until no evidence is found of further reactions at a
constant temperature and can be obtained from iso-
thermal scanning measurement. The ultimate heat of
reaction H

is the amount of heat generated during


dynamic scanning till the completion of the chemical
reactions. The material properties used in this numerical
simulation were taken from the experimental results of
Kang et al. [7]. The resin was vinylester (825 and 280 by
National Korea, Inc. with the mixing ratio of 7:3). The
values of the parameters required for Eqs. (20)(23) are
shown in Table 1. Other properties of the resin and ber
reinforcement used in this study are listed in Table 2.
5. Verication of the numerical results
The validity of the numerical code was veried with
simple problems for which the exact solutions were known.
In order to verify the pressure eld, a steady resin ow
through a cube where resin was injected from the front
surface at constant pressure P
o
= 1 ( ) is considered. On
the other ve surfaces, the pressure was maintained at
zero. In this case, the momentum equation [Eq. (1)] and
the continuity equation [Eq. (2)] can be combined to
yield
Table 1
Curve t variables for the viscosity and cure model [see Eqs. (20)(23)]
[7]
Viscosity model Cure model
j

5.41910
5
Pa s A
1
1.248310
10
min
1
E
R
3636.45 K A
2
2.043310
11
min
1
k 26.89
E1
R
100048.4 K
E2
R
9505.58 K
m
1
0.693
m
2
1.327
T
C
100

C
C
1
0.02639
C
2
8.8466
Table 2
Properties of resin and ber mat used in RTM simulation [7]
Resin Fiber
,
r
1030 kg/m
3
,
f
2540 kg/m
3
c
pr
1900 J/kg K c
pf
835 J/kg K
k
r
0.193 W/m K k
f
0.76 W/m K
K
xx
110
9
m
2
K
yy
110
9
m
2
K
zz
110
9
m
2
Fig. 5. Three-dimensional pressure distribution along the centre line for
a cube. Comparison between the analytical solution and the numerical
results for three-dimensional steady ow. Pressure of unity is applied
on one surface while the pressure is kept zero on other surfaces.
S.T. Lim, W.I. Lee / Composites Science and Technology 60 (2000) 961975 965
V
K [ ]
j
VP
_ _
= 0 (PR)
If the permeability is isotropic and the viscosity is
constants Eq. (24) can be simplied as
V
2
P = 0 (PS)
Eq. (25) can be solved for a cube of unit size and the
solution can be found elsewhere as [23]:
P =
16

o
i=0

o
j=0
sinh 1 x ( ) sin 2i 1 ( )y [ ] sin 2j 1 ( )z [ ]
2i 1 ( ) 2j 1 ( ) sinh 1 ( )
_ _
(PT)
The exact solution was compared with the numerical
solution in Fig. 5 along the center line. Both solutions
agreed well, as shown in the gure.
In order to validate the ow front advancement
scheme, a one-dimensional advancement of the resin
front was considered. The resin was injected at one sur-
face of the cube and air was ventilated at the opposite
surface. In this case, the resin ow becomes one-dimen-
sional. The exact solution for the ow front location can
be obtained as [24]
x =

2KP
0
j
t
_
(PU)
Fig. 6. Verication of the ow front advancement scheme. One-
dimensional ow in the cartesian coordinate system.
Fig. 7. Principle of the optical ber sensor. Before the resin reaches
the sensing spot, light can be transmitted through the optical ber but
leaks as soon as the resin reaches the spot.
Fig. 8. Typical signal from the optical ber sensor. The intensity of
the infrared light decreases sharply as the resin front reaches the sen-
sing bare spots on an optical ber.
Table 3
Refractive indices of air, resin and glass for l = 632.8 nm
Material Refractive
index
Mismatch with
glass ber (%)
Air 1.000293 35.6
Water 1.3307 14.4
Polyester resin 1.5556 0.13
Glass ber (silica core) 1.55365 0
Fig. 9. Experimental setup for the three-dimensional RTM mold ll-
ing process.
966 S.T. Lim, W.I. Lee / Composites Science and Technology 60 (2000) 961975
where P
0
is the injection pressure, j is the viscosity, K is the
permeability, x is the location of the owfront and t is time.
In Fig. 6, the exact solution was compared with the numer-
ical result. As can be seen, the agreement is very close.
6. Experiments
In order to further verify the numerical results, experi-
ments were performed. There are several experimental
techniques to nd out the resin front location during
the lling process such as ow visualization techni-
ques [2527], the dielectric method [29], and a
method using thermocouples [30]. In these techniques,
sensors are mainly installed on the mold surface. As a
consequence, resin ow only along the mold wall can be
monitored and hence these methods are inappropriate
to nd out the ow front location inside the ber pre-
form. In this study, the optical ber was used to moni-
tor the ow of the resin inside the thick ber preform
[28].
6.1. Principle of the ber-optic sensor
Optical ber consists of core and cladding. The core is
the path of light and the cladding is the mechanical
protection from outside impact. In this study, optical
bers with silica core and polymer cladding were used.
First, a very shod section of polymer cladding was
removed from the optical ber by burning or chemical
etching. The length of the bare spot is as short as 1 mm.
Three or four bare spots were made consecutively along
a single ber, each bare spot serving as a sensor. The
optical ber thus prepared was positioned inside the
ber preform. An infrared light signal was transmitted
through the optical ber from one end and the intensity
of the light signal was monitored on the other end. Before
the resin reaches the bare spots along the optical-ber
sensor, a relatively large light signal can be transmitted
through the optical ber (see Fig. 7). However, as the
resin reaches the bare spots, the light leaks through
these bare spots because the refractive indices of the
resin and the silica are close (Table 3) [31]. Therefore,
the intensity of the transmitted light signal drops sig-
nicantly and the arrival of the resin front can be
detected. A typical signal from the sensor is demon-
strated in Fig. 8.
Fig. 10. The locations of sensing planes where optical ber sensors
were installed, between stacked ber preform inside the mold. The
inlet gate was placed at the centre of the bottom surface and the posi-
tions of the sensing spots are given in Table 4.
Fig. 11. Installation of the optical ber sensors on each plane shown
in Fig. 10. The positions of the sensing spots are given in Table 4.
Table 4
The locations of the optical ber sensors embedded inside the ber preform for the measurement of the three-dimensional ow front location (see
Figs. 911)
Sensor 1 Sensor 2 Sensor 3 Sensor 4
x (mm) y (mm) x (mm) y (mm) x (mm) y (mm) x (mm) y (mm)
1st plane (z = 0.0 mm) 23.3 0.0 0.0 23.3 23.3 0.0 0.0 23.3
46.6 0.0 0.0 46.6 46.6 0.0 0.0 46.6
70.0 0.0 0.0 70.0 70.0 0.0 0.0 70.0
2nd plane (z = 10.2 mm) 0.0 0.0 0.0 0.0
23.3 0.0 0.0 23.3
46.6 0.0 0.0 46.6
3rd plane (z = 19.6 mm) 0.0 0.0 0.0 0.0
23.3 0.0 0.0 23.3
46.6 0.0 0.0 46.6
4th plane (z = 29.0 mm) 0.0 0.0
S.T. Lim, W.I. Lee / Composites Science and Technology 60 (2000) 961975 967
6.2. Three-dimensional permeability measurement
The three-dimensional permeability must be known
for the three-dimensional RTM mold-lling analysis.
The three-dimensional permeability was measured using
the optical-ber sensors described above.
First, the optical-ber sensors were embedded in the
ber preform at designated locations (see Figs. 10 and
11). The optical ber used in this experiment was a
multimode ber (CeramOptec, HWF 200/230/500T)
which can transmit the light of a wavelength between
0.4 and 2.4 mm. The cubic mold cavity was closed and
then resin was injected from the inlet gate at the center
of the bottom surface. As the permeability is aniso-
tropic, the shape of the resin front is known to be ellip-
soidal [28]. As the resin reached the sensor point, the
signal from the photo detector changed sharply and the
time to the sensing point was recorded. Once the times for
the resin front to reach the specic locations are measured,
the three-dimensional permeability can be estimated
from curve-tting to the following equation [28,29].
Fig. 12. Typical signals from the optical ber sensors. The voltage
outputs are obtained from photo-detector sensors after proper signal
conditioning.
Fig. 13. Dimensionless front locations along x, y and z direction versus the modied time. Experimental data are curve-tted to a line to yield the
permeability values [see Eq. (27)].
968 S.T. Lim, W.I. Lee / Composites Science and Technology 60 (2000) 961975
1
3
,
3
fi

1
2
,
2
fi

1
6
=
i
i = x. y. z (PV)
where ,
fi
=
r
i
r
0i
and
i
= K
i
P
0
t
jcr
2
0i
. r
i
is the resin-front loca-
tion and r
0i
is the radius of the inlet gate. P
0
is the inlet
pressure and t is the time for the resin front to reach a
specic sensor location. j is the resin viscosity and c is
the porosity of the ber preform.
The experimental setup is shown in Fig. 9. A halogen
lamp was used for the infrared light source. A photo-
transistor (Opto Electronics, ST-1KLA) was used for
the detection of the intensity of the infrared light. The
voltage signal from the photo transistor was amplied
and then recorded by a data acquisition system
(Advantech, PCL-812PG). In this study, three bare
spots per optical ber were prepared. Fig. 10 illustrates
how the optical bers were placed inside the preform.
Ten glass-ber mats were stacked between each sensing
plane. A total of 54 chopped-strand mats were laid up.
In the rst sensing plane, four optical bers were placed
in the x and y directions. Thus, there were 12 sensing
points in the rst plane. In the second and third planes,
two optical bers were installed and, in the fourth plane,
only one optical ber was used with one bare spot.
Therefore, the number of sensing points was 25 in one
experiment. The sensing point locations are shown in
Table 4 and Figs. 10 and 11.
In order to measure the pressure at the inlet gate, a
pressure transducer (Sea Systems, Model C208) was
used. The box-shape mold cavity was constructed of 30
mm thick aluminum plates. The mold was designed to
change the cavity height so that the ber-volume frac-
tion can be changed. The resin was pressurized by
compressed nitrogen and the inlet-gate pressure was
controlled by a pressure regulator.
Automobile engine oil (LG Caltex, SAE 7.5W/30,
viscosity=0.29 Pa s at 25

C) was used for the perme-


ability measurement. The pressure at the inlet gate was
0.142 MPa. The ber reinforcement used in this study
was a glass-ber chopped-strand mat (LG Owens-
Corning, CM450-723, 450 g/m
2
). The ber-volume
fraction was 20.9%. The inlet-gate diameter was 1.70
mm. In the chopped-strand mat, the bers are randomly
oriented, but all in the same plane. Thus the in-plane
permeability can be assumed to be isotropic and the
out-of-plane principal permeability is perpendicular to
the plane of the bers [28].
Fig. 14. Locations of the sensor points. At these locations, results of
numerical simulation and the experimental data are compared in Table 5.
Fig. 15. Three-dimensional RTM mold lling pattern at dierent
times. Result of numerical simulation to be used for the comparison
with the experimental data (see Table 5).
Table 5
Comparison between numerical simulation and experimental results
by the optical ber sensor
a
Sensing points
number
Experiment
(s)
Simulation
(s)
Error
(s)
Error
(%)
Sensor 1 Sensor 2
1 2.5 2.5 2.6 0.1 4
2 21.9 23.5 23.3 0.21.4 0.96.4
3 76.1 83.3 79.8 3.53.7 4.24.9
4 3.8 4.7 3.3 0.51.4 13.229.8
5 31.3 32.6 29.3 2.03.3 6.410.1
6 96.1 93.6 100.6 4.57.0 4.77.5
7 1.6 1.6 0.7 0.9 56.3
8 6.9 5.9 1 14.5
9 33.5 29.7 3.8 11.3
10 10.3 7 3.3 32
11 45.1 36.4 8.7 19.3
12 10.3 9.7 9.5 0.20.8 2.17.8
13 18.8 18.6 0.2 1.1
14 45.1 47.7 2.6 5.8
15 20.1 20 0.1 0.5
16 64.8 55.4 9.4 14.5
17 34.4 35.1 0.7 2
a
The locations of the sensing points are shown in Fig. 14.
S.T. Lim, W.I. Lee / Composites Science and Technology 60 (2000) 961975 969
Fig. 16. Results of mold lling simulation for a cubic shape. The inlet gate was located at the centre of the front surface. Pressure distributions (b),
temperature distributions (c), degree of cure distributions (d) and lling time (e) are shown at three dierent cross-sectional planes along the thick-
ness as described in (a).
970 S.T. Lim, W.I. Lee / Composites Science and Technology 60 (2000) 961975
Fig. 17. Results of mold lling simulation for the thick slab with a pit and a projection. The inlet gate was located at the centre of the hollow pit.
Pressure distributions (b), temperature distributions (c), degree of cure distributions (d) and lling time (e) are shown at dierent cross-sectional
planes along the y and z directions as described in (a).
S.T. Lim, W.I. Lee / Composites Science and Technology 60 (2000) 961975 971
Typical output from the sensor during the experiments
is shown in Fig. 12. From the voltage-output data, the
time of resin-front arrival at each sensing point can be
monitored. Based on Eq. (23), the three-dimensional
permeability was obtained (see Fig. 13). As was expec-
ted, k
xx
is identical to k
yy
, assuring that the in-plane
permeability is isotropic.
6.3. Comparison between experimental and numerical
results
It is noted that once the resin front reaches the mold
boundary, the exact solution as used in the permeability
measurement is not valid. Using the measured perme-
ability, numerical simulation was performed for the
same geometry described above. Sensor location are
shown in Fig. 14. Only a quarter is displayed in this
gure because of the symmetry of the geometry. The
sensing points 16, 7, 12 have extra experimental results
because there were two sensors at the same location.
The numerical results are displayed in Fig. 15. Com-
parisons between the two results are shown in Table 5.
The absolute error in lling time was within 9.4 s where
the time to ll the entire cavity was 100.6 s. Compara-
tively close agreements are found.
7. Practical applications of the computer code
Numerical calculations were done for four dierent
geometries to demonstrate the eectiveness of the com-
puter code developed in this study. First, a simple cubic-
shaped preform (10 cm10 cm10 cm) was lled with
an injection gate at the center of the front surface. The
total numbers of nodes and elements were 1331 and
5000, respectively. The injection pressure was 1.0 MPa
and the injection temperature was 25

C. The preform
was preheated to 70

C. The locations of the air vents


were expected to be at the four vertices of the opposite
surface as these were the farthest points from the gate.
However, the results showed that the vent hole was
required at the center of the opposite surface. This was
because the heat transferred from the heated mold
lowered the resin viscosity near the wall and thus made
Fig. 18. Results of mold lling simulation for automobile head lamp bezel. Resin was injected at two points. Pressure distribution (a) and lling time
(b) are shown along the boundary surface.
972 S.T. Lim, W.I. Lee / Composites Science and Technology 60 (2000) 961975
the ow along the wall easier [see Fig. 16(e)]. The tem-
perature distribution is illustrated in Fig. 16(c). As can
be seen, the temperature near the wall is higher.
The second geometry considered was a thick slab (20
cm10 cm4 cm) with a pit (6 cm4 cm2 cm) and a
projection (6 cm6 cm2 cm) as illustrated in Fig. 17a.
The number of discretized cubic nite elements was
4120. Numerical results are shown in Fig. 17(b)(e). The
resin was injected at the location where the thickness
is the least and the inlet pressure was 0.7 MPa. The
temperature of the injected resin was 25

C and rose
immediately to the mold temperature 70

C after
injection, as injection was done at the thinnest part of
the cavity and the heat from the hot mold wall was
transferred well to the resin [see Figure 17(c)]. This
means that the resin viscosity was lowered and the ll-
ing time was shortened.
To check the eectiveness of this computer code fur-
ther, an automobile headlamp bezel was considered. In
this example, resin was injected at two points. As shown
in Fig. 18(b), the locations of the ``weld lines'' as well as
the air vents could be predicted using the code devel-
oped in this study. Next, the insert was replaced with a
thin hollow cavity without ber preform to allow easy
resin ow. Therefore, in place of a hole at the insert, a
thin resin membrane was formed after the molding. The
thin cavity allowed the ow resistance to be lowered,
resulting in a faster lling time compared with the case
with a hole [see Fig. 19(b)]. In the two example cases,
the inlet gate pressure was 0.7 MPa, the inlet tempera-
ture was also 25

C and the wall temperature was 70

C.
The last numerical example was molding of a
centrifugal-pump casing (Fig. 20). This complicated
geometry was discretized into 2035 nodes and 7874 ele-
ments. The resin was injected from the inlet surface
(surface gating) at 0.6 MPa. The temperature of the
resin and mold was 70

C. The marks in Fig. 20(b) indi-


cate the expected air-vent locations where two ow
fronts met each other. These simulation results can be
used for the design of the RTM mold.
Fig. 19. Results of mold lling simulation for automobile headlamp bezel with membrane. Resin was injected at the centre of the hollow membrane.
Pressure distribution (a) and lling time (b) are shown along the boundary surface.
S.T. Lim, W.I. Lee / Composites Science and Technology 60 (2000) 961975 973
8. Conclusions
A numerical code for the RTM process was developed.
Three-dimensional analyses of resin ow were performed
including the non-isothermal eect and conversion dis-
tribution. In the simulation of the RTMprocess for thicker
parts, the necessity of three-dimensional analysis was illu-
strated by sample runs. Using optical-ber sensors, the
ow-front location was monitored and the three-dimen-
sional permeability was measured. Numerical and experi-
mental results were also compared. Close agreements were
found. Numerical simulations were done for some practical
cases to illustrate the eectiveness of the numerical code.
Acknowledgements
This work was supported by the Turbo and Power
Machinery Research Center and the Ministry of Science
and Technology.
References
[1] Coulter JP, Gu c eri SI. Resin impregnation during the manu-
facturing of composite materials. CCM report no. 88-07, Uni-
versity of Delaware, 1988.
[2] Chan AW, Hwang ST. Modeling of impregnation process during
resin transfer molding. Polym Eng Sci 1991;31:114956.
[3] Um MK, Lee WI. A study on mold lling process in resin trans-
fer molding. Polym Eng Sci 1991;31:76571.
[4] Yoo YE, Lee WI. Numerical simulation of the resin transfer
mold lling process using the boundary element method. Polym
Compos 1996;17:36874.
[5] Bruschke MV, Advani SG. Filling simulation of complex three
dimensional shell-like structures. SAMPE Quart 1991; October;
211.
[6] Lin R, Lee LJ, Liou M. Non-isothermal mold lling and curing
simulation in thin cavities with preplaced ber mats. Intern
Polymer Processing VI 1991;4:35669.
[7] Kang MK. Mold lling and curing simulation in non-isothermal
resin transfer molding. Master's thesis of mechanical engineering,
Seoul National University, 1993.
[8] Young WB. 3-Dimensional nonisothermal mold lling simula-
tions in resin transfer molding. Polym Compos 1994;15:118
27.
Fig. 20. Filling time distribution for a centrifugal pump casing. The resin was injected at the rear surface. The dot marks represent the expected vent
locations.
974 S.T. Lim, W.I. Lee / Composites Science and Technology 60 (2000) 961975
[9] Phelan FR. Simulation of the injection process in resin transfer
molding. Polym Compos 1997;18:46076.
[10] Varma RR, Advani SG. Three-dimensional simulations of lling
in resin transfer molding. Advances in nite element analysis in
uid dynamics (ASME), 1994;200:217.
[11] Lam YC, Joshi SC, Liu XL. Application of a general-purpose
nite element package for numerical modelling of resin ow
through brous media. International Conference on Advances in
Materials and Processing Technologies (AMPT-4) Proceedings,
August 1998. p. 92633.
[12] Dullien FAL. Porous media uid transport and pore structure.
Academic Press, 1979. p. 1579.
[13] Kamal MR, Sourour S. Kinetics and thermal characterization of
thermoset resin. Polym Eng Sci 1973;13:59.
[14] Baliga BR, Patankar SV. Handbook of numerical heat transfer.
John Wiley & Sons, Inc., 1988. p. 42161.
[15] Hirt CW, Nicholas BD. Volume of uid (VOF) method for the
dynamics of free boundaries. J Computational Physics,
1981;39:20125.
[16] Broyer E, Tadmor Z, Gutnger C. Israel J Technology
1973;11:189.
[17] Tadmor Z, Broyer E, Gutnger C. Polym Eng Sci 1974;14:660.
[18] Broyer E, Gutnger C, Tadmor Z. Trans Soc Rheol 1975;19:423.
[19] Tucker CL. Fundamentals of computer modeling for polymer
processing. New York: Hanser Publishers, Oxford University
Press, 1989.
[20] Ferziger JH. Numerical methods for engineering application.
Wiley-Interscience, 1981.
[21] Lee WI, Loos AC, Springer GS. Heat of reaction, degree of cure
and viscosity of Hercules 3501-6 resin. J Compos Mater
1982;16:51020.
[22] Dusi MR, Lee WI, Ciriscioli PR, Springer GS. Cure kinetics and
viscosity of Fiberite 976 resin. J Compos Mater 1987;21:24361.
[23] Carslaw HS, Jaeger JC. Conduction of heat in solids. Oxford
University Press, 1959.
[24] Cai Z. Analysis of mold lling in RTM process. J Compos Mater
1992;26:131038.
[25] Adams KL, Miller B, Rebenfeld L. Forced in-plane ow of epoxy
resin in brous networks. Polym Eng Sci 1986;26:143441.
[26] Adams KL, Rebenfeld L. Permeability characteristics of multi-
layer ber reinforcements. Part I: experimental observations.
Polym Compos 1991;12:17985.
[27] Chan AW, Hwang ST. Anisotropic in-plane permeability of fab-
ric media. Polym Eng Sci 1991;31:122339.
[28] Ahn SH, Lee WI, Springer GS. Measurements of the three
dimensional permeability of ber preforms using embedded ber
optic sensors. J Compos Mater 1995;29:71433.
[29] Aklonis JJ, MacKnight WJ. Introduction to polymer viscoelas-
tieity. John Wiley & Sons, Inc., 1983.
[30] Beckwith TG, Marangoni RD, Lienhard VJH. Mechanical mea-
surements. Addison-Wesley Publishing Company, 1995.
[31] Hecht E. Optics. Addison-Wesley Publishing Company, 1987.
S.T. Lim, W.I. Lee / Composites Science and Technology 60 (2000) 961975 975

Você também pode gostar