Vous êtes sur la page 1sur 17

Curing simulation by autoclave resin infusion

D.C. Blest
a
, S. McKee
b,
*, A.K. Zulkie
b
, P. Marshall
c
a
School of Mathematics and Physics, University of Tasmania, Box 1214, Launceston, Tasmania, 7250 Australia
b
Department of Mathematics, University of Strathclyde, Livingstone Tower, 26 Richmond Street, Glasgow G1 IXH, UK
c
British Aerospace, Sowerby Research Centre, Bristol, UK
Received 31 July 1997; received in revised form 16 April 1999; accepted 19 July 1999
Abstract
This paper deals with the modelling and simulation of resin ow, heat transfer and the curing of a multilayer thermoset compo-
site by the resin lm infusion process. For approximately isothermal ows, the model is based on Darcy's Law and Stoke's equa-
tions where a similarity solution is obtained and subsequently used in a two-dimensional convection-diusion heat equation
coupled with a rate of cure equation. A nite dierence scheme is applied to the energy equation on a moving grid and simulations
for varying laminate thicknesses and number of plies are performed. # 1999 Elsevier Science Ltd. All rights reserved.
Keywords: Heat transfer; Thermoset composites; Resin infusion; Computer model
1. Introduction
Resin lm infusion is an alternative method to the
prepreg technique of fabricating composite materials.
The lay-up of the infusion process is similar to the pre-
preg process where dry bre plies, interspersed with
resin, are enclosed in a vacuum bag and placed in an
autoclave for curing to take place. Heat and pressure are
simultaneously applied to the set-up; this initiates an
exothermic chemical reaction and forces the resin to
infuse through the dry bres. As the resin completely
impregnates the dry bres and consolidation takes place,
it gels and solidies into the desired nished product.
Several investigators have studied and proposed
models for the RFI process used in composites manu-
facturing. Resin ow through the dry bres is con-
ventionally modelled as an unsaturated ow through
porous media, where Darcy's Law is employed. The
determination of the exact location of the ow front is
an important feature in the analysis. When high pres-
sure gradients are applied, it is necessary to treat the
bre layers as deformable, and a recent study of this
uidsstructure interaction problem has been treated by
Ambrosi and Preziosi [22]. Another important issue is
the elimination of trapped air where minimizing the
presence of voids can signicantly improve the quality
of the composite. As in the prepreg process, proper
selection of the cure cycle of the autoclave is important
in order to produce a high quality part where full resin
wet-out and complete, uniform curing are achieved at
the end of the process.
For steady ow through a wall-bounded porous
medium, Givler and Altobelli [9] have used the Brink-
manForchheimer equation to model the ow. They
noted that when a porous ow domain contains an
interface, the Brinkman term (j
e
\
2
v) is an important
determinant in predicting the development of boundary
layers which emanate from the interface. However,
inaccurate determination of j
e
, the eective viscosity,
has in the past tended to hamper the usage of this term.
Nassehi and Petera [10] presented a nite-element
method for linking the NavierStokes and Darcy equa-
tions along a porous inner boundary in a ow regime
governed by both of these equations. Ni et al. [20]
investigated a two-regional ow introducing an equiva-
lent permeability parameter for the ow in the bre free
region. A comprehensive study of single uid ow in
porous media with application to cylindrical beds of
brous mats has been presented by Liu and Masliyah
[15]. A discussion of ow in porous media with appli-
cation to composite processing has also been given by
Tucker and Dessenger (see Ref. [7]).
Both nite-element and nite-dierence approaches
have been employed to solve the macroscopic and
microscopic models. Coulter and Gu ceri [2] developed
a numerical code, TGIMP, for computing a two-
dimensional Darcy isothermal resin ow based on a
nite-dierence method using boundary-tted coordi-
nates with numerical grid generation. They determined
0266-3538/99/$ - see front matter # 1999 Elsevier Science Ltd. All rights reserved.
PI I : S0266- 3538( 99) 00084- 6
Composites Science and Technology 59 (1999) 22972313
* Corresponding author.
the resin front by relocating the computational nodes
using the resultant surface velocities and a pre-
determined time increment. Coulter and Gu ceri [3] also
performed physical experiments and found reasonable
agreement with results predicted by the TGIMP code.
Ahn et al. [4] proposed a model for the RFI vacuum
process which predicts the nal resin content of the
laminate and the desired autoclave processing cycle
required to eliminate voids. They dened a dimension-
less parameter, the Infusion Flow Number, which
describes the degree of impregnation of a preform based
on the total pressure, resin viscosity, resin velocity, pre-
form permeability and thickness of bre preform. They
controlled the resin content in the nal composite by
changing the initial amounts of resin in the bleeder and
they detected no voids using this process. This com-
pared well with the prepreg process which displayed
voids due to air pockets trapped in between the prepreg
layers during the lay-up. Bruschke and Advani [8] pre-
sented a non-isothermal viscous ow model, using a
nite element control volume method, to predict the free
surface of a shear-thinning resin injected through a bre
preform on the in-plane direction of a thin part.
Wymer and Engel [11] developed a numerical model to
study the ow of a thermoset resin through, and parallel
to, a heated unidirectional bre array, with temperature-
dependent viscosity. They considered a micro-model of
the nonisothermal incompressible ow of the resin,
employing the CrankNicolson nite dierence scheme
on the steady one-dimensional convective energy equa-
tion of the RTM process. Kang et al. [12] presented
numerical and experimental studies of the simulation of
resin transfer moulding (both non-isothermal mould
lling and curing) using a nite-element control-volume
technique. Malkin et al. [13] proposed a model for the
impregnation of liquid above a porous layer and applied
the model to moulding of low viscosity resin. Sadiq et
al. [14] investigated experimentally the transverse ow
through aligned cylinders and provided data for the
progress of the ow front and the formation of voids
through a heterogeneous bre bed. Loos and MacRae
[16] developed an analytical model to simulate the non-
isothermal inltration of resin in the resin lm infusion
process for manufacturing a blade-stiened panel. Their
model predicts the temperature, resin viscosity, and extent
of cure during inltration of an anisotropic bre preform
using a nite element/control volume technique.
Williams et al. [1] presented a comprehensive review
of the `resin infusion under exible tooling' process
(RIFT) which is a variant of the vacuum-driven RTM
in which one of the solid tool faces is replaced by a
exible polymeric lm. This process potentially is a safer
and more economical method of production, where
resin is drawn into the dry reinforcement in an evac-
uated vacuum bagged tool using only a partial vacuum
to drive the resin. Mogavero and Advani [19] performed
ow experiments through preforms composed of multi-
ple layers of reinforcement material, and investigated
the eect of varying the order of the lay-up of a xed
number of plies and the impact of varying the thickness
of individual layers of a thick preform. They found that
the weighted average scheme provided a reasonable esti-
mate for the eectiveness of the preforms. Yu and Young
[21] proposed an RTM simulation model integrated with
genetic algorithms to search for the process parameters
(the mould heating rate, mould temperature, resin lling
and curing temperatures) that could reduce cycle time
and enhance the uniformity of the nal product.
Ambrosi and Preziosi [22] proposed a model of resin
ow under isothermal conditions for an injection
moulding process. Their model allows for deformation
of the reinforcing network of the dry and the wetted
part of the preform generated by the inltration process.
Pillai and Advani [23] performed simulations of unsa-
turated ow of resin in woven and stitched bre mats
used in RTM using an adaptation of the Finite Ele-
ment/Control Volume (FEM/CV) technique. A dual-
scale porous medium was modelled and the inlet pres-
sures, inlet ll times, and mat saturation were studied.
Lekakou and Bader [24] proposed a macro- and a
micro-inltration model based Darcy's law incorporat-
ing mechanical, capillary and vacuum pressures. Pearce
et al. [25] conducted experiments to investigate ow
behaviour inside moulds where ow fronts converged
due to multiport injection in RTM processing. They
found that when ow fronts meet at a mould edge, they
merge and act as a single front and when the ow fronts
meet head-on, voids can be formed. Abraham and
McIlhagger [26] investigated experimentally the eect of
a gating arrangement and the use of vacuum and posi-
tive pressure transfer to maximize fabric wet-out and
minimize void content.
In our work, a similarity solution is developed for the
approximately isothermal ow of the resin, where Stokes'
slow ow equations are used in the bre-free regions and
Darcy's Law in the saturated bre regions. A nite dier-
ence scheme is then employed in the convection-diusion
heat equation which includes the heat source term from
the exothermic reactions to calculate the temperature and
cure proles of the laminate during the process.
2. Modelling the resin ow
The schematic diagram for the set-up, with the chosen
coordinate axis, is shown in Fig. 1. A one-dimensional
model is developed where the resin is assumed to be
Newtonian with constant viscosity and the ow is in the
direction of the applied vertical force normal to the dry
bre plies. The ow through the dry bre plies is con-
sidered as ow through a porous medium where Darcy's
law is employed.
2298 D.C. Blest et al. / Composites Science and Technology 59 (1999) 22972313
Initially there are 2n+1 layers, n dry bre plies inter-
spersed with n+1 resin layers where the resin is considered
to be in liquid form. A typical k dry bre ply, which is
considered as non-deformable and of constant thickness
d=h
2k
h
2k1
, is therefore the 2k layer of the lay-up, for
k=1, 2, . . ., n.
As external pressure is applied at the top plate,
y=h
2n+1
(t), the laminate is compressed, resulting in the
ow of the resin into the dry bre plies, thus wetting the
bres and consolidating the laminate while heat is
simultaneously applied in order to cure the resin into a
solid composite part.
The velocity and pressure in the resin region [L,
L][h
2k
, h
2k+1
] are denoted by v
r
2k1
and p
r
2k1
, while in
the lower wetted bre region [L, L][h
2k1
, o
2k1
] the
velocity and pressure are denoted by v
w.l
2k1
and p
w.l
2k1
and
in the upper wetted bre region [L, L][o
2k
, h
2k
] the
velocity and pressure are denoted by v
w.u
2k
and p
w.u
2k
for
k=1, 2,. . ., n.
In a typical (2k1) bre-free layer, k=1, 2,. . ., n+1,
indicated by the superscript r, resin is assumed to ow
only in the y-direction. Hence, the incompressiblity
condition can be written as
ov
r
2k1
oy
= 0 (1)
and the momentum equation in the x direction yields
op
r
2k1
ox
= 0 (2)
Given the high viscosity of the resin and considering a
quasi-steady-state assumption, the steady Stokes' equation
is employed to model the ow in the y-direction, so that

o
2
v
r
2k1
ox
2
=
op
r
2k1
oy
(3)
In the dry bre plies, as the resin can ow from above
or below each ply, an upper and lower wetted region is
identied for the pressure and velocity of each ply. The
wetted regions of the 2k layer are dened by
h
2k1
t ( )4y4o
l
2k1
t ( ) and o
u
2k
t ( )4y4h
2k
t ( ). (4)
k = 1. 2. . . . . n
where o
2k1
l
(t) and o
2k
u
(t) are the lower and upper free
surfaces respectively. The ow in the wetted regions,
denoted by the superscript w, can be described by the
continuity equation and the momentum equations in
the x and y directions of an amended Darcy's law, tak-
ing account of the vertical motion of the dry bre plies,
as
ov
w.i
2k
oy
= 0 (5)
op
w.i
2k
ox
= 0 (6)
v
w.i
2k
h
.
2k
t ( ) =
k
j
op
w.i
2k
oy
(7)
for k=1, 2,. . ., n where i=l, u denotes the lower or
upper wetted region respectively, h
.
2k
(t) is the velocity of
each dry bre ply, k is the permeability of the bre ply
and j is the dynamic viscosity of the resin. In this work,
both the parameters k and j are taken to be constant.
The boundary conditions for the velocities in the
above set-up are
. At the top and bottom plate, y=h
2n+1
(t) and y=0
respectively, the kinematic condition is imposed.
Thus
v
r
2n1
x. h
2n1
t ( ). t ( ) = h
.
2n1
t ( ) (8)
v
r
1
x. 0. t ( ) = 0 (9)
. At the resin-wet bre interfaces, a continuity con-
dition is imposed whereby
v
r
2k1
x. h
2k1
t ( ). t ( ) = v
w.l
2k
x. h
2k1
t ( ). t ( ). (10)
v
r
2k
x. h
2k
t ( ). t ( ) = v
w.u
2k
x. h
2k
t ( ). t ( ). (11)
Fig. 1. Set-up for n dry bre plies.
D.C. Blest et al. / Composites Science and Technology 59 (1999) 22972313 2299
for k=1,2,. . .,n
. At the free surfaces y=o
k
(t), k=1, 2,. . ., 2n, the
kinematic condition is applied whereby
v
w.l
2k
h
.
2k1
t ( ) = o
.
2k1
t ( ) h
.
2k1
t ( )
_ _
. (12)
v
w.u
2k
h
.
2k
t ( ) = o
.
2k
t ( ) h
.
2k
t ( )
_ _
(13)
for k=1, 2,. . . n, where is the porosity of the bre
ply.
The boundary conditions for the pressure are:
. At the top bounding plate, the external force F
a
acts
normal to the plate at y=h
2n+1
(t) in the negative y-
direction. Thus, applying a force balance over the
plate of length 2L gives the pressure condition as
p
r
2n1
x. h
2n1
t ( ). t ( ) =
F
a
2L
(14)
where L is the half-length of the ply.
. At the resin-wet bre interfaces y=h
k
(t), k=1,
2,. . ., 2n, the continuity condition is imposed
whereby
p
r
2k1
x. h
2k1
t ( ). t ( ) = p
w.l
2k
x. h
2k1
t ( ). t ( ) (15)
p
r
2k
x. h
2k
t ( ). t ( ) = p
w.u
2k
x. h
2k
t ( ). t ( ). (16)
for k=1, 2,. . ., n.
. At the free surfaces y=o
k
(t), k=1, 2,. . ., 2n, the
pressures are taken to be zero. Thus
p
w.l
2k
x. o
2k1
t ( ). t ( ) = 0. (17)
p
w.u
2k
x. o
2k
t ( ). t ( ) = 0. (18)
for k=1, 2,. . ., n.
A similarity solution is sought for the pressure and
velocity in both the bre-free and the wetted bre or
saturated regions (see Blest[28]). In the bre-free layers,
the pressure is in terms of t only (see Appendix A)
p
r
1
t ( ) = p
r
3
t ( ) = . . . = p
r
2n1
t ( ) =
F
a
2L
(19)
In a typical dry bre ply of the 2k layer, the pressure in
the upper wetted region is
p
w.u
2k
y. t ( ) =
j
k
h
.
2k
o
.
2k
_ _
y o
2k
( ) (20)
and in the (2k+1) resin layer is
p
r
2k1
t ( ) = p
w.u
2k
h
2k
t ( ). t ( )
=
j
k
h
.
2k
t ( ) o
2k
t ( )
_ _
h
2k
t ( ) o
2k
t ( ) ( ) (21)
In the upper wetted region of the 2k dry bre ply,
imposing the free-surface kinematic condition gives
v
w.u
2k
t ( ) = h
.
2k
t ( ) o
.
2k
t ( ) h
.
2k
(t)
_ _
(22)
At the bottom resin layer, imposing the kinematic
boundary condition at y=0 yields
v
r
1
t ( ) = 0.
The pressure in the rst resin layer is
p
r
1
t ( ) =
j
k
h
.
1
t ( ) h
1
t ( ) h
1.0
_ _
. (23)
The position of the bottom bre ply is
h
1
t ( ) = h
1.0

kF
a
t
Lj
_
. (24)
In the lowermost bre ply, the free surface of the lower
wetted region is
o
1
t ( ) =
1

h
1
t ( )
1

h
1.0
. (25)
and eliminating h
1
(t) in (25) using (24) yields
o
1
t ( ) = h
1.0
1
1

_ _

kF
a
t
Lj
_
. (26)
At the top resin layer, the velocity is
v
r
2n1
t ( ) = h
.
2n1
t ( ) (27)
and from (21) with k=n, the pressure is
p
r
2n1
t ( ) =
j
k
h
2n
t ( ) o
2n
t ( ) ( ) h
.
2n1
t ( ) h
.
2n
t ( )
_ _
. (28)
In terms of its initial position, the position of the top
plate is
h
2n1
t ( ) = h
2n1.0
2n 1 ( )

kF
a
t
Lj
_
. (29)
From the initial condition o
2k
(0)=h
2k
(0)=h
2k,0
, the
position of the free surfaces of the upper wetted regions
in the 2k bre ply can be found to be (see Appendix A)
2300 D.C. Blest et al. / Composites Science and Technology 59 (1999) 22972313
o
2k
t ( ) = h
2k
t ( )

kF
a
t
Lj
_
. k = 1. 2. . . . . n. (30)
The position of the free surfaces in the lower wetted
regions in the 2k bre plies are
o
2k1
t ( ) = h
2k1
t ( )

kF
a
t
Lj
_
. k = 0. 1 . . . . n 1. (31)
Hence, the heights of the resin-bre interfaces and the
free-surface ow fronts for a typical k layer, are
h
2k1
= h
2k1.0
2k 1 ( )

kF
a
t
Lj
_
. k = 0. 1. . . . . n (32)
h
2k
= h
2k1
d. k = 1. 2. . . . . n (33)
o
2k1
= h
2k1.0
2k 1
1

_ _

kF
a
t
Lj
_
.
k = 0. 1. . . . . n 1
(34)
o
2k
= h
2k.0
2k 1
1

_ _

kF
a
t
Lj
_
.
k = 1. 2. . . . . n
(35)
where we recall that n is the total number of bre layers
and d is the constant thickness of a bre layer.
Thus, dierentiating (32)(35), we obtain the velocities
of both the bre layers and the free-surface fronts:
h
.
2k1
= k
1
2
_ _

kF
a
Ljt
_
. k = 0. 1. . . . . n (36)
h
.
2k
= h
.
2k1
. k = 1. 2. . . . . n (37)
o
.
2k1
=
1
2
2k 1
1

_ _

kF
a
Ljt
_
. k = 0. 1. . . . . n (38)
o
.
2k
=
1
2
2k 1
1

_ _

kF
a
Ljt
_
. k = 1. 2. . . . . n (39)
3. Heat transfer model
The heat applied in the autoclave initiates an exo-
thermic heat reaction of the resin. As the resin ows
into the dry bre layers there are three dierent regions
through which the heat transfer process will occur,
namely the bre-free resin regions, the wetted bre
regions and the dry bre regions.
In the bre-free resin layers, the two-dimensional
convection-diusion heat equation can be expressed as
oT
r
ot
v
r
oT
r
oy
= K
r
o
2
T
r
ox
2

o
2
T
r
oy
2
_ _

H
R
c
r
oo
ot
. (40)
while in the wetted bre layers, which is considered as
saturated porous layers, the heat transfer is described by
oT
w
ot
v
w
oT
w
oy
= K
w
o
2
T
w
ox
2

o
2
T
w
oy
2
_ _

H
R
c
w
oo
ot
(41)
and in the dry bre layers, in the absence of resins, only
the conduction mode occurs, and hence
oT
f
ot
= K
f
o
2
T
f
ox
2

o
2
T
f
oy
2
_ _
(42)
where T
i
=T(x,y,t), i=r,w,f denote the temperature, v
i
,
i=r,w denote the vertical velocity ow component,
o x. Y. t ( ) is the degree of cure of the resin and the suf-
ces r,w and f denote the resin, wetted bre and dry
bre layers respectively. K
i
, i=r,w,f are the thermal dif-
fusivities, c
i
, i=r,w,f are the thermal specic heats, is
the voidage of the bre layer and H
R
is the heat of
reaction of the resin.
The thermal diusivities are dened by
K
i
=
k
i
,
i
c
i
for i = r. w. f (43)
where k
i
are the thermal conductivities and ,
i
are the
densities of the respective regions.
In the wetted bre regions, the values of the thermal
conductivity and heat capacity are expressed by the rule
of mixtures [31]
k
w
= k
r
1 ( )k
f
. (44)
c
w
,
w
= c
r
,
r
1 ( )c
f
,
f
. (45)
where the suces r and f again denote the resin and dry
bre properties respectively. The expressions for the rate
of degree of cure for the thermoset resin are obtained
from Lee et al. [27]; they are
oo
ot
=
a
1
a
2
o ( ) 1 o ( ) 0.47 o ( ) o40.3
a
3
1 o ( ) o > 0.3
_
(46)
where the temperature-dependent a
i
, i=1, 2, 3 are
a
i
= A
1
exp
E
i
RT
_ _
. i = 1. 2. 3.
D.C. Blest et al. / Composites Science and Technology 59 (1999) 22972313 2301
A
i
are specied pre-exponential factors, E
i
are the
known activation energies, and R is the universal gas
constant. The degree of cure takes the value zero initi-
ally and the value 1 when fully cured.
3.1. Boundary and initial conditions
The initial temperature, T
ini
, is assumed to be uniform
across the dierent regions and also assumed known,
T
r
x. y. 0 ( ) = T
w
x. y. 0 ( ) = T
f
x. y. 0 ( ) = T
ini
(47)
The prescribed autoclave temperature, T
a
, is assumed
known and is imposed at the top and bottom bounding
plates of the laminates at y=h
2n+1
(t) and y=0 respec-
tively. Thus
T
r
x. 0. t ( ) = T
r
x. h
2n1
t ( ). t ( ) = T
a
. (48)
The laminates are also assumed to be insulated at the
ends x=L; thus
oT
r
ox
L. y. t ( ) =
oT
w
ox
L. y. t ( ) =
oT
f
ox
L. y. t ( ) = 0.
(49)
oT
r
ox
L. y. t ( ) =
oT
w
ox
L. y. t ( ) =
oT
f
ox
L. y. t ( ) = 0. (50)
Continuity conditions for the temperature and ux
are imposed at the interface of the saturated bre and
the bre-free layers:
T
r
x. h
k
t ( ). t ( ) = T
w
x. h
k
t ( ). t ( ).
k
r
oT
r
oy
x. h
k
t ( ). t ( ) = k
w
oT
w
oy
x. h
k
t ( ). t ( )
_
k = 1. 2. . . . . 2n
and at the free surfaces:
T
w
x. o
k
t ( ). t ( ) = T
f
x. o
k
t ( ). t ( ).
k
w
oT
w
oy
x. o
k
t ( ). t ( ) = k
f
oT
f
oy
x. o
k
t ( ). t ( )
_
k = 1. 2. . . . . 2n.
4. Solution techniques
The following non-dimensionalised variables are intro-
duced to facilitate the numerical solution procedure:
x~ =
x
d
. y~ =
y
d
. v~
i
=
v
i
V
. T
~
i
=
T
i
T
ini
_ _
T
a
T
ini
( )
and t
~
=
tK
r
d
2
for i=r,w,f, where d is the thickness of a dry bre layer
and V is the velocity of the uppermost plate, h

2n+1
(t).
In the bre-free resin layers, the non-dimensional heat
equation becomes
oT
r
ot
Pev
r
oT
r
oy
=
o
2
T
r
ox
2

o
2
T
r
oy
2
_ _
J
r
oo
ot
(51)
while in the wetted bre layers, Eq. (41) transforms to
oT
w
ot
Pev
w
oT
w
oy
= D
w
o
2
T
w
ox
2

o
2
T
w
oy
2
_ _
J
w
oo
ot
(52)
and in the dry bre layers, Eq. (42) manifests itself as
oT
f
ot
= D
f
o
2
T
f
ox
2

o
2
T
f
oy
2
_ _
(53)
Note that the tilde has been omitted. The dimensionless
numbers are
Pe =
Vd
K
r
. (54)
J
r
=
H
R
c
r
T
a
T
ini
( )
. (55)
J
w
=
,
r
H
R
c
w
,
w
T
a
T
ini
( )
. (56)
D
f
=
K
f
K
r
. (57)
D
w
=
K
w
K
r
. (58)
The degree of cure Eq. (46) becomes
oo
ot
=
_
C
1
C
2
o ( ) (1 o) 0.47 o ( ) o40.3.
C
3
1 o ( ) o > 0.3
(59)
where
C
i
=
d
2
a
i
K
r
. i = 1. 2. 3. (60)
The solutions to the coupled, non-dimensionalised
temperature and degree of cure Eqs. (51), (52), (53) and
(59) are determined in a domain which changes with time.
The modied Euler method is used to discretize the
rst-order degree of cure Eq. (59). The predictor-cor-
rector method can be expressed as
2302 D.C. Blest et al. / Composites Science and Technology 59 (1999) 22972313
o
+
n1
= o
n
tf o
n
( ) predictor (61)
o
n1
= o
n

1
2
t f o
n
( ) f o
+
n1
_ _ _ _
corrector (62)
The alternating direction implicit (ADI) nite dier-
ence scheme with the appropriate upwinding, proposed
by McKee et al. [32], is employed to discretize the
energy Eqs. (51), (52), (53). The ADI method in split
form is given by
1
ra
2
_ _
T
l
( )
m1
+
i.j
= 1 sao
2
y
qv
l
PeV
y

ra
2
o
2
x
_ _
T
l
( )
m
i.j
tJ
l
g o
m
i.j
. T
l
( )
m
m.j
_ _
(63)
1
sa
2

qv
l
Pe
2
V
y
_ _
T
l
( )
m1
i.j
= T
l
( )
m1
+
i.j

sa
2
o
2
y

qv
l
Pe
2
V
y
_ _
T
l
( )
m
i.j
(64)
where (T
l
)
i,j
m
, l=r,w,f represents the approximate tem-
perature in the respective l layer of resin, wet bre or
dry bre at the mth time level for the node (i,j), where
x
i
=x
0
+ix. i N and y
j
=y
0
+jy. j N.
The dierence notations used are
V
y
T
l
( )
m
i.j
= T
l
( )
m
i.j
T
l
( )
m
i.j1
o
2
x
T
l
( )
m
i.j
= T
l
( )
m
i1.j
2 T
l
( )
m
i.j
T
l
( )
m
i1.j
o
2
y
T
l
( )
m
i.j
= T
l
( )
m
i.j
T
l
( )
m
i.j
T
l
( )
m
i.j1
and
a =
_
D
w
for the wetted fibre region
D
f
for the dry fibre region
1 for the resin layer
p, q, r and s are dened as
p = t,x. q = t,y. r = t,x
2
and
s = t,y
2
(65)
and g(o
m
i.j
. T
l
( )
m
i.j
) is the exothermic reaction function
from (59).
5. Numerical results
The simulation to obtain the temperature distribution
and the degree of cure distribution during the process is
performed for 16, 32 and 52 plies of bres which are
initially dry. The thermal material constants employed
can be found in Loos and Springer [29] and are given in
Table 1. Other data constants for the ow properties
used for the simulation are given in Table 2 (Blest [28]).
The value of constants used in the curing equations are
obtained from Lee et al. [27] and are given in Table 3.
Fig. 2 shows the computed change in the non-dimen-
sional thickness of the resin layers with time as pressure
is applied to laminates consisting of three dierent
numbers of plies. It is observed that for the given mag-
nitude of force and viscosity (j=1, constant), the resin
inltration into the dry bres and the consolidation of
the saturated plies are almost instantaneous for all the
thicknesses.
Figs. 3 and 4 show temperature versus time at the
centre of a composite consisting of dierent numbers of
plies at heating rates of 2.8 K/min and 11.1 K/min
respectively. At both rates of heating, exotherms are
observed to occur for all the composites and the
magnitude of the exotherms is seen to increase for
thicker composites.
Fig. 5 shows the degree of cure versus time at the
centre of a composite consisting of dierent numbers of
plies at a heating rate of 2.8 K/min. It can be seen that
the predicted time for maximum cure to be achieved
increases as the laminate thickness increases.
Fig. 6 compares the temperature at the centre of a 16-
ply composite for heating rates of 2.8 K/min and 11.1
K/min respectively. It can be seen that for the same
thickness, higher exotherms are predicted for higher
heating rates.
Fig. 7 compares the cure at the centre of a 16-ply
composite for two dierent heating rates and it can be
seen that the time to achieve maximum cure is increased
as the heating rates increases.
Table 1
Material properties of Hercules 3501-6 resin and bre
Resin density, ,
r
1.2610
3
kg/m
3
Specic heat of resin, c
r
1.2610
3
J/(kg K)
Thermal conductivity of resin, k
r
1.6710
1
W/(m.K)
Heat of reaction of resin, H
R
474 J/g
Fiber density, ,
f
1.7910
3
kg/m
3
Specic heat of bre, c
f
7.1210
2
J/(kg K)
Thermal conductivity of bre, k
f
2.6010
1
W/(m.K)
Gas constant, R 8.31435 J/Kmol
Initial ambient temperature, T
ini
300 K
Applied cure temperature, T
C
450 K
Permeability of porous mat, k 10
16
m
2
Applied force, F
a
10
3
N
Table 2
Flow parameters from Blest [28]
Resin viscosity, j 1.0 Pas
Porosity, 0.5
Half-length of composite, L 0.5 m
D.C. Blest et al. / Composites Science and Technology 59 (1999) 22972313 2303
Figs. 8 and 9 show a one-dimensional prole of tem-
perature through the thickness at dierent times for a
16-ply and a 32-ply composite at a heating rate of 11.1
K/min. From the graphs, the model predicts a fairly
uniform temperature distribution for the selected cure
cycle throughout the process. The gures also display
the eects of the exotherms as the composite thickness
increases.
6. Experimental investigation
In an experimental investigation 4 layers (equiva-
lently, 8 plies) of stitched fabric were used. The plies in
each layer were oriented alternatively at +45

and 45

to obtain a composite with uniform ``strength'' proper-


ties in both in-plane directions. Two separate cases were
investigated; in the rst, resin was distributed at the top
and bottom of the stack of bers, and in the second the
Table 3
Curing constants of Hercules 3501-6 resin
Pre-exponential constant, A
1
2.101 10
9
min
1
Pre-exponential constant, A
2
2.104 10
9
min
1
Pre-exponential constant, A
3
1.960 10
5
min
1
Activation Energy, E
1
8.0710
4
J/mol
Activation Energy, E
2
7.7810
4
J/mol
Activation Energy, E
3
5.6610
4
J/mol
Fig. 2. Resin thickness vs time.
Fig. 3. Temperature vs time at the centre of the composites at heating
rate 2.8 K/min.
Fig. 4. Temperature vs time at the centre of the composites at heating
rate 11.1 K/min.
Fig. 5. Cure vs time at the centre of the composites at heating rate 2.8
K/min.
2304 D.C. Blest et al. / Composites Science and Technology 59 (1999) 22972313
same amount of resin was applied at the top only. The
applied pressure was 85 psi (5.8610
5
Nm
2
), and the
temperature was quickly ramped to 175

C and held for


a 30 min period. Here it may be remarked that the resin
utilised begins to gel after such a length of time at this
temperature. In each case the nal thickness of the
laminate was approximately 3.5 mm.
For the rst case, Fig. 10 shows complete impregna-
tion of all plies although the 5th ply from the top indi-
cates some small defects. Fig. 11 represents an increased
magnication of the right-hand micrograph in Fig. 10.
However, when the same experimental regime was
applied to the stack with resin at the top only, only 6
out of 8 plies were impregnated, as is indicated in Fig.
12. It is clear that when the resin was placed at both
ends of the stack of bers, the infusion process is more
successful and achieved full penetration in a shorter
time; this was of course predicted by the model, and
entirely to be expected.
For a one resin, one ber set-up, Eqs. (34) and (35)
for the impermeation depth reduce to
o =

kPt
j
_
(66)
where P denotes the pressure.
Additional experimentation on the resin alone, over a
cure cycle equivalent to that of the infusion process,
shows that the viscosity is below 10 Pas. Given the very
large changes in the viscosity over the cure cycle (as
Fig. 6. Comparison of temperature vs time at the centre of the com-
posites at dierent heating rates.
Fig. 7. Comparison of degree of cure vs time at the centre of the
composite at dierent heating rates.
Fig. 8. Temperature vs normalized height at dierent times during the
process for n=16 plies.
Fig. 9. Temperature vs normalized height at dierent times during the
process for n=32 plies.
D.C. Blest et al. / Composites Science and Technology 59 (1999) 22972313 2305
much as from 1 to 10
5
Pas), an estimated viscosity of 10
Pas with an impermeation period of 15 min was adop-
ted for the purpose of representing the cycle. The per-
meability, k, more dicult to obtain, is estimated to lie
between 510
3
and 510
4
m
2
, and 10
3
was adopted
for the order of magnitude calculation. Typical values
of the porosity lie near 0.5. With these values,
o = 3.25mm (67)
Given the average values taken for the parameters this
is a reasonable match for the measured value of 3.5(6/
8)=2.625 mm.
Finally, there was no evidence of distinctly diering
resin distributions away from the centres of the micro-
graphed sections, allowing some support for the one-
dimensional model of the uid ow oered in this paper.
Appendix A
Derivation of RFI ow equations
We rst note that an ``internal'' layer, whether it be
bre or resin, has (geometrical) symmetry. The bottom
resin layer (k=1) and the top layer (k=2n+1) must be
considered separately.
Fig. 10. Section through centre of 8 ply composite made with resin lm at top and bottom.
Fig. 11. Section through centre of 8 ply composite made with resin
lm at top and bottom x50.
2306 D.C. Blest et al. / Composites Science and Technology 59 (1999) 22972313
We begin, however, by making a number of observa-
tions for all layers. Firstly, in the bre-free (2k1) layer of
resin, integrating the incompressibility condition (1) yields
v
r
2k1
= v
r
2k1
x. t ( ) (i.e. v
r
2k1
is not a function of y)
and integrating the momentum Eq. (2) yields
p
r
2k1
= p
r
2k1
y. t ( ) i.e. p
r
2k1
is not a functionof x
_ _
. (A1)
Hence, dierentiating (A1) with respect to x, reduces
Eq. (3) to
o
3
v
r
2k1
ox
3
= 0
which, upon integration, yields
v
r
2k1
x. t ( ) = A
2k1
t ( )x
2
B
2k1
t ( )x C
2k1
t ( ). (A2)
Similarly, in a typical (either upper or lower) wetted
region of the 2k bre layer, dened by (4), integrating
(5) and (6) respectively yields
v
w.i
2k
= v
w.i
2k
x. t ( ) (A3)
p
w.i
2k
= p
w.i
2k
y. t ( ) (A4)
where i=u,l denotes upper or lower regions respectively.
Hence dierentiating (7) with respect to y gives:
o
2
oy
2
p
w.i
2k
y. t ( ) = 0
which on integrating yields the pressure
p
w.u
2k
y. t ( ) = Q
u
2k
y R
u
2k
. (A5)
p
w.l
2k
y. t ( ) = Q
l
2k
y R
l
2k
. (A6)
where Q and R are functions of t.
Further, the velocity expressions in the lower and
upper wetted regions may be obtained from (7):
v
w.u
2k
t ( ) = h
.
2k
t ( )
k
j
Q
u
2k
. (A7)
v
w.l
2k
t ( ) = h
.
2k1
t ( )
k
j
Q
l
2k
. (A8)
The dependence of p
r
2k1
y. t ( ) and v
r
2k1
x. t ( ) on t only
will now be established. Recall that
v
r
2k1
x. t ( ) = A
2k1
t ( )x
2
B
2k1
t ( )x C
2k1
t ( ) (A9)
Fig. 12. Section through centre of 8 ply composite made with resin lm on top only.
D.C. Blest et al. / Composites Science and Technology 59 (1999) 22972313 2307
and
v
w.l
2k
x. t ( ) = h
.
2k1
t ( )
k
j
Q
l
2k
t ( ). (A10)
At y = h
2k1
t ( ) the continuity of the normal velocity is
assumed, that is
v
r
2k1
x. t ( ) = v
w.l
2k
x. t ( ) (A11)
which implies, upon equating coecients of powers of x,
A
2k1
= B
2k1
= 0 (A12)
and
C
2k1
= h
.
2k1
t ( )
k
j
Q
l
2k
t ( ). (A13)
Thus v
r
2k1
x. t ( ) is a function of t only, and henceforth
will be denoted by v
r
2k1
t ( ). Note that this argument
holds for all k=1, 2,. . ., n.
The assumption of continuity of normal velocity is
really valid only for high porosity (i.e. ~ 1) a more
accurate requirement might be continuity of ux, that is
1

v
r
2k1
x. t ( ) = v
w.l
2k
x. t ( ) etc.
Now consider the pressure in the general resin layer.
From (3)
o
oy
p
r
2k1
=
o
2
ox
2
v
r
2k1
= 0 (A14)
using the fact that v
r
2k1
has now been shown to be a
function of t only. This demonstrates that p
2
2k1
is inde-
pendent of y. However, from (2), p
2
2k1
is also indepen-
dent of x, and so can only be a function of t. Henceforth,
we shall write
p
r
2k1
y. t ( ) = p
r
2k1
t ( ).
Note again that this is true for all resin layers, k=1,
2,. . ., n.
The fact that the p
r
2n1
y. t ( ) is also a function of t only
is established later [e.g. see (A39)].
Now from (A5), applying the zero pressure condition
of the upper free surface y = o
2k
t ( ) yields
R
u
2k
= Q
u
2k
o
2k
t ( ). (A15)
The pressure is assumed to be continuous on the inter-
face y = h
2k
t ( ), that is,
p
r
2k1
t ( ) = p
w.u
2k
h
2k
t ( ). t ( ) = Q
u
2k
h
2k
t ( ) R
u
2k
(A16)
using (A5). However, from (A15) we have
p
r
2k1
t ( ) = Q
u
2k
h
2k
t ( ) o
2k
t ( ) ( ). (A17)
(Note that since we have shown p
r
2k1
is independent of
y, this is true for all y h
2k
. h
2k1
[ [.) But, imposing the
free surface kinematic condition, (13) in the upper wet-
ted 2k brous layer gives
v
w.u
2k
t ( ) = h
.
2k
o
.
2k
h
.
2k
_ _
(A18)
and comparing this with (A7) yields
Q
u
2k
=
j
k
o
.
2k
h
.
2k
_ _
. (A19)
Thus, inserting (A19) into (A17) gives
p
r
2k1
t ( ) = p
w.u
2k
h
2k.
t
_ _
=
j
k
h
.
2k
o
.
2k
_ _
h
2k
o
2k
( ). (A20)
Also, inserting (A15) and (A19) into (A5) gives
p
w.u
2k
y. t ( ) =
j
k
h
.
2k
o
.
2k
_ _
(y o
2k
). (A21)
We shall nowconsider k=1 separately. At y=0 we have
v
r
1
t ( ) = 0. (A22)
and since we have demonstrated that v
r
is not a function
of y (i.e. the resin in the bottom layer is static) we have
v
r
1
t ( ) = 0 for all y c 0. h
1
t ( ) [ ].
At y=h
1
(t) we assume continuity of ow, that is
v
r
1
t ( ) = v
w.l
2
t ( ) = h
.
1
t ( )
k
j
Q
l
2
which implies
h
.
1
t ( ) =
k
j
Q
l
2
. (A23)
However, we require the continuity of pressure across
y=h
1
(t), that is
p
r
1
t ( ) = p
w.l
2
h
1
t ( ). t ( ) = Q
l
2
h
1
t ( ) R
l
2
. (A24)
2308 D.C. Blest et al. / Composites Science and Technology 59 (1999) 22972313
But the pressure p
1
r
(t) has been shown to be independent
of y and so
p
r
1
t ( ) = Q
l
2
h
1
t ( ) R
l
2
for all yc 0. h
1
t ( ) [ ]
=
j
k
h
.
1
t ( )h
1
t ( ) R
l
2
from(A23) (A25)
Furthermore on the free surface y=o
1
(t), there is zero
pressure:
p
w.l
2
o
1
t ( ). t ( ) = 0
which implies that
R
l
2
= Q
l
2
o
1
t ( )
so that
p
r
1
t ( ) =
j
k
h
.
1
t ( ) h
1
t ( ) o
1
t ( ) ( ). (A26)
At y=o
1
(t) there is also the kinematic condition, that is
v
w.l
2
h
.
1
t ( ) = o
.
1
t ( ) h
.
1
t ( )
_ _
. (A27)
But this expression is independent of y and so must hold
for all y h
1
t ( ). o
1
t ( ) [ [. Equating (A27) with the velocity
v
r
1
= 0 ( )at y = h
1
t ( ) gives
h
.
1
t ( ) = o
.
1
t ( ) h
.
1
t ( )
_ _
which, after rearranging and integrating, gives
1 ( )h
1
t ( ) o
1
t ( ) = constant.
The constant may be determined by applying the initial
condition for the free surface:
o
1
0 ( ) = h
1
0 ( ) = h
1.0
yielding
o
1
t ( ) =
1

h
1
t ( )
1

h
1.0
. (A28)
and so eliminating o
1
(t) in (A26) yields
p
r
1
t ( ) =
j
k
h
.
1
h
1
h
1.0
_ _
. (A29)
We consider the bre layer k=n where the velocity
and pressure in the upper wetted layer are
v
w.u
2n
t ( ) = h
.
2n
t ( )
k
j
Q
u
2n
. (A30)
p
w.u
2n
y. t ( ) = Q
u
2n
y R
u
2n
. (A31)
From the continuity conditions at the interface
y = h
2n
t ( ) the pressure is
p
r
2n1
h
2n.
t
_ _
= Q
u
2n
h
2n
R
u
2n
. (A32)
Applying the zero-pressure free-surface condition at
y = o
2n
gives
0 = Q
u
2n
o
2n
R
u
2n
. (A33)
Hence,
R
u
2n
= Q
u
2n
o
2n
. (A34)
Applying the kinematic condition (13) at y = o
2n
gives
Q
u
2n
=
j
k
o
.
2n
h
.
2n
_ _
. (A35)
Hence, the velocity and pressure in the 2n layer are
v
w.u
2n
t ( ) = h
.
2n
o
.
2n
h
.
2n
_ _
. (A36)
p
w.u
2n
y. t ( ) =
j
k
o
.
2n
h
.
2n
_ _
y o
2n
( ). (A37)
We now consider the topmost resin layer where the
velocity of the top plate is v
r
2n1
t ( ) = h
.
2n1
t ( ). Matching
this velocity with the velocity of the upper wetted bre
layer (A7), at y=h
2n
, yields
Q
u
2n
=
j
k
h
.
2n1
h
.
2n
_ _
. (A38)
In (A5), applying the zero pressure condition at y =
o
2n
t ( ) and substituting into (A38), k=n, gives
p
r
2n1
t ( ) = p
w.u
2n
h
2n
t ( ). t ( )
=
j
k
h
2n
o
2n
( ) h
.
2n1
h
.
2n
_ _
. (A39)
We now turn our attention to the top resin layer,
y=h
2n1
, where the applied force, F
a
, is opposed by
the viscosity of the resin. A force balance over the top
plate of length 2L gives
F
a
=
_
L
L
p
r
2n1
t ( )dx (A40)
and on integration yields
D.C. Blest et al. / Composites Science and Technology 59 (1999) 22972313 2309
p
r
2n1
t ( ) =
F
a
2L
. (A41)
Since the pressure in this layer is a function of t only,
p
r
2n1
t ( ) =
F
a
2L
for y c h
2n
. h
2n1
[ ]. (A42)
Furthermore, since the bre layers are assumed incom-
pressible, the next resin layer will experience the same
pressure and so on. We can, therefore, conclude that
p
r
1
t ( ) = p
r
3
t ( ) = . . . = p
r
2n1
t ( ) =
F
a
2L
. (A43)
Equation (A20) now becomes a dierential equation in
h
2k
o
2k
:
h
.
2k
o
.
2k
_ _
h
2k
o
2k
( ) =
kF
a
2Lj
. (A44)
Integrating and applying the initial condition o
2k
(0) =
h
2k
(0) = h
2k.0
results in
o
2k
= h
2k

kF
a
t
Lj
_
. k = 1. 2. . . . . n. (A45)
From the geometry (see Fig. 1), o
2k
- h
2k
and conse-
quently the minus sign has been chosen. This relates the
moving front o
2k
t ( ) to the interface h
2k
t ( ). However, we
also require an expression for o
2k1
t ( ). To obtain this we
return to p
r
2k1
t ( ) and p
w.l
2k2
(t). Requiring continuity at
y = h
2k1
t ( ) implies
p
r
2k1
t ( ) = p
w.l
2k2
h
2k1
t ( ). t ( )
= Q
l
2k2
h
2k1
t ( ) R
l
2k2
. (A46)
From (A6), applying the zero-pressure surface-condition
at the lower surface y = o
2k1
implies
R
l
2k2
= Q
l
2k2
o
2k1
t ( ). (A47)
Again we have the kinematic condition (13):
v
w.l
2k2
t ( ) = h
.
2k1
o
.
2k1
h
.
2k1
_ _
(A48)
which implies
Q
l
2k2
=
j
k
o
.
2k1
h
.
2k1
_ _
. (A49)
Thus
p
r
2k1
t ( ) =
j
k
o
.
2k1
h
.
2k1
_ _
h
2k1
o
2k1
( ) (A50)
Using (A43) we obtain the dierential equation
o
.
2k1
h
.
2k1
_ _
o
2k1
h
2k1
( ) =
kF
a
2Lj
. (A51)
Integrating the above yields
1
2
o
2k1
h
2k1
( )
2
=
kF
a
t
2Lj
constant. (A52)
But the constant is 0, since h
2k1
0 ( ) = o
2k1
0 ( ) = h
2k1.0
.
Therefore
o
2k1
h
2k1
=

kF
a
Lj
_
.
(A53)
Rearranging the above gives
o
2k1
= h
2k1

kF
a
t
Lj
_
. k = 0. 1. . . . . n 1. (A54)
For the general 2k+1 resin and 2k bre regions we
have
v
r
2k1
= v
w.l
2k2
= h
.
2k1
o
.
2k1
h
.
2k1
_ _
(A55)
v
w.u
2k
= h
.
2k
o
.
2k
h
.
2k
_ _
. (A56)
At y = h
2k
we have v
r
2k1
= v
w.u
2k
so that
h
.
2k1
h
.
2k
o
.
2k1
h
.
2k1
_ _
o
.
2k
h
.
2k
_ _
= 0
(A57)
or, upon rearranging,
1 ( ) h
.
2k1
h
.
2k
_ _
= o
.
2k1
o
.
2k
_ _
= [ h
.
2k1

1
2

kF
a
Ljt
_ _ _
(A58)
h
.
2k

1
2

kF
a
Ljt
_ _ _
] (A59)
= h
.
2k1
h
.
2k
_ _

kF
a
Ljt
_ _ _
. (A60)
2310 D.C. Blest et al. / Composites Science and Technology 59 (1999) 22972313
Thus
h
.
2k1
h
.
2k
=

kF
a
Ljt
_
. (A61)
Integrating and using
h
2k1
= h
2k1.0
and h
2k
= h
2k.0
when t = 0 (A62)
gives
h
2k1
h
2k
= h
2k2.0
h
2k.0
2

kF
a
t
Lj
_
. (A63)
Consider (A27), where now
F
a
2L
= p
r
1
t ( ) =
j
k
h
.
1
h
1
h
1.0
_ _
. (A64)
Integrating this gives
h
1
t ( ) = h
1.0

kF
a
t
jL
_
. (A65)
We now prove the following:
h
2k1
t ( ) = h
2k1.0
2k 1 ( )

kF
a
t
jL
_
(A66)
o
2k1
t ( ) = h
2k1.0
2k 1
1

_ _

kF
a
t
jL
_
(A67)
o
2k
t ( ) = h
2k.0
2k 1
1

_ _

kF
a
t
jL
_
(A68)
with h
2k
=h
2k+1
+d. Consider
h
2k1
t ( ) = h
2k1.0
2k 1 ( )

kF
a
t
jL
_
.
k = 0. 1. . . . . n.
(A69)
Setting k=0, we obtain
h
1
t ( ) = h
1.0

kF
a
t
jL
_
(A70)
which is none other than (A65). Assume (A69) is true
for k=v1:
h
2v1
= h
2v1.0
2v 1 ( )

kF
a
t
jL
_
(A71)
Thus (A63) for k=v becomes
h
2v1
= h
2v
h
2v1.0
h
2v.0
2

kF
a
t
jL
_
= h
2v1
d ( ) h
2v1.0
h
2v1.0
d
_ _
2

kF
a
t
jL
_
= h
2v1.0
2v 1 ( )

kF
a
t
jL
_ _ _
h
2v1.0
h
2v1.0
2

kF
a
t
jL
_
= h
2v1.0
2v 1 ( )

kF
a
t
jL
_
(A72)
which completes the induction argument.
From (A54) we have
o
2k1
t ( ) = h
2k1
t ( )

kF
a
t
jL
_
. (A73)
Therefore
o
2k1
t ( ) = h
2k1.0
2k 1 ( )

kF
a
t
jL
_ _ _

1

kF
a
t
jL
_
= h
2k1.0
2k 1
1

_ _

kF
a
t
jL
_
.
(A74)
Similarly
o
2k
t ( ) = h
2k

kF
a
t
jL
_
= h
2k1
d

kF
a
t
jL
_
= h
2k1.0
2 k 1 ( ) 1 ( )

kF
a
t
jL
_
d
1

kF
a
t
jL
_
= h
2k.0
2k 1
1

_ _

kF
a
t
jL
_
.
(A75)
Thus (A66)(A68) have been demonstrated to be true.
We now consider the top layer resin layer. Eliminating
the pressure in (A39) using (A43) and eliminating o
2n
t ( )
using (A45) with k=n results in
h
.
2n1
h
.
2n
=

kF
a
4Ljt
_
. (A76)
Integrating the above yields
D.C. Blest et al. / Composites Science and Technology 59 (1999) 22972313 2311
h
2n1
= h
2n
h
2n1.0
h
2n.0

kF
a
t
Lj
_
. (A77)
But
h
2n
= h
2n.0
2n

kF
a
t
Lj
_
. (A78)
Hence
h
2n1
= h
2n.0
2n

kF
a
t
Lj
_ _ _
h
2n1.0
h
2n.0

kF
a
t
Lj
_
= h
2n1.0
2n 1 ( )

kF
a
t
Lj
_
.
(A79)
In summary, the heights of the resin-bre interfaces
and the free-surface ow fronts for a typical k layer, are
h
2k1
= h
2k1.0
2k 1 ( )

kF
a
t
Lj
_
. k = 0. 1. . . . . n
(A80)
h
2k
= h
2k1
d. k = 1. 2. . . . . n (A81)
o
2k1
= h
2k1.0
2k 1
1

_ _

kF
a
t
Lj
_
.
k = 0. 1. . . . . n 1
(A82)
o
2k
= h
2k.0
2k 1
1

_ _

kF
a
t
Lj
_
.
k = 1. 2. . . . . n.
(A83)
where we recall that n is the total number of bre layers
and d is the constant thickness of a bre layer.
Thus, dierentiating (A80)(A82), we then obtain the
velocities of both the bre layers and the free-surface
fronts:
h
.
2k1
= k
1
2
_ _

kF
a
Ljt
_
. k = 0. 1. . . . . n (A84)
h
.
2k
= h
.
2k1
. k = 1. 2. . . . . n (A85)
o
.
2k1
=
1
2
2k 1
1

_ _

kF
a
Ljt
_
. k = 0. 1. . . . . n (A86)
o
.
2k
=
1
2
2k 1
1

_ _

kF
a
Ljt
_
. k = 1. 2. . . . . n. (A87)
Acknowledgements
References 5, 6, 17, 18 and 30 were deleted in proof.
References
[1] Williams C, Summerscales J, Grove S. Resin infusion under ex-
ible tooling (RIFT): a review. Composites Part A, 1996; 27A:
51724.
[2] Coulter JP, Gu ceri SI. Resin impregnation during the manu-
facture of thermoplastic matrix composite materials. In: Newaz
GM, editor. Advances in thermoplastic matrix composite materi-
als, American Society for Testing and Materials, 1989. p. 1432.
[3] Coulter JP, Gu ceri SI. Resin impregnation during composite
manufacturing: theory and experimentation. Comp Sci Technol
1989;35:31730.
[4] Ahn KJ, Seferis JC, Letterman L. Autoclave resin infusion pro-
cess: analysis and prediction of resin content. Sampe Quarterly
1990;21(2):310.
[7] Advani SG. Flow and rheology in polymer composites manu-
facturing. The Netherlands: Elsevier Science B.V, 1994.
[8] Bruschke MV, Advani SG. A numerical approach to model
nonisothermal viscous ow through brous media with free sur-
faces. Int J Numer Meth in Fluids 1994;19:575603.
[9] Givler RC, Altobelli SA. A determination of the eective viscos-
ity for the Brinkman-Forchheimer ow model. J Fluid Mech
1994;258:35570.
[10] Nassehi V, Petera J. A new least-squares nite element model for
combined Navier-Stokes and Darcy ows in geometrically com-
plicated domains with solid and porous boundaries. Int J Numer
Meth Eng 1994;37(9):160920.
[11] Wymer SA, Engel RS. A numerical study of nonisothermal resin
ow in RTM with heated uniaxial bres. J Comp Mater
1994;28(1):5365.
[12] Kang MK, Lee WI, Yoo JY, Cho SM. Simulation of mold lling
process during resin transfer molding. J Mater Proc Manufac Sci
1995;3:279313.
[13] Malkin AY, Baranov AV, Timifeev SV. Molding of resins with
impregnation of a reinforcing layer. J Polym Eng 1995;14(2
3):14159.
[14] Sadiq TAK, Advani SG, Parnas RS. Experimental investigation
of transverse ow through aligned cylinders. Int J Multiphase
Flow 1995;21(5):75574.
[15] Liu S, Masliyah JH. Single uid ow in porous media. Chem Eng
Comm 1996;148150:653732.
[16] Loos AC, MacRae JD. A process simulation model for the man-
ufacture of a balde-stiened panel by resin lm infusion. Comp
Sci Technol 1996;56:27389.
[19] Mogavero J, Advani SG. Experimental investigation of ow
through multi-layered preforms. Polym Comp 1997;18(5):649
55.
[20] Ni J, Zhao Y, Lee LJ, Nakamura S. Analysis of two-regional ow
in liquid composite molding. Polym Comp 1997;18(2):25469.
[21] Yu HW, Young WB. Optimal design of process parameters for
resin transfer molding. J Comp Mater 1997;31(11):111339.
[22] Ambrosi D, Preziosi L. Modelling matrix injection through elas-
tic porous preforms. Composites Part A 1998;29A:518.
[23] Pillai KM, Advani SG. Numerical simulation of unsaturated ow
in woven bre preforms during the resin transfer molding pro-
cess. Polym Comp 1998;19(1):7180.
[24] Lekakou C, Bader MG. Mathematical modelling of macro- and
micro-inltration in resin transfer moulding (RTM). Composites
Part A 1998;29A:2937.
2312 D.C. Blest et al. / Composites Science and Technology 59 (1999) 22972313
[25] Pearce N, Guild F, Summerscales J. A study of the eects of
convergent ow fronts on the properties of bre reinforced com-
posites produced by RTM. Composites Part A 1998;29A:14152.
[26] Abraham D, McIlhagger R. Investigations into various methods
of liquid injection to achieve mouldings with minimum void
contents and full wet out. Composites Part A 1998;29A:5339.
[27] Lee WI, Loos AC, Springer GC. Heat of reaction, degree of cure,
and viscosity of Hercules 3501-6 resin. J Comp Mater
1982;16:51020.
[28] Blest DC. Fluid mechanical aspects of the manufacture of lami-
nar composites, MSc Thesis, University of Strathclyde, 1993.
[29] Loos AC, Springer GS. Curing of an epoxy matrix. J Comp
Mater 1982;17:13569.
[31] Nield DA, Bejan A. Convection in porous media. New York:
SpringerVerlag, 1992.
[32] McKee S, Wall D, Wilson S. An alternating direction implicit
scheme for parabolic equations with mixed derivative and con-
vective terms. J Comput Phys 1996;126:6476.
D.C. Blest et al. / Composites Science and Technology 59 (1999) 22972313 2313