Composites: Part A 37 (2006) 1316–1324

www.elsevier.com/locate/compositesa

Development of a process model for the vacuum assisted resin

transfer molding simulation by the response surface method
Chensong (Jonathan) Dong *

Rakon Limited, One Pacific Rise, Mt. Wellington, Private Bag 99943, Newmarket, Auckland 1031, New Zealand

Received 22 April 2005; received in revised form 26 August 2005; accepted 28 August 2005

Abstract

The vacuum assisted resin transfer molding process (VARTM) offers many advantages over the traditional resin transfer molding
such as lower tooling cost, room temperature processing. In the VARTM process, complete filling of the mold with adequate wetting
of the fibrous preform is critical to the product quality. Computer simulation has become a powerful tool for liquid composite molding
process design and optimization. However, in the VARTM process, since the presence of high permeable media, which has a much higher
permeability than fiber, 3-D models are necessary and the extensive computation limits the usage of simulation in the process design and
optimization.
With the permeability, porosity, and thickness of the fiber preform and the RTM mold filling time as references, the dimensionless
VARTM process variables and mold filling time are introduced in this paper. The significant process variables were identified by using
the design of experiments (DOE). A quadratic regression model was developed by the RSM. The model was validated against the 3-D
VARTM simulation and experiments. The results show that the accuracy is within 15% for most commonly used cases while the com-
putation time saving is over 99%. The approach presented in this paper provides a general guideline for the VARTM process design and
optimization. The process variables and flow media can be quickly chosen by using the developed regression model, which is extremely
useful for composite part design in the early stage.
 2005 Elsevier Ltd. All rights reserved.

Keywords: E. Resin transfer molding (RTM); E. Resin flow; C. Computational modeling; C. Statistical properties/methods

1. Introduction film. The cavity is sealed, e.g., by vacuum tapes. Vacuum

is created in the mold cavity to draw the resin into the fiber
The vacuum assisted resin transfer molding process mats. After the cavity is filled with resin, the resin cures and
(VARTM) offers many advantages over the traditional re- solidifies into the composite part. Finally, the solidified
sin transfer molding such as lower tooling cost, room tem- composite is taken out of the mold. Although this process
perature processing. This process has been employed to appears simple, in actual fabrication, the procedure can be
manufacture many large components ranging from turbine quite complicated. The locations of the inlets and outlets
blades and boats to rail cars and bridge decks. must be carefully selected so that the mold can be com-
The VARTM process can be divided into five steps. pletely filled. The mold and resin temperature must be
First, in pre-molding, the mold surface is cleaned. Then, monitored to avoid resin gelling during resin infusion (see
mold release agent and gel coat are sprayed onto the sur- Fig. 1).
face. Next, during reinforcement loading, dry fiber mats This study focuses on one of the common VARTM pro-
are mounted into the mold and covered by a flexible bag cesses: the Seemann composite resin infusion molding pro-
cess (SCRIMP), which was invented and patented in the
*
Tel.: +64 9 573 5554x389; fax: +64 9 573 5559. late 1980s by Bill Seemann. In this process, a highly perme-
E-mail address: [email protected] able distribution medium is incorporated into fiber preform

1359-835X/$ - see front matter  2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compositesa.2005.08.012
 Chensong (Jonathan) Dong / Composites: Part A 37 (2006) 1316–1324 1317

of complex shape composite structures. Mohan et al. [13]

modeled and characterized the flow in channels using
equivalent permeability. The equivalent permeability is
used as input for numerical simulation of the mold filling
process. The numerical simulations are based on a pure fi-
nite element based methodology. The mold filling in the
VARTM was investigated by Sun et al. [14] based on a high
permeable medium and Ni [15] et al. based on grooves. A
Fig. 1. Vacuum assisted resin transfer molding process.
3-D control volume finite element method was adopted
to solve the flow governing equations. Based on experimen-
tal observations and CVFEM simulation, a simplified leak-
as a surface layer. During infusion, resin flows preferen- age flow model was presented, where they considered the
tially across the surface and simultaneously through the preform and the peel ply as a sink for the resin, while mod-
preform thickness, which enables large parts to be eling the flow in the distribution layer. Tari et al. [16] de-
fabricated. rived a closed form model for vacuum bag resin transfer
Complete filling of the mold with adequate wetting of molding under several simplifying assumptions. They as-
the fibrous preform is critical in the VARTM. Incomplete sumed that the resin velocity in the saturated fiber preform
impregnation in the mold leads to defective parts contain- is negligible. Hsiao et al. [17] avoided this assumption and
ing dry spots. In order to achieve good quality, processing hence the velocity for the resin, as well as the shape of the
parameters such as the locations and numbers of gates and flow front through the thickness of the fiber preform, was
vents need to be properly set. accurately captured. Han et al. [18] proposed a hybrid
Traditionally, trial-and-error techniques are widely ap- 2.5-D and 3-D flow model. Dong [19] presented an equiv-
plied in the composite industry, which largely depend on alent medium method for improving computation effi-
the experience and skills of operators. It is very costly ciency of the VARTM simulation.
and time consuming. With the development of computing From the literature survey, most of the studies focused
technology, simulation has become a powerful tool for on the development of simplified models due to the exten-
the process design and optimization. The control volume fi- sive computations involved in the 3-D CVFEM method.
nite element method (CVFEM) has been the predominant However, the 3-D nature of VARTM simulation is not
method for process simulation [1–6]. It forms and solves avoided. Thus, the simulation is more time-consuming than
a set of equations for nodal control volumes as if they were 2-D RTM simulation. In addition, the 3-D finite element
finite elements. Mesh regeneration is not required, which modeling and material defining process are more compli-
makes the computation more efficient. cated due to the anisotropic nature of flow media. The fiber
In the conventional RTM process, fiber is the only flow direction needs to be specified for each element to relate the
medium. The part can often be regarded as a shell and sim- material orientation to the coordinate system, as shown in
ulation in 2-D domain. The simulation techniques are quite Fig. 2. In view of the industrial application, the mold filling
developed and several commercial simulation software process and time need to be predicted in a timely manner to
packages are available [7–10]. In the VARTM process, reduce the lead time. The 3-D simulation is incapable of
due to the existence of two distinct flow media, fiber pre- meeting this requirement. Thus, it is necessary to develop
form and high permeable medium (HPM), usually 3-D a more convenient tool for VARTM simulation.
models are required for simulation. Dimensionless VARTM process variables including
When a 3-D model is used, for large parts, the VARTM dimensionless permeability, porosity, and thickness have
simulation requires a large number of nodes and elements been introduced in this paper. The dimensionless VARTM
during meshing. In addition, the distribution medium is mold filling time was also derived by correlating the
usually much thinner than the preform. Therefore, a finer VARTM process with the RTM process. The significant
mesh is needed to avoid the high aspect ratio, which may process variables were identified by using the design of
result in poor conditioning in simulation, as well as from
discretization errors. This uses a large amount of computer
hardware resources and increases the computer load. The
simulation time increases significantly and makes the simu-
lation not feasible.
VARTM simulation has been studied extensively.
Mathur et al. [11] developed an analytical model, which
predicts the flow times and flow front shapes as a function
of the properties of the preform, distribution media and re-
sin. Further, they formulated a performance index to give a
measure of the process efficacy. Loos et al. [12] developed a
3-D model to simulate the VARTM manufacturing process Fig. 2. Resin flow in VARTM process.
1318 Chensong (Jonathan) Dong / Composites: Part A 37 (2006) 1316–1324

8 9 2 38 9
experiments (DOE). A quadratic regression model was
<u>
> = K xx K xy K xz > < op=ox >
=
developed by the RSM. The VARTM mold filling time 6 7
v ¼ ð1=lÞ4 K yx K yy K yz 5 op=oy ð1Þ
can be obtained by multiplying the RTM mold filling time : >
> ; >
: >
;
of the same composite part by the dimensionless VARTM w K zx K zy K zz op=oz
mold filling time. The model was validated against the 3-D
where Kij (i, j = x, y, or z) are the components of the per-
VARTM simulation and experiments. The results show
meability tensor. op/ox, op/oy and op/oz are the pressure
that the accuracy is within 10% for most commonly used
gradients in the three directions, respectively (see Fig. 3).
cases, while the computation time saving is over 99%.
For an incompressible fluid, the mass conservation
The original contributions of this research are:
equation can be reduced to the form:
1. A guideline for VARTM process design is provided. The ou=ox þ ou=oy þ ou=oz ¼ 0. ð2Þ
process parameters can be selected per this guideline. Eq. (2) can be integrated over a control volume and leads
2. An efficient and effective approach is developed for the to
VARTM process optimization. By using this approach, Z Z Z
the process parameters can be found to meet the objec- ðou=ox þ ou=oy þ ou=ozÞ dV ¼ 0. ð3Þ
tive in a time-efficient way.
V
3. Since variations exist for the process parameters [20], the
approach presented in this paper can also be used as a Using the divergence theorem (GaussÕs theorem), the con-
robust design tool to reduce the sensitivity to the varia- trol volume integral can be transformed into a control sur-
tions of process parameters. face integral. Thus, Eq. (3) can be written as,
8 9
Z Z <u>
> =
The rest of the paper is structured as follows. In Section
b nx ny nz c v dS ¼ 0; ð4Þ
2, the approach to develop a VARTM process regression S : >
> ;
model is presented. The model validation is presented in w
Section 3. Section 4 discusses the application of the pre- where nx, ny and nz are the normal components of the sur-
sented model. Conclusions are drawn in Section 5. face vector of the control volume. Substituting Eq. (1) into
Eq. (4) yields,
2. Approach 2 38 9
Z Z K xx K xy K xz > < op=ox >
=
2.1. Control volume finite element method (CVFEM) 6 7
ð1=lÞb nx ny nz c4 K yx K yy K yz 5 op=oy dS ¼ 0.
S >
: >
;
K zx K zy K zz op=oz
The flow of a viscous fluid through an anisotropic,
homogenous, porous medium is represented by DarcyÕs ð5Þ
law [21], Eq. (5) is the working equation for solving the problems of
flow through anisotropic porous media and is a combina-
tion of the mass and momentum equations, while the
momentum equation is represented by using the DarcyÕs
law.
In order to solve such moving boundary problems as the
resin flow front advances using the traditional finite ele-
ment method, it requires the computation domain redefini-
tion and mesh regeneration. Mesh regeneration needs a
large amount of computation time as the domain becomes
complicated. Alternatively, the control volume finite ele-
ment method, which forms and solves a set of equations
for nodal control volumes as if they were finite elements,
does not require mesh regeneration. Thus, the computation
is more efficient.
The boundary conditions for mold filling simulation are
as follows:
At the flow front:
p ¼ 0. ð6Þ
At the inlet gates:

For constant pressure : p ¼ p0 ; ð7Þ

Fig. 3. 3-D anisotropy of composites. For the constant flow rate : v ¼ v0 . ð8Þ
 Chensong (Jonathan) Dong / Composites: Part A 37 (2006) 1316–1324 1319

At the mold boundaries: If the reference variables are /ref, lref, p0ref and Kref, and
op=on ¼ 0. ð9Þ the reference mold filling time is tref, Eq. (11) can be written
as
It is assumed that at the beginning of mold filling, the  
control volumes enclosing the inlet nodes are filled with re- t ¼ tref ½/l=ðp0 K Þ=b/ref lref = p0ref K ref c. ð12Þ
sin. At the flow front, a parameter f is used to represent the When the material properties of fiber are considered as the
status of each control volume in the flow domain. If the only variables, Eq. (12) becomes
control volume has not been occupied by the fluid, f is
t ¼ tref ð/=KÞ=ð/ref =K ref Þ. ð13Þ
equal to zero. If the control volume is partially filled, f is
equal to the volume fraction of the fluid occupying the con- In the VARTM process, two types of flow media – fiber
trol volume. f factor is set to 1 if the volume is completely preform and HPM exist. If the properties of fiber preforms
filled by advancing fluid. The control volumes with f values and the RTM mold filling time are regarded as the refer-
varying between 0 and 1 are considered flow front ele- ence variables and mold filling time, respectively, the
ments. The pressure in these partially filled flow front con- VARTM mold filling time can be derived by using Eq.
trol volumes is set to the ambient pressure. With the (13) as
aforementioned boundary conditions, the set of linear alge-
tVARTM ¼ tRTM bF 1 ð/f ; /H ; hf ; hH Þ=F 2 ðK f ; K H ; hf ; hH Þc
braic equations can be solved to determine the pressure
field at each time step during mold filling. Based on the cal- =ð/f =K f Þ. ð14Þ
culated pressure field, the velocity field can then be com- Eq. (14) shows that the VARTM mold filling time is depen-
puted using DarcyÕs law. dent on the permeability and porosity of fiber preform and
The time increment is selected in such a way that a con- HPM, and the composition of fiber preform and HPM. It
trol volume will be fill at each time step. Sometimes, several can be further written as
control volumes can be filled simultaneously. After f values
are updated, another pressure computation is performed tVARTM ¼ tRTM bG1 ð/H =/f ; hH =hf Þ=G2 ðK H =K f ; hH =hf Þc
for all the fully filled control volumes. The process is re- ð15Þ
peated until the whole mold is filled [2].
If the dimensionless process variables for the VARTM pro-
cess are derived as K* = KH/Kf, /* = /H//f, h* = hH/hf,
2.2. RTM and VARTM simulation
and the dimensionless VARTM mold filling time is given
by the VARTM–RTM mold filling time ratio as
In the traditional RTM process, fiber preform is the
t* = tVARTM/tRTM, Eq. (15) becomes
only flow medium. The through-thickness resin flow can
often be neglected and thus a 2-D model can be applied. tVARTM ¼ tRTM ½G1 ð/ ; h Þ=G2 ðK  ; h Þ. ð16Þ
In the VARTM process, however, another flow medium
– high permeable medium (HPM) presents. Considering a As a preliminary study, several RTM and VARTM
1-D flow in the VARTM process, the resin flow front is mold filling process simulation for the same part by both
plotted in Fig. 2. The HPM is much thinner than the pre- linear and port injection was simulated. For RTM simula-
form. The flow is assumed to be well developed and can tion, 2-D models were used and for VARTM simulation 3-
be divided into two regions: saturated region and flow front D models were used. The permeability was considered as
region. In the saturated region, no cross-flow exists, while the only variable. The thickness is 4 mm for the fiber pre-
in the flow front region the resin is infiltrating into the pre- form and 1 mm for the HPM. The porosity is 0.5 for the
form from the HPM. fiber preform and 0.8 for the HPM. The injection pressure
When the inputs in the CVFEM mold filling simulation is 1 · 105 Pa. When the permeability of the fiber preform
are considered, they can be divided into: and HPM is 100 Darcy and 3000 Darcy, respectively, the
RTM and VARTM mold filling process is shown in
 Geometric properties: molding geometry Fig. 4. The complete result for various permeability values
 Material properties: permeability and porosity is shown in Table 1. Thus, it is confirmed that the dimen-
 Condition properties: injection method and pressure, sionless mold filling time is dependent on /*, h* and K* .
viscosity of resin For a VARTM manufacturer, the microscopic level
through-thickness flow is usually of less importance than
The closed form solution for the 1-D flow through a the in-plane resin flow front development and mold filling
homogeneous fiber preform can be derived from DarcyÕs time. Thus, it is possible to use 2-D RTM simulation and
law as dimensionless VARTM process variables to simulate the
mold filling time.
t ¼ l2 /l=ð2p0 KÞ. ð10Þ The approach is illustrated in Fig. 5. For any given com-
Expanded to the general case, it can be derived that posite part design, the mold filling process can be simulated
in 2-D using the RTM assumption. The dimensionless
t / /l=ðp0 KÞ. ð11Þ VARTM process variables /* , h* and K* are calculated
1320 Chensong (Jonathan) Dong / Composites: Part A 37 (2006) 1316–1324

Table 1
Mold filling simulation of RTM and VARTM processes
Permeability Mold VARTM–RTM
(darcy) filling time (s) mold filling
time ratio
K11f K11H RTM VARTM
Linear 50 1500 98.90 16.80 0.170
100 3000 49.50 8.41 0.170
200 6000 24.70 4.20 0.170
Port 50 1500 343.00 66.00 0.192
100 3000 171.00 33.00 0.193
200 6000 85.70 16.50 0.193

h* and K*. The ANOVA also indicates that a significant

curvature exists.

Fig. 4. RTM vs. VARTM mold filling. 2.4. Model development by response surface method (RSM)

The result from the screening design shows that /* is an

and the dimensionless VARTM mold filling time t* is cal- insignificant variable so that only h*, K* are used in the
culated by the developed VARTM process model. The ac- further analysis.
tual VARTM mold filling time can be given by The significant effect of curvature indicates that a linear
tVARTM ¼ t tRTM . ð17Þ model is not sufficient. A second-order response surface
model needs to be developed. One effective way to develop
such a model is the central composite design (CCD). Thus,
2.3. Screening design
the design was augmented into a 22 design with four axial
runs as shown in Fig. 7.
First, a screening design was conducted to uncover the
The Box–Cox Method [22] was used to stabilize the var-
individual contributions to the VARTM mold filling time
iance. An inverse transformation of the flow time was cho-
of hH, /H andKH. A factorial design was conducted to
sen. The final fitted regression model is
identify the significant factors.
Normally, the permeability is 50–200 Darcy for fiber t ¼ 1=ða0 þ a1 h þ a2 K  þ a11 h2 þ a22 K 2
preforms and 1000–5000 Darcy for HPM. The porosity is þ a12 h K  Þ; ð18Þ
0.4–0.6 for fiber preforms and 0.7–0.9 for HPM. The thick-
ness of HPM is less than 1 mm. In order to derive the where a0 = 1.51 · 10 ; a1 = 8.62; a2 = 4.50 · 102;
1

dimensionless variables, Kf, /f and hf were fixed at 100 a11 = 1.84 · 101; a22 =  5.35 · 105; a12 = 5.01 · 101.
Darcy, 0.5, and 4 mm, respectively. The levels of hH, /H Its corresponding response surface is shown in Fig. 8.
and KH are shown in Table 2. These were chosen to cover
the normal application range. The dimensionless VARTM 3. Model validation
process variables were derived and the dimensionless
VARTM mold filling time is the response. A boat hull as shown in Fig. 9 was simulated to validate
A full 23 factorial design with center points were chosen. the regression model. Its length is 2 m.
After data analysis, the half normal plot, the main effects The permeability is 80 Darcy for the fiber preform and
and interaction plots are shown in Fig. 6. It shows that 3000 Darcy for the HPM. The porosity is 0.5 for the fiber
the significant factors are: h*, K*, and the interaction of preform and 0.9 for the HPM the porosity.

Part design
(CAD)

Material 2-D RTM RTM mold

parameters simulation filling time
(fiber, VARTM
HPM) mold filling
VARTM
time
Dimensionless process Dimensionless
Processing variables regression mold filling time
parameters model
(viscosity,
vacuum)

Fig. 5. Approach flowchart.

 Chensong (Jonathan) Dong / Composites: Part A 37 (2006) 1316–1324 1321

Table 2 Permeability
Level selection
Low High (0.25, 58.28)
hH (mm) 0.4 1.6 (0.10, 50.00) (0.40, 50.00)
/H 0.7 0.9
KH (darcy) 1000 5000 (0.25,
h* 0.1 0.4
(0.04, 30.00) 30.00) (0.46, 30.00)
/* 1.4 1.8 Thickness
K* 10 50

(0.10, 10.00) (0.40, 10.00)

First, the thickness of the fiber preform was assumed to (0.25, 1.72)
be 20 mm. The processing of the boat hull by the RTM
process was simulated by using a 2-D CVFEM model. Half
of the structure was modeled due to its symmetry, as shown Fig. 7. Central composite design (CCD).
in Fig. 10(a). The injection pressure was 1 · 105 Pa. The
viscosity was 200 cP. The mold filling process is shown in Second, the thickness of the fiber preform was assumed
Fig. 10(b). The mold filling time is 13,903 s. to be 10 mm. The mold filling time from 2-D simulation is
When the VARTM processing of the boat hull was con- 13,903 s. When the VARTM processing of the boat hull
sidered, the thickness of the HPM was assumed to be was considered, the thickness of the HPM was also as-
1 mm. The dimensionless variables were calculated as sumed to be 1 mm. The dimensionless variables were calcu-
lated as
h ¼ 0:05;
K  ¼ 37:5. h ¼ 0:1;
K  ¼ 37:5.
The mold filling time was calculated as
tVARTM ¼ t tRTM ¼ 4993 s. The mold filling time was calculated as
tVARTM ¼ t tRTM ¼ 3461 s.
The VARTM mold filling process was also simulated by
using a 3-D CVFEM model. The result is shown in Fig. 11. The mold filling time from 3-D simulation is 3608 s. The
The mold filling time is 5597 s. The computation time is 2 h computation time is 1 h 41 0 . The relative error is 4% and
50 0 . The relative error is 11% and the time saving is >99%. the time saving is >99%.

Fig. 6. Half normal plot, main effects and interaction plots.

1322 Chensong (Jonathan) Dong / Composites: Part A 37 (2006) 1316–1324

of the hood at the pressure of 0.6 vacuum and the viscosity

was 500 cP. The mold filling process simulated by the 2-D
CVFEM is shown in Fig. 12. The RTM mold filling time is
4545 s. The thickness of the fiber preform and the HPM is
2.5 and 0.65 mm, respectively. Thus, the dimensionless
variables can be derived as h* = 0.26 and K* = 5.56. The
dimensionless VARTM mold filling time can be calculated
as t* = 0.55 and the VARTM mold filling time can be cal-
culated as tVARTM = 2500 s. The actual mold filling process
is shown in Fig. 13. The mold filing time was 2910 s. The
relative error is 14%.

4. Model application

In the liquid composite molding, the viscosity increases

with increasing curing time and temperature. The rate of
viscosity increase is low at the early stages of curing. After
a threshold degree of cure is achieved, the resin viscosity in-
Fig. 8. Response surface of the fitted regression model. creases at a very rapid rate. The time at which this occurs is
called the gel time. It is necessary to complete the mold fill-
ing process before the gel time since the flow of resin in the
mold becomes increasingly difficult at the end of this time
period.
From the modeling by the RSM, a general HPM selec-
tion chart has been developed as shown in Fig. 14. The hor-
izontal and vertical axes are dimensionless thickness and
permeability, respectively, and the contour is the dimen-
Fig. 9. A boat hull.
sionless mold filling time. This chart can assist in selecting
the HPM to achieve desirable mold filling time.
As an example, from 2-D simulation we know that the
tRTM = 80 min. The gel time of the resin being used is
The developed model was further experimentally vali- 30 min. Thus, it is desirable that tVARTM 6 25 min, i.e.
dated through the VARTM processing of a car hood.
t ¼ tVARTM =tRTM 6 0:31.
The approximate dimensions are: 1300 · 900 mm. The per-
meability and porosity was determined from measurement Thus, the dimensionless thickness and permeability
as Kf = 90 Darcy, /f = 0.52, KH = 500 Darcy, /H = 0.82. should be selected from the shaded area shown in
The vacuum was approximated 0.6 · 105 Pa and the vis- Fig. 15. The corresponding thickness and permeability of
cosity was 435 cP. The resin was injected from the center the HPM can be easily calculated.

Fig. 10. RTM boat hull processing.

Fig. 11. VARTM boat hull processing.

 Chensong (Jonathan) Dong / Composites: Part A 37 (2006) 1316–1324 1323

Fig. 12. RTM simulation of car hood.

Fig. 13. VARTM processing of car hood.

Fig. 15. Example of HPM selection.

5. Conclusion

In this paper, a VARTM process model developed by

2-D CVFEM simulation, the design of experiments, and
the response surface method (RSM) is presented. With
the permeability, porosity, and thickness of the fiber pre-
form and the RTM mold filling time as references, the
dimensionless VARTM process variables and mold filling
time are introduced in this paper. The significant process
variables were identified by using the design of experi-
ments (DOE). A quadratic regression model was devel-
oped by the RSM. The model was validated against the
3-D VARTM simulation and experiments. The results
show that the accuracy is within 15% for most commonly
used cases while the computation time saving is over 99%.
Fig. 14. General HPM selection chart. The approach presented in this paper provides a general
1324 Chensong (Jonathan) Dong / Composites: Part A 37 (2006) 1316–1324

guideline for the VARTM process design and optimiza- [4] Lin RJ, Lee LJ, Liou MJ. Non-isothermal mold filling and curing
tion. The process variables and flow media can be quickly simulation in thin cavities with preplaced fiber mats. Int Polym
Process 1991;VI(4):356–69.
chosen by using the developed regression model, which is [5] Young WB. Three-dimensional nonisothermal mold filling simulation
extremely useful for composite part design in the early in resin transfer molding. Polym Compos 1994;15(2):118–27.
stage. It is also noticed that for thick composite parts, this [6] Lee LJ, Young WB, Lin RJ. Mold filling and curing modeling of
model yields a larger error because of the neglecting of the RTM and SCRIM processes. Compos Struct 1994;27(1–2):109–20.
through-thickness resin flow. [7] Simacek P, Advani SG. Desirable features in mold filling simulations
for liquid molding processes. Polym Compos 2004;25(4):355–67.
The advantages of this model are that the computation [8] Bruschke MV, Advani SG. A finite element/control volume approach
time is reduced significantly. The complicated 3-D Finite to mold filling in anisotropic porous media. Polym Compos
Element modeling and material property defining process 1990;11(6):398–405.
are avoided, which makes this model very convenient for [9] ESI Group. PAM-RTM. Available from: <www.esi-group.com>.
practical applications. For example, the mold filling time [10] Polyworx. Available from: www.polyworx.com.
[11] Mathur R, Heider D, Hoffmann C, Gillespie Jr JW, Advani SG, Fink
can be obtained promptly when the permeability and/or BK. Flow front measurements and model validation in the vacuum
the thickness of the HPM have changed. The range for assisted resin transfer molding process. Polym Compos
developing the RSM based regression model have covered 2001;22(4):477–90.
the normal application range. Thus, the model is adequate [12] Loos AC, Sayre J, McGrane R, Grimsley B. VARTM process model
in predicting the mold filling time of the VARTM process. development. In: Proceedings of international SAMPE symposium
and exhibition, vol. 46, no. I, Long Beach, CA; May, 2001. p. 1049–
The HPM selection chart presented in this paper provides a 61.
convenient tool for flow media selection and process opti- [13] Mohan RV, Shires DR, Tamma KK, Ngo ND. Flow channels/fiber
mization, which is extremely useful for composite part de- impregnation studies for the process modeling/analysis of complex
sign in the early stage. engineering structures manufactured by resin transfer molding.
The original contributions of this research are: Polym Compos 1998;19(5):527–42.
[14] Sun XD, Li SJ, Lee LJ. Mold filling analysis in vacuum-assisted resin
transfer molding. part I: SCRIMP based on a high-permeable
1. A guideline for VARTM process design is provided. The medium. Polym Compos 1998;19(6):807–17.
process parameters can be selected per this guideline. [15] Ni J, Li SJ, Sun XD, Lee LJ. Mold filling analysis in vacuum-assisted
2. An efficient and effective approach is developed for the resin transfer molding. Part II: SCRIMP based on grooves. Polym
VARTM process optimization. By using this approach, Compos 1998;19(6):818–29.
[16] Tari MJ, Imbert JP, Lin MY, Lavine AS, Hahn HT. Analysis of resin
the process parameters can be found to meet the objec- transfer molding with high permeability layers. Trans ASME J
tive in a time-efficient way. Manuf Sci Eng 1998;120(3):609–16.
3. Since variations exist for the process parameters, the [17] Hsiao KT, Mathur R, Advani SG. A closed form solution for flow
approach presented in this paper can also be used as a during the vacuum assisted resin transfer molding process. Trans
robust design tool to reduce the sensitivity to the varia- ASME J Manuf Sci Eng 2000;122(3):463–75.
[18] Han K, Jiang SL, Zhang C, Wang B. Flow modeling and simulation
tions of process parameters. of SCRIMP for composite manufacturing. Compos Part A
2000;31(1):79–86.
[19] Dong CS. An equivalent medium method for the vacuum assisted
References resin transfer molding process simulation. J Compos Mater [to
appear].
[1] Bruschke MV, Advani SG. Filling simulation of complex three- [20] Dong CS. Uncertainty analysis for fiber permeability measurement.
dimensional shell-like structures. SAMPE Quart 1991;23(1):2–11. Trans ASME J Manuf Sci Eng 2005;127(4).
[2] Young WB, Han K, Fong LH, Lee LJ. Flow simulation in molds [21] DÕArcy H. Les Fontaines Publiques de la villa de Dijon. Paris: Dal-
with preplaced fiber mats. Polym Compos 1991;12(6):391–404. mont; 1856.
[3] Liu XL. Isothermal flow simulation of liquid composite molding. [22] Box GEP, Cox DR. An analysis of transformations. J Roy Stat Soc B
Compos Part A 2000;31(12):1295–302. 1964;26:211–43.