The International Journal of Advanced Manufacturing Technology (2021) 115:345–365

https://doi.org/10.1007/s00170-021-07164-6

ORIGINAL ARTICLE

Characterization and simulation of electromagnetically induced

preform resting (EIPR) process
Mohsen Poorzeinolabedin 1 & Kemal Levend Parnas 2,3

Received: 13 October 2020 / Accepted: 26 April 2021 / Published online: 10 May 2021
# The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2021

Abstract
Electromagnetically induced preform resting (EIPR) process is a new version of vacuum-assisted resin transfer molding
(VARTM) process which allows manipulation of resin flow during the filling step. The EIPR process enhances the permeability
of preform locally in case of undesirable flow front situations. This technique ensures the perfect filling despite the existence of
permeability variation. To utilize the EIPR process in a better way, its comprehensive characterization to find the optimum value
of key factors and filling simulation of it is necessary. Key factors of the EIPR process (i.e., amplitude, frequency of vibration)
and permeability of preform as a material index are recognized as independent factors that must be considered to characterize the
process. These factors are considered to establish a mathematical model for the permeability of preforms. Maximum and
minimum values of the frequency and amplitude are determined based on the observations in acceptable composite material
manufacturing and in-plane permeability characterization. Response surface methodology is used to model the permeability of
EIPR process to find the optimum response values of the key factors for selected preforms. To assess the process numerically,
EIPR process simulation with the optimum values is conducted on two case studies. These case studies involving two different
permeability zones are designed in order for evaluation. In each case, a low permeability preform is employed in the middle of
high permeability one to create an artificial problem. Simulation results demonstrate an acceptable accuracy.

Keywords Resin flow control . EIPR process . Permeability . Simulation

1 Introduction deal with important issue [1–4]. Nowadays, for example,

Fiber-reinforced composite structures have been utilized in-
creasingly in high-technology applications such as aerospace,
automotive, marine, and wind energy industries. High specific
stiffness and strength properties of these materials make them
favorable candidates for new applications. However, compos-
ite materials have been facing strong competition from tradi-
tional materials in terms of cost. So, developing cost-effective
manufacturing approaches of these materials are required to
* Mohsen Poorzeinolabedin
poorzeinolabedin_m@yahoo.com
1
Department of Mechanical Engineering, Shabestar Branch, Islamic
Azad University, Shabestar, Iran
2
Department of Mechanical Engineering, TED University,
Ankara, Turkey
3
Leopar Composite Engineering and Consulting Services Ltd.,
Ankara, Turkey
high-tech industry have been using out-of-autoclave process
to manufacture composite parts [5–7]. Liquid composite
molding (LCM) is a composite material manufacturing pro-
cess which allows the production of strong lightweight struc-
tures. The well-known LCM process is vacuum-assisted resin
transfer molding (VARTM) in which preform is first placed in
the mold, and then, resin is transformed to saturate the dry
preform. In this process, defects such as dry spots, voids,
and welding lines often arise during the mold filling and there-
fore the mechanical properties of the produced parts are de-
creased [8, 9]. These defects may often occur due to unexpect-
ed resin flow in the current vacuum bags [10]. To evaluate
VARTM process with rival approaches, i.e., autoclave and
RTM process, elimination of limitations caused by such de-
fects is required.
One of the major reasons of dry spots is formations of
undesirable flow front due to inherent variation in the preform
permeability. According to the ideal gas law (Pressure ×
Volume = constant), pressure increases in this dry spot area
and the volume continually decreases until the pressure in the
346
air trap becomes close to the infusion pressure. As a result, the
pressure gradient gets very small and the resin can no longer
moves and forms dry spots [11, 12].
Diversified approaches have been proposed to modify the
filling in LMC process. These approaches mainly involve
closing/opening of gate/vent, controlling the pressure, heating
the resin to increase the viscosity, increasing the porosity of
preform by vacuum-induced preform relaxation, and resin de-
livery by a flow flooding chamber.
An intelligent open/close gate/vent approach for RTM and
VARTM processes with correct timing based on flow sensing
sensors in the mold is presented by Devillard [13]. This ap-
proach uses a combination of a genetic algorithm and filling
simulation to prevent unexpected distribution failures during
the mold-filling stage. A model based on a resin flow control
approach in liquid composite molding was developed by
Johnson et al. [14] and Nielsen et al. [15, 16]. For controlling
the gate pressure and gate flow rate in real time, authors utilized
an intelligent neural network. Nalla et al. [17] utilized a multi
segment injection lines. In this system, each segment/line
operates independently and a closed-loop controller is used
for delivering the resin to different locations. An active control
system capable of the flow front detection, flow disturbances
identification, and implementing real-time corrective actions by
computer-controlled ports was suggested by Modi et al. [18]. In
spite of these improvements in modify the filling in LMC pro-
cess, there are some limitation in application of these ap-
proaches. First, these approaches are not generally cost-
effective methods due to multi gate/vent ports. In addition, there
is a possibility of entrapping voids by these solutions. Finally,
due to compaction of preform in atmospheric pressure, these
gate ports accumulate resin in the infusion points.
Another active control method is presented in Johnson
et al. [14, 19]. In this method, resin viscosity is reduced by
local heating locally which leads to enhancement of preform
permeation and therefor elimination of entrapment void and
dry spots. An active flow control approach in which folw front
pattaren is progressively forcated by numerical simulations of
VARTM is proposed by Matsuzaki et al. [20]. Using these
forecasted patterns, the flow front is modified by decreasing
the viscosity and therefor increasing the flow speed. The re-
duction of viscosity is accomplished by dielectric heating at
specific target points. Since the viscosities are low, the pro-
vided control action by heating is limited. Also, the cure pro-
cess of the thermoset resins may be accelerated by heating that
results in the increasing of the viscosity.
In order to reduce the filling time and material waste, a new
resin delivery method called the flow flooding chamber is in-
vestigated by Justin Alms et al. [21]. In this method, the vacu-
um bag is stretched by a chamber above the bag using vacuum
to rest the preform. Using this chamber, the resin is accumulat-
ed into the rested preform and is driven into preform by atmo-
spheric pressure by releasing vacuum. In another approach
Int J Adv Manuf Technol (2021) 115:345–365
called the vacuum induced preform relaxation developed by
Justin Alms [22], an external local vacuum pressure is utilized
to reduce the compaction of preform and manipulation of per-
meability. In real time, control of the flow in an automated
feature is used by this system. The main disadvantages of these
methods are creation of a gap between the preform and vacuum
bag and accumulation of resin in this gap for resin delivery in
the problematic area. Therefore, resin flow cannot be precisely
controlled in these approaches. Though this problem is solved
by this process to some extent, since sealing of the chamber
become an abstraction once the region under the chamber is
saturated, this method is applicable over the injection ports.
In order to increase the reliability of the LCM processes, it is
essential to have a well understanding about the resin flow
saturation in fiber preform. Numerical analysis enables one to
optimize the filling process and evaluate the resin filling pro-
cesses. Recently, various VARTM process simulations have
been conducted. Simulations for the liquid composite molding
require defining boundary conditions, in addition to preform
characteristics. There are several programs for simulation of
composite laminate molding mostly with an acceptable approx-
imation. To simulate resin infusion, commercial software such
as LIMS from the University of Delaware [23], RTM-Worx
from the Polyworx [24], FLUENT from Ansys, Abaqus CFD
from Simulia Abaqus, and PAM-RTM from ESI group [25]
have been extensively used. The resin flow simulation of LCM
process in 2D, 2.5D with a layer of shell, and 3D perspective
has been studied by several authors [26–31]. Permeability and
compressibility of preform are the essential ingredient for gen-
erating successful simulation in filling process.
The previous work [32] presented an active and real-time
resin flow control approach. This method is called electromag-
netically induced preform resting (EIPR) process which pinches
and vibrates the upper flexible mold to rest the preform and
increase the permeability locally. This approach detects the prob-
lematic region by monitoring the flow pattern during the filling
process. To implement this process, a thorough understanding
and full characterization of it is required to effectively predict the
resin flow and permeability of the preform. To do that, in this
study, various factors that may significantly affect the process are
recognized. The amplitude and frequencies of vibration beside
the preform property are found as the main factors of EIPR
process. The experimental design and response surface method
are applied to refine models using factorial designs, especially for
curvature in the response surface. This method is applied to
model their response that uses a few tests to obtain the relation-
ship among the major factors and the response of the system.
Permeability of the preform with different parameters is calcu-
lated to optimize and characterize the preform permeability as a
function of main factors for three different preforms. Using these
factor values, two case studies are designed to simulate this pro-
cess. These cases, involving two different permeability zones, are
designed for process evaluation.
Int J Adv Manuf Technol (2021) 115:345–365
2 EIPR process
The EIPR approach is a new method of resin delivery process,
which one can influence the permeability of preforms in real
time manner and hence manipulate the resin flow front. This
process incorporates a creative reusable self-sealing vacuum
bag with embedded ferromagnetic elements to pitch and vi-
brate upper flexible mold using an electromagnetic field
source. Besides the upper flexible mold, the EIPR process
uses an automated gantry system carrying the electromagnetic
and resin flow front detection system. Resin flow control is
possible almost anywhere in the infusion mold by the ferro-
magnetic elements which are scattered in vacuum bag during
the preparation of it. The principle of this approach is locally
resting the preform and decreasing the resistance against the
resin flow at problematic locations. In this process, the crea-
tive upper flexible mold is dropped over the whole preform
surface which must be saturated to produce a composite part.
This configuration considers the flow front in online manner
and, after detection of an unwanted disturbance in the flow
pattern, calls the control system to move the electromagnetic
field source over the problematic zone to influence this pat-
tern. Invoking the upper flexible mold at that region makes the
preform more permeable. This correction action reduces the
compaction pressure and increases the fabric preform porosi-
ty. Resin is delivered faster in these locations where an unpre-
dictable permeability variation occurs to compensate the resin
flow perturbation in real time [32]. The process is schemati-
cally demonstrated in Fig. 1.
Fig. 1 Schematic of EIPR process [32]
347
3 EIPR process characterization methodology
Process characterization is essential in order to fully under-
stand the EIPR process before applying it in resin flow con-
trolling. In this respect, the ferromagnetic element shape, fre-
quency of the vibration, and amplitude for each preform type
are investigated.
The upper flexible mold has a fundamental task in resin
delivery at the problematic zone using embedded elements.
The embedded elements consist of an aluminum element
and a ferromagnetic part. The shape of the aluminum ele-
ment has an important role on flow manipulation and con-
siderably affects the process performance. In order to select
an appropriate shape for the aluminum element, experi-
ments with two different shapes are conducted. For this
purpose, a circular aluminum element with a radius of
45 mm and a square with a side length of 45 mm are
studied. Preform relaxation is provided by both shapes,
and the corresponding flow front progresses are inspected.
Figure 2 a shows the contribution of circular element to the
flow front. The circular element causes a flow front in the
form of a curve or a semicircle, to be precise, in which the
fluid speed under the center of the element is the highest. It
is found more suitable for a case where the race-tracking is
likely to occur, since it can reduce the race tracking effect.
In contrast, the square element generates a straight flow
front as shown Fig. 2b. Therefore, the second one is pre-
ferred for EIPR characterization study that can compensate
and deliver the fluid uniformly.
348
Fig. 2 The corrected flow front patterns for a circular shape and b square shape elements
After preparation of the flexible upper mold and selecting
the shape of element, the full specifications of the EIPR pro-
cess must be characterized. Two critical factors of the EIPR
process, known as amplitude and the frequency of vibration,
were revealed during the experimental evaluation. As depicted
in Fig. 1, amplitude is the distance between the external elec-
tromagnet and the corresponding embedded element that can
be lifted by the electromagnetic force while tapping frequency
denotes the rate of electromagnetic force engagement on the
corresponding vacuum bag location.
In order to achieve the proper ranges of amplitude and
frequency, some experiments are performed. These experi-
ments showed that the most effective ranges for amplitude
and frequency are 0.2–0.8 mm and 1–10 Hz, respectively.
There would be an inordinate gap between the vacuum bag
and upper surface of the preform if the amplitudes are over 0.8
mm. This gap not only causes an undesirable filling because
of providing a very high permeability channel over the pre-
form that is significantly different from the in-plane perme-
ability of the preform, but also cumulates the resin under the
vacuum bag and decreases the fluid velocity. On the other
hand, if the frequencies were over 10 Hz, the electromagnet
practically clings to the element, where the vibration ceases.
To manipulate these independent factors and measure their
effects on the permeability of the three different preforms, it is
needed to create a test procedure using experimental design.
A statistical method called response surface method [33] is
used for experiment design and modeling the permeability as
a response, mathematically. This method is used to relate the
response, i.e., the permeability of the preform to the signifi-
cant parameters with a first- or second-order polynomial. The
Int J Adv Manuf Technol (2021) 115:345–365
quadratic equation is used to model the curvature response. It
takes the following form:
k k
Y ¼ β 0 þ ∑ βi X i þ ∑ βii X 2i þ ∑ βij X i X j þ ε ð1Þ
i¼1 i¼1 i< j
where β0, βi, βii, and βij are the constant, linear, quadratic, and
interaction coefficients, respectively. Also, Xi and ε are param-
eters and the error term, respectively.
This modeling is conducted on the selected three different
preforms using frequency and amplitude as significant param-
eters with three levels. Central composite design is utilized for
the design experiment of surface response method. Full facto-
rial design with center point is used for design of two con-
tinues factors (amplitude and frequency) and one categorical
factor (preform type). A total of 39 experiments are designed
for three factors with four replications at the center point of
each categorical factor.
4 Material
4.1 Material specifications
Three different preforms are selected to explore the permeabil-
ity feature. The criterion for selection is their different perme-
ability values. E-glass multiaxial 600 g/m2, E-glass twill 300
g/m2, and chapped (mat) E-glass EMAT1 450 g/m2 fiber pre-
forms according to the size of the prepared mold are cut care-
fully. Number of layers for twill, multiaxial, and mat samples
Int J Adv Manuf Technol (2021) 115:345–365 349

are 12, 5, and 5, respectively. Prepared samples are shown in where xff is the position of flow front at time t, Δp is the
Fig. 3. pressure difference between infusion and vent line, μ is the
Since viscosity of thermoset resins varies during the pro- fluid viscosity, and φ is the preform porosity. Here
cess and regarding their Newtonian behavior before the gela-
φ ¼ 1−v f ð3Þ
tion, instead of these resin types, a test fluid (Motor oil
20W50) is used to have repeatable and reliable measurement, and vf is the fiber volume fraction.
as recommended in [35]. The test fluid viscosity is 0.165 Pa s To simulate and predict the filling time and flow front
with a density of 900 kg/m3. pattern, a complete characterization of material property is
necessary. Since permeability is known to be anisotropic
4.2 Fiber-reinforced preform’s permeability property in porous media, then a second-order tensor de-
scribes it as follows:
Fiber-reinforced permeability is an anisotropic property, 2 3
which characterizes the ease of establishing a resin flow K xx K xy K xz
K ¼ 4 K yx K yy K yz 5 ð4Þ
through the preform in LCM processes. In LCM, resin flow
K zx K zy K zz
depends on the filling condition, resin property, and preform
characteristics. Permeability is a way to represent the resis- This tensor is diagonalized to obtain what is known as the
tance of preform against the resin flow. Therefore, obtaining principal permeability. It is assumed that the first two principal
the permeability of the preforms under EIPR process is of permeability values lie in the fabric plane while the third one is
critical importance in characterization of this process. orthogonal to the fabric plan. To find the principal permeabil-
Unidirectional infusion experiments are performed to de- ities, three permeability measurements are required at 0°, 45°,
termine in-plane permeability. For the unsaturated rectilinear and 90° directions of the preform. Once these values are ob-
flow at constant pressure, the permeability is calculated from tained, calculation of principal permeabilities (K1, K2) is pos-
linear infusion of fluid as inlet line into the fiber preform. The sible as follows [35]:
flow front is assumed uniform and oriented parallel to the inlet
line. Darcy’s law [34] gives the permeability as follows: ° α1 −α2
K 1 ¼ K 0xx α2 ð5Þ
xff μφ α1 −
K xx ¼ − ð2Þ cosð2β Þ
2ΔPt

Fig. 3 E-glass fiber preform

samples: (a) twill, (b) multiaxial,
and (c) mat
350
and
° α1 þ α2
K 2 ¼ K 90
xx α2 ð6Þ
α1 þ
cosð2βÞ
where α1and α2 are as follows:
° °
K 0xx þ K 90
xx
α1 ¼ ð7Þ
2 hðPcom Þρ
° °
K 0 −K 90
α2 ¼ xx xx ð8Þ
2 and h is the preform thickness.
!
1 α1 α21 −α22
β ¼ tan−1 − ð9Þ
2 α2 α2 :K 45
xx
° from the Darcy’s equation has a gradient component along the
° ° °
where K 0xx ; K 45
xx ; and K xx and β are permeability of the pre-
90
form along 0 , 45 , and 90° orientations and the angle between
° °
the elliptic pattern of flow and warp direction of fabric.
4.3 Fiber-reinforced preform’s compressibility
Another important factor in all VARTM processes is fab-
ric compressibility which affects the properties related to
both material and process of the part. Fiber volume frac-
tion is increased by compaction due to compression of
fluid pressure or mold surface. Thus, the thickness of
the part is decreased and, consequently, the porosity
which results in the decrease of permeability. The fabric
compaction may lead to several important phenomena in
this type of processes. In the processes such as VARTM
where the flow is in the fabric plane, since the net com-
paction pressure varies throughout the mold, a region with
non-uniform thickness can be created.
According to the Darcy’s law, increasing the pressure will
result in an increase in the fluid velocity through fabric. On the
other hand, the compaction pressure will also be elevated by
an increase in the pressure of the fluid and hence the perme-
ability is lowered. It could be possible in certain cases for an
increase in pressure to have a decrease in injection time, al-
though this is not common. In most fabrics, the thickness
decrease tends to compensate the permeability decrease in
through-thickness flow. The effect of compaction on perme-
ability is not the same for all fabrics, and since it highly de-
pends to the fabric architecture, some fibers are affected more
than the others.
Fiber compaction also influences the fabric porosity which
affect the saturation time of an unsaturated flow, introducing
another complexity to the problem. Although compaction de-
creases permeability, the decrease in porosity can increase the
velocity of the fluid through a preform. Since this parameter
has an important effect on the permeability of the preform and
processability of the composite parts, calculation of preform’s
compressibility values are required for getting accurate values
Int J Adv Manuf Technol (2021) 115:345–365
for fabric compaction. It is defined as a function of preform
thickness (h) of the pressure applied on the preform surface:
h ¼ f ðPcom Þ ð10Þ
In LCM processes, the compressibility of the preform di-
rectly defines the fiber volume fraction by
 
m
V f ðPcom Þ ¼ ð11Þ
where m is areal density of the fiber preform, ρ is fiber density,
During the infusion process, the resin pressure calculated
saturation path from the atmospheric pressure at infusion line
side toward the vacuum pressure at vent line side. This pres-
sure difference (Patm − Pvac) is constant and taken as external
pressure. Thickness (h) of the mold is not constant due to
balancing property of vacuum bags on the pressure gradient
against the summation of resin pressure and reinforcement
compaction pressure that is computed from compressibility
curve [11, 26, 36].
Pext ¼ Presin þ Pcom ð12Þ
5 Experimental
5.1 Test setup
An experimental setup capable of automatic correction of
the flow pattern is designed and constructed for in order
to implement EIPR process and its control methodology.
The upper flexible mold with the embedded elements and
gate/vent ports is the important part of this setup. An
automatic gantry system controlled/actuated by step mo-
tors is used to deploy the electromagnetic field source on
the embedded elements of the upper mold. These move-
ments are achieved with a written program in ARDUINO.
In order to monitor the flow pattern during the filling
process, a flow front monitoring system is mounted under
the transparent mold. The flow pattern/front is continu-
ously analyzed by a code/program in MATLAB [32].
Two experiment sets are applied to characterize EIPR pro-
cess and its simulation. To characterize the EIPR process, an
upper flexible mold for a rectangle plate of 60 × 180 mm is
prepared as illustrated in Fig. 4.
To evaluate the EIPR process numerically, two case studies
[32] are simulated. The plates used in these tests have a rect-
angular geometry of 140 mm × 240 mm which are stiffened in
the middle of the plate to disturb flow front pattern as Fig. 5.
To simulate the process material property of preforms, i.e.,
Int J Adv Manuf Technol (2021) 115:345–365 351

Fig. 4 Upper flexible mold with

an embedded element for process
characterization

permeability and compressibility must be defined. Beside the
permeability, compressibility of the preform must be tested.
For this purpose, the out-of-plane compression response of
preforms is tested on a displacement-controlled Instron ma-
chine. Three specimens of preform with the size of 60 ×
60 mm are prepared to run the tests as shown in Fig. 6.
5.2 Test procedure
In this study, two test procedures are designed to characterize
and evaluate EIPR process separately.
EIPR process characterization The test series for EIPR charac-
terization are conducted according to the design of

Fig. 5 Case studies (a) case 1 and

(b) case 2 where a high
permeability preforms stiffened
with a low permeability one
352 Int J Adv Manuf Technol (2021) 115:345–365

Fig. 6 Schematic of
compressibility test setup

experiment. In this regard, prepared preforms are placed on

the transparent mold and then covered with the flexible mold
and a vacuum pressure of 66.7 kPa (500 mmHg) is applied
into the mold. To have an in-plane permeability measurement
and avoid infusion through-the-thickness, distribution fabric
or mesh fabric is not used.
Next, frequency of the EIPR process is set in the control
program as the design experiment plan. Before commencing
the infusion, the image processing unit that is placed under the
transparent mold is calibrated and set to take images at desired
interval. Sequentially, the EIPR system is located over the
element and the amplitude is set on the instrument. Finally,
the EIPR process is lunched to measure the permeability.
Flow front position (the average flow front distance from the
infusion line) versus time is obtained, and permeability value
of the preform is calculated.
Additionally, non-control (VARTM) infusion process for
each fiber preform is also done to calculate the permeability as
a comparison tool.
EIPR process evaluation For the planned evaluation tests, the
optimum values of key factors of the EIPR process are given
to the controller program as an input to start vibrating the
elements whenever necessary and increase the permeability
locally. It was programmed to record the flow front during
the filling process with regular intervals. Finally, the EIPR
system is initiated. To compare the filling results, no-control
process is done for each case. All these tests are done under
the same condition of EIPR characterization tests.
6 Process simulation
Simulation of EIPR process can incorporate the resin flow
control in real time. The equivalent permeability of the
preform to model the filling process, the fluid flowing through
the preform is assumed to be isothermal and incompressible
Newtonian. Continuity of the incompressible flow gives the
following:
K
V ¼− ∇p ð14Þ
μ
Substituting V in Eq. (13) yields the Laplacian for the pres-
sure:
∇2 p ¼ 0 ð15Þ
where ∇p is the pressure gradient (Pa/m) and μ is the fluid
viscosity (Pa s).
EIPR process simulation is run for the selected case studies
(Fig. 4) with the use of PAM-RTM program. The model of
case studies is 140 × 240 mm plate with a thickness of
about 2 mm, where the middle of the plate is stiffened with
a low permeability preform, i.e., twill preform. Boundary con-
ditions are defined as atmospheric pressure at the infusion line
and vacuum pressure, 66.7 kPa (500 mmHg), at the flow front
or the vent line of the models. The models are meshed using
triangular 2D elements. Resin viscosity and physical proper-
ties of the preforms (principle permeabilities an compressibil-
ity) are applied as input data of the program. Numerical study
is done for both non-control and EIPR processes to see the
efficiency of the presented method.
7 Results and discussion
7.1 EIPR process characterization results
For characterization of the EIPR process on the three different
fiber preforms that have low, medium, and high permeability
values, design experiment test is carried out with three levels
of the process key parameters. In addition, to compare the
EIPR process with non-control process, normal VARTM test
Table 1 Actual and coded levels of parameters in this study
Factors Levels
Low (1) Medium (2) High (3)

∇:V ¼ 0 ð13Þ Frequency (F) 1 5.5 10

Amplitude (A) 0.2 0.5 0.8
where flow velocity vector (m/s) through the preform is
Int J Adv Manuf Technol (2021) 115:345–365 353

Table 2 Permeability and filling time of the preforms for non-control is also conducted for each preform. The studied levels of the
process (i.e., F0A0)
frequency are 1.0, 5.5, and 10.0 Hz and the amplitudes are 0.2,
Material Permeability (m2) Filling time (s) 0.5, and 0.8 mm. These frequency values are coded as 1–3 for
each level, respectively. Note that parameters exceeding the
Mat 8.476E−11 244 threshold values are not desirable for acceptable filling.
Multiaxial 1.05E−10 178 Table 1 demonstrates the coded and actual level values for
Twill 1.72E−11 1099 design parameters and experiments. The value after latter F

Table 3 Experimental design and

corresponding response Experiment Experiment Frequency Hz Amplitude mm Material Permeability
code no. (Freq.) (Amp.) (Mat.) (×10−11 m2)

F1A1 1 1 0.2 Mat 11.56

F3A1 2 10 0.2 Mat 12.60
F1A3 3 1 0.8 Mat 12.89
F3A3 4 10 0.8 Mat 12.10
F1A2 5 1 0.5 Mat 12.40
F3A2 6 10 0.5 Mat 13.29
F2A1 7 5.5 0.2 Mat 12.80
F2A3 8 5.5 0.8 Mat 14.50
F2A2 9 5.5 0.5 Mat 13.77
F2A2 10 (C) 5.5 0.5 Mat 14.50
F2A2 11 (C) 5.5 0.5 Mat 13.00
F2A2 12 (C) 5.5 0.5 Mat 14.50
F2A2 13 (C) 5.5 0.5 Mat 13.95
F1A1 14 1 0.2 Multiaxial 16.070
F3A1 15 10 0.2 Multiaxial 18.900
F1A3 16 1 0.8 Multiaxial 20.450
F3A3 17 10 0.8 Multiaxial 23.780
F1A2 18 1 0.5 Multiaxial 18.750
F3A2 19 10 0.5 Multiaxial 24.100
F2A1 20 5.5 0.2 Multiaxial 20.380
F2A3 21 5.5 0.8 Multiaxial 24.810
F2A2 22 5.5 0.5 Multiaxial 24.800
F2A2 23 (C) 5.5 0.5 Multiaxial 24.000
F2A2 24 (C) 5.5 0.5 Multiaxial 25.100
F2A2 25 (C) 5.5 0.5 Multiaxial 24.700
F2A2 26 (C) 5.5 0.5 Multiaxial 24.400
F1A1 27 1 0.2 Twill 3.610
F3A1 28 10 0.2 Twill 4.301
F1A3 29 1 0.8 Twill 5.790
F3A3 30 10 0.8 Twill 5.460
F1A2 31 1 0.5 Twill 4.802
F3A2 32 10 0.5 Twill 5.448
F2A1 33 5.5 0.2 Twill 4.810
F2A3 34 5.5 0.8 Twill 6.360
F2A2 35 5.5 0.5 Twill 6.300
F2A2 36 (C) 5.5 0.5 Twill 6.700
F2A2 37 (C) 5.5 0.5 Twill 5.962
F2A2 38 (C) 5.5 0.5 Twill 6.100
F2A2 39 (C) 5.5 0.5 Twill 6.000
354
Fig. 7 Flow front progression
through mat preform for different
key parameters of EIPR process
shows the frequency level, and the one after the latter A shows
the amplitude level of the system. According to these codes,
the F0A0 one shows the non-control experiment result.
All experiments are run in the same mold geometry, test
conditions, and test fluid. Table 2 presents the non-control
(F0A0) permeability of preforms which are obtained as a ref-
erence. Each test is repeated three times and the average of
them is presented. Table 3 gives the permeability values of
EIPR process tests. Flow front position versus time is detected
and recorded by the program written in MATLAB. This pro-
gram calculates the flow front distance from the infusion line
by using image processing. The results for all preform are
shown in Figs. 7, 8, and 9.
Fig. 8 Flow front progression
through multiaxial preform for
different key parameters of EIPR
process
Int J Adv Manuf Technol (2021) 115:345–365
Figure 7 represents the results of mat preform. In this figure,
one can see the effect of the EIPR process in reducing the filling
time. It reduces the filling time from 34 to 44% regarding differ-
ent process parameters. Similarly, the flow progression is pre-
sented for multiaxial one in Fig. 8. By comparing, it is clear that
there is a reduction from 20 to 44% in filling time for this case.
Sequentially, recorded flow front with respect to time for twill
preform is given in Fig. 9. Twill fabric has the lowest permeabil-
ity among the preforms. In this case, there is a 42 to 60% reduc-
tion in filling time. The lowest filling time belongs to F2A2 test
series. By considering the filling time of the experiments, it can
be concluded that the EIPR process is more effective for the low
permeability preforms rather than high permeability ones.
Int J Adv Manuf Technol (2021) 115:345–365 355

Fig. 9 Flow front progression

through twill preform for different
key parameters of EIPR process

7.2 Response surface analysis and regression model Results of RSM analysis are given in terms of permeability
for EIPR process permeability are shown in Table 4. The probability values for frequency
and amplitude are more than 99%. Therefore, these factors are
To obtain a mathematical model of EIPR process permeability effective and significant. Also, P-value for material effect is
response, a statistical analysis is required. As mentioned be- more than 99% which shows its significance. It is observable
fore, the tapping frequency and amplitude are taken as inde- that the interactions of factor squares are effective and signif-
pendent factors to evaluate their effects on the corresponding icant. P-values of all are more than 99% expect for frequency/
response. In this analysis, results are investigated based for a amplitude and frequency/material interactions; these values
confidence factor of 95% by taking P-value α = 0.05 . are 96%. Figure 10 shows Pareto charts that illustrate the
Therefore, when the probability of factors is more than 95% effectiveness of the independent factors on the preform
or α ≤ 0.05, it is considered as a significant factor. For P- permeability.
values more than 0.05, the factors are rejected as insignificant. Lack-of-fit is an index that shows a regression model
is not significantly describe the extracted model between
the factors and response. If lack-of-fit is significant, it
Table 4 Analysis of variance for transformed response can be due to exclude of quadratic terms or exist of
unusually large residual results from the fitting the mod-
Source DF Adj SS Adj MS F-value P-
el. Moreover, in this analysis, lack-of-fit is not signifi-
value
cant with P-value 0.122, so the extracted model fits the
Model 11 1.61980 0.147255 593.47 0.000 experiment data satisfactorily.
Linear 4 1.57425 0.393562 1586.14 0.000 The transformed regression models with the determined
Freq. 1 0.00575 0.005748 23.17 0.000 coefficients for forecasting response of permeability of all
Amp. 1 0.02128 0.021280 85.76 0.000 preforms take the following form:
Mat. 2 1.54722 0.773610 3117.82 0.000 Regression model for mat:
Square 2 0.03624 0.018121 73.03 0.000
Permeability0:208834 ¼ 1:5614 þ 0:03057 F þ 0:3551 A
Freq. × freq. 1 0.01847 0.018470 74.44 0.000
−0:002332 F  F−0:2759 A  A
Amp. × amp. 1 0.00511 0.005110 20.60 0.000
−0:00743 F  A
2-way interaction 5 0.00932 0.001863 7.51 0.000
ð16Þ
Freq. × amp. 1 0.00121 0.001207 4.86 0.036
Freq. × mat. 2 0.00334 0.001669 6.72 0.004 Regression model for multiaxial:
Amp. × mat. 2 0.00477 0.002386 9.62 0.001
Error 27 0.00670 0.000248 Permeability0:208834 ¼ 1:6656 þ 0:03754 F þ 0:4622 A
Lack-of-fit 15 0.00459 0.000306 1.74 0.170 −0:002332 F  F−0:2759 A  A
Pure error 12 0.00211 0.000176 −0:00743 F  A
Total 38 1.62650 ð17Þ
356 Int J Adv Manuf Technol (2021) 115:345–365

Fig. 10 Effectiveness of all terms

in form of Pareto chart for
permeability (α = 0.05)

Table 5 Model summary for

transformed response S R-sq R-sq (adj) R-sq (pred)

0.0157520 99.59% 99.42% 98.99%

Regression model for twill: where F and A in these equations are frequency and
amplitude.
Permeability0:208834 ¼ 1:2044 þ 0:03189 F þ 0:4769 A R2 is another criterion to evaluate the regression model in
−0:002332 F  F−0:2759 A  A predicting results. The more this value is close to 100%, which
−0:00743 F  A is the more accurate prediction result. Table 5 presents the R2
ð18Þ values for the transformed response.

Fig. 11 Residual plots for permeability. a Normal probability. b Versus fits. c Histogram. d Versus order
Int J Adv Manuf Technol (2021) 115:345–365
Normal probability of residuals for the preform permeability
response is illustrated in Fig. 11 a and b. This type of plots
reveals whether a particular distribution fits to the collected data
and allows the comparison of sample distributions. Points close
to the distribution line and close together mean good fitness of
the selected distribution. It is observable that the points are very
close to the fit line, i.e., the normality assumption is valid in this
analysis. The histogram plot (Fig. 11c) also shows the normal
distribution without skew and existence of outliers. Residuals
versus fitted values plot show a random pattern of residuals on
both sides of zero as illustrated in Fig. 11d. And this confirms
the constant variance assumption in experiment data. From the
residuals versus order of data, it is noticeable that the residuals
are uncorrelated with each other.
To show the effect of the independent variables, 3D
surface and 2D contour plots are utilized, while holding
Fig. 12 a 3D surface and b 2D
contour plot for mat preform on
permeability
357
the material at fixed level at the same time. Figures 12,
13, and 14 display 3D surface and 2D contour plots of
mat, multiaxial, and twill fiber preforms, respectively. It is
observable as Fig. 15; the preform permeability increases
and then decreases with increasing frequency. However, it
increases with increasing amplitude.
Finally, an optimization study is performed using the
response optimizer where the aim is to maximize the
permeability. The optimum values depend on the pre-
forms for frequency and amplitude that are found to
b e 5 . 6 – 5 .7 H z a n d 0 . 5 – 0 . 7 m m , r es p e c t i v e l y .
Obviously, these values are the optimum values for
the selected range of these parameters. Corresponding
permeabilities of mat, multiaxial, and twill preforms
are 14.580 × 10−11, 24.93 × 10−11, and 6.183 ×
10−11 m2, respectively.
358
Fig. 13 a 3D surface and b 2D
contour plot for multiaxial
preform on permeability
7.3 Evaluation and simulation results of EIPR process
The simulations of two planned tests are conducted using
PAM-RTM software for both EIPR and non-control process-
es. Principal permeability of twill preform for both processes
is calculated based on the test data at 0°, 45°, and 90° directions
to define the material property in the software. Since mat
preform has a homogenous in-plane property, only 0° direc-
tion permeability test is done to define its property. It must be
noted that since control action is done on the low permeability
section, the property of mat preform under EIPR process is not
needed to define in simulation software. Permeability property
of the preforms is given in Table 6.
Compressibility of preform is another property that must be
defined in PAM-RTM software. The pressure response of mat
Int J Adv Manuf Technol (2021) 115:345–365
(5 layers) and twill (12 layers) preforms is demonstrated as a
function of fiber volume fraction (based on power law) in
Figs. 16 and 17.
To evaluate the EIPR process numerically, a comparison
study is conducted for the filling process considering two
different cases. EIPR and non-control processes are simulated,
and flow front pattern, filling time, and dry spot formation of
each case are compared with experiments.
Figure 18 shows the forecasted filling pattern of case 1
for EIPR process in comparison with two frames of exper-
imental results at 300 and 550 s of infusion. In the figure,
the low permeability zone is illustrated with a black rect-
angle. These frames clearly show that simulation of EIPR
process predicts flow patterns and filling time correctly. As
it is expected, the EIPR process is shown to avoid the
Int J Adv Manuf Technol (2021) 115:345–365
Fig. 14 a 3D surface and b 2D
contour plot for twill preform on
permeability
formation of dry spots both in simulations and in
experiments.
For comparison purposes, one can easily see the dry spot
formation for the no-control process, as shown in Fig. 19. The
flow starts from the left side of the plate in the EIPR process,
and it progresses towards the low permeability section in the
center. When flow reaches the low permeability zone which is
at the center, it faces a resistance in this region and it flows
faster around this section through two high permeability chan-
nels. Finally, the fluid flow in these two channels joins in the
right-hand side of the low permeability zone and a dry spot is
formed there. The simulation also predicts the unsaturated
zone in non-control process correctly. In the presented figure,
a frame of flow pattern before air entrapment and formed dry
359
spot at the end of filling process, experimentally and numeri-
cally, is exemplified.
Figure 20 illustrates the filling pattern of case 2 in the EIPR
process in comparison with the experimental one. Two frames
of mold filling at times 357 and 459 s of process are selected to
validate the simulation results. Flow front geometries are very
like the simulation patterns corresponding to similar filling
times.
Figure 21 shows the simulation and experimental results
for non-control process. Similarly, reduction in the speed of
flow in the low permeability section causes an air trap because
of the fact that fluid flows faster in the high permeability
sections surrounding this low permeability section. Non-
control process simulation of case 2 predicts the dry spot
360 Int J Adv Manuf Technol (2021) 115:345–365

Fig. 15 Mean effects plot of frequency, amplitude, and material on preform permeability fitted mean

Table 6 Effective and principle permeability and orientation of elliptical flow

Permeability (10−11 m2) Ellipse

orientation
° ° °
Process Material K 0xx K 45
xx K 90
xx K1 K2 β

EIPR Twill 6.29 6.78 7.74 7.77 6.28 96.2°

VARTM Mat 8.47 - - 8.47 8.47 0°
Twill 1.72 1.86 2.42 2.47 1.68 102.1

Fig. 16 Pressure response of mat

preform
Int J Adv Manuf Technol (2021) 115:345–365 361

Fig. 17 Pressure response of twill

preform

formation as well, and the shape of dry spot zone reveals the
good agreement between simulation and experiment.
Mold filling time is one of the important parameters in the
production of composite parts since the gel time of resins does
not significantly vary. Thus, it can be used as an index for
comparison or evaluation purposes. Filling time values for
the simulations and tests are given in Table 7. Results show
that the simulation predictions are quite satisfactory. It can be
observed that the EIPR process in both simulations and
experiments reduces the filling time (approximately 10–
12%) for each case.
8 Conclusion
EIPR process is an active control methodology that detects the
flow front. It devises a relaxation strategy if there is an unex-
pected development in the flow front and stimulates a

Fig. 18 Case 1: filling pattern of (a) simulation and experiment at (b) 300 s and (c) 550 s
362 Int J Adv Manuf Technol (2021) 115:345–365

Fig. 19 Case 1: flow front at 445 s of VARTM process for (a) simulation and (b) experiment; dry spot formation in (c) simulation and (d) experiment

ferromagnetic element to rest the preform and pulsate it to that controls preform permeability is characterized. Leading
increase the local permeability. In this study, the EIPR process factors of the process are identified including frequency and

Fig. 20 Case 2: filling pattern of (a) simulation and experiment at (b) 357 s and (c) 459 s
Int J Adv Manuf Technol (2021) 115:345–365 363

Fig. 21 Case 2: flow front at 304 s of VARTM process for (a) simulation and (b) experiment; dry spot formation in (c) simulation and (d) experiment

amplitude of vibration. In order to evaluate this process and permeability first increases with increasing frequency and
obtain a confident model, three preform types with low, mean, then reduces. Meanwhile, it has an ever-increasing trend with
and high permeability properties values are selected. Central increasing amplitude. Simulation of this process, using the
composite design method is used to design of experiment. A experimental data, is evaluated. Two case studies which are
total of 39 experiments are conducted to study the effect of the stiffened in the middle using low permeability preform are
factors on the preform permeability. The filling time of pre- modeled for both EIPR and non-control processes. The ability
form infusion of EIPR process shows a significant reduction of the EIPR process to fill the mold completely without any
from 20 to 60% depending on the material and factor levels. dry spots is illustrated numerically. There is a good agreement
High reduction values relate to the low permeability preform. between the simulation and experimental findings as the dif-
The response surface methodology is utilized to analyze and ference between results is below 3%.
model the permeability response. Mathematical models are
obtained as a function of frequency and amplitude for each
material. It shows that the response defined in terms of Author contribution M. Poorzeinolabedin designed and performed the
experiments, derived the models, and analyzed the data. M.
Poorzeinolabedin wrote the manuscript in consultation with K.L.
Parnas. K.L. Parnas supervised the project.
Table 7 Filling time in simulations and experiments with/without the
EIPR process Availability of data and materials Not applicable.
Filling time (s)
Declarations
No control EIPR
Ethics approval Not applicable.
Simulation Experiment Simulation Experiment
Consent to participate Not applicable.
Case 1 674 680 591 604
Case 2 557 557 510 500 Consent for publication Not applicable.
364
Competing interests Not applicable.
References
1. Teoh KJ, Hsiao KT (2011) Improved dimensional infidelity of
curve-shaped VARTM composite laminates using a multi-stage
curing technique - experiments and modeling. Compos Part A
Appl Sci Manuf 42:762–771. https://doi.org/10.1016/j.
compositesa.2011.03.003
2. Khan LA, Mehmood AH (2016) Cost-effective composites
manufacturing processes for automotive applications. Elsevier
Ltd. https://doi.org/10.1016/B978-1-78242-325-6.00005-0
3. Fan J, Njuguna J (2016) An introduction to lightweight composite
materials and their use in transport structures. Elsevier Ltd. https://
doi.org/10.1016/B978-1-78242-325-6.00001-3
4. Ortiz de Mendibil I, Aretxabaleta L, Sarrionandia M, Mateos M,
Aurrekoetxea J (2016) Impact behaviour of glass fibre-reinforced
epoxy/aluminium fibre metal laminate manufactured by vacuum
assisted resin transfer moulding. Compos Struct 140:118–124.
https://doi.org/10.1016/j.compstruct.2015.12.026
5. Menta VK. Advanced composites using non-autoclave processes:
manufacturing and characterization 2011.
6. Advani SG (2017) Role of process models in composites
manufacturing. Ref Modul Mater Sci Mater Eng Elsevier. https://
doi.org/10.1016/B978-0-12-803581-8.09898-2
7. Holmes M (2017) Aerospace looks to composites for solutions.
Reinf Plast 61:237–241. https://doi.org/10.1016/j.repl.2017.06.079
8. Matsuzaki R, Shiota M. Data assimilation through integration of
stochastic resin flow simulation with visual observation during
vacuum-assisted resin transfer molding: a numerical study.
Compos PART A 2016;84:43–52. https://doi.org/10.1016/j.
compositesa.2016.01.006
9. Sreekumar PA, Joseph K, Unnikrishnan G, Thomas S. A compar-
ative study on mechanical properties of sisal-leaf fibre-reinforced
polyester composites prepared by resin transfer and compression
moulding techniques 2007;67:453–461. https://doi.org/10.1016/j.
compscitech.2006.08.025.
10. Hsiao K-T, Heider D (2012) 10 - vacuum assisted resin transfer
molding (VARTM) in polymer matrix composites. Woodhead
Publishing Limited. https://doi.org/10.1533/9780857096258.3.310
11. George A (2011) Optimization of resin infusion processing for
composite materials: simulation and characterization strategies.
University of Stuttgart
12. Sayre JR (2000) Vacuum-assisted resin transfer molding
(VARTM) model development, verification, and process analysis.
Analysis 179
13. Devillard M, Hsiao KT, Advani SG (2005) Flow sensing and con-
trol strategies to address race-tracking disturbances in resin transfer
molding - part II: automation and validation. Compos Part A Appl
Sci Manuf 36:1581–1589. https://doi.org/10.1016/j.compositesa.
2004.04.009
14. Johnson RJ, Pitchumani R (2008) Active control of reactive resin
flow in a vacuum assisted resin transfer molding (VARTM) pro-
cess. J Compos Mater 42:1205–1229. https://doi.org/10.1177/
0021998308091264
15. Nielsen DR, Pitchumani R (2002) Closed-loop flow control in resin
transfer molding using real-time numerical process simulations.
Compos Sci Technol 62:283–298. https://doi.org/10.1016/S0266-
3538(01)00213-5
Int J Adv Manuf Technol (2021) 115:345–365
16. Nielsen DR, Pitchumani R (2002) Control of flow in resin transfer
molding with real-time preform permeability estimation. Polym
Compos 23:1087–1110. https://doi.org/10.1002/pc.10504
17. Nalla AR, Fuqua M, Glancey J, Lelievre B (2007) A multi-segment
injection line and real-time adaptive, model-based controller for
vacuum assisted resin transfer molding. Compos Part A Appl Sci
Manuf 38:1058–1069. https://doi.org/10.1016/j.compositesa.2006.
06.021
18. Modi D, Correia N, Johnson M, Long A, Rudd C, Robitaille F
(2007) Active control of the vacuum infusion process. Compos
Part A Appl Sci Manuf 38:1271–1287. https://doi.org/10.1016/j.
compositesa.2006.11.012
19. Johnson RJ, Pitchumani R (2003) Enhancement of flow in
VARTM using localized induction heating. Compos Sci Technol
63:2201–2215. https://doi.org/10.1016/S0266-3538(03)00179-9
20. Matsuzaki R, Kobayashi S, Todoroki A, Mizutani Y (2013) Flow
control by progressive forecasting using numerical simulation dur-
ing vacuum-assisted resin transfer molding. Compos Part A Appl
Sci Manuf 45:79–87. https://doi.org/10.1016/j.compositesa.2012.
09.014
21. Alms J, Advani SG (2007) Simulation and experimental validation
of flow flooding chamber method of resin delivery in liquid com-
posite molding. Compos Part A Appl Sci Manuf 38:2131–2141.
https://doi.org/10.1016/j.compositesa.2007.06.011
22. Alms JB, Advani SG, Glancey JL (2011) Liquid composite mold-
ing control methodologies using vacuum induced preform relaxa-
tion. Compos Part A Appl Sci Manuf 42:57–65. https://doi.org/10.
1016/j.compositesa.2010.10.002
23. P. Simacek, E.M. Sozer, S.G. Advani user manual for LIMS 4.0,
Center for Composite Materials, University of Delaware (1998) n.d.
24. Polyworx inc. n.d. http://www.polyworx.com/doc/.
25. ESI Group International Ltd. n.d. https://www.esi-group.com.
26. Song X. Vacuum assisted resin transfer molding (VARTM): model
development and verification. Dr Thesis 2003:161.
27. de Oliveira IR, Amico SC, Souza JÁ, de Lima AGB (2013) Resin
transfer molding process: a numerical and experimental investiga-
tion. Int J Multiphys 7:125–136. https://doi.org/10.1260/1750-
9548.7.2.125
28. Kozioł M. Simplified simulation of vari process using PAM-RTM
software 2016.
29. Isoldi LA, Oliveira CP, Rocha LAO, Souza JA, Amico SC. Three-
dimensional numerical modeling of RTM and LRTM processes. J
Brazilian Soc Mech Sci Eng 2012;XXXIV:105–11. 10.1590/
S1678-58782012000200001.
30. Letzow M, Amico SC, Souza JA, Isoldi LA. Computational model-
ing of RTM and LRTM 2012;11:93–99.
31. Poorzeinolabedin M, Parnas L, Dashatan SH (2014) Resin infusion
under flexible tooling process and structural design optimization of
the complex composite part. Mater Des 64:450–455. https://doi.
org/10.1016/j.matdes.2014.08.008
32. Poorzeinolabedin M, Parnas KL (2019) Flow correction control
with electromagnetically induced preform resting process. Adv
Manuf 7:199–208. https://doi.org/10.1007/s40436-019-00257-2
33. Box AGEP, Wilson KB, Journal S, Statistical R, Series S. On the
experimental attainment of optimum conditions Published by :
Wiley for the Royal Statistical Society Stable URL : http://www.
jstor.org/stable/2983966 2017;13:1–45.
34. Darcy H. Les fontaines publiques de la ville de Dijon. Recherche
1856.
35. Vernet N, Ruiz E, Advani S, Alms JB, Aubert M, Barburski M,
Barari B, Beraud JM, Berg DC, Correia N, Danzi M, Delavière T,
Dickert M, di Fratta C, Endruweit A, Ermanni P, Francucci G,
Garcia JA, George A, Hahn C, Klunker F, Lomov SV, Long A,
Int J Adv Manuf Technol (2021) 115:345–365 365

Louis B, Maldonado J, Meier R, Michaud V, Perrin H, Pillai K, permeability and compressibility measurements. Polym Compos
Rodriguez E, Trochu F, Verheyden S, Wietgrefe M, Xiong W, 12:20–29. https://doi.org/10.1002/pc.750120105
Zaremba S, Ziegmann G (2014) Experimental determination of
the permeability of engineering textiles: Benchmark II. Compos Publisher’s note Springer Nature remains neutral with regard to jurisdic-
Part A Appl Sci Manuf 61:172–184. https://doi.org/10.1016/j. tional claims in published maps and institutional affiliations.
compositesa.2014.02.010
36. Trevino L, Rupel K, Young WB, Liou MJ, Lee LJ (1991) Analysis
of resin injection molding in molds with preplaced fiber mats. I: