Composites: Part B 57 (2014) 47–55

Contents lists available at ScienceDirect

Composites: Part B

journal homepage: www.elsevier .com/locate /composi tesb

Process simulation for a large composite aeronautic beam by resintransfer molding

1359-8368/$ - see front matter � 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.compositesb.2013.09.039

⇑ Corresponding author. Tel.: +39 06 49 919 756; fax: +39 06 49 919 757.E-mail address: susanna.laurenzi@uniroma1.it (S. Laurenzi).

S. Laurenzi a,⇑, A. Grilli a, M. Pinna a, F. De Nicola b, G. Cattaneo c, M. Marchetti a

a Department of Astronautic Electrical and Energy Engineering, Sapienza Università di Roma, Italyb Advanced Materials and Technologies Laboratory, Centro Italiano Ricerche Aerospaziali S.c.p.A., Italyc Research & Innovation Technology, Alenia Aermacchi S.p.A, Italy

a r t i c l e i n f o a b s t r a c t

Article history:Received 21 November 2012Received in revised form 9 September 2013Accepted 16 September 2013Available online 26 September 2013

Keywords:E. Resin transfer molding (RTM)A. Polymer–matrix composites (PMCs)Numerical analysis

This paper presents the numerical process analysis and the experimental investigation for the manufac-turing of a reinforced carbon-fiber demonstrator of a large aeronautic beam by resin transfer molding(RTM). The component is a primary structure characterized by several thick sections with abrupt changesin shape that complicates the resin impregnation of the preform. Process simulations based on a finiteelement method-modified control volume (FEM-CV) were conducted to investigate the resin flow frontpatterns and find the injection scheme that guarantees both a good impregnation of the preform and afilling time compatible with the resin gel time. The beam component was successfully manufactured,and a good agreement between the numerical analysis and the fabrication process was demonstrated.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

Aerospace industries always investigate new technology solu-tions responding to the market pressure and the technology de-mands. In order to comply with the environmental directions onCO2 emissions, the next global objective is to reduce 50% of con-sumed fuel by 2020 and a further 20% by 2025. These objectivescan be reached in several ways, the use of lightweight structuresbeing the most cost effective from the industrial point of view.For these reasons, aerospace companies, which are traditionallybased on the use of metal alloys, have turned to the research anddevelopment of composite polymeric materials. The main advanta-ges of polymeric composites with respect to metals, such as resis-tance to corrosion and fatigue, and high performance/weight ratio,are a set of qualities for winning the current and future applica-tions. In this context, resin transfer molding (RTM) is a cost-com-petitive process to manufacture composite structures foraeronautics [1–5]. Many parts manufactured using RTM in theaerospace field have been mostly related to non-critical structures,whereas the development of large critical structures by RTM stillrequires large research efforts. A good design for RTM leads to fab-rication of three-dimensional near-net-shape parts, offering pro-duction of cost-effective structural parts in medium volumequantities [2,6].

Also, so far RTM process have been used for the manufacture ofthin laminate structures and very little work on thick composites

can be found in the literature. As a matter of fact, manufacturinga thick-sectioned structure by RTM is challenging and severalimportant processing considerations have to be accounted for.The highly exothermic nature of thermoset resins and the limitedtemperature control make it difficult to avoid detrimental thermaland cure gradients within the composite [7]. Moreover, the resintransverse flow front, due to the preform permeability throughthe section, becomes relevant. The resin flow patterns are difficultto predict, making the gates and vents location analysis even moredifficult to perform [6]. In the case of thick-section composites,therefore, the process features are a serious limitation for the man-ufacture of composite materials for critical structures. Processmodeling can accelerate the path from conception to prototype,thus reducing industrial costs and time.

In this work, numerical and experimental studies were per-formed to manufacture an aeronautic beam demonstrator in com-posite material that is traditionally made by metal alloy. Thecomponent under investigation is a beam of a thrust reverser,which is a large primary structure with complex geometry(Fig. 1). The composite beam was redesigned from the metallicand thicknesses ranging from 3 mm to 33 mm. An inverse engi-neering approach was applied to determine the permeability val-ues needed to perform the simulation process. Filling simulationsbased on finite element-modified control volume method wereconducted in order to find the injection scheme that guaranteesboth a good impregnation of the preform and a filling time compat-ible with the resin gel time. The beam component was manufac-tured using a RTM mold designed according to the results of theprocess simulations.

Fig. 1. Beam element of thrust reverser in nacelle (top); the beam is typically 1000 mm long and it has complex 3D shape with several thick sections as shown in CAD model(bottom).

48 S. Laurenzi et al. / Composites: Part B 57 (2014) 47–55

2. Materials and methods

2.1. Materials

Hexcel G0926 carbon fiber reinforcements were used for thepermeability experiments and for the component fabrication. Theweave was a 5H Satin with areal weight of 360 g/m2. The HenkelEpsilon 99900 binder was used for the preforming. The binderwas dispersed at 10 wt% loading with respect to the fiber reinforce-ment, and the preform was maintained under vacuum at 120 �C for1 h.

The resins used for the experimental tests were the epoxy Hex-cel RTM6 and the benzoxazine Henkel Epsilon II 99110, whereasthe beam demonstrator was manufactured using the benzoxazinesystem. This resin was selected because of its low exothermicbehavior that guarantees a large processing window, thus avoidingpremature resin gelling that can occur as a consequence of thelarge dimensions of the component. The two resin systems wereused in according to the process cycles recommended by the pro-ducer. Table 1 summarizes the processing conditions adopted forthe two resin systems during the experimental and numericalinvestigations.

2.2. Manufacturing

The resin selected to manufacture the beam requires that theentire apparatus (injection machine, dispensing resin and mold)

Table 1Processing conditions adopted for experimental tests and numerical analysis. The demonreported in this table.

Resin Density (g/cm3) Pre-heating T (�C) Injection T (�

HexFlow RTM 6 1.12 80 80Epsilon II 99110 1.22 110 110

is pre-heated to 80–90 �C before starting the impregnation phase.In our experiments, we used a commercial Hyperjet machine formono-component polymers, which injects the resin at a constantpressure. The resin was loaded, degassed and heated inside theRTM equipment. When the set temperatures of both the resinand the mold were reached, the resin was pumped inside the molduntil it came out from the vent. After the fiber preform was com-pletely saturated with the resin, curing reactions were allowed tocontinue past the gel point to form a cross-linked polymericstructure.

Due to the complex geometry of the beam, several performpieces were assembled to achieve the final part shape. Binder pow-der was applied to stabilize the layers. The lay-up of each part wasdesigned using the commercial FiberSim software, in order toavoid fiber angle deviation and waste of material during the pre-form preparation, and for the reproducibility of the process. Theplies were cut using an automatic machine and then placed inthe mold. Finally the preform is consolidated under vacuum.Fig. 2 shows the process steps from the design to the perform fab-rication. In particular, two consecutive ply creation steps areshown in Fig. 2a and b, where the green lines represent the bound-ary layers. Fig. 2c shows the layer after the automated cutting andthe mold used to pre-shape the part, finally Fig. 2d shows a pictureof the consolidation phase.

The beam was designed with quasi-isotropic laminations con-sidering constant fiber volume fraction. The stacking sequenceadopted for the characteristic thicknesses of the preform is

strator was fabricated with the process parameter of the system Epsilon II 99110 as

C) Viscosity at Tinj (mPa s) Gel time (min) Cure cycle

180 240 75 min at 160 �C100 240 90 min at 180 �C

Fig. 2. Steps showing from concept to realization of a part of the preform; first, the shape and paths of each layer is studied by simulating the laying and the overlapping (aand b). When the ply-book is completed, the layers are cut by an automatic machine (c), finally the layers are positioned on the mold and the preform is consolidated undervacuum.

S. Laurenzi et al. / Composites: Part B 57 (2014) 47–55 49

reported in Table 2. The beam mold was designed in steel alloywith an optimal distribution of electrical heating points to achievethe uniform temperature required for the process. The heating sys-tem was controlled by a unit connected with thermocouples insidethe mold. During the injection phase, the temperature gradient

Table 2Characteristics of stacking sequence adopted for the typical thicknesses of thepreform; these features were also used for the experimental part related topermeability calculation.

Thickness(mm)

Fiber volumefraction

Number ofplies

Lamination sequence

3 0.55 8 [0�/+45�/�45�/90�]S

10 0.55 26 [0�/(0�3/+45�3/�45�3/90)]S

20 0.55 52 [0�/(0�3/+45�3/�45�3/90)2/90�]S

30 0.55–0.53 76 [0�/(0�3/+45�3/�45�3/90)3/90�]S

was also monitored by a thermal camera. In order to avoid racetracking of the resin between the preform and the mold wall, theends of the preforming tools were sealed using silicon blocks. Amultipart mold design was adopted to allow for an easy de-molding.

The RTM apparatus described above was also used to measurethe experimental filling time presented in Section 4.1. In this case,the mold was a flat metal tool with dimensions of400 mm � 400 mm, and the gap-height was adjusted in the rangebetween 3 mm and 30 mm by spacer frames. The resin was injectedthroughout the mold using a circular inlet while the outlet vent waspositioned in opposite part. The inlet provided an internal linearchannel in order to produce a uniform preform impregnation. Theseexperimental parts had thicknesses specified in Table 2, using the fi-ber reinforcement and the resins selected for the beam (Table 1) inorder to determine the filling time for the permeability analysis andto investigate possible problems related to the use of the materials.The thicknesses and the related lay-up sequence used for each platewere identified as the most representative for the beam.

Fig. 3. View of the FEM model adopted in the simulation process of the beam withthe indication of the regions for the injection scheme.

50 S. Laurenzi et al. / Composites: Part B 57 (2014) 47–55

3. Flow analysis

The numerical analysis was performed using the commercialcode MoldFlow. The modeling of the resin flow allows investiga-tion of the resin flow patterns during the impregnation processand strategical designing of the gates and vents locations [8]. Acritical issue is the optimal filling of the composite part whileavoiding dry spots and premature gelling of the resin.

The RTM process is simulated using numerical methods basedon finite element-modified control volume (FEM-CV) analysis. Thismethod uses the porous media flow approach, solving the Darcyequation coupled with the continuity and energy equations atany instant during the filling process for given boundary condi-tions. The control volume consists of a fabric unit cell formed byfibers and porosity.

The continuity equation is derived from the mass balance of theflow rate in the control volume:

@q@tþ U � rqþ qr � U ¼ s ð1Þ

where q is the actual density of the resin, that is the mass of resindivided by the control volume containing both resin and fiber, andmay not be constant within the control volume; U

�is the resin veloc-

ity, and s are the sink effects due to the porous medium.The Darcy’s law Eq. (2) describes the flow of a viscous fluid

through an anisotropic, homogenous, porous medium:

ux

uy

uz

0@

1A ¼ � 1

lð1� mf Þ

Kxx Kxy Kxz

Kyx Kyy Kyz

Kzx Kzy Kzz

0@

1A @P

@x@P@y@P@z

0@

1A ð2Þ

where l is the resin viscosity, vf is volume fraction of fiber, and1 � vf is the porosity of the control volume that is occupied by theresin. Eq. (2) is written in components along the three principalaxes, and gives the average velocity profile that is directly propor-tional to the permeability tensor K

��

and the pressure gradient rP.The boundary conditions are given by the part geometry, the pres-sure value (or flow rate) at gate and vent positions.

Usually, the numerical analysis is assumed to be isothermal inorder to reduce both the cost-time of the runs and the number ofphysical parameters. In particular, this assumption greatly simpli-fies the model, drastically reducing the computational time toreach the convergence, which in our simulations took approxi-mately 320–340 s of CPU time. Further, in this case, the materialproperties necessary for the simulation are limited to the resin vis-cosity and the permeability tensor.

The FEM model adopted for the analysis of the full scale beamconsisted of 35,042 dual-domain triangular elements, as shownin Fig. 3. A sensitive mesh analysis performed on the optimal injec-tion, showed that with this number of elements, the simulationconverges to constant values in terms of filling time. Increasingthe number of elements leads only to an increase in the computa-tional time. For the set up where the mold is maintained at con-stant temperature for the entire impregnation phase, simulationswere carried out isothermally considering for the resin viscosityvalue at the mold temperature as reported in the material data-sheets (Table 1). Material properties, such as permeability value, fi-ber volume fraction and part thickness, were assigned to eachelement allowing for non-uniform properties and variablethickness.

Regarding the process parameters, the value of the injectionpressure was considered constant at 3 bars for the entire numericalanalysis. This value was extrapolated from experiments conductedon representative flat laminates to measure the filling time for thepermeability determination. In particular, it was observed that an

injection pressure above 3 bars produces fiber washing phenom-ena and local reinforcement deformations.

The main objective of the simulations was to determine theinjection scheme that guaranteed a good impregnation of the beampreform in agreement with the process window. In particular, thetotal filling time of the mold had to be lower than the resin geltime. This requirement is strongly influenced by the injection strat-egy. For components with simple geometry, the choices for theinjection gate positions are limited and can be easily adjustedusing simple simulations. In our case, the beam geometry wasrather complex and the choice of the initial injection scheme wasfundamental to reduce the number of possible gate locations. Forthis reason, we performed the flow analysis considering a potentialinjection scheme established by taking into account some physicaland technological observations. First of all, the flow rate through aporous medium drops during the impregnation phase, and in par-ticular tends to decrease drastically in the proximity of abrupt geo-metrical changes such as T-shape edges and increased thicknesses.In order to compensate for the pressure drop it would be necessaryto increase the injection pressure. On the other hand, this solutioncan produce a local fiber deformation and fiber washing phenom-ena, which affect the quality of the final product. Alternatively, theuse of schemes with multiple gates can be adopted, even though itis not recommended for complex shapes. The running of severalflow fronts can create dry spots and introduce operational difficul-ties, because each of them needs to be controlled and opened at theright time. Based on these considerations, there was a clear prefer-ence for the injection channels scheme (Fig. 3). In particular, thestarting injection scheme was linear and positioned along the thickpart of the beam corresponding to the elongated edge. This wouldallow for one single resin inlet port, thus simplifying the resin dis-pensing system. In order to compensate for the pressure aroundthe ribs on the bottom side, additional gates were considered onthe short edge. The vent points were positioned on opposite sidesand close to the torque box to push the resin through it. This initialinjection scheme was adjusted until satisfactory results in terms offilling times and preform impregnation were reached. The optimalinjection scheme was then used to design the mold for the beam.

4. Results and discussion

4.1. Determination of permeability

In general, preforms show different permeability value in differ-ent directions due to the anisotropy of the fiber architecture. As thethickness of the composite part is typically negligible comparedwith the in plane-dimensions, the permeability assumes a

Fig. 4. Example of typical curves of the square of flow front position xf as function ofthe filling time tf obtained for two different injection pressure Pinj. The dashed linerepresents the linear approximation used to calculate the permeability.

S. Laurenzi et al. / Composites: Part B 57 (2014) 47–55 51

two-dimensional form, where the in plane-directions flow resis-tances are considered predominant. On the other hand, in the caseof thick preforms, the permeability value in the out-of-plane direc-tion cannot be neglected due to the flow front through the thick-ness. This value is difficult to determine experimentally. Severalauthors relied on laborious mathematical models that often re-quire a large amount of experimental data [2,9–17]. From anindustrial point of view, these approaches are not cost-effectiveand a faster method to obtain input for the design of the injectionscheme is required. In order to overcome these issues, in our workwe introduced an effective in-plane permeability value, which alsotakes into account the transverse flow through the thickness. Thisapproach started from the observation that the permeability K isindirectly determined by experimental data. Usually, the in-planemeasurements consist of recording the flow front position at eachtime, and then fitting the data by the integral of the Darcy equationbetween the position of the injection point at time t0 (x0, y0) andthe position of the flow front at time tf(xf, yf). The integral of theDarcy’s law for the one-dimensional experimental approach isgiven by the following expression:

xf ¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2KXrPtf

lu

sð3Þ

where the viscosity l, the pressure gradient and porosity u areknown, the xf and tf are experimentally determined. The permeabil-ity value is calculated as the slope of the curve obtained from Eq.(3):

Kx ¼lu

2rP

x2f

tfð4Þ

The permeability values determined experimentally suffer gen-erally from lack of reproducibility, and a large number of experi-ments are required to obtain an average value. We should stressthat typically the permeability is calculated from the linear partof the curve assuming that the resin flow can be considered qua-si-stationary. In reality, the curve can be not linear due to thedual-scale porosity of the preform, and therefore the permeabilityvalue is an approximation of several factors that relate to the differ-ent resin flows through the preform. An example of typical trend ofthe squared flow front position (x2

f ) as a function of the filling time(tf) is presented in Fig. 4. In the graph, the dashed line is the linearapproximation of the curve used to determine the permeability. Theplotted data refer to the real case of a carbon fiber-reinforced lam-inate for permeability experiments that were conducted using twodifferent injection pressures, as indicated in the picture. Startingfrom these considerations, we considered a different approachbased on the simple measurement of the total filling time necessaryto impregnate a rectangular plate. The experimental set-up waspreviously described in Section 2.2. After determining the preformfilling time through experiments, an iterative numerical analysiswas run by varying the permeability value until the same fillingtime obtained experimentally was reached. The numerical analysiswas based on a FEM-CV model reproducing exactly the part geom-etry as the experiment, and with the material data (viscosity and fi-ber volume fraction) as those adopted in the experiments. In thisapproach, the permeability value is resolved along the total geo-metrical dimensions and the total filling time, and it includes theinteraction of fiber reinforcement and resin that are used in reality.In fact, the traditional experimental characterization of the perme-ability requires the use of a liquid (not resin) that interacts differ-ently with the fibers. In our case, the effects of the flow throughthe thickness are included in a global permeability value. Thismethod is particularly suitable for quasi-isotropic laminates, wherethe permeability values are almost the same. In our procedure thein-plane permeability values Kxx and Kyy were assumed to be equal

following the quasi-isotropic lay-up sequence preform. Followingthe laminate characteristics summarized in Table 2, the permeabil-ity values that we determined with this global method were6 � 10�12 m2 for thickness of 30 mm, 3.6 � 10�11 m2 for 20 mm,4.73 � 10�11 m2 for 10 mm, and 12.4 � 10�9 for 3 mm.

4.2. Simulation process

The flow analysis was used to simulate the mold filling processduring the impregnation phase. The resin flow front position foreach filling time, the pressure distribution at the end of the pro-cess, and the total mold filling time were the output of the simula-tions. From these simulations, potential macrovoids and problemsrelated to the impregnation of the preform can be extrapolated.Three different injection schemes were selected based on theconsiderations discussed earlier. The process conditions for thesimulations are the following:

� Case 1: linear injection channel along the entire edge of thetransverse beam.� Case 2: linear injection channel along the length of the beam

except for the torque box area.� Case 3: linear injection channel further reduced with respect to

simulation Case 2.

Results of the simulations show that varying the length of themain linear channel strongly affects the resin flow front positionand the filling time of the impregnation step. Fig. 5 shows the casewhere the injection scheme is positioned along the entire length ofthe beam (simulation Case 1). Here, the flow front reaches the ventposition before finishing the impregnation of the torque box re-gion, which remains dry because the resin flow circumvents thiszone. Upon reducing the length of the main channel as in simula-tion Case 2, the flow slows down in the torque box area and theimpregnation is guaranteed (Fig. 6). A further reduction of theinjection channel (simulation Case 3) causes a drastic pressuredrop around the torque box (Fig. 7). By comparing the flow frontposition of simulation Cases 2 and 3 at the same time (Fig. 8), wecan observe that in simulation Case 3 the torque box region is stillnot impregnated, unlike the situation in simulation Case 2. In sim-ulation Case 3 to obtain a total impregnation of the preform, a timeof 3 h 40 min is necessary, whereas in simulation Case 2 the re-quired time is about 2 h. Furthermore, as previously noted, this isthe result of a numerical analysis conducted under isothermalassumptions. We have to stress that in the actual process the resin

Fig. 5. Simulation Case 1; numerical results of injection scheme with the injection channel along the thick edge and on the short one. In particular, the picture shows the flowfront position as a function of the filling time and the pressure gradient distribution inside the mold at the end of the filling. The total filling time is �1 h 54 min.

Fig. 6. Simulation Case 2; numerical results of the scheme with the injection channel along the thick edge having a reduced length respects to the Case 1. In particular, thepicture shows the flow front position as a function of the filling time and the pressure gradient distribution inside the mold at the end of the filling phase. The total filling timeis 2 h 6 min.

52 S. Laurenzi et al. / Composites: Part B 57 (2014) 47–55

is exothermic and a significant viscosity increase occurs during thistime. As reported in the material datasheet, the viscosity at 110 �Ccan be considered constant for about 3 h. After this time the viscos-ity increases, and at approximately 4 h, as in the simulation Case 3,the viscosity becomes five times the initial value. Therefore, theactual time of the impregnation phase is larger than the numericalone, with a high of probability to run into premature gelation of the

resin and the formation of dry regions due to the pressure drop.Fig. 9 compares the flow patterns of simulation Cases 1 and 2 atthe same time step before reaching the vent. It is clear that in Case1 the resin reaches the vent before impregnating the entire pre-form. The expected dry zones are indicated in Fig. 9.

Fig. 10 gives an enlarged view of the torque box area for thecases considered here. In particular, the picture shows the flow

Fig. 7. Simulation Case 3; numerical results of the scheme with the injection channel along the thick edge having a reduced length respects to the Case 2. In particular, thepicture shows the flow front position as a function of the filling time and the pressure gradient distribution inside the mold at the end of the filling phase. The total filling timeis 3 h 40 min.

Fig. 8. Comparison of flow patterns for simulation Cases 2 and 3 at a given filling time.

S. Laurenzi et al. / Composites: Part B 57 (2014) 47–55 53

front position at the filling time when the resin reaches the vent forthe injection scheme of simulation Case 1. In this case, a large dryarea is formed in proximity of the torque box by the time the resinhas reached the vent (Fig. 10a). This is due to fact that the injectionscheme is along the entire length of the beam, and the resin injec-tion channel is near to the vent. Therefore, the resin reaches thevent before finishing the impregnation of the entire torque box

area. In addition, the resin infuses fast through theT-shaped edges with higher permeability circumventing the zonewith larger thickness. In simulation Case 2, the length of the injec-tion channel goes up to the torque box region (Fig. 10b). In this sit-uation, the resin has time to impregnate the preform and proceedsuniformly towards the vent. Continuing to reduce the length of

Fig. 9. Comparison of flow patterns for Case 1 and Case 2 at a given filling time. In Case 1, the red arrows point to the expected dry zone caused by the resin reaching the ventbefore the impregnation of the preform is completed.

Fig. 10. Comparison of the flow front position at the given filling time around the torque box areas for the three simulation cases under consideration. In simulation Case 1, adry area occurs in correspondence of the torque box when the resin has already reached the vent (a). In simulation Case 2, the resin flow front proceeds uniformly towards thevent (b). In simulation Case 3, the torque box area shows a large dry zone when the resin is in proximity of the vent (c).

54 S. Laurenzi et al. / Composites: Part B 57 (2014) 47–55

Fig. 11. View of the transverse beam manufactured by RTM.

S. Laurenzi et al. / Composites: Part B 57 (2014) 47–55 55

injection channel, as in Case 3 (Fig. 10c), the torque box area showsregions that are not completed filled.

Based on these analyses, the injection scheme of the simulationCase 2 was adopted to design and fabricate the mold. This injectionscheme represents a good compromise between the process win-dow and the material behavior. We found that the experimentalfilling time to manufacture the beam was in agreement with thenumerical values. The injection time was about 2 h 20 min, veryclose to the predicted value with an injection pressure of 3 bars.The final piece shows a complete impregnation of all the requiredareas with no macroscopic defects on the surface (Fig. 11). Itshould be underlined that the use of effective permeability values,as we reported in this work, was an efficient way to reduce thenumber of experimental investigations required for such a com-plex geometry, resulting in a good agreement between the exper-imental and numerical filling times.

5. Conclusions

In this work a lightweight structural beam was successfullymanufactured and prototyped in polymeric composite materialsas a replacement for aluminum based ones. The composite compo-nent has paintable and aesthetical high grade surfaces, and offers a20% weight saving over the aluminum part. The injection schemewas established following the numerical flow analyses whileadopting a new methodology to include the out-plane flow effectstypical of thick sections in effective in-plane permeability values.This method is particularly useful in industrial settings whereexperimental and trial error methods should be minimized. Thistechnological demonstrator is a proof of concept for resin transfermolded structural composite parts to replace metal in primarystructures for aeronautic applications.

Acknowledgements

This project was funded by the Italian Ministry of Education,University and Research (MIUR). The authors acknowledge DEMAS.p.A. (Italy) for contribution to the design and construction ofthe beam mold. Dr. M.G. Santonicola (MESA+ Institute forNanotechnology, University of Twente, the Netherlands) and

G. Attolini (Sapienza Università di Roma, Italy) provided usefulcomments on the manuscript.

References

[1] Laurenzi S, Marchetti M. Advanced composite materials by resin transfermolding for aerospace applications. In: Ning H, editor. Composites and theirproperties. Rijeka, Croazia: Intech; 2012. p. 127–226.

[2] Kruckenberg T, Paton R. Resin transfer moulding for aerospacestructures. Netherlands and Norwell: Kluwer Academic Publishers; 1998.

[3] Laurenzi S, Di Nallo D, Marchetti M, Lalia Morra E, Anamateros E.Manufacturing approach to realize a prototype of a helicopter transmissioncomponent by composite materials. In: Proceedings of ERF31. Florence, Italy;September, 2005. p. 1683–96.

[4] Laurenzi S, Marchetti M, Anamateros E. Liquid composite molding foraeronautical application. In: Proceedings of 25 ICAS. Hamburg, Germany;September, 2006. p. 2565–72.

[5] Soutis C. Carbon fiber reinforced plastics in aircraft construction. Mater Sci EngA-Struct 2005;412(1–2):171–6.

[6] Advani SG, Sozer EM. Process modeling in composites manufacturing. MarcelDekker Press; 2002.

[7] Ruiz E, Trochu F. Numerical analysis of cure temperature and internal stressesin thin and thick RTM parts. Compos Part A-Appl S 2005;36(6):806–26.

[8] Brouwer WD, van Herpt ECFC, Labordus M. Vacuum injection moulding forlarge structural applications. Compos Part A-Appl S 2003;34(6):551–8.

[9] Amico S, Lekakou C. An experimental study of the permeability and capillarypressure in resin-transfer moulding. Compos Sci Technol2001;61(13):1945–59.

[10] Barlow D, Howe C, Clayton G, Brouwer S. Preliminary study on costoptimisation of aircraft composite structures applicable to liquid mouldingtechnologies. Compos Struct 2002;57(1–4):53–7.

[11] Luo Y, Verpoest I, Hoes K, Vanheule M, Sol H, Cardon A. Permeabilitymeasurement of textile reinforcements with several test fluids. Compos PartA-Appl S 2001;32(10):1497–504.

[12] Parseval YD, Pillai KM, Advani SG. A Simple model for the variation ofpermeability due to partial saturation in dual scale porous media. TransportPorous Med 1997;27(3):243–64.

[13] Simacek P, Advani SG. A numerical model to predict fiber tow saturationduring liquid composite molding. Compos Sci Technol 2003;63(12):1725–36.

[14] Slade J, Pillai KM, Advani SG. Investigation of unsaturated flow in woven,braided and stitched fiber mats during mold-filling in resin transfer molding.Polym Compos 2001;22(4):491–505.

[15] Saouab A, Bréard J, Lory P, Gardarein B, Bouquet G. Injection simulations ofthick composite parts manufactured by the RTM process. Compos Sci Technol2001;61(3):445–51.

[16] Turner DZ, Hjelmstad KD. Determining the 3D permeability of fibrous mediausing the Newton method. Compos Part B-Eng 2005;36(8):609–18.

[17] Samir J, Echaabi J, Hattabi M. Numerical algorithm and adaptive meshing forsimulation the effect of variation thickness in resin transfer molding process.Compos Part B-Eng 2011;42(5):1015–28.