The fibre reinforcements used to manufacture composites with FDM Molds were carbon fiber (93 g/m²), glass fiber (280 g/m²) and coarse twill flax fiber (550 g/m²). All fiber reinforcements were twill type. Figure 1 shows the three fiber reinforcements. Coarse twill flax reinforcement was chosen because of its twill architecture and because it was expected to form better doubly-curved composites as established by Kazmi and co-workers [22] [30].

The fiber reinforcements used to manufacture composites with non-FDM molds were weaved glass fiber reinforcements. All reinforcements used in this research were twill weaved with alternating (±45˚) weft and warp.

Figure 2 shows the weaved glass fibers to take the shape of a complicated structure with multiple hollow cylinders.

Larit epoxy resin system was used in all experiments. The ratio of the resin (L-160) and hardener (502) was 100:40 by mass. Gelling time was approximately

50 minutes after careful mixing and degassing for 15 minutes. The epoxy resin system with a gel time of 35 minutes was chosen because of the smaller size of the molds used. A lower gel time meant that the resin would be in its liquid state for a longer period of time and that was not advised because the filling time of the mold was approximately 20 minutes.

An FDM printer was used to create singly-curved and doubly-curved molds. Figure 3 shows the picture and schematic of the printer. PLA was used in the form of a filament of diameter 1.75 mm wound on a spool as shown in Figure 3. The nozzle of diameter 0.35 mm was used at a temperature of 200˚C. The feed rate of liquid PLA coming out through the nozzle on the bed was 38 mm/s and the bed was at 60˚C.

In order to print the infusion molds, the “Mendelmax FM Pro” in version 1.5 has been used. The printer in modular principle is a cost-effective option in comparison to an industrial FDM printer. In Table 1 the printer settings are displayed. A nozzle diameter of 0.35 mm was used to gain a reasonable printing quality. With a nozzle temperature of 200˚C, the PLA has shown proper melt behaviour and printing accuracy. Due to the staircase effect, the molds have been printed in perpendicular direction to the front side. As a result, the front side itself shows the best surface quality, the printer is able to. The solid infill ratio of 35% has been suitable to withstand the applied pressure and kept the mold from collapsing by the use of vacuum pressure. If printing time would not have been the primary issue, a higher degree of filling could have been achieved resulting in a stronger mold with a greater resistance to compaction pressures. For a higher degree of filling, the printing time will increase noticeably.

The previously explained FDM printer was used to create single and double sided molds to manufacture composites with the VARTM process. The FDM molds were primarily used to compare the geometries of the VARTM epoxy composites. An infusion pressure of about 0.8 bar will be used so that the mold can withstand the pressure of the applied compaction force. Negative and positive molds were produced to “muffle” the reinforcement material. It was expected to achieve a better surface finish with the double-sided mold upgrading the idea of Professor Summerscales [31], who suggested to use a flexible vacuum bag instead of a double-sided mold. He was referring to the costly metallic molds but in this case, significantly cheaper PLA molds were used. Hence, using double-sided PLA molds will not be a threat to the overall cost of the composite part in question.

Figure 4 shows the male and female parts of the singly and doubly curved composites produced by the FDM Printer. The female singly-curved mold had a cavity which was a semi-cylinder of radius 25 mm and length of 100 mm. The cavity for the doubly-curved mold was a hemisphere of radius 25 mm.

The manually prepared molds were created to manufacture the hollow cylindrical composites and the structure with multiple hollow cylinders. These molds were created to induce the best possible cylindrical shape and surface finish in the manufactured composites and this was one of the most challenging parts of this research. The reason for this was the de-molding process. The outer surface of the mold was made soft enough to remove the cylindrical composite after manufacturing but hard enough so that the composite could take a good cylindrical shape and surface finish. Readily available PVC pipes were used and rubber tape

was used on the pipes to create the required diameter so that the closed loop weaved glass fiber reinforcements can fit on and removed from the pipes conveniently. For the special structure with the hollow cylinders (Figure 2) smaller PVC pipes were fitted inside the holes for the flat structure and flexible rubber pipes were fitted for the curved structure.

Figure 5 shows the schematic of experimental setup employed for the resin infusion (VARTM) process. The fiber reinforcement was kept inside the single-sided or double-sided mold. The mold was covered with a nylon bag so that it could be reused and the manufactured composite part could be easily removed after the experiment. A finely perforated fabric known as peel ply was used to create vacuum inside the mold by removing the air inside the cavity through a polyethylene port as shown in Figure 5. The vent catch pot was used in case any excess resin was drawn out of the mold, to avoid any damage to the vacuum pump. Moreover, just before the port connected to the vent catch-pot, cotton was inserted between the mold and the vent port to decrease the filling speed of the resin and to block the resin from going to the vent catch-pot or the pump. When the nylon bag was vacuum taped on glass, the vacuum pump was turned on. When all the air inside the cavity was extracted through the vent catch pot, vacuum was created and this vacuum drove the resin from the inlet pot through the connecting tube into the mold cavity. When the resin filled all the coarse twill reinforcement it was blocked by the cotton. After the resin solidified, the vacuum pump was left turned on overnight and the next day the de-molding of the composites was followed by its post-curing. The standard industrial VARTM process was followed where first the resin fills the mold and when the resin hits the break, the inlet is clamped and left overnight for curing. During curing overnight, the vacuum pump keeps running and the vent stays at full vacuum which is around 900 mbars of compaction pressure in this case.

This research aims to explore the potential of VARTM to manufacture curved composites using cost-effective molds. Three types of molds were manufactured using the FDM printer. The structural integrity of the flat composites was tested

by using tensile, bending, and charpy testing. Table 2 presents the manufacturing plan for manufacturing and testing of the composites. Manufacturing was aimed to start from flat to simpler curvatures and then to the more complicated structures. All experiments were repeated to cross check the results. Number of layers of the fibers infused with resin was varied to keep the areal mass of the fiber reinforcements constant at 300 g/cm². All the manufacturing details will be kept constant for repeatable results. Bending tests were carried out to show the structural integrity of the curved composite parts. Thickness at different positions of the curved composites was measured to determine the dimensional consistency of the composite parts.

The hollow composites needed special molding technique because of having a closed loop. There were two challenges. The first was to create the desired hollow cylindrical shape and the second was to de-mold the composite from the cylindrical mold. Figure 6(a) shows how a cylindrical mold was used with other consumables and the weaved glass fiber reinforcement before resin was infused through the assembly. First, a soft rubber core (vacuum tape) was wrapped around the solid cylindrical mold. This was then covered by a vacuum bag sealed by a vacuum tape in order to create vacuum. The glass fiber reinforcement covered from both sides by a breather (peel ply) was then rolled over the vacuum bag. A distribution medium was wrapped around the breather for a faster flow of the resin through the reinforcement. In the end everything was sealed using a vacuum bag and vacuum tape.

Two types of structure with multiple hollow cylinder network were manufactured. One was flat and the other curved. Figure 6(b) shows how the flat structure was prepared before resin infusion. Multiple PVC hollow pipes were used as shown so that the structure could be formed. For the parts were the pipes could not reach, smaller sections were cut and placed but those sections were not expected to come out of the composite structure. After placing all the hollow pipes inside the appropriate place in the weaved glass fiber reinforcement as shown in Figure 6(b), the reinforcement was wrapped by a breather (peel ply). A distribution

medium was placed on the reinforcement and then everything was vacuum bagged and connected to the inlet and vent for VARTM (Figure 5).

For the curved structure, instead of the solid PVC pipes, flexible rubber pipes were used but in this case most of the rubber pipes stayed inside the composite because there was no way to bring them out after infusion. The curved structure was placed on a curved surface to achieve the required angle and then the whole mold was infused with resin.

Coarse twill flax reinforcement was used with epoxy resin. Four singly-curved and four doubly-curved composites were manufactured. Two composites each were manufactured using the single-sided female molds and two composites each were manufactured using double-sided singly and doubly-curved molds shown in Figure 4. Three layers of coarse twill reinforcement were used to manufacture all the composites. The cutting of the reinforcement before manufacturing the doubly-curved composites was executed as described in by Kazmi and co-workers [22]. Thickness of both the singly-curved and doubly-curved composites was measured at five different points. A total of five specimens were cut from each composite part to be used for measuring their thickness and testing them under shear. These measurements were executed for all the composites. The choice of these points was made to find the trend of thickness and shear strength across the curve for both the singly and doubly-curved composites. Figure 7 shows the positions where the thickness of the singly and doubly-curved composites were measured. These are also positions from where the specimens for the shear tests were obtained.

Short beam shear and tensile tests were conducted on samples to evaluate the effect of variation in the fiber and curvature on the mechanical strength of the composites. Samples taken from parts manufactured according to the manufacturing plan were tested for short beam shear strength under ASTM-D2344 standard and for tensile strength under ASTM-3039. A total of five specimens were cut out of each composite part to be used for testing. The specimens were cut out from the manufactured composite parts. Instron 5567 testing machines were used to perform the bending test. The cross-head speed was set to 1 mm/minute to prevent abrupt failures. The criterion for specimen failure was a flexural deformation of 8%. All other measurements were according to the standards.

This research also aimed to compare the manufacturing of the composites using single sided PLA molds with the double sided counterparts. Single-sided and double-sided FDM molds were used so that the dimensional consistency of the VARTM manufactured epoxy composites using each mold type could be compared. Figure 8 shows the thickness across the curve of singly-curved and doubly curved coarse twill composites manufactured using both mold types. The composites manufactured using the double-sided molds were thinner than those manufactured using the single-sided molds at every position. The thickness across the curve for both composite types does not show a clear trend but the thickness range is different for the two types in Figure 8. The range is 2.77 to 2.89 mm (4.3%) for the singly-curved double-sided molded composites and 2.91 to 3.15 mm (7.8%) for the single-sided counterparts. This implies that thickness

of the singly-curved double-sided molded composites is more consistent along the curve.

Unlike the singly-curved composites, the doubly-curved counterparts show a slight trend in the thickness across the curve. Maximum thickness was recorded for both single-sided and double-sided molded composites at Position 3. As shown in Figure 7, this position is the central part of the doubly-curved composite and thickness at this point was the greatest for both composites types as shown in Figure 8. The trend is that the thickness is the least at the ends of the composite (Position 1 and 5) and maximum at the central part (Position 3). The range of thickness was 3.28 to 3.40 mm (3.6%) for the double-sided molded composites and 3.40 to 3.60 mm (5.7%) for the single-sided counterparts.

The parts produced with single-sided molds have a larger scatter of thickness values (7.8% and 5.7%) than the ones produced with the double-sided arrangement (4.3% and 3.6%). To produce parts with a more stable thickness over the surface, double sided molds looked promising and maintained a good consistency of thickness over the surface of the manufactured parts.

The shear strengths of the singly-curved and doubly-curved coarse twill composites manufactured using single-sided and double-sided PLA molds are presented in Figure 9. The range of shear strength values across the curve of the singly-curved composites manufactured using the single-sided molds was from 17.9 to 18.2 MPa and the range for those manufactured using the double-sided molds was from 16.6 to 17.3 MPa. The range of shear strength for the single-sided molded composites was from 10 to 10.15 MPa and for the double-sided ones was from 9.8 to 9.9. It can be deduced that the average shear strength of both the

singly-curved and doubly-curved composites manufactured using the single-sided molds was greater than those manufactured using the double-sided molds.

Position 3 was the strongest under shear for both the singly-curved and doubly-curved composites manufactured using single-sided molds and the weakest for the double-sided mold counterparts. Considering Figure 8, Position 3 was the thickest for the single-sided molds and the thinnest for the double-sided molds. For an epoxy composite, a thinner composite is not an indication of the quality of consolidation of the composite. The thickness in the epoxy composites is only due to the compaction pressure during the infusion process and after the resin has completely filled the reinforcement. In this case, since Position 3 is the thickest part in the single-sided composite, it has more mass and hence more resistance under shear. Position 3 on the double-sided molded composite was thinnest and hence the weakest part under shear because it had a lower mass of resin.

Double-sided molds were expected to induce higher initial imperfections and voids in the parts than those manufactured using single-sided molds. The PLA molds used in this research were considered significantly weaker than any other molds being used in the industry to produce epoxy resin composites. They were only good enough to withstand a compaction pressure less than or equal to the atmospheric pressure, so the imperfections and voids induced by the double-sided molds were almost negligible and difficult to quantify. The low compaction pressure is the reason why the thickness of the double-sided molded composites was more consistent across the curve than the single-sided counterparts. Further research can be carried out by subjecting the double-sided molds to varying compaction pressures and then investigating the effect on the dimensional stability of the part and voids induced in it.

According to Zhu and co-workers [32], under a compaction pressure of over 6000 mbars there will be residual stress or strain in curved parts that will decrease with an increase in the thickness of the part. This phenomenon is known as spring forward in which the curved manufactured part will have a lower radius of curvature than the mold. For a 4 mm thick part, this strain was 1.05˚ according to the viscoelastic model and 0.69˚ according to the elastic model. Bogetti and co-workers [33] ; Miller and Ramani [34] and Ruiz and Trochu [35] established that the residual stress causing part deformation was dependent on reaction kinetics, specimen thickness and the cure cycle. Golestanian and El-Gizawy [36] found that the finished part was 1.2 to 1.38 mm shorter than the mold geometry over a 50 cm dimension.

All of these researchers used double-sided molds in their models and experiments. The compaction pressures involved were much more than the atmospheric pressure. In this research, there was no temperature gradient in the mold. The use of single-sided molds further reduced the probability of inducing residual stresses in the manufactured composites. Hence, the shear strength of the composites manufactured using the single-sided molds was better for both singly-curved and doubly-curved composites.

The composites parts manufactured via the resin infusion process were tested for the shear strengths. Thickness at six points on all the parts was measured and the average thickness was determined for each type of part. Figure 10 shows the average thicknesses of all the types of glass fiber composites manufactured during this research using various types of molds. The flat composites are the thinnest because of the post-filling pressure which was not enough due to the curvature of the curved composites. This implies that the composite thickness for a part with a more complicated curve will be less compared to the one with a less complicated curve. As the curve gets complicated, it is more difficult for vacuum to press the reinforcement on the mold and hence more resin flows and impregnates the fiber reinforcement. This can be reduced by pulling vacuum from the vent and inlet in the post filling stage rather than just clamping the inlet. It will be interesting to see how close the thickness values can be brought for the given three curvatures under these conditions. On the other hand, by pulling full vacuum from both the inlet and the vent in the post-filling stage, voids will be

induced in the composite part, which can considerably reduce the mechanical strength of the part. Hence it is a trade-off between the dimensional accuracy and the mechanical strength of the part. If a designer wants to achieve a specific type of thickness then there are many other factors including the compaction pressure to be considered. The average thickness values are only a comparison between the compressibility of each mold type keeping all the other variables constant. Also it needs to be understood that the variation in thickness will increase if the number of fiber layers are increased. Hence, the designer must know all the variables including the compaction pressure, number of fiber layers, part geometry and part size to choose a specific type of mold.

The short beam shear strengths of all the types of glass fiber composites manufactured in this research are shown in Figure 11. The flat composite is by far the strongest in shear because it is the most suitable shape for a VARTM process. Composites manufactured using single sided molds were stronger in shear than those manufactured using the double sided molds. This is mainly because there was more impregnated resin in the single sided molded composites. If a comparison is made between the different curvatures, then it is seen that there is a marked difference in the shear strengths of single and double curved composite part. This low shear strength is mainly because of the complicated curvature itself and partly because of the indirect effect of this double curve on the manufacturing process. When the part geometry gets more complicated the compaction of the fiber reinforcement due to vacuum is weaker. The hollow cylindrical composite is although weaker in shear than the flat composite but slightly stronger in shear compared to the singly curved composites. This may be because the hollow composites had a slightly larger diameter than the singly curved composites. The cylinder web composite structures were the weakest because of their complicated geometry. The obtained specimens from the webs were also not very perfect according to the ASTM standards.

When the short beam shear strengths were compared to previous research; the flat glass fiber epoxy composites were 10% stronger than those manufactured by Fan and co-workers [37] using VARTM, 61% stronger than the woven glass fiber

epoxy composites manufactured by Yang and co-workers [38] using RTM, 19% stronger than the twill glass fiber epoxy composites manufactured by Almeida Jr. and co-workers [39] using VARTM and83% stronger than the unidirectional glass fiber epoxy composites manufactured by Selmy and co-workers [40] using the wet hand layup technique. These comparisons cannot be absolute due to the difference in experimental conditions such as fiber orientation, manufacturing method and the fiber volume fraction. However, this comparison was necessary to show that the shear results of this research are promising. Most of the singly curved composites are as strong in shear as other glass fiber composites manufactured. The doubly curved composites are although weaker in shear but still comparable to the other glass fiber composites manufactured by previous researchers.

The tensile strengths of all the types of glass fiber composites manufactured in this research are shown in Figure 12. The carbon fiber flat composites were the strongest in tension while the cylinder web and singly curved composites were the weakest. The trend is very similar to the one in Figure 11. The only comparison here is between the carbon, flax and glass fibers where the carbon fiber composites were the strongest and the flax fiber composites were the weakest.

When the tensile strengths were compared to previous research; the carbon fiber epoxy composites were 5% stronger than the ones manufactured by Godara and co-workers [41] using a vacuum autoclave. The flax epoxy composites were 25% stronger than the ones manufactured by Yan and Chouw [42] using the wet hand layup technique. The glass fiber epoxy composites were 50% weaker than the unidirectional glass fiber composites tested by Devendra and Rangaswamy [43]. Again, since the manufacturing conditions are different, these comparisons are only to show that the tensile results of this research are also promising.

The challenge in this research was to achieve the targeted shapes. Many experiments were unsuccessful because of the complicated process. The PLA molds used were considered significantly weaker than any other molds being used in

the industry to produce epoxy resin composites. However, since the compaction pressure was not more than the atmospheric pressure, the weak PLA molds were able to withstand this stress and convert that to shape the composite part. Figure 13 shows pictures of a simple single curved carbon fiber reinforced epoxy composite part, a singly curved one and a double curved flax composite. The shape and dimensions are quite in line with what was targeted. Especially for the simple singly curved parts, the shape and dimensions were very consistent for all fiber reinforcements. The shape and dimensions for the one shown in Figure 13(b) were also quite in line with the targeted ones. Both the parts in Figure 13(a) and Figure 13(b) were manufactured using a double sided mold so that is also a reason why the shape dimensions are quite consistent. Although, parts manufactured by the single sided molds were considerably consistent in the shape and dimensional accuracy. The curve geometry of the finished parts was very much what was expected given the geometry of the mold. The double curved geometry as shown in Figure 13(c), which is actually a hemispherical shape in this research, was not absolutely achieved due to wrinkling of the reinforcement layers. This wrinkling can be reduced by tailored cutting of the reinforcement to adjust with the hemispherical shape but then there will be a major reduction of mechanical strength in the area where the reinforcement is cut.

Figure 14 shows the pictures of the closed loop composites where the first one is a hollow cylinder and the second one is an assembly of a flat cylinder web and a curved cylinder web. The shape conformance of the very complicated web was not as bad as expected but there is still room for improvement in the management of the VARTM process to better facilitate the manufacturing of the curved composites. Controlling the temperature and humidity will be very useful in case of environments with varying conditions.

The summary of tensile, shear and thickness results with standard deviations are shown in Table 3. This is actually a summary of the results shown in Figures 10-12.