Micro scale flow behavior and void formation mechanism during impregnation through a unidirectional stitched fiberglass mat.
Liquid composite molding (LCM) processes such as resin transfer molding (RTM) and structural reaction injection molding (SRIM) have emerged in recent years as important methods for the near-net-shape production of a wide range of composite parts (1-3). LCM is a matched mold process in which liquid thermosetting resin is introduced into a mold cavity containing dry fiber preform. Resin expels air from the mold and impregnates the fibers. After the resin injection is complete, the resin is allowed to cure. This is followed by demolding and trimming to obtain the finished product. Relatively short cycle times with good surface quality can be easily achieved in LCM. Since it is not limited by the size of the autoclave or the press, fabrication of large complex structure is possible. Being a closed mold process, it eliminates problems of volatile chemical emissions. Other benefits of LCM over competing methods include low investment costs for equipment and tooling, the ability to manufacture complex parts with a core in a single step, and improved control of product shape, weight, reinforcement volume fraction, and quality (3-6).
Traditionally, LCM processes have been low volume manufacturing processes. However, recently growing interest in high volume structural applications has led to a need to thoroughly analyze the LCM technology. Although a number of studies are available in the literature on mold filling and curing stages of LCM processes, much less attention has been paid to fiber impregnation and void formation.
The reinforcements used in LCM are initially in an unimpregnated form, and it is the complete impregnation of these fibrous networks that is the ultimate goal. The resin injection step involves two types of flow which occur simultaneously. One is mold filling, that is the advancement of bulk flow front, and the other is resin impregnation, which involves local penetration of resin into the fiber tows (7, 8). During the injection of resin into the mold, resin must quickly fill the mold and wet all the individual fibers before much reaction occurs. The mold filling is completed in a few seconds in SRIM and a few minutes in RTM. This gives very little time for the resin-fiber interactions and often leads to void formation and poor wetting.
The presence of voids and poor wetting are highly detrimental to the performance of composite parts, since they adversely affect physical and mechanical properties, as well as finish of the product. Several researchers have studied the effects of porosity and wetting on composite properties, such as tens fie strength, shear strength, and transverse flexural strength (9-11). The possible reasons for void formation include: mechanical entrapment of air already present in the mold, partial evaporation of mold release agent, and volatilization of dissolved gases and volatile chemical species during curing. The void formation and changes are affected by various factors, such as injection pressure, molding temperature, pressure during curing, resin properties (viscosity, surface tension, etc.), reinforcement characteristics (type and orientation of fibers, surface treatment, etc.), and contact angle between resin and fibers. There appears to be no proven way to eliminate voids completely. Vacuum bag type of methods for manufacturing composite parts reduce the occurrence of voids but can not totally solve the problem of void formation and poor wetting (12, 13). Thus, the understanding of the flow behavior and the void formation mechanism is critical in LCM.
Problem of void formation in autoclave processing of composites has been studied extensively by Ahn et al. (14), Dave et al. (15) and Kardos et al. (16). In LCM, several studies are available on resin flow behavior at the micro scale. The interaction of competitive micro and macro flows during resin injection was studied by Patel et al. (17). They observed that during the resin injection process, the maximum inlet pressure was not obtained at the end of mold filling. After the mold filing was complete, if the resin injection was continued, the inlet pressure continued to rise and reached a steady state value after a period of time. Flow visualization studies by Molnar et al. (18) have shown that, at a low flow rate, flow in the fiber tows led flow between the fiber tows, whereas the opposite results were obtained at a high flow rate.
Recently visualization studies to characterize resin flow and void formation have also begun. Morgan et al. (19) studied macroscopic flow behavior of low viscosity liquids into glass and carbon fiber mats as a function of fiber mat lay-up and geometry. They found that carbon fiber tows wet more slowly than glass tows. Peterson and Robertson (20, 21) reported that although bulk flow characteristics could be adequately described by a Darcy's law type of equation, resin flow along the fiber tows is better described by flow through noncircular channels. They also investigated the origin of interstitial voids in aligned fibers. Wang et al. (22) carried out preliminary study on the effects of fiber mat structure and liquid viscosity on bubbles purged from the fiber mats during injection process. Mahale et al. (23) have quantified void formation in continuous random fiberglass mat by image analysis technique. Stabler et al. (13) have studied the effects of surface waxing, initial bubble content of the resin, vibration frequency of the mold during filing, and fill pressure on void formation in graphite-epoxy composites. The effect of evacuating the mold on fiber wetting and voidage has been investigated by Hayward and Harris and Lundstrom et al. (1, 12). Interactions between sized fiber surface and reacting resin, which are responsible for spreading and wetting in LCM have been characterized by Larson et al. (24).
Several investigators have carried out theoretical analysis of void formation during flow through fibrous porous media. Parnas and Phelan (25) and Chan and Morgan (26, 27) developed a model based on Darcy's law that predicted entrapment of air in the fiber tows. Capillary forces were neglected and permeability of fiber tows was assumed to be much smaller than that of global porous medium in their studies. Preliminary experiments carried out by Sadiq et al. (28) on periodic array of fiber tows verified qualitatively the air entrapment mechanism of void formation assumed in the model of Parnas and Phelan (25). Elmendorp and During (29) modeled transverse impregnation of aligned hexagonal array of fibers from the point of view of void creation. They also carried out visualization of void formation in a hexagonal array of stainless steel rods mounted perpendicular to the flow direction.
Although a few studies exist concerning micro mechanics aspect of LCM technology, very little information is available on mechanisms of flow front progression and void form in LCM. The objective of this study was to carry out a systematic investigation of micro scale flow behavior of liquids through fiber reinforcement in liquid composite molding environment in order to understand the void formation mechanism and relate it to process variables. Flow visualization experiments were carried out to study the effects of flow rate, fiber mat structure, and liquid properties on the formation and elimination of voids.
Several nonreactive liquids: water, DOP oil (diphenyl-octyl-phthalate), various silicone oils (Dow Corning 200 fluid, dimethylpolysiloxane), ethylene glycol and glycerine, and one reactive resin: an unsaturated polyester (UP) resin were used in this study. The unsaturated polyester resin used was a 1:1 mixture of propylene glycol and maleic anhydride containing 35% by weight styrene (Q6586, Ashland Chemical). Additional styrene was added to make molar ratio of styrene to unsaturated polyester equal to 2.0. Both unsaturated polyester and styrene monomer were used as received without removing the inhibitor. They were weighed and mixed in a flask with appropriate weight ratio. The prepared resin was stored in a refrigerator.
There are numerous reinforcement choices for LCM processes, such as random chopped and continuous fiberglass mats, and stitched, knitted and woven glass, graphite, and aramid fiber reinforcements. In order to minimize experimental effort, this study concentrated on a stitched unidirectional fiberglass mat (CoFab A0108). it is a nonwoven type fiber reinforcement with fiber tows oriented in one direction only. The tows are bound together by a continuous stitch. On one side of the fiber mat, stitches loop back on themselves to form double stitches. The other side has only one stitch. The single stitched side has, in addition to single stitches, relatively thicker stitches made up of several filaments, possibly to maintain fiber mat structure. These thicker stitches are oriented perpendicular to the fiber tows, but are much thinner than the fiber tows. Thus, this fiber mat is not truly unidirectional in nature. Further work is being carried out in our laboratory using other types of fiber reinforcements to study the effects of fiber preform architecture on micro scale flow pattern and void formation.
Instrumentation and Experimental Procedure
Measurements of Liquid Properties, Solid Surface Energy, and Contact Angle
The viscosities of liquids used in the study were either obtained from handbooks or measured by a Haake viscometer. The surface tensions of the test liquids and the fiber-liquid-air contact angles were measured by Wilhelmy technique. In this technique, a solid is partially immersed in a liquid, thereby, the liquid either rises or depresses along the vertical wall of the solid. The force exerted by liquid on the solid due to surface tension is the wetting force ([F.sub.w]) and is given by,
[F.sub.w] = [P.sub.er] [[Gamma].sub.LV] cos [Theta] (1)
where [P.sub.er] is the perimeter of the solid along the three-phase boundary line, [[Gamma].sub.LV] is the surface free energy of the liquid-vapor interface or the surface tension of the liquid, and [Theta] is the contact angle (30, 31).
The wetting force was measured by a dynamic contact angle analyzer (Cahn Model DCA-322). It consists of a traveling stage, which can be lowered or raised either manually or by a motor to provide dynamic advancing or receding force measurements, but the velocity must be kept low to prevent viscous deformation of the contact angle (31). To measure contact angle, the fiber sample was suspended from the balance arm via a hangdown wire. Liquid was contained in a beaker that was placed on the traveling stage. The liquid was slowly raised at a carefully controlled speed and force readings were taken at a specified time interval. At the first contact of the liquid with the fiber, the wetting force jumped. As more and more area of the fiber sample contacted the liquid, the wetting force decreased due to buoyancy. After desirable length of filament was traversed, the direction of liquid movement was reversed to obtain data for receding wetting in a similar manner. The measured force was then plotted as a function of depth of immersion and the data were analyzed using a computer program based on Eq 1.
This method of measuring the contact angle requires the perimeter of the fiber sample. The same principle was used to measure the perimeter. Hexane and hexadecane were used as wetting liquid for this purpose. These liquids are known to give nearly zero contact angle (32). To measure surface tension of the liquid, a clean glass cover slip (Wilhelmy plate) was used. It gives zero contact angle for nearly all liquids (32). Application of Eq 1 then gave the surface tension of the liquid.
Several methods have been proposed in literature to calculate solid surface energy. The following four methods were used to measure surface energy of mold surface used in this study in order to analyze its effect on void formation. In the method by Fox and Zisman (33) cosine of the advancing contact angles made by the solid of interest with a series of liquid are plotted vs. the surface tensions of the liquids. The surface tension value obtained by extrapolating to cos [Theta] = 1 the best fit line through this data set is known as the critical surface tension. This value corresponds to minimum value of the liquid surface tension required for complete spreading. The critical surface tension thus yields a single value that is useful in assessing the general wetting characteristics of a solid surface (32). The geometric mean method based on the theories developed by Girifalco and Good (34), Fowkes (35), and Owen and Wendt (36) allows calculations of both the dispersive and polar components of the solid surface energy based on the following equation.
[Mathematical Expression Omitted]
where [[Gamma].sub.SL] is the surface energy of the solid-liquid interface, [[Gamma].sub.S] and [[Gamma].sub.L] are the surface tension of the solid and liquid interfaces and the subscripts d and p represents the dispersive and polar components, respectively. This equation permits the calculation of solid surface energy components from the contact angle measurements with two liquids of known surface tension components. A similar approach has also been described by Kaeble (37), which allows the calculation of solid surface energy based on contact angle measurements with a number of liquids. This approach is referred to as geometric multiliquid method in this paper. The harmonic or reciprocal mean equation developed by Wu (38) differs from the geometric mean approach only in the calculation of the dispersive and polar interaction terms.
[Mathematical Expression Omitted]
Again this method requires contact angles measurements with two liquids. The contact angles made by the test solid with various standard liquids were measured using the dynamic contact angle analyzer. The solid surface energy was then obtained by the four methods described above. The details of theories behind these methods are reported elsewhere (32-38). The test solid consisted of a glass plate coated with a mold release agent (Frekote 1711). The surface energy of virgin glass and PMMA were obtained from handbooks.
Micro Scale Flow Visualization and Void Formation Studies
The equipment set-up used for the flow visualization experiments is described as follows. A transparent PMMA mold was used for flow visualization unless otherwise stated. A glass mold was used to study the effect of mold surface on void formation. The mold dimensions were 15.5 cm x 8 cm. Inlet and outlet holds were drilled in the upper platen 11.5 cm apart.
A rectangular mold cavity was created by a Teflon gasket. The cavity dimensions were 13 cm x 5 cm. Strips of fiber mat were cut to snug fit into the rectangular cavity. The strips were 5 cm wide and 9 cm long. One layer of fiber mat was used for each experiment (porosity = 57%). Liquid was injected using a Harvard apparatus infusion/withdrawal pump (Model 919). It has a piston cylinder type mechanism for injection. Thus, being a positive displacement pump, it ensured delivery of liquid at a constant flow rate. Care was taken to remove all the bubbles from the source and the tube connecting the pump to the mold. The liquid was injected at various flow rates. The flow pattern and void formation were viewed using a CCD video camera (Cohu Model 4915-2001) and recorded using a S-VHS recorder (Panasonic Model AG 1960). Thus the voids considered in this study are the surface voids.
RESULTS AND DISCUSSION
Measurements of Liquid Properties, Solid Surface Energy, and Contact Angle
The room temperature properties of the test liquids are listed in Table 1. Table 2 shows the results of single fiber contact angle measurement experiments. Experiments for each liquids were repeated at least three times and average values of contact angles are reported in Table 2. Table 3 lists the standard liquids used in the solid surface energy measurements, their properties of interest, and the advancing contact angle made by mold release agent coated glass plate with these liquids. Again average values are reported in the table. Some or all of these results were used in the calculations of surface energy. Table 4 shows the surface energies of various solids of interest. Surface energy measured by the geometric mean, the harmonic mean and the geometric multi-liquid methods are very close to each other. Zisman plot method, however, gives lower surface energy than the other three methods. It has been pointed out that Zisman plot method is often inconsistent and unreliable, primarily due to the scatter and nonlinearity of the experimental data (38). Nevertheless, all four methods indicate that mold release agent reduces surface energy of the mold surface.
Table 1. Room Temperature Properties of the Test Liquids.
Surface Liquid Viscosity Tension (mPa.S) (mN/m)
Silicone oil, 1 cS 0.82 17.4 Silicone oil, 10 cS 9.35 20.1 Silicone oil, 100 cS 96.4 20.9 Silicone oil, 200 cS 193.4 21.0 DOP oil 43.3(*) 25.4(*) Water 1.0 72.3(*) Ethylene glycol 19.8 48.4 Glycerine 1499 63.4(*) Hexane 0.33 18.4 Hexadecane 3.34 26.7 Unsaturated polyester resin 54.62(*) 34.5(*)
* Measured; the rest are from handbook or provided by manufacturer.
[TABULAR DATA FOR TABLE 3 OMITTED]
Micro Scale Flow Visualization and Void Formation Studies
Results Obtained Using Unidirectional Stitched Fiberglass Mat With Fiber Tows in the Direction of Global Flow
(a) Micro scale flow pattern
Figure 1 shows the flow front progression when water was injected into the unidirectional stitched fiberglass mat at a relatively low flow rate (superficial velocity, [u.sub.s] [similar to] 0.8 cm/s). The double stitched side was at top. The figure elucidates complex nature of flow. Two different flow fronts were observed inside the fiber mat. A clear lead-lag or fingering could be seen in the main flow front with flow leading inside the fiber tows. This main flow front is referred to as "the primary flow front". In addition to the primary flow front, liquid was drawn forward within the fiber tows. This second flow front was due to the wicking of liquid and is referred to as "the wick flow front." The shape of the wick flow front could not be clearly established because the fiber tows were only partially wetted. The fingering at the primary flow front and the presence of wick flow front resulted into lateral dispersion from wet areas to relatively dry areas. Figure 2 shows the flow front progression when water was injected at a higher flow rate ([u.sub.s] [similar to] 3.9 cm/s). As can be seen from this figure, flow between the fiber tows led the flow within the fiber tows in the primary flow front. Also note the absence of wick flow front. Again, there was lateral dispersion from wet areas to relatively dry areas. Experiments carried out using a range of flow rates revealed that the extent of fingering in the primary flow front and the wicking depend on the flow rate.
Table 2. Equilibrium Advancing Fiber-Liquid-Air Contact Angles.
Liquid Fiber-Liquid-Air Contact Angle
Silicone oil, 1 cS [similar to] 0 [degrees] Silicone oil, 10 cS [similar to] 0 [degrees] Silicone oil, 100 cS [similar to] 0 [degrees] Silicone oil, 200 cS [similar to] 0 [degrees] DOP oil [similar to] 0 [degrees] Water [similar to] 66 [degrees] Ethylene glycol [similar to] 56 [degrees] Glycerine [similar to] 67 [degrees] Hexane 0 [degrees] Hexadecane 0 [degrees] Unsaturated polyester resin [similar to] 38 [degrees] Table 4a. Room Temperature Surface Energy (Critical Surface Tension) of Solids.
Solid Surface Energy (mN/m)(*)
Clean PMMA 41.1 Clean glass and metal 49 Paraffin wax 23
* Literature values. Table 4b. Room Temperature Surface Energy of Mold Release Agent Coated Glass Plate.
Measurement Method Surface Energy (mN/m)(**)
Zisman plot method 18.6 Geometric mean 2 liquid method 25.8 Harmonic mean method 24.4 Geometric mean multi-liquid method 26.7
When experiments were carried out using other liquids similar results were obtained. The shape of the primary flow front and the extent of wicking were found to depend not only on the flow rates, but also on the liquid properties, notably the viscosity and the surface tension. At relatively low flow rates, liquid flowed faster inside the fiber tows. The fingering and wicking were very strong for water, and silicone oils of kinematic viscosity 1 cS and 10 cS. At relatively high flow rates liquid flowed faster in the gap between the tows. Figure 3 shows flow front progression when DOP oil was injected at [u.sub.s] [similar to] 0.2 cm/s and 0.8 cm/s. A short wick flow front can be seen in Fig. 3a. Figure 4 shows flow front progression for glycerine at [u.sub.s] [similar to] 0.2 cm/s. Severe fingering with flow leading between the fiber tows can be seen clearly. Also, as can be seen from this figure, wetting of tows took place after the primary flow front had already passed. When silicone oil of kinematic viscosity 1 cS was injected into unidirectional fiber mat, flow led within the fiber tows even for the highest available flow rate ([u.sub.s] [similar to] 3.9 cm/s). These results elucidate the importance of dynamics of wicking and the relative importance of viscosity and surface tension in the micro scale flow in LCM preform.
The following mechanism is proposed to explain the flow pattern inside the fiber mat. There are two different driving forces for the flow in the fiber preform: hydrodynamic pressure and capillary pressure. In a fiber reinforcement, each fiber tow has a large number of fibers and the interstitial space within the fiber tows is much smaller than the gap between the tows. Thus, there would be a much stronger capillary action within the fiber tows as compared to between the fiber tows. That is, the flow between the fiber tows depends largely on the applied hydrodynamic pressure, whereas, the flow within the fiber tows is governed primarily by the capillary pressure. Depending on the relative magnitude of these forces at a particular location, the flow pattern may be different. This can be explained using a schematic diagram shown in Fig. 5. The hydrodynamic pressure is highest at the inlet. It drops and eventually becomes zero at the primary flow front. Liquid within the fiber tows may be drawn forward beyond the primary flow front due to the capillary action. This is the origin of the wick flow front. In the fiber mats, there are local inhomogenities. Also liquid wicks into the single stitches and the relatively thicker stitches on the single stitched side. All these may be responsible for the irregular shape of the wick flow front.
Beyond the wick flow front, the fibers are dry or, in other words, saturation is zero (saturation is defined as the fractional volume of pores in the porous media occupied by the wetting phase). The capillary pressure is a function of saturation and is highest for zero saturation. Moving towards the primary flow front, saturation increases and hence, the capillary pressure decreases. At the primary flow front, saturation is still not 1.0, that is, the fibers are not completely wet. Hence, there is a finite capillary pressure. The shape of the primary flow front depends on the relative magnitudes of the capillary and the hydrodynamic pressures. If the capillary pressure is larger than the hydrodynamic pressure, liquid flows faster within the fiber tows than in channels between the tows [ILLUSTRATION FOR FIGURES 1 AND 3A OMITTED]. If, on the other hand, the hydrodynamic pressure is larger than the capillary pressure, flow between the tows leads the flow within the tows [ILLUSTRATION FOR FIGURES 2, 3B, AND 4 OMITTED]. This velocity difference accounts for the lead-lag phenomenon near the flow front (also called fingering). The extent of lead-lag depends on the relative magnitudes of the capillary and the hydrodynamic pressures. If the two pressures are comparable in magnitude, a relatively flat flow front will be observed. As the saturation approaches 1.0, the capillary pressure becomes negligibly small and the only driving force is the hydrodynamic pressure.
Thus, the flow through porous media can be divided into three regions. In Zone I (wick from zone), only the capillary pressure is operative. In Zone II (primary front zone), both the capillary and the hydrodynamic pressures are important. In Zone III (fully developed flow zone), only the hydrodynamic pressure is the driving force. The length of Zone I depends on the wicking rate and the rate at which the primary flow front propagates. The wicking rate depends on liquid properties (surface tension and viscosity) and the solid-liquid-air contact angle, and is independent of the flow rate. At low flow rates, the length of Zone I is large as compared to at high flow rates. At a high enough flow rate, Zone I may shrink to almost zero length, as was observed in Fig. 2. At a high flow rate and/or for liquids with poor wetting characteristics, wetting may take place behind the primary flow front [ILLUSTRATION FOR FIGURE 4 OMITTED].
(b) Void formation studies
Figure 6 shows voids formed when DOP oil was injected into the unidirectional fiber mat at [u.sub.s] [similar to] 0.2 cm/s. This photograph was taken after the liquid injection process was completed. As can be seen there were large number of voids trapped inside the fiber mat. There were two distinct locations where the voids were trapped: between the fiber tows and within the double stitches in the neck area. The voids in the gap between the tows were elongated, with the dimension in the direction of fibers being larger. As flow rate increased the elongated voids shrank in the flow direction, with no apparent lateral decrease in size. The voids in the double stitches were relatively circular in shape. Further injection of liquid at the same flow rate after the mold filling was complete did not result in any movement of these voids. when liquid was injected at a higher flow rate ([u.sub.s] [similar to] 3.9 cm/s), the voids started to move and eventually all the voids could be driven out (i.e. bleeding). when DOP oil was injected at relatively high flow rates no voids of these types were found trapped in the fiber mat. Similar results were obtained for other liquids and unsaturated polyester resin also. One major difference in the results obtained using water was that, the trapped voids could not be eliminated by pumping water at a high flow rate ([u.sub.s] [similar to] 3.9 cm/s) during the bleeding stage. Also bleeding could not be carried out for silicone oil of kinematic viscosity 1 cS, since voids were formed even at the highest available flow rate ([u.sub.s] [similar to] 3.9 cm/s). It has been observed that voids formed during injection of unsaturated polyester resin were stable, or in other words they retained their shape even after the resin was cured (39).
In this study it was found that injection at higher flow rates helped in eliminating or reducing voids. However, although no voids were formed between the tows and in the double stitch area when injection was carried out at high flow rates, the composite parts were not defect-free. Injection at high flow rates resulted in incomplete wetting due to the formation of microvoids in the fiber tows. Figure 7 shows photographs of fiber mat when unsaturated polyester resin was injected at (a) 0.05 cm/s and (b) 3.9 cm/s. Voids were formed between the fiber tows and within the double stitches at low flow rate [ILLUSTRATION FOR FIGURE 7A OMITTED]. Also comparison of Figs. 7a and 7b shows that wetting was poor at high flow rate. Patel et al. (40) have also reported similar results. They observed that lower injection pressure, and hence lower flow rate resulted in better fiber wetting. This was because at high flow rates, wetting of fiber tows took place after the flow front had already passed. This often resulted in trapping of air between the fiber filaments leading to poor wetting. Thus, high flow rates reduced voids, while low flow rates favored wetting. There seems to be a specific flow rate or a range of flow rates for each liquid at which both void and microvoid contents are minimized. In this paper only the analysis of void formation in the gap between the tows and in the double stitches is presented. Detailed analysis of microvoid formation within the fiber tows is reported elsewhere (39).
The void formation mechanism can be understood by examining the detailed flow front progression at low flow rates. Figure 8 shows a sequence of photographs illustrating the nature of flow as the primary flow front moved past one double stitch and progressed to the next double stitch. As the primary flow front moved towards a double stitch, the flow within the tows led the flow between the fiber tows [ILLUSTRATION FOR FIGURE 8A OMITTED]. When the flow front within the tow reached the double stitch, there was a transverse flow (flow perpendicular to the main flow direction) [ILLUSTRATION FOR FIGURE 8B OMITTED]. This then led to the two adjacent flow fronts meeting each other. The flow between the tows had not yet reached the double stitch. Hence, air was trapped in the gap between the tows [ILLUSTRATION FOR FIGURE 8C OMITTED]. The shape of this type of voids would depend on the relative rates of the transverse flow and the flow between the tows, i.e. the amount of fingering. If the fingering is severe, a large elongated void would be trapped. However, if the lead-lag is relatively small, the flow front between the fiber tows would reach the double stitch before the transverse flow could be completed. As a result, no voids would be trapped.
The occurrence of circular voids within the double stitches can be explained by examining the flow in the double stitch area [ILLUSTRATION FOR FIGURE 9 OMITTED]. When the flow reached the first limb of the double stitch, it slowed down. After the liquid moved past the limb, it flowed faster again within the fiber tow [ILLUSTRATION FOR FIGURE 9A OMITTED]. Since the distance between the two limbs of the double stitch was shorter away from the neck area, the flow reached the other limb first away from the neck region [ILLUSTRATION FOR FIGURE 9B OMITTED]. Hence, there was a transverse flow that led to the trapping of air in the neck area [ILLUSTRATION FOR FIGURE 9C OMITTED]. Again, whether these circular voids would form or not depends on the amount of lead-lag and transverse flow within the double stitch area.
Thus for both types of voids the mechanism of formation is the same. Two microflows should be present, one is the lead-lag at the flow front and the other is the transverse flow. The lead-lag or fingering has to be large enough so that the two adjacent leading flow fronts can meet before the lagging flow front has time to catch up and drive the air out. The extent of fingering depends on the balance of the capillary and the hydrodynamic pressures. This, in turn, depends on liquid properties, pressure profile, and the fiber mat structure. The cross flow or transverse flow may take place due to the presence of stitches, mat pore structure, or local inhomogeneities.
During the bleeding stage, when liquid was injected at a higher flow rate, the voids could be purged from the fiber mat. This can be explained as follows. After the voids are formed the liquid hydrodynamic pressure is balanced by the air pressure inside the voids plus the capillary pressure, or in other words voids are stabilized by the capillary pressure. In addition, the void may be pushed into a corner with a very small opening. Thus, in order to move a particular void, the surrounding hydrodynamic pressure gradient has to overcome the capillary pressure as well as the resistance due to the steric hindrance. When liquid was pumped at a higher flow rate during the bleeding stage, the applied pressure gradient was large enough to generate sufficient viscous drag to drive the voids out of the mat. The larger voids were elongated by the large viscous drag force acting at high pressure and were broken into smaller voids. These smaller voids were then driven out of the fiber mat by the liquid due to lesser steric hindrance. As mentioned earlier, in the case of water, trapped voids could not be eliminated by bleeding at higher flow rates. This can be explained as follows. Capillary forces stabilizing the voids were high in the case of water due to its high surface tension. Also due to low viscosity of water enough viscous drag forces could not be generated to drive the voids out of the fiber mat.
(c) Quantification of void formation
It was pointed out earlier that the size of voids in the flow direction decreased with increasing flow rate. It was of interest to see whether the void fraction could be correlated to the flow rates. Hence, investigation of void formation was carried out for different liquids by systematically varying the flow rates. Each experiment was carried out at least twice to minimize experimental error. Photographs of the fiber mats were taken after the liquid injection was completed and the area of voids was computed. Thus void fraction is reported as percentage aerial voidage. In this analysis only the voids formed between the fiber tows and in the double stitches were considered. Figure 10 shows the results obtained for various liquids. In the case of glycerine, no voids were observed in the double stitches or in the gap between the fiber tows for the range of superficial velocities investigated ([u.sub.s] [similar to] 0.004 cm/s-3.9 cm/s). As can be seen from these figures, no voids were formed above a certain critical velocity. Below this critical velocity void fraction increased logarithmically. The critical velocity was different for different liquids. Thus, void fraction is not only a function of velocity but also the liquid properties. It was proposed earlier that the voids are formed due to fingering at the flow front. The fingering in turn depends on the balance of the capillary and viscous forces. Capillary number (Ca) is considered to give the measure of the ratio of the viscous and the capillary forces in the flow field (41).
Ca = [Eta][u.sub.s]/[[Gamma].sub.LV] (4)
where, [Eta] is the viscosity of liquid.
When the void fraction data for various liquids were plotted against capillary number three curves were obtained as shown in Fig. 11. The void fraction data for the DOP oil and various silicone oils fell on one curve, whereas water and ethylene glycol gave separate curves. For each group of liquids there was a distinct value of critical cap filmy number ([Ca.sub.cri]) above which no voids were formed. This can be explained as follows. Capillary number as defined by Eq 4 does not take into account the liquid-solid-air contact angle. The DOP off and all the silicone oils have contact angle close to zero, whereas the contact angles for water and ethylene glycol are 66 [degrees] and 56 [degrees] respectively. Thus, corresponding to three distinct values of contact angle three curves were obtained when void fraction was plotted against capillary number. The equations relating the voids fraction (V) with the capillary number are:
V = -35.346 - 14.344 Log(Ca)
For Group I ([Theta] [similar to] 0 [degrees]) and [Ca.sub.cri] = 3.43 x [10.sup.-3] (5)
V = -41.451 - 14.706 Log(Ca)
For Group II ([Theta] [similar to] 56 [degrees]) and [Ca.sub.cri] 1.52 x [10 .sup.-3] (6)
V = -78.332 - 21.680 Log(Ca)
For Group III ([Theta] [similar to] 66 [degrees]) and [Ca.sub.cri] = 2.44 x [10.sup.-4] (7)
Unsaturated polyester resin has a surface tension of 34.5 mN/m and gives an equilibrium contact angle of 38 [degrees] with glass fibers. Thus if a curve of void fraction vs. capillary number were to be generated for UP resin, it would fall between that for Groups I and II. Three experiments were carried out with UP resin to verify this argument. The results obtained are shown in Table 5. As expected, at the same capillary number, UP resin gave void fraction between that for Groups I and II.
A modified capillary number ([Ca.sup.*]) can be defined by dividing the capillary number by the contact angle.
[Ca.sup.*] = [Eta][u.sub.s]/[[Gamma].sub.LV] cos [Theta] (8)
When the void fraction was plotted vs. the modified capillary number the curves for the three groups of liquids came closer to each other [ILLUSTRATION FOR FIGURE 12 OMITTED] The critical modified capillary number above which no voids are formed in the gap between the tows and in the double stitches was in the range of 6.00 x [10.sup.-4] - 3.43 x [10.sup.-3]. However, a master curve was not obtained. The possible reasons for this could be as follows. Contact angle is a function of velocity, resin viscosity, and surface tension. In general it remains constant at sufficiently low values of capillary number, then increases as capillary number increases. The dependence of contact angle on capillary number has been reported to begin in the range of [10.sup.-6] [less than] Ca [less than] [10.sup.-5] (42). Since the flow rates used in this study gave capillary numbers higher than this critical range, dynamic contact angles were developed. If dynamic contact angle is used in Eq. 8 instead of equilibrium contact angle, the curves for the three groups of liquids may come closer to each other. The equipment used for measuring contact angle in this study was incapable of generating the velocities used in flow visualization. In addition to the dynamics of contact angle, the modified capillary number defined by Eq. 8 does not take into account the surface energy of the mold surface. Liquids in the three groups may interact differently with the mold surface thus affecting surface void content differently.
(d) Effect of mold surface on void formation
In order to assess the effect of mold surface on void formation it was necessary to carry out experiments using molds of varying surface properties, the main property of interest being surface energy or the critical surface tension. Survey of the relevant literature revealed that surface energy of PMMA, which was used as mold in this study, is [similar to] 41.1 mN/m, whereas that of clean metal and glass is [similar to] 49 mN/m. It is well known that use of wax or mold release agent lowers the surface energy of mold surfaces (Table 4a). When experiments were carried out using glass plate coated with mold release agent, the surface energy was indeed found to be lower (Table 4b).
Table 5. Comparison of Void Fractions for Unsaturated Polyester Resin and Groups I and II Liquids (Fiber Tows in the Direction of Global Flow).
Percentage Aerial Voidage Capillary UP Group I Group II Number Resin Liquids(a) Liquids(b)
3.17 x [10.sup.-3] 0.00 0.51 0.0 1.58 x [10.sup.-3] 2.20 4.82 0.0 0.79 x [10.sup.-3] 7.19 9.14 4.16
a Predictions by Eq 5. b Predictions by Eq 6.
Three flow visualization experiments were carried out using glass mold coated with mold releases agent. DOP off was used as the wetting liquid. The results are shown in Table 6. Indeed, the void formation was affected by the mold surface. The reduction in mold surface energy due to application of mold release agent resulted in an increase in void fraction at the same flow rate. The reason for this may be that the mold release agent provided sites for the attachment of voids. Similar results have been reported by Stabler et al. (13). In this study clean PMMA mold platens were used for void formation studies for majority of experiments. It should, however, be noted that in actual production of composite parts by LCM techniques. mold release agent has to be applied to the mold surfaces in order to permit easy demolding of the parts and to prevent damaging the composite part surfaces. Thus the surface void content in actual composite part would be more than predicted by this study.
Table 6. Effect of Mold Surface on Void Formation for Injection of DOP Oil (Fiber Tows in the Direction of Global Flow).
Percentage Aerial Voidage
Capillary Clean PMMA Glass Coated with Mold Number Mold(a) Release Agent(b)
1.36 x [10.sup.-3] 5.75 7.70 0.67 x [10.sup.-3] 10.20 14.33 0.34 x [10.sup.-3] 14.36 20.09
a Predictions by Eq 5. b Experimental data.
Results Obtained Using Unidirectional Stitched Fiberglass Mat With Fiber Tows Transverse to the Direction of Global Flow
When liquids were injected into the unidirectional fiber mat with fiber tows perpendicular to the global flow direction, similar results were obtained. Figure 13 shows a sequence of photographs illustrating the micro scale flow pattern and void formation mechanism when DOP oil was injected into the unidirectional fiber mat at a low flow rate ([u.sub.s] [similar to] 0.05 cm/s). Again, the double stitched side was at the top. Note that here, double stitches and the relatively thicker stitches on the single stitched side were in the direction of the main flow. Flow was leading in the double stitches [ILLUSTRATION FOR FIGURE 13A OMITTED]. However, the fingering was not as strong as in the case of flow in the fiber direction. The mechanism of flow front progression proposed earlier was found to be applicable here too. In this case, the main capillary action was in the direction perpendicular to the primary flow, although some wicking did exist in the direction of primary flow due to the presence of stitches. At higher flow rates, the flow front was relatively flat. The mechanism of void formation was also the same as in the case of flow in the direction of tows. As liquid reached double stitches, it flowed faster in the limbs of the double stitches. As flow continued, liquid also started to flow along the fibers in the fiber tows due to the capillary action. Thus, the presence of stitches was the source of Fingering and the presence of fiber tows led to the transverse flow. Voids were formed if the adjacent leading flow fronts met before air could be driven out by the lagging flow front. At low flow rates, both the elongated voids in the gap between the fiber tows and the circular voids in the double stitches were formed. Voids of these types were not formed at high flow rates. However, injection at high flow rates resulted in incomplete wetting. It seemed that the problem of poor wetting was more severe for the flow transverse to the fiber tows as compared to the flow in the directing of the tows. Again voids formed at low flow rates could be eliminated during the bleeding stage by injecting DOP oil at high flow rates. When experiments were carried out using other liquids, similar mechanism was observed. The same general trend of increase in void fraction with decrease in velocity was observed for all liquids [ILLUSTRATION FOR FIGURE 14 OMITTED]. When void fractions for different liquids were plotted against the capillary number, three curves corresponding to three different values of contact angles were obtained [ILLUSTRATION FOR FIGURE 15 OMITTED]. Thus, there was a logarithmic relationship between the void fraction and the capillary number given by the following equations.
V = -27.003 - 14.382 Log(Ca)
For Group I ([Theta] [similar to] 0 [degrees]) and [Ca.sub.cri] = 1.33 x [10.sup.-2] (9)
V = - 57.817 - 18.857 Log (Ca)
For Group II ([Theta] [similar to] 56 [degrees]) and [Ca.sub.cri] = 8.59 x [10.sup.-4] (10)
V = -45.512 - 15.527 Log (Ca)
For Group III ([Theta] [similar to] 66 [Theta]) and [Ca.sub.cri] = 1.17 x [10.sup.-3] (11)
Again, on plotting the void fraction data vs. modified capillary number the curves for these three groups of liquids came close to each other [ILLUSTRATION FOR FIGURE 16 OMITTED]. Master curve was not obtained since the dependence of contact angle on velocity and the effect of mold surface energy were not taken into account in the present analysis. The critical modified capillary number above which no voids were formed between the tows and in the double stitches was in the range of 1.53 x [10.sup.-3] - 1.33 x [10.sup.-2]. Comparison of the critical modified capillary numbers for the global flow in the direction of fibers and transverse to the fibers shows that voids are more likely to form when flow is transverse to the fiber direction. Thus, the void formation depends on the orientation of fibers relative to the flow direction in addition to the capillary number and the contact angle. Again, as was observed in the case of global flow in the direction of fiber tows, voids formed by injection of water could not be eliminated during the bleeding stage by injecting water at higher flow rates, and bleeding could not be carried out for silicone oil of kinematic viscosity 1 cS since voids were formed even at the highest available flow rate ([u.sub.s] [similar to] 3.9 cm/s).
Mechanisms of micro scale flow behavior and void formation in unidirectional stitched fiberglass mat were proposed in this study. In a fiber mat, fingering at flow front takes place because the permeabilities in the fiber tows and in the gap between the fiber tows are different. The extent of fingering depends on the relative magnitudes of the capillary and the hydrodynamic pressures. In addition to this primary flow front, liquid may wick along the fibers by capillary action. The extent of wicking depends on liquid properties and flow rate. At low flow rates the voids are formed at the flow front. Two types of microflows are necessary for void formation: one is fingering at the flow front and the other is the transverse or cross flow. In general, voids are trapped in the fiber mat where inhomogeneities, such as stitches, obstruct the pathway of voids and prevent them from being swept away by the liquid flow. Use of mold release agent enhances void formation by lowering solid surface energy. At high flow rates, although voids crossing several fiber filaments are not formed, the wetting of fiber tows is incomplete.
This work was supported by the National Science Foundation (Grant DDM-9112990) and the Engineering Research Center for Net Shape Manufacturing at The Ohio State University. The authors would like to thank Yulu Ma for helping in quantifying void formation. Material donations from Ashland Chemical, Dow Chemical, and Dow Corning are greatly appreciated.
1. J. S. Hayward and B. Harris, Plast. Rubber Process. Applications, 11, 191 (1989).
2. C. D. Rudd, M. J. Owen, V. Middleton, K. N. Kendall, and I. D. Revill, Proc. 6th ASM/ESD Adv. Composites Conf., 301 (1990).
3. M. Schlack. Plastics World, May 1990, p. 61.
4. M. V. Bruschke and S. G. Advani, SAMPE Quart, 23, 2 (1991).
5. R. S. Hansen, SME Technical Papers, EM90-214 (1990).
6. C. F. Johnson, "Resin Transfer Molding," in Composite Materials Technology - Processes and Properties, Ch 5, P. K. Mallick and S. Newman, eds, Hanser Publishers, New York (1990).
7. S. G. Damani and L. J. Lee, Polym. Compos., 11, 174 (1990).
8. C. W. Macosko, in RIM: Fundamentals of Reaction Injection Molding, Hanser Publishers, New York (1989).
9. N. C. W. Judd and W. W. Wright, SAMPE J., 14, 10 (1978).
10. S. Feldgoise, M. F. Foley, D. Martin, and J. Bohan, Proc. 23rd Intl. SAMPE Tech. Conf., 259 (1991).
11. S. R. Ghiorse and R. M. Jurta, Composites, 22, 3 (1991).
12. T. S. Lundstrom, B. R. Gebart, and C. Y. Lundemo, Proc. 47th Ann. SPI Compos. Inst. Conf., Session 16-F (1992).
13. W. R. Stabler, G. B. Tatterson, R. L. Sadler, and A. H. M. El-Shiekh, SAMPE Quart., 23, 38 (1992).
14. K. L. Ahn, J. C. Seferis, and L. Letterman, SAMPE J., 27, 19 (1991).
15. R. Dave, J. L. Kardos, S. J. Choi, and M. P. Dudukovic, Proc. 32nd Intl. SAMPE Symp., 325 (1987).
16. J. L. Kardos, R. Dave, and M. P. Dudukovic, Proc. ASME Manufacturing Intl. '88, IV, 41 (1988).
17. N. Patel, L. J. Lee, W. Young, and M. J. Liou, SPE ANTEC Tech. Papers, 37, 1985 (1991).
18. J. A. Molnar, L. Trevino, and L. J. Lee, Polym. Compos., 10, 414 (1989).
19. R. J. Morgan, D. E. Larive, Y. H. Huang, K. P. Battjes, and Z. Yuan, Proc. 6th ASM/ESD Adv. Composites Conf., 229 (1990).
20. R. C. Peterson and R. E. Robertson, Proc. 8th ASM/ESD Adv. Composites Conf., 63 (1992).
21. R. C. Peterson and R. E. Robertson, Proc. 7th ASM/ESD Adv. Composites Conf., 203 (1991).
22. T. J. Wang, M. J. Perry, and L. J. Lee, SPE ANTEC Tech. Papers, 38, 756 (1992).
23. A. D. Mahale, R. K. Prud'homme, and L. Rebenfeld, Polym. Eng. Sci., 32, 319 (1992).
24. B. K. Larson and L. T. Drzal, SPE ANTEC Tech. Papers, 38, 752 (1992).
25. R. S. Parnas and F. R. Phelan Jr., SAMPE Quart., 22, 53 (1991).
26. A. W. Chan and R. J. Morgan, SAMPE Quart., 23, 48 (1992).
27. A. W. Chan and R. J. Morgan, SPE ANTEC Tech. Papers, 39, 844 (1993).
28. T. Sadiq, R. Parnas and S. Advani, Proc. 24th Intl. SAMPE Tech. Conf., 660 (1992).
29. J. J. Elmendorp and F. During, SPE ANTEEC Tech. Papers, 36, 1361 (1990).
30. B. Miller, in Surface Characteristics of Fibers and Textiles, Part II, pp. 417-45, M. J. Schick, ed., Marcel Dekker, Inc., New York (1977).
31. J. C. Berg, "Composite Systems from Natural and Synthetic Polymers," in Material Science Monograph, 26, L. Salem, A. De Ruvo, J. C. Seferis, and E. B. Stark, eds., Elsevier (1986).
32. Dynamic Contact Angle Analyzer, DCA-322, Instruction Manual, Cahn Instruments, Inc., Cerritos, Calif.
33. H. W. Fox and W. A. Zisman, J. Colloid. Sci., 5, 514 (1950).
34. L. A. Girifalco and R. J. Good, J. Phys. Chem., 61, 904 (1957).
35. F. M. Fowkes, J. Phys. Chem., 66, 382 (1962).
36. D. K. Owens and J. Wendt, Appl. Polym. Sci., 13, 1741 (1969).
37. D. H. Kaeble, J. Adhesion, 2, 66 (1970).
38. S. Wu, J. Polym. Sci., 34, 19 (1971).
39. V. Rohatgi, N. Patel, and L. J. Lee, Proc. 9th ASM ESD Adv. Composites Conf. (1993).
40. N. Patel, V. Rohatgi, and L. J. Lee, Polym. Compos., 14, 161 (1993).
41. F. A. L. Dullien, in Porous Media Fluid Transport and Pore Structure, Academic Press, Inc., San Diego, Calif. (1992).
42. K. J. Ahn, J. C. Seferis, and J. C. Berg, Polym. Compos., 12, 146 (1991).
|Printer friendly Cite/link Email Feedback|
|Author:||Patel, N.; Rohatgi, V.; Lee, L. James|
|Publication:||Polymer Engineering and Science|
|Date:||May 1, 1995|
|Previous Article:||Dilatometric study of low profile unsaturated polyester resins.|
|Next Article:||Analysis of volumetric changes of unsaturated polyester resins during curing.|