- Email: [email protected]
The viscoelastic compression behavior of liquid composite molding preforms
Composites: Part A 34 (2003) 431–444 www.elsevier.com/locate/compositesa
The viscoelastic compression behavior of liquid composite molding preforms S. Bickerton*, M.J. Buntain, A.A. Somashekar Department of Mechanical Engineering, Centre for Advanced Composite Materials, University of Auckland, Private Bag 92019, Auckland, New Zealand Received 26 September 2002; revised 27 February 2003; accepted 5 March 2003
Abstract The compressive deformation of reinforcements is an important consideration for many composites manufacturing processes. The apparent viscoelastic behavior of three common Liquid Composite Molding reinforcements has been studied, and the implications for modeling mold filling processes discussed. Though no resin was present, these fibrous structures have displayed complex time-dependent response, including loading hysteresis, stress relaxation, and strain rate dependent loading behavior. Stress relaxations have been observed to be as high as 64% of the peak stress, with the magnitude of relaxation being dependent on the preform fiber volume fraction. Increasing peak stresses are generated with faster strain rates, and two of the materials studied displayed stress relaxation sensitive to the initial strain rate. Two mold filling experiments are also presented, demonstrating the significant influence of preform viscoelasticity, and providing motivation for further study of these effects. q 2003 Elsevier Science Ltd. All rights reserved. Keywords: Liquid composite molding (LCM); A. Fabrics/textiles; B. Stress relaxation; E. Tooling
1. Introduction The majority of fiber reinforced plastic manufacturing processes involve compressive deformation of the reinforcing material being included in the product. For some processes, development of accurate process models will depend on good knowledge of the stress –strain behavior of the reinforcement. In this paper we describe an experimental investigation into the apparent viscoelastic behavior of three common reinforcements used in Liquid Composite Molding (LCM) processes. The time-dependent behavior demonstrated has the potential to significantly influence the mold filling phases of all LCM processes. The term LCM encompasses a growing list of thermoset composite manufacturing processes, including Resin Transfer Molding (RTM), Injection/Compression Molding (I/CM), and the Seeman Composite Resin Infusion Molding Process (SCRIMP). All LCM variations involve placement of a reinforcing structure, or preform, within some form of * Corresponding author. Tel.: þ 64-9-373-7599x88194; fax: þ 64-9-3737479. E-mail address: [email protected] (S. Bickerton).
closed mold. A thermosetting resin is introduced into the mold cavity, completely saturating the porous preform. Compressive deformation of the preform occurs prior to, or in many cases during resin impregnation. For some situations resin flow and preform deformation are highly coupled due to cavity thickness changes during mold filling, a feature that is inherent to a number of LCM processes. The various LCM processes can be distinguished by the nature of the mold filling phase. Molds are either rigid, flexible, or formed from thin flexible membranes, while resin can be introduced using a combination of positive pressure and vacuum driven infusion, and possibly aided by compression driven flow. Several examples are presented schematically in Fig. 1. Significant effort has been placed into numerical simulation of RTM filling, for which we have the luxury of assuming a completely rigid tool, with the preform compressed to the final dimensions prior to injection [1 –3]. These studies have focused on specification of gates and vent placement, determination of mold filling times and internal resin pressures, while little attention has been given to the prediction of internal forces and total mold clamping required during the process. Several efforts to model filling for the SCRIMP [4 – 6] and I/CM [7 – 11]
1359-835X/03/$ - see front matter q 2003 Elsevier Science Ltd. All rights reserved. doi:10.1016/S1359-835X(03)00088-5
432
S. Bickerton et al. / Composites: Part A 34 (2003) 431–444
Fig. 1. Schematic comparison of mold filling for (a) RTM, (b) I/CM, and (c) SCRIMP.
processes have been presented. These processes require consideration of the total pressure applied to the reinforcement, and the resulting coupled flow and reinforcement deformation problem. These initial studies have utilised non-linear elastic models for reinforcement compression, an approach that we will show does not capture the complex deformation behavior of fibrous reinforcing materials. In this paper we present a systematic study of three common reinforcements, and discuss the implications viscoelastic behavior has for improved modeling of LCM filling processes. Results from two sets of LCM filling experiments are presented, providing motivation for the study of viscoelastic reinforcement deformation. The clamping force required through an RTM and I/CM filling process are presented, demonstrating the influence of these effects.
2. Viscoelastic deformation behavior 2.1. Previous work The current study has been motivated by our experimental observations, and other authors descriptions in the literature
of the viscoelastic behavior of LCM reinforcements. The stress –strain behavior in transverse compression exhibits classic viscoelastic characteristics, including stress relaxation, strain rate dependency, and hysteresis. Fig. 2 demonstrates schematically these observed behaviors for a variety of specified deformation histories. In general, stress may be a function of strain, strain rate, and time. Conceptually simple experiments are required to investigate reinforcement behavior under compression, requiring controlled closure of two rigid, parallel plates, while monitoring the required applied force. Many previously published composites process models have utilised nonlinear elastic models for compressive deformation behavior, stress assumed a function of local fiber volume fraction only [4 –13]. The application of an elastic model may be justified if the time scale of the manufacturing process considered is long enough. These models have been successfully applied to simulate manufacturing processes involving prepreg consolidation [12,13]. Several compressive deformation models have been developed in the textiles literature, focusing on assemblies of randomly oriented fibers, or to bundles of aligned fibers. The initial treatment was provided by van Wyk, his model
Fig. 2. Schematic representations of observed viscoelastic behavior. Mold cavity thickness, h; is specified above, with resulting compressive stress shown below. (a) Stress relaxation, (b) rate dependent behavior, (c) hysteresis (stress plotted versus cavity thickness).
S. Bickerton et al. / Composites: Part A 34 (2003) 431–444
addressing a three-dimensional network of randomly oriented fibers [14]. Van Wyk assumed that deformation of the network is caused solely by bending of the fiber segments between fiber – fiber contact points. Several extensions of this work have been presented, including changing fiber orientation during compression, and the use of fiber orientation density functions [15]. Carnaby and Pan introduced fiber slippage to their model, to account for stress hysteresis observed during repeated loading and unloading cycles [16]. Curiskis and Carnaby have also addressed deformation of bundles of aligned fibers [17]. A variety of models have appeared in the literature regarding compression of fibrous reinforcements. Gutowski and coworkers have presented models based upon bundles of initially wavy fibers [18], later considering the influence of fluid lubrication [19]. Sˇima´cˇek and Kharbari have extended Gutowski’s work to include frictional resistance to both in-plane and transverse shear [20], and also considered the potential lubricating effect of fiber sizing [21]. Both Lomov and Verpoest [22], and Chen and Chou [23] have included geometry of the full bundles in their models, addressing unit cells of plain weave fabrics. Chen et al. have extended their model to include interaction of multiple layers, addressing layer nesting [24]. While a brief review is presented here, Robitaille and Gauvin have provided an excellent discussion of analytical compression models in their recent paper [25]. It should be noted that none of the models mentioned above can generate stress relaxation. Noting a clear departure from purely elastic behavior, several authors have reported stress hysteresis and stress relaxation during compaction experiments on dry reinforcing fabrics [26 –28]. Robitaille and Gauvin have presented a thorough review of compression deformation data available in the literature, quantifying response during compression and relaxation phases [25]. This was followed by a well structured experimental program, serving to identify important parameters governing compression and relaxation [29]. They have identified rate of compaction, and repeated application of load to be the most significant parameters. In all of these studies the observed stress relaxation occurred over time scales on the same order as that required to complete a typical LCM process, therefore demonstrating the significant influence of this phenomenon. Few analytical models have been presented to account for stress relaxation within the reinforcement. Several authors have provided separate empirical fits for stress increase during constant speed compression, and then relaxation at constant strain [25,27]. Kim et al. have provided a model based on a parallel grouping of five Maxwell viscoelastic elements [26], but applied this only to simple stress relaxation. Saunders et al. have added a viscous term to a power law compression model [30], but this is only valid during compression, and will not display stress relaxation when a constant strain is held. It is clear that a more comprehensive model is required to address
433
the variety of deformation paths that may be encountered during LCM filling. While no effort is made to model the apparent viscoelastic behavior of preforms in this paper, a set of experimental data is established from which a comprehensive model may be defined. 2.2. Implications for the modeling of LCM filling processes Reinforcement viscoelasticity has a variety of implications to the full range of LCM processes. Considering RTM, as the mold is assumed rigid and is closed before resin injection, resin flow is decoupled from the preform compaction process. Reinforcement deformation behavior therefore has no effect on the progression of mold filling. If existing flow simulations are to be extended, and utilised to predict the total clamping force applied to a mold, or to accurately model the total pressure distribution applied to an RTM mold, an accurate representation of reinforcement deformation is required. The I/CM process includes an initial injection phase into a partially open mold, followed by a controlled closure of the mold to complete resin infusion. The mold tooling can typically be assumed rigid, and we may consider constant speed, or constant force closing in the compression phase. During the compression phase the total force applied to the tooling is balanced by the internally generated resin pressure, and the resistance to compaction offered by the preform. For constant speed closing, a good knowledge of the compressive behavior of the preform is required for prediction of total force applied to the mold. If a constant force closing condition is specified, accurate compressive behavior is a necessity to predict the evolution of mold filling, and hence mold filling time. Several authors have presented mold filling analyses of the SCRIMP process, addressing the coupled flow and preform deformation problem [4 – 6]. If accurate predictions are necessary, a good model for reinforcement behavior is required to predict the evolution of mold cavity thickness. To date, all reported SCRIMP flow models have assumed a non-linear elastic response. It seems for SCRIMP and all other LCM processes, there is significant motivation for detailed study of viscoelastic reinforcement behavior, and subsequent development of predictive models.
3. Sample LCM filling experiments A series of LCM experiments are presented here to demonstrate the influence of viscoelastic preform deformation behavior on mold filling. A simple mold geometry has been installed within an Instron testing machine, which allows for measurement of the required mold clamping force during processing. Several RTM and I/CM filling experiments were performed, and the clamping force traces reported here. The full study has been presented elsewhere,
434
S. Bickerton et al. / Composites: Part A 34 (2003) 431–444
including measurement of internally generated fluid pressures [31]. 3.1. Experimental equipment A cast aluminium mold was used, the part geometry being a 0.15 m radius circular disc of constant thickness. The mold is depicted in Fig. 3a, and consists of two pieces. The upper platen is a simple flat disc, while the lower platen is essentially a high walled basin. A resin or test fluid can be injected at the center of the mold. The mold was mounted on an Instron 1186 testing machine, with the upper platen attached to a 200 kN load cell, as shown in Fig. 3b. Using the Instron to control the speed and position of the female half of the mold, it was possible to consistently specify the compaction phases of RTM and I/CM cycles, while logging the clamping force to a computer running LabVIEW data acquisition software. A pressure transducer and one-way check valve were included in the injection line, as shown in Fig. 3b. The check valve prevents back flow of fluid during the compression phase of I/CM filling, effectively closing the injection gate. Fluid injection was controlled via a purpose built injection device capable of maintaining a wide range of constant flow rates. This device was directly controlled by the data acquisition computer, automatically initiating injection at the completion of the initial preform compression phase.
study was mixed to a viscosity of 0.178 Pa. s, at 18 8C on a Physica Rheolab viscometer. For the RTM experiments performed, the preform was compressed from 50 mm at a constant speed of 2.0 mm/min until the final cavity thickness of 4.7 mm was reached. Resin injection was initiated at this time, a constant flow rate being maintained. A series of four tests were completed, at injection flow rates of 3.06, 5.80, 8.83, and 11.43 cm3/s. For the I/CM experiments, the preform was compressed at 2.0 mm/min, until the cavity thickness reached 6.5 mm (a volume fraction of 28%). Resin was then injected at a constant flow rate of 18.7 cm3/s. Once the required volume of fluid was injected, a second phase of constant speed compression was initiated, and stopped once the cavity thickness reached 4.7 mm. A series of four experiments were completed, at compression speeds of 2.0, 8.0, 18.0, and 28.0 mm/min. For all experiments, total clamping force was measured after completion of filling, documenting further relaxation of this force. For each experimental series, a second compression force trace is presented. Two additional preform samples have been prepared, and exposed to the same compression history, without any fluid injection. This was done to demonstrate that significant viscoelastic deformation of the reinforcement occurs without the presence of a resin, and to highlight any further effects due to the presence of the fluid. 3.3. Results
3.2. Experimental procedure A 450 g/m2 E-glass fiber, continuous filament mat (CFM) was used as the reinforcement in these experiments. Circular specimens were cut from CFM with an outer diameter of 290 mm (^ 2 mm), and a central 19 mm diameter hole. Ten specimens were stacked into each preform sample, each having an initial thickness of approximately 48 mm. All experiments were initiated at a cavity thickness of 50 mm. The intended final thickness of the part was set to 4.7 mm, producing a fiber volume fraction of 38%. A non-reactive corn syrup solution was used as the test fluid for both experiments. This material was chosen as it is easily tailored to a desired Newtonian viscosity. The solution used in this
Fig. 3. (a) Upper and lower mold platens. (b) The mold installed within the Instron testing machine.
Clamping force histories for the RTM and I/CM experiments are presented in Figs. 4 and 5, respectively. Also presented are force histories for compression of the dry preform samples. A single experiment is depicted in Figs. 4a and 5a, including compression prior to fluid injection. All experiments are displayed in Figs. 4b and 5b, focusing on the period following initiation of fluid injection. Due to small variations in compression response, each curve has been normalised to the force required at the end of the initial compaction. Fluid injection was then initiated at this point, time being set equal to zero in each experiment. For the RTM experiments, the injection pressure reached maximums of 275.0, 490.0, 765.0 and 985.0 kPa. For the I/CM experiments, pressure at the center of the mold was maximum in each case at the completion of injection, being approximately 280.0 kPa. The dry compression curve in Fig. 4 shows rapid increase as cavity thickness approaches 4.7 mm at t ¼ 0: This curve demonstrates significant load relaxation after this time, the majority occurring over the first 100 s. The RTM experiment with Q ¼ 5:80 cm3 =s is also provided on this plot. The RTM curve is identical prior to injection, but dips significantly below the dry curve over the 37.0 s injection period. Assuming that the reinforcement carries the same load when dry and wet, we had expected the RTM data to rise above the dry curve due to the force generated by internal fluid pressure. This additional force component
S. Bickerton et al. / Composites: Part A 34 (2003) 431–444
435
Fig. 4. Normalised clamping force histories for the RTM experiments, and related dry compression test. Measured clamping forces at t ¼ 0 s were 36.8, 35.0, 33.2 and 28.6 kN for the RTM experiments with increasing flow rate, and 37.5 kN for the dry compression test. (a) 2 400 , t , 400 s, (b) 0 , t , 120 s.
could be determined by integrating the fluid pressure field over the wetted surface area of the mold at any time. The actual behavior demonstrates a balance between the force due to fluid pressure, and accelerated stress relaxation within the reinforcement due to the presence of fluid. The presence of a fluid may serve to lubricate fiber – fiber connections, allowing the reinforcement to settle into a state
requiring lower compressive stress. At the completion of injection the force trace shows a sudden drop, as the force generated by internal fluid pressure is removed. Shortly afterwards, the RTM data adopts the same relaxation rate as the dry curve, the clamping force settling to a significantly lower value. The series of RTM force traces provided in Fig. 4b all exhibit similar behavior. As injection flow rate is
436
S. Bickerton et al. / Composites: Part A 34 (2003) 431–444
Fig. 5. Normalised clamping force histories for the I/CM experiments, and related dry compression test. Measured forces at t ¼ 0 s were 11.0, 10.6, 10.9 and 10.6 kN for the I/CM experiments with increasing compression speed, and 12.1 kN for the dry compression test. (a) 2100 , t , 100 s, (b) 0 , t , 100 s.
increased, the clamping force is seen to rise more rapidly during injection, increasing above the dry compression curve for the faster flow rate. In each case, the clamping force drops significantly below the dry curve once fluid injection is complete. The dry compression curve in Fig. 5a peaks initially at t ¼ 0; and then exhibits load relaxation as the cavity
thickness is held constant. The load rises rapidly during the second phase of compaction, before demonstrating significant relaxation once the final cavity thickness is reached. The I/CM experiment with compression speed of 18 mm/ min is also provided on this plot. As was found for the RTM experiments, during the injection phase, the I/CM clamping force does not exceed the dry curve. For the applied
S. Bickerton et al. / Composites: Part A 34 (2003) 431–444
437
Fig. 6. Close-up images of studied reinforcement materials, (a) CFM, (b) CSM, and (c) PW.
injection flow rate, the force generated by fluid pressure has been balanced by the fiber lubrication effect. At the conclusion of injection ðt ¼ 16:0 sÞ; the fluid force component falls away, being approximately 2.2 kN. At the initiation of the secondary compression phase, the clamping force rises at a similar rate as for the dry experiment. For these conditions the I/CM curve is maintained below the dry curve, again a balance being made between the generated fluid force and the fiber lubrication effect. At the completion of filling ðt ¼ 23:0 sÞ the fluid force component falls away, and the I/CM curve assumes the same relaxation rate as the dry data. The series of I/CM force traces provided in Fig. 5b all display a similar drop relative to the dry curve at the completion of mold filling. As compression speed is increased, the clamping force is seen to rise more rapidly during injection, and an increasing trend is noted in the maximum force obtained. This is expected due to the greater fluid pressures generated by increasing compression speeds. The experiments presented in Figs. 4 and 5 demonstrate the influential role played by the viscoelastic behavior of reinforcements. For both situations presented, viscoelasticity will influence prediction of local and total clamping forces exerted on molds. Such behavior will be even more important for modeling of the constant force version of I/CM, and other LCM processes utilising flexible molds. The presence of a fluid within the reinforcement has a strong effect on compression behavior. For the remainder of this study we have concentrated on the response of dry reinforcements.
4. Experimental characterisation program A detailed study has been undertaken into compressive deformation behavior of the CFM (450 g/m2), a chopped strand mat (CSM, 430 g/m2), and a glass plain weave (PW, 600 g/m2). Images of each reinforcement are presented in Fig. 6. The behavior of each has been considered at varying compaction levels, and at various compression speeds. 4.1. Experimental procedure Two types of experiment have been carried out with each preform style. Samples of each material have been placed between a set of rigid, parallel plates setup within our Instron testing machine. A predetermined compressive strain history can be prescribed, while monitoring the required compaction force. The distance between the plate surfaces at any time is defined as the cavity thickness, h: h0 is defined as the initial thickness of the preform sample before any load is applied, and is determined prior to an experiment. The basic experiment is described schematically in Fig. 7, being a period of constant speed compression to a predetermined cavity thickness, then a holding period at that thickness. A typical load response curve is also presented. Load data was converted to an applied stress, all preform samples being cut into 0.2 m squares. The measured load curves are presented here as either stress – time, or stress – strain
Fig. 7. (a) Schematic description of cavity thickness evolution during a compression characterisation test. (b) Example of resulting load curve.
438
S. Bickerton et al. / Composites: Part A 34 (2003) 431–444
curves. For this study the measure of strain, 1; is defined below: 1¼
h0 2 h : h0
ð1Þ
Our initial studies have exposed a significant amount of permanent deformation that occurs within these materials when repeated loading cycles are applied. Preform samples displayed reductions in the force required to achieve a particular volume fraction with every loading cycle applied to them. Under manufacturing conditions preforms are cut from rolls, and placed into a mold without any significant previous compressive deformation. The characterisation experiments described here have been performed on freshly cut preform samples, removing any influence due to previous deformation. 4.2. Varying compaction level experiments For each preform style four samples have been compacted at 2.0 mm/min, to increasing final fiber volume fractions. These tests have been completed to assess preform relaxation at various compaction levels. The applied load has been recorded during this initial compaction period, and then during a period of constant cavity thickness. Six, eight, and eight layers of the CFM, CSM, and PW have been used in each preform sample. 4.3. Varying compression speed experiments Four or five samples of each preform style have been compressed to the same final fiber volume fraction, at a range of different compression speeds (from 0.5 to
10.0 mm/min). Otherwise, these tests are completed in an identical manner to those described above, using the same number of layers for each preform style. These tests have been completed to assess the effect of varying strain rate on compression response. 4.4. Continuous filament mat response The stress – time behavior recorded during the compaction level tests is presented in Fig. 8. During the constant speed phase of these experiments, each successive curve falls on the previous, such repeatability providing some assurance that the material is behaving in a possibly predictable manner. Measured peak stress was found to increase with increasing final volume fraction, which was varied from 29.6 to 51.9%. The relaxations were 40.3, 35.2, 27.4, and 19.5% of the peak stresses, from lowest to highest final volume fraction. These results are also presented as stress –strain data in Fig. 9. An additional curve has been plotted on this figure, representing the possible long term behavior after complete stress relaxation. By considering the stress reached in the material after long times at constant cavity thickness, it is possible to define the minimum compaction stress required as a function of strain. This stress –strain curve might be applicable for manufacturing processes occurring over long time scales, an elastic model being acceptable in these situations. The stress – time behavior recorded during the compression speed tests is presented in Fig. 10. In each case a final volume fraction of 41.5% was reached. Larger peak stresses are displayed with increasing speed, demonstrating significant rate dependent behavior. Once the peak stress has been reached, stress relaxation occurs to the same level,
Fig. 8. Stress versus time behavior recorded during compaction level tests for CFM.
S. Bickerton et al. / Composites: Part A 34 (2003) 431–444
439
Fig. 9. Stress versus strain behavior recorded during compaction level tests for CFM. The dotted line has been added to demonstrate possible long term behavior of the material.
independent of the initial compression speed. The same results are presented as stress –strain data in Fig. 11. In this format the rate dependent effect is clearly demonstrated, the stress – strain curve followed during compression being influenced by the rate at which the reinforcement is deformed. 4.5. Chopped strand mat response The recorded stress – time behavior for the CSM compaction level tests is presented in Fig. 12. The maximum
fiber volume fraction has been varied between 28.0 and 56.0%. Similar behavior has been observed as for CFM, with good repeatability demonstrated during the initial loading section. Again, decreasing amounts of relaxation were noticed with increasing final volume fraction, being 63.8, 60.6, 33.3, and 13.4%. Data collected for CSM during the compression speed tests is presented in Fig. 13. The final volume fraction for each test was 42.0%. This material also shows clear strainrate dependency in the peak stress. However, unlike CFM, the measured stress did not relax to the same level, though
Fig. 10. Stress versus time behavior recorded during compression speed tests for CFM. The final volume fraction achieved in each test was 41.5%.
440
S. Bickerton et al. / Composites: Part A 34 (2003) 431–444
Fig. 11. Stress versus strain behavior recorded during compression speed tests for CFM.
the final fiber volume fraction was identical in each case. It appears that this material shows a greater potential for stress relaxation the faster it is initially deformed. While in this study we have focused on the macroscopic behavior, consideration must be given to the microscopic architecture if this fundamental difference in the response of two glass random mats is to be explained. Significant differences in structure include fiber length (CFM being continuous, CSM formed from chopped strands 30 – 40 mm in length), and binder loading. The CSM includes a much larger amount of binder, essentially bonding the chopped strands together as a mat.
4.6. Plain weave response Data recorded during the compaction level tests is presented in Fig. 14, maximum volume fraction being varied from 55.0 to 65.0%. Qualitatively similar behavior is noted to the other materials studied, while the PW shows an increased potential for stress relaxation. The relaxations were 57.6, 49.2, 45.0, and 39.1% of the peak stresses, from lowest to highest final volume fraction. It is noticeable that the majority of stress relaxation occurs over a shorter time period than for CFM and CSM. Increased variation was found during the loading period, demonstrating
Fig. 12. Stress versus time behavior recorded during compaction level tests for CSM.
S. Bickerton et al. / Composites: Part A 34 (2003) 431–444
441
Fig. 13. Stress versus time behavior recorded during compression speed tests for CSM. The final volume fraction achieved in each test was 42.0%.
the influence of stacking variations, or nesting of layers. In particular the data for Vf ¼ 57:5% shows an increased rate of increase in measured stress. The stress versus time behavior recorded during the compression speed tests is presented in Fig. 15a, the final volume fraction being 52.5% in each case. Three of the curves display an increasing peak stress with increasing compression speed, behavior demonstrated by the other materials studied. The slowest compression speed breaks this trend, demonstrating the possible influence of stacking
and nesting with woven materials. This may also be caused by variations in reinforcement areal weight, though this is thought to be unlikely, as similar behavior should be expected from the two random mats studied. Similar to CSM, the PW shows an increased potential for stress relaxation with faster compression speeds. To further investigate the possible influence of layer nesting, a second set of tests was completed, the final volume fraction being set to 62.5%. The generated stress data is plotted in Fig. 15b, demonstrating stronger influence due to nesting at this
Fig. 14. Stress versus time behavior recorded during compaction level tests for PW.
442
S. Bickerton et al. / Composites: Part A 34 (2003) 431–444
Fig. 15. Stress versus time behavior recorded during compression speed tests for PW. The final volume fraction achieved in each test was (a) 52.5%, and (b) 62.5%.
higher volume fraction. From this data it is more difficult to identify any clear trends. It should be noted that the plane weave under consideration is a heavy fabric, requiring relatively few layers to generate the necessary part
thickness. Under such conditions, misalignment of 1 or 2 layers can have significant effect. A more in depth study is required to identify the magnitude of scatter due to nesting, considering a variety of woven and stitched fabrics.
S. Bickerton et al. / Composites: Part A 34 (2003) 431–444
5. Conclusion The apparent time dependent viscoelastic deformation of LCM reinforcements must be an important consideration for the simulation of these mold filling processes. Two LCM filling experiments were described, demonstrating the strong influence viscoelastic reinforcement deformation has on required mold clamping force. Models for deformation behavior are required to accurately predict generated clamping forces and total pressure distributions within a mold, and for some LCM variations, are necessary to accurately model filling. The initial characterisation study presented here has highlighted a variety of complex phenomenon displayed by three common LCM reinforcements. Each material showed significant stress relaxation after compression to a constant volume fraction. It is important to note that stress relaxation occurs over time periods on the same order as the full molding processes. The magnitude of stress relaxation is strongly dependent on the level of compaction, the amount of which is reduced with increased final fiber volume fraction for all three materials studied. The response of all three materials was dependent on the speed of compression, or strain rate. Increasing the strain rate produced greater peak stresses though samples were compressed to the same fiber volume fraction. The CFM subsequently relaxed to the same final stress independent of compaction speed, being the simplest behavior noted. CSM showed a clear increase in stress relaxation with increasing closing speed. This effect is possibly due to the presence of significant binder in the structure, providing a tougher challenge than the CFM to model. It was more difficult to identify trends in the behavior of the PW due to the influence of layer nesting. This phenomenon may affect the deformation behavior of other woven and stitched fabrics, the extent of which requires further investigation. The data presented in this paper will be used to develop empirical models for viscoelastic response, which will be necessary to address processes such as I/CM and SCRIMP. While this study has highlighted complex behavior, definite trends have been identified, and the authors are optimistic that useful deformation models can be produced.
Acknowledgements The authors would like to acknowledge the generous support of the Foundation for Research Science and Technology, New Zealand.
References [1] Bruschke MV, Advani SG. A finite element/control volume approach to mold filling in a anisotropic porous media. Polym Compos 1990; 11(6):398–405.
443
[2] Young WB, Rupel K, Han K, Lee LJ, Liou MJ. Analysis of resin injection molding in molds with preplaced fiber mats II: numerical simulation and experiments of mold filling. Polym Compos 1991; 12(1):20– 9. [3] Trochu F, Gauvin R, Gao D-M. Numerical analysis of the resin transfer molding process by the finite element method. Adv Polym Technol 1993;12(4):329–42. [4] Sun X, Li S, Lee LJ. Mold filling analysis in vacuum-assisted resin transfer molding. Part I: SCRIMP based on a high-permeable medium. Polym Compos 1998;19(6):807–17. [5] Han K, Jiang S, Zhang C, Wang B. Flow modeling and simulation of SCRIMP for composites manufacturing. Composites, Part A 2000; 31(1):79– 86. [6] Kang M-K, Lee W-I, Hahn H-T. Analysis of vacuum bag resin transfer molding process. Composites, Part A 2001;32(11):1553– 60. [7] Han K, Ni J, Toth J, Lee LJ. Analysis of an injection/compression liquid composite molding process. Polym Compos 1998;19(4):487–96. [8] Pham X-T, Trochu F, Gauvin R. Simulation of compression resin transfer molding with displacement control. J Reinforced Plast Compos 1998;17(17):1525–56. [9] Kang M-K, Lee W-I. Analysis of resin transfer/compression molding process. Polym Compos 1999;20(2):293–304. [10] Pham X-T, Trochu F. Simulation of compression resin transfer molding to manufacture thin composite shells. Polym Compos 1999; 20(3):436– 59. [11] Pillai KM, Tucker II CL, Phelan Jr FR. Numerical simulation of injection/compression liquid composite molding. Part 2. Preform compression. Composites, Part A 2001;32(2):207 –320. [12] Gutowski TG, Cai Z, Bauer S, Boucher D, Kingery J, Wineman S. Consolidation experiments for laminate composites. J Compos Mater 1987;21(7):650–69. [13] Dave R, Kardos JL, Dudukovic´ MP. A model for resin flow during composites processing: Part 1—general mathematical development. Polym Compos 1987;8(1):29–38. [14] van Wyk CM. Note on the compressibility of wool. J Textile Inst 1946;37:T285–92. [15] Komori T, Makishima K. Numbers of fiber-to-fiber contacts in general fiber assemblies. Textile Res J 1977;57:T15–T23. [16] Carnaby GA, Pan N. Theory of the compression hysteresis of fibrous assemblies. Textile Res J 1989;59(5):275–84. [17] Curiskis JI, Carnaby GA. Continuum mechanics of the fiber bundle. Textile Res J 1985;16(4):58–64. [18] Gutowski TG. A resin flow/fiber deformation model for composites. SAMPE Q 1985;16(4):58–64. [19] Cai Z, Gutowski TG. 3-D deformation of a lubricated fiber bundle. J Compos Mater 1992;26(8):1207–37. [20] Sˇima´cˇek P, Karbhari VM. Notes on the modeling of preform compaction: I micromechanics at the fiber bundle level. J Reinforced Plast Compos 1996;15(1):86–122. [21] Karbhari VM, Sˇima´cˇek P. Notes on the modeling of preform compaction: II effect of sizing on bundle level micromechanics. J Reinforced Plast Compos 1996;15(8):837–61. [22] Lomov SV, Verpoest I. Compression of woven reinforcements: a mathematical model. J Reinforced Plast Compos 2000;19(16):1329–50. [23] Chen B, Chou T-W. Compaction of woven-fabric preforms in liquid composite molding processes: single-layer deformation. Compos Sci Technol 1999;59:1519– 26. [24] Chen B, Leng AH-D, Chou T-W. A nonlinear compaction model for fibrous preforms. Composites, Part A 2001;32(5):701–7. [25] Robitaille F, Gauvin R. Compaction of textile reinforcements for composites manufacturing. I: review of experimental results. Polym Compos 1998;19(2):198 –216. [26] Kim YR, McCarthy SP, Fanucci JP. Compressibility and relaxation of reinforcements during composite processing. Polym Compos 1991; 12(1):13– 19. [27] Pearce N, Summerscales J. The compressibility of a reinforcement fabric. Compos Manufact 1995;6(1):15–21.
444
S. Bickerton et al. / Composites: Part A 34 (2003) 431–444
[28] Saunders RA, Lekakou C, Bader MG. Compression and microstructure of fibre plain woven cloths in the processing of polymer composites. Composites, Part A 1998;29(4):443–54. [29] Robitaille F, Gauvin R. Compaction of textile reinforcements for composites manufacturing. II: compaction and relaxation of dry and H2O-saturatedwovenreinforcements.PolymCompos1998;19(5):543–57.
[30] Saunders RA, Lekakou C, Bader MG. Compression in the processing of polymer composites. 2. Modeling of the viscoelastic compression of resin-impregnated fibre networks. Compos Sci Technol 1999;59(10):1483–94. [31] Buntain MJ, Bickerton S. Prediction of tooling forces in liquid composite molding processes. SAMPE J 2003;39(1):8– 15.
The viscoelastic compression behavior of liquid composite molding preforms S. Bickerton*, M.J. Buntain, A.A. Somashekar Department of Mechanical Engineering, Centre for Advanced Composite Materials, University of Auckland, Private Bag 92019, Auckland, New Zealand Received 26 September 2002; revised 27 February 2003; accepted 5 March 2003
Abstract The compressive deformation of reinforcements is an important consideration for many composites manufacturing processes. The apparent viscoelastic behavior of three common Liquid Composite Molding reinforcements has been studied, and the implications for modeling mold filling processes discussed. Though no resin was present, these fibrous structures have displayed complex time-dependent response, including loading hysteresis, stress relaxation, and strain rate dependent loading behavior. Stress relaxations have been observed to be as high as 64% of the peak stress, with the magnitude of relaxation being dependent on the preform fiber volume fraction. Increasing peak stresses are generated with faster strain rates, and two of the materials studied displayed stress relaxation sensitive to the initial strain rate. Two mold filling experiments are also presented, demonstrating the significant influence of preform viscoelasticity, and providing motivation for further study of these effects. q 2003 Elsevier Science Ltd. All rights reserved. Keywords: Liquid composite molding (LCM); A. Fabrics/textiles; B. Stress relaxation; E. Tooling
1. Introduction The majority of fiber reinforced plastic manufacturing processes involve compressive deformation of the reinforcing material being included in the product. For some processes, development of accurate process models will depend on good knowledge of the stress –strain behavior of the reinforcement. In this paper we describe an experimental investigation into the apparent viscoelastic behavior of three common reinforcements used in Liquid Composite Molding (LCM) processes. The time-dependent behavior demonstrated has the potential to significantly influence the mold filling phases of all LCM processes. The term LCM encompasses a growing list of thermoset composite manufacturing processes, including Resin Transfer Molding (RTM), Injection/Compression Molding (I/CM), and the Seeman Composite Resin Infusion Molding Process (SCRIMP). All LCM variations involve placement of a reinforcing structure, or preform, within some form of * Corresponding author. Tel.: þ 64-9-373-7599x88194; fax: þ 64-9-3737479. E-mail address: [email protected] (S. Bickerton).
closed mold. A thermosetting resin is introduced into the mold cavity, completely saturating the porous preform. Compressive deformation of the preform occurs prior to, or in many cases during resin impregnation. For some situations resin flow and preform deformation are highly coupled due to cavity thickness changes during mold filling, a feature that is inherent to a number of LCM processes. The various LCM processes can be distinguished by the nature of the mold filling phase. Molds are either rigid, flexible, or formed from thin flexible membranes, while resin can be introduced using a combination of positive pressure and vacuum driven infusion, and possibly aided by compression driven flow. Several examples are presented schematically in Fig. 1. Significant effort has been placed into numerical simulation of RTM filling, for which we have the luxury of assuming a completely rigid tool, with the preform compressed to the final dimensions prior to injection [1 –3]. These studies have focused on specification of gates and vent placement, determination of mold filling times and internal resin pressures, while little attention has been given to the prediction of internal forces and total mold clamping required during the process. Several efforts to model filling for the SCRIMP [4 – 6] and I/CM [7 – 11]
1359-835X/03/$ - see front matter q 2003 Elsevier Science Ltd. All rights reserved. doi:10.1016/S1359-835X(03)00088-5
432
S. Bickerton et al. / Composites: Part A 34 (2003) 431–444
Fig. 1. Schematic comparison of mold filling for (a) RTM, (b) I/CM, and (c) SCRIMP.
processes have been presented. These processes require consideration of the total pressure applied to the reinforcement, and the resulting coupled flow and reinforcement deformation problem. These initial studies have utilised non-linear elastic models for reinforcement compression, an approach that we will show does not capture the complex deformation behavior of fibrous reinforcing materials. In this paper we present a systematic study of three common reinforcements, and discuss the implications viscoelastic behavior has for improved modeling of LCM filling processes. Results from two sets of LCM filling experiments are presented, providing motivation for the study of viscoelastic reinforcement deformation. The clamping force required through an RTM and I/CM filling process are presented, demonstrating the influence of these effects.
2. Viscoelastic deformation behavior 2.1. Previous work The current study has been motivated by our experimental observations, and other authors descriptions in the literature
of the viscoelastic behavior of LCM reinforcements. The stress –strain behavior in transverse compression exhibits classic viscoelastic characteristics, including stress relaxation, strain rate dependency, and hysteresis. Fig. 2 demonstrates schematically these observed behaviors for a variety of specified deformation histories. In general, stress may be a function of strain, strain rate, and time. Conceptually simple experiments are required to investigate reinforcement behavior under compression, requiring controlled closure of two rigid, parallel plates, while monitoring the required applied force. Many previously published composites process models have utilised nonlinear elastic models for compressive deformation behavior, stress assumed a function of local fiber volume fraction only [4 –13]. The application of an elastic model may be justified if the time scale of the manufacturing process considered is long enough. These models have been successfully applied to simulate manufacturing processes involving prepreg consolidation [12,13]. Several compressive deformation models have been developed in the textiles literature, focusing on assemblies of randomly oriented fibers, or to bundles of aligned fibers. The initial treatment was provided by van Wyk, his model
Fig. 2. Schematic representations of observed viscoelastic behavior. Mold cavity thickness, h; is specified above, with resulting compressive stress shown below. (a) Stress relaxation, (b) rate dependent behavior, (c) hysteresis (stress plotted versus cavity thickness).
S. Bickerton et al. / Composites: Part A 34 (2003) 431–444
addressing a three-dimensional network of randomly oriented fibers [14]. Van Wyk assumed that deformation of the network is caused solely by bending of the fiber segments between fiber – fiber contact points. Several extensions of this work have been presented, including changing fiber orientation during compression, and the use of fiber orientation density functions [15]. Carnaby and Pan introduced fiber slippage to their model, to account for stress hysteresis observed during repeated loading and unloading cycles [16]. Curiskis and Carnaby have also addressed deformation of bundles of aligned fibers [17]. A variety of models have appeared in the literature regarding compression of fibrous reinforcements. Gutowski and coworkers have presented models based upon bundles of initially wavy fibers [18], later considering the influence of fluid lubrication [19]. Sˇima´cˇek and Kharbari have extended Gutowski’s work to include frictional resistance to both in-plane and transverse shear [20], and also considered the potential lubricating effect of fiber sizing [21]. Both Lomov and Verpoest [22], and Chen and Chou [23] have included geometry of the full bundles in their models, addressing unit cells of plain weave fabrics. Chen et al. have extended their model to include interaction of multiple layers, addressing layer nesting [24]. While a brief review is presented here, Robitaille and Gauvin have provided an excellent discussion of analytical compression models in their recent paper [25]. It should be noted that none of the models mentioned above can generate stress relaxation. Noting a clear departure from purely elastic behavior, several authors have reported stress hysteresis and stress relaxation during compaction experiments on dry reinforcing fabrics [26 –28]. Robitaille and Gauvin have presented a thorough review of compression deformation data available in the literature, quantifying response during compression and relaxation phases [25]. This was followed by a well structured experimental program, serving to identify important parameters governing compression and relaxation [29]. They have identified rate of compaction, and repeated application of load to be the most significant parameters. In all of these studies the observed stress relaxation occurred over time scales on the same order as that required to complete a typical LCM process, therefore demonstrating the significant influence of this phenomenon. Few analytical models have been presented to account for stress relaxation within the reinforcement. Several authors have provided separate empirical fits for stress increase during constant speed compression, and then relaxation at constant strain [25,27]. Kim et al. have provided a model based on a parallel grouping of five Maxwell viscoelastic elements [26], but applied this only to simple stress relaxation. Saunders et al. have added a viscous term to a power law compression model [30], but this is only valid during compression, and will not display stress relaxation when a constant strain is held. It is clear that a more comprehensive model is required to address
433
the variety of deformation paths that may be encountered during LCM filling. While no effort is made to model the apparent viscoelastic behavior of preforms in this paper, a set of experimental data is established from which a comprehensive model may be defined. 2.2. Implications for the modeling of LCM filling processes Reinforcement viscoelasticity has a variety of implications to the full range of LCM processes. Considering RTM, as the mold is assumed rigid and is closed before resin injection, resin flow is decoupled from the preform compaction process. Reinforcement deformation behavior therefore has no effect on the progression of mold filling. If existing flow simulations are to be extended, and utilised to predict the total clamping force applied to a mold, or to accurately model the total pressure distribution applied to an RTM mold, an accurate representation of reinforcement deformation is required. The I/CM process includes an initial injection phase into a partially open mold, followed by a controlled closure of the mold to complete resin infusion. The mold tooling can typically be assumed rigid, and we may consider constant speed, or constant force closing in the compression phase. During the compression phase the total force applied to the tooling is balanced by the internally generated resin pressure, and the resistance to compaction offered by the preform. For constant speed closing, a good knowledge of the compressive behavior of the preform is required for prediction of total force applied to the mold. If a constant force closing condition is specified, accurate compressive behavior is a necessity to predict the evolution of mold filling, and hence mold filling time. Several authors have presented mold filling analyses of the SCRIMP process, addressing the coupled flow and preform deformation problem [4 – 6]. If accurate predictions are necessary, a good model for reinforcement behavior is required to predict the evolution of mold cavity thickness. To date, all reported SCRIMP flow models have assumed a non-linear elastic response. It seems for SCRIMP and all other LCM processes, there is significant motivation for detailed study of viscoelastic reinforcement behavior, and subsequent development of predictive models.
3. Sample LCM filling experiments A series of LCM experiments are presented here to demonstrate the influence of viscoelastic preform deformation behavior on mold filling. A simple mold geometry has been installed within an Instron testing machine, which allows for measurement of the required mold clamping force during processing. Several RTM and I/CM filling experiments were performed, and the clamping force traces reported here. The full study has been presented elsewhere,
434
S. Bickerton et al. / Composites: Part A 34 (2003) 431–444
including measurement of internally generated fluid pressures [31]. 3.1. Experimental equipment A cast aluminium mold was used, the part geometry being a 0.15 m radius circular disc of constant thickness. The mold is depicted in Fig. 3a, and consists of two pieces. The upper platen is a simple flat disc, while the lower platen is essentially a high walled basin. A resin or test fluid can be injected at the center of the mold. The mold was mounted on an Instron 1186 testing machine, with the upper platen attached to a 200 kN load cell, as shown in Fig. 3b. Using the Instron to control the speed and position of the female half of the mold, it was possible to consistently specify the compaction phases of RTM and I/CM cycles, while logging the clamping force to a computer running LabVIEW data acquisition software. A pressure transducer and one-way check valve were included in the injection line, as shown in Fig. 3b. The check valve prevents back flow of fluid during the compression phase of I/CM filling, effectively closing the injection gate. Fluid injection was controlled via a purpose built injection device capable of maintaining a wide range of constant flow rates. This device was directly controlled by the data acquisition computer, automatically initiating injection at the completion of the initial preform compression phase.
study was mixed to a viscosity of 0.178 Pa. s, at 18 8C on a Physica Rheolab viscometer. For the RTM experiments performed, the preform was compressed from 50 mm at a constant speed of 2.0 mm/min until the final cavity thickness of 4.7 mm was reached. Resin injection was initiated at this time, a constant flow rate being maintained. A series of four tests were completed, at injection flow rates of 3.06, 5.80, 8.83, and 11.43 cm3/s. For the I/CM experiments, the preform was compressed at 2.0 mm/min, until the cavity thickness reached 6.5 mm (a volume fraction of 28%). Resin was then injected at a constant flow rate of 18.7 cm3/s. Once the required volume of fluid was injected, a second phase of constant speed compression was initiated, and stopped once the cavity thickness reached 4.7 mm. A series of four experiments were completed, at compression speeds of 2.0, 8.0, 18.0, and 28.0 mm/min. For all experiments, total clamping force was measured after completion of filling, documenting further relaxation of this force. For each experimental series, a second compression force trace is presented. Two additional preform samples have been prepared, and exposed to the same compression history, without any fluid injection. This was done to demonstrate that significant viscoelastic deformation of the reinforcement occurs without the presence of a resin, and to highlight any further effects due to the presence of the fluid. 3.3. Results
3.2. Experimental procedure A 450 g/m2 E-glass fiber, continuous filament mat (CFM) was used as the reinforcement in these experiments. Circular specimens were cut from CFM with an outer diameter of 290 mm (^ 2 mm), and a central 19 mm diameter hole. Ten specimens were stacked into each preform sample, each having an initial thickness of approximately 48 mm. All experiments were initiated at a cavity thickness of 50 mm. The intended final thickness of the part was set to 4.7 mm, producing a fiber volume fraction of 38%. A non-reactive corn syrup solution was used as the test fluid for both experiments. This material was chosen as it is easily tailored to a desired Newtonian viscosity. The solution used in this
Fig. 3. (a) Upper and lower mold platens. (b) The mold installed within the Instron testing machine.
Clamping force histories for the RTM and I/CM experiments are presented in Figs. 4 and 5, respectively. Also presented are force histories for compression of the dry preform samples. A single experiment is depicted in Figs. 4a and 5a, including compression prior to fluid injection. All experiments are displayed in Figs. 4b and 5b, focusing on the period following initiation of fluid injection. Due to small variations in compression response, each curve has been normalised to the force required at the end of the initial compaction. Fluid injection was then initiated at this point, time being set equal to zero in each experiment. For the RTM experiments, the injection pressure reached maximums of 275.0, 490.0, 765.0 and 985.0 kPa. For the I/CM experiments, pressure at the center of the mold was maximum in each case at the completion of injection, being approximately 280.0 kPa. The dry compression curve in Fig. 4 shows rapid increase as cavity thickness approaches 4.7 mm at t ¼ 0: This curve demonstrates significant load relaxation after this time, the majority occurring over the first 100 s. The RTM experiment with Q ¼ 5:80 cm3 =s is also provided on this plot. The RTM curve is identical prior to injection, but dips significantly below the dry curve over the 37.0 s injection period. Assuming that the reinforcement carries the same load when dry and wet, we had expected the RTM data to rise above the dry curve due to the force generated by internal fluid pressure. This additional force component
S. Bickerton et al. / Composites: Part A 34 (2003) 431–444
435
Fig. 4. Normalised clamping force histories for the RTM experiments, and related dry compression test. Measured clamping forces at t ¼ 0 s were 36.8, 35.0, 33.2 and 28.6 kN for the RTM experiments with increasing flow rate, and 37.5 kN for the dry compression test. (a) 2 400 , t , 400 s, (b) 0 , t , 120 s.
could be determined by integrating the fluid pressure field over the wetted surface area of the mold at any time. The actual behavior demonstrates a balance between the force due to fluid pressure, and accelerated stress relaxation within the reinforcement due to the presence of fluid. The presence of a fluid may serve to lubricate fiber – fiber connections, allowing the reinforcement to settle into a state
requiring lower compressive stress. At the completion of injection the force trace shows a sudden drop, as the force generated by internal fluid pressure is removed. Shortly afterwards, the RTM data adopts the same relaxation rate as the dry curve, the clamping force settling to a significantly lower value. The series of RTM force traces provided in Fig. 4b all exhibit similar behavior. As injection flow rate is
436
S. Bickerton et al. / Composites: Part A 34 (2003) 431–444
Fig. 5. Normalised clamping force histories for the I/CM experiments, and related dry compression test. Measured forces at t ¼ 0 s were 11.0, 10.6, 10.9 and 10.6 kN for the I/CM experiments with increasing compression speed, and 12.1 kN for the dry compression test. (a) 2100 , t , 100 s, (b) 0 , t , 100 s.
increased, the clamping force is seen to rise more rapidly during injection, increasing above the dry compression curve for the faster flow rate. In each case, the clamping force drops significantly below the dry curve once fluid injection is complete. The dry compression curve in Fig. 5a peaks initially at t ¼ 0; and then exhibits load relaxation as the cavity
thickness is held constant. The load rises rapidly during the second phase of compaction, before demonstrating significant relaxation once the final cavity thickness is reached. The I/CM experiment with compression speed of 18 mm/ min is also provided on this plot. As was found for the RTM experiments, during the injection phase, the I/CM clamping force does not exceed the dry curve. For the applied
S. Bickerton et al. / Composites: Part A 34 (2003) 431–444
437
Fig. 6. Close-up images of studied reinforcement materials, (a) CFM, (b) CSM, and (c) PW.
injection flow rate, the force generated by fluid pressure has been balanced by the fiber lubrication effect. At the conclusion of injection ðt ¼ 16:0 sÞ; the fluid force component falls away, being approximately 2.2 kN. At the initiation of the secondary compression phase, the clamping force rises at a similar rate as for the dry experiment. For these conditions the I/CM curve is maintained below the dry curve, again a balance being made between the generated fluid force and the fiber lubrication effect. At the completion of filling ðt ¼ 23:0 sÞ the fluid force component falls away, and the I/CM curve assumes the same relaxation rate as the dry data. The series of I/CM force traces provided in Fig. 5b all display a similar drop relative to the dry curve at the completion of mold filling. As compression speed is increased, the clamping force is seen to rise more rapidly during injection, and an increasing trend is noted in the maximum force obtained. This is expected due to the greater fluid pressures generated by increasing compression speeds. The experiments presented in Figs. 4 and 5 demonstrate the influential role played by the viscoelastic behavior of reinforcements. For both situations presented, viscoelasticity will influence prediction of local and total clamping forces exerted on molds. Such behavior will be even more important for modeling of the constant force version of I/CM, and other LCM processes utilising flexible molds. The presence of a fluid within the reinforcement has a strong effect on compression behavior. For the remainder of this study we have concentrated on the response of dry reinforcements.
4. Experimental characterisation program A detailed study has been undertaken into compressive deformation behavior of the CFM (450 g/m2), a chopped strand mat (CSM, 430 g/m2), and a glass plain weave (PW, 600 g/m2). Images of each reinforcement are presented in Fig. 6. The behavior of each has been considered at varying compaction levels, and at various compression speeds. 4.1. Experimental procedure Two types of experiment have been carried out with each preform style. Samples of each material have been placed between a set of rigid, parallel plates setup within our Instron testing machine. A predetermined compressive strain history can be prescribed, while monitoring the required compaction force. The distance between the plate surfaces at any time is defined as the cavity thickness, h: h0 is defined as the initial thickness of the preform sample before any load is applied, and is determined prior to an experiment. The basic experiment is described schematically in Fig. 7, being a period of constant speed compression to a predetermined cavity thickness, then a holding period at that thickness. A typical load response curve is also presented. Load data was converted to an applied stress, all preform samples being cut into 0.2 m squares. The measured load curves are presented here as either stress – time, or stress – strain
Fig. 7. (a) Schematic description of cavity thickness evolution during a compression characterisation test. (b) Example of resulting load curve.
438
S. Bickerton et al. / Composites: Part A 34 (2003) 431–444
curves. For this study the measure of strain, 1; is defined below: 1¼
h0 2 h : h0
ð1Þ
Our initial studies have exposed a significant amount of permanent deformation that occurs within these materials when repeated loading cycles are applied. Preform samples displayed reductions in the force required to achieve a particular volume fraction with every loading cycle applied to them. Under manufacturing conditions preforms are cut from rolls, and placed into a mold without any significant previous compressive deformation. The characterisation experiments described here have been performed on freshly cut preform samples, removing any influence due to previous deformation. 4.2. Varying compaction level experiments For each preform style four samples have been compacted at 2.0 mm/min, to increasing final fiber volume fractions. These tests have been completed to assess preform relaxation at various compaction levels. The applied load has been recorded during this initial compaction period, and then during a period of constant cavity thickness. Six, eight, and eight layers of the CFM, CSM, and PW have been used in each preform sample. 4.3. Varying compression speed experiments Four or five samples of each preform style have been compressed to the same final fiber volume fraction, at a range of different compression speeds (from 0.5 to
10.0 mm/min). Otherwise, these tests are completed in an identical manner to those described above, using the same number of layers for each preform style. These tests have been completed to assess the effect of varying strain rate on compression response. 4.4. Continuous filament mat response The stress – time behavior recorded during the compaction level tests is presented in Fig. 8. During the constant speed phase of these experiments, each successive curve falls on the previous, such repeatability providing some assurance that the material is behaving in a possibly predictable manner. Measured peak stress was found to increase with increasing final volume fraction, which was varied from 29.6 to 51.9%. The relaxations were 40.3, 35.2, 27.4, and 19.5% of the peak stresses, from lowest to highest final volume fraction. These results are also presented as stress –strain data in Fig. 9. An additional curve has been plotted on this figure, representing the possible long term behavior after complete stress relaxation. By considering the stress reached in the material after long times at constant cavity thickness, it is possible to define the minimum compaction stress required as a function of strain. This stress –strain curve might be applicable for manufacturing processes occurring over long time scales, an elastic model being acceptable in these situations. The stress – time behavior recorded during the compression speed tests is presented in Fig. 10. In each case a final volume fraction of 41.5% was reached. Larger peak stresses are displayed with increasing speed, demonstrating significant rate dependent behavior. Once the peak stress has been reached, stress relaxation occurs to the same level,
Fig. 8. Stress versus time behavior recorded during compaction level tests for CFM.
S. Bickerton et al. / Composites: Part A 34 (2003) 431–444
439
Fig. 9. Stress versus strain behavior recorded during compaction level tests for CFM. The dotted line has been added to demonstrate possible long term behavior of the material.
independent of the initial compression speed. The same results are presented as stress –strain data in Fig. 11. In this format the rate dependent effect is clearly demonstrated, the stress – strain curve followed during compression being influenced by the rate at which the reinforcement is deformed. 4.5. Chopped strand mat response The recorded stress – time behavior for the CSM compaction level tests is presented in Fig. 12. The maximum
fiber volume fraction has been varied between 28.0 and 56.0%. Similar behavior has been observed as for CFM, with good repeatability demonstrated during the initial loading section. Again, decreasing amounts of relaxation were noticed with increasing final volume fraction, being 63.8, 60.6, 33.3, and 13.4%. Data collected for CSM during the compression speed tests is presented in Fig. 13. The final volume fraction for each test was 42.0%. This material also shows clear strainrate dependency in the peak stress. However, unlike CFM, the measured stress did not relax to the same level, though
Fig. 10. Stress versus time behavior recorded during compression speed tests for CFM. The final volume fraction achieved in each test was 41.5%.
440
S. Bickerton et al. / Composites: Part A 34 (2003) 431–444
Fig. 11. Stress versus strain behavior recorded during compression speed tests for CFM.
the final fiber volume fraction was identical in each case. It appears that this material shows a greater potential for stress relaxation the faster it is initially deformed. While in this study we have focused on the macroscopic behavior, consideration must be given to the microscopic architecture if this fundamental difference in the response of two glass random mats is to be explained. Significant differences in structure include fiber length (CFM being continuous, CSM formed from chopped strands 30 – 40 mm in length), and binder loading. The CSM includes a much larger amount of binder, essentially bonding the chopped strands together as a mat.
4.6. Plain weave response Data recorded during the compaction level tests is presented in Fig. 14, maximum volume fraction being varied from 55.0 to 65.0%. Qualitatively similar behavior is noted to the other materials studied, while the PW shows an increased potential for stress relaxation. The relaxations were 57.6, 49.2, 45.0, and 39.1% of the peak stresses, from lowest to highest final volume fraction. It is noticeable that the majority of stress relaxation occurs over a shorter time period than for CFM and CSM. Increased variation was found during the loading period, demonstrating
Fig. 12. Stress versus time behavior recorded during compaction level tests for CSM.
S. Bickerton et al. / Composites: Part A 34 (2003) 431–444
441
Fig. 13. Stress versus time behavior recorded during compression speed tests for CSM. The final volume fraction achieved in each test was 42.0%.
the influence of stacking variations, or nesting of layers. In particular the data for Vf ¼ 57:5% shows an increased rate of increase in measured stress. The stress versus time behavior recorded during the compression speed tests is presented in Fig. 15a, the final volume fraction being 52.5% in each case. Three of the curves display an increasing peak stress with increasing compression speed, behavior demonstrated by the other materials studied. The slowest compression speed breaks this trend, demonstrating the possible influence of stacking
and nesting with woven materials. This may also be caused by variations in reinforcement areal weight, though this is thought to be unlikely, as similar behavior should be expected from the two random mats studied. Similar to CSM, the PW shows an increased potential for stress relaxation with faster compression speeds. To further investigate the possible influence of layer nesting, a second set of tests was completed, the final volume fraction being set to 62.5%. The generated stress data is plotted in Fig. 15b, demonstrating stronger influence due to nesting at this
Fig. 14. Stress versus time behavior recorded during compaction level tests for PW.
442
S. Bickerton et al. / Composites: Part A 34 (2003) 431–444
Fig. 15. Stress versus time behavior recorded during compression speed tests for PW. The final volume fraction achieved in each test was (a) 52.5%, and (b) 62.5%.
higher volume fraction. From this data it is more difficult to identify any clear trends. It should be noted that the plane weave under consideration is a heavy fabric, requiring relatively few layers to generate the necessary part
thickness. Under such conditions, misalignment of 1 or 2 layers can have significant effect. A more in depth study is required to identify the magnitude of scatter due to nesting, considering a variety of woven and stitched fabrics.
S. Bickerton et al. / Composites: Part A 34 (2003) 431–444
5. Conclusion The apparent time dependent viscoelastic deformation of LCM reinforcements must be an important consideration for the simulation of these mold filling processes. Two LCM filling experiments were described, demonstrating the strong influence viscoelastic reinforcement deformation has on required mold clamping force. Models for deformation behavior are required to accurately predict generated clamping forces and total pressure distributions within a mold, and for some LCM variations, are necessary to accurately model filling. The initial characterisation study presented here has highlighted a variety of complex phenomenon displayed by three common LCM reinforcements. Each material showed significant stress relaxation after compression to a constant volume fraction. It is important to note that stress relaxation occurs over time periods on the same order as the full molding processes. The magnitude of stress relaxation is strongly dependent on the level of compaction, the amount of which is reduced with increased final fiber volume fraction for all three materials studied. The response of all three materials was dependent on the speed of compression, or strain rate. Increasing the strain rate produced greater peak stresses though samples were compressed to the same fiber volume fraction. The CFM subsequently relaxed to the same final stress independent of compaction speed, being the simplest behavior noted. CSM showed a clear increase in stress relaxation with increasing closing speed. This effect is possibly due to the presence of significant binder in the structure, providing a tougher challenge than the CFM to model. It was more difficult to identify trends in the behavior of the PW due to the influence of layer nesting. This phenomenon may affect the deformation behavior of other woven and stitched fabrics, the extent of which requires further investigation. The data presented in this paper will be used to develop empirical models for viscoelastic response, which will be necessary to address processes such as I/CM and SCRIMP. While this study has highlighted complex behavior, definite trends have been identified, and the authors are optimistic that useful deformation models can be produced.
Acknowledgements The authors would like to acknowledge the generous support of the Foundation for Research Science and Technology, New Zealand.
References [1] Bruschke MV, Advani SG. A finite element/control volume approach to mold filling in a anisotropic porous media. Polym Compos 1990; 11(6):398–405.
443
[2] Young WB, Rupel K, Han K, Lee LJ, Liou MJ. Analysis of resin injection molding in molds with preplaced fiber mats II: numerical simulation and experiments of mold filling. Polym Compos 1991; 12(1):20– 9. [3] Trochu F, Gauvin R, Gao D-M. Numerical analysis of the resin transfer molding process by the finite element method. Adv Polym Technol 1993;12(4):329–42. [4] Sun X, Li S, Lee LJ. Mold filling analysis in vacuum-assisted resin transfer molding. Part I: SCRIMP based on a high-permeable medium. Polym Compos 1998;19(6):807–17. [5] Han K, Jiang S, Zhang C, Wang B. Flow modeling and simulation of SCRIMP for composites manufacturing. Composites, Part A 2000; 31(1):79– 86. [6] Kang M-K, Lee W-I, Hahn H-T. Analysis of vacuum bag resin transfer molding process. Composites, Part A 2001;32(11):1553– 60. [7] Han K, Ni J, Toth J, Lee LJ. Analysis of an injection/compression liquid composite molding process. Polym Compos 1998;19(4):487–96. [8] Pham X-T, Trochu F, Gauvin R. Simulation of compression resin transfer molding with displacement control. J Reinforced Plast Compos 1998;17(17):1525–56. [9] Kang M-K, Lee W-I. Analysis of resin transfer/compression molding process. Polym Compos 1999;20(2):293–304. [10] Pham X-T, Trochu F. Simulation of compression resin transfer molding to manufacture thin composite shells. Polym Compos 1999; 20(3):436– 59. [11] Pillai KM, Tucker II CL, Phelan Jr FR. Numerical simulation of injection/compression liquid composite molding. Part 2. Preform compression. Composites, Part A 2001;32(2):207 –320. [12] Gutowski TG, Cai Z, Bauer S, Boucher D, Kingery J, Wineman S. Consolidation experiments for laminate composites. J Compos Mater 1987;21(7):650–69. [13] Dave R, Kardos JL, Dudukovic´ MP. A model for resin flow during composites processing: Part 1—general mathematical development. Polym Compos 1987;8(1):29–38. [14] van Wyk CM. Note on the compressibility of wool. J Textile Inst 1946;37:T285–92. [15] Komori T, Makishima K. Numbers of fiber-to-fiber contacts in general fiber assemblies. Textile Res J 1977;57:T15–T23. [16] Carnaby GA, Pan N. Theory of the compression hysteresis of fibrous assemblies. Textile Res J 1989;59(5):275–84. [17] Curiskis JI, Carnaby GA. Continuum mechanics of the fiber bundle. Textile Res J 1985;16(4):58–64. [18] Gutowski TG. A resin flow/fiber deformation model for composites. SAMPE Q 1985;16(4):58–64. [19] Cai Z, Gutowski TG. 3-D deformation of a lubricated fiber bundle. J Compos Mater 1992;26(8):1207–37. [20] Sˇima´cˇek P, Karbhari VM. Notes on the modeling of preform compaction: I micromechanics at the fiber bundle level. J Reinforced Plast Compos 1996;15(1):86–122. [21] Karbhari VM, Sˇima´cˇek P. Notes on the modeling of preform compaction: II effect of sizing on bundle level micromechanics. J Reinforced Plast Compos 1996;15(8):837–61. [22] Lomov SV, Verpoest I. Compression of woven reinforcements: a mathematical model. J Reinforced Plast Compos 2000;19(16):1329–50. [23] Chen B, Chou T-W. Compaction of woven-fabric preforms in liquid composite molding processes: single-layer deformation. Compos Sci Technol 1999;59:1519– 26. [24] Chen B, Leng AH-D, Chou T-W. A nonlinear compaction model for fibrous preforms. Composites, Part A 2001;32(5):701–7. [25] Robitaille F, Gauvin R. Compaction of textile reinforcements for composites manufacturing. I: review of experimental results. Polym Compos 1998;19(2):198 –216. [26] Kim YR, McCarthy SP, Fanucci JP. Compressibility and relaxation of reinforcements during composite processing. Polym Compos 1991; 12(1):13– 19. [27] Pearce N, Summerscales J. The compressibility of a reinforcement fabric. Compos Manufact 1995;6(1):15–21.
444
S. Bickerton et al. / Composites: Part A 34 (2003) 431–444
[28] Saunders RA, Lekakou C, Bader MG. Compression and microstructure of fibre plain woven cloths in the processing of polymer composites. Composites, Part A 1998;29(4):443–54. [29] Robitaille F, Gauvin R. Compaction of textile reinforcements for composites manufacturing. II: compaction and relaxation of dry and H2O-saturatedwovenreinforcements.PolymCompos1998;19(5):543–57.
[30] Saunders RA, Lekakou C, Bader MG. Compression in the processing of polymer composites. 2. Modeling of the viscoelastic compression of resin-impregnated fibre networks. Compos Sci Technol 1999;59(10):1483–94. [31] Buntain MJ, Bickerton S. Prediction of tooling forces in liquid composite molding processes. SAMPE J 2003;39(1):8– 15.