Large thermoplastic composite wind turbine blades

Transcription

1 Large thermoplastic composite wind turbine blades J. Teuwen (TU-Delft) project )

2

3 REPORT Large thermoplastic composite wind turbine blades Process optimization for vacuum infusion of large blades Julie Teuwen December 2009

4 Outline List of Figures... iii List of tables...v 1 Introduction Thin laminates: process optimization Introduction Experimental Pot life Infusion strategy Upscaling Cure Post processing of APA-6 composites Conclusions Thick laminates: Heating methods Introduction Experimental Heat distribution Comparison of heating methods Conclusions Reduction of cure shrinkage induced voids Introduction Experimental Resin constitution Pure resin Laminates Conclusions Optimisation of fibre-to-matrix bond Introduction Experimental Effect of fibre sizing Characterisation of glass rovings Vacuum infusion set-up for UD laminates Characterisation of the UD-laminates Conclusions Conclusions and recommendations Conclusions Recommendations References ii

5 List of Figures Figure 1. Anionic polymerisation of caprolactam into polyamide Figure 2. Aminosilane sized E-glass fibres... 3 Figure 3. Vacuum infusion set-up (from left to right: Mini Mixing Unit, buffer vessel, laminate and heating device, resin trap and vacuum pump) Figure 4. Infrared panel as heating device for the laminate Figure 5. Viscosity in time of a reactive mixture with an initial buffer vessel residence temperature (110 or 130 C) and time (0, 5 or 13.5 min)... 6 Figure 6. Degree of crystallinity of the inlet and the outlet of a vacuum-infused APA-6 composite 7 Figure 7. Temperature logging of the infusion (t=0sec) and reaction of a laminate at 160 C at three locations: resin inlet, middle and resin outlet... 7 Figure 8. De-blocking of carbamoylcaprolactam (step 1) and the subsequent formation of fibre-tomatrix bond (step 2)... 8 Figure 9. C-Scan images of APA-6 laminates at different infusion temperatures ( C) with the inlet at the bottom side of the picture... 9 Figure 10. Degree of conversion at different infusion temperatures Figure 11. Average short-beam strength values at different infusion temperatures Figure 12. The effect of the degassing procedure on the inter laminar shear strength (ILSS) of the composites Figure 13. Flow front position during infusion and their corresponding flow rate at different inlet tube cross-sections and infusion pressures Figure 14. C-scan results for different infusion pressures - post-fill pressures respectively, with an inlet tube cross-section of 20mm Figure 15. Degree of conversion near the inlet and outlet of the laminates at different inlet tube cross-sections [mm 2 ] - infusion pressures [mbar] - cure pressures [mbar] Figure 16. C-scan results for infusion temperature of 110 C when inlet is not clamped during heating (test 2) Figure 17. C-scan results for infusion temperature of 130 C when inlet is not clamped during heating (test 3) Figure 18. C-scan results for inlet clamping after infusion (test 4) Figure 19. Evolution of degree of crystallinity of pure resin in time Figure 20. Evolution of degree of conversion of pure resin in time Figure 21. Degree of crystallinity for different cure times Figure 22. Short-beam strength for different cure times Figure 23. Change in crystallinity at different annealing temperatures (averaging the results for different annealing times and cooling rates at the different annealing temperatures) Figure 24. Change in crystallinity at different annealing times (averaging the results for different annealing temperatures and cooling rates at the different annealing times) Figure 25. Cooling rates at different annealing temperatures Figure 26. Change in crystallinity for different cooling rates Figure 27. Short-beam strength for different annealing temperatures Figure 28. Short-beam strength for different annealing time s Figure 29. Vacuum infusion bag for thick laminates with the coarse fabric weave Figure 30. (Kapton ) Thermofoil Heater Figure 31. Position of the resistive mesh in the middle of the fabric lay-up and its connection to the power supply Figure 32. Connection of the carbon fabric via two copper plate to the copper wire of the power supply Figure 33. Position of the thermocouples through the thickness of the lay-up Figure 34.Glass beaker with cover lid and supporting plate Figure 35.Viscosity measurements at different mould temperatures Figure 36. Adiabatic temperature measurements with an initial pot life of 5 minutes at 115 C, cured at different temperatures Figure 37. Heating profiles through the thickness of the laminate during heating from 110 to 180 C with the heated platen iii

6 Figure 38. Heating profiles of different mould plates using the infrared panels as heating device30 Figure 39. Heating profiles through the thickness of the laminate during heating from 110 to 180 C with the infrared panels Figure 40. Heating profiles through the thickness of the laminate during heating from 110 to 180 C with a thermofoil on the mould side and an infrared panel at the top Figure 41. Comparison of heating profiles of laminates heated by double infrared, or by thermofoil in combination with the top infrared panel Figure 42. Cross-section of a cured 24 mm thick laminate heated by thermofoil in combination with the top infrared panel Figure 43. Heating profiles through the thickness of the laminate during heating from 110 to 180 C with a resistive mesh in the middle of the fibre lay-up and the infrared heating as surface heating Figure 44. Cross-section of laminates heated by a metal mesh or a carbon fabric in combination with the intrared panels Figure 45. Heating profiles through the thickness of the laminate during heating from 110 to 180 C with a resistive carbon fabric in the middle of the fibre lay-up and the infrared heating as surface heating Figure 46. Heating profiles of the heat input layers in the carbon fabric experiment Figure 47. Heating distribution through the thickness in the carbon fabric experiment Figure 48. Heating profiles through the thickness of the laminate during heating from 110 to 180 C with a resistive carbon fabric in the middle of the fibre lay-up and the infrared heating as surface heating Figure 49. Heat distribution for different heating methods, before start of the chemical reaction. 35 Figure 50. Reactivity of the resin with different amounts of rubber. T1 stands for the infusion window, T2 for the start of the polymerisation and T3 for the start of the crystallisation. The system has been compared to the reaction times of the normal system Figure 51. Viscosity versus time for different material systems: C-M and R-M at different testing temperatures: and 180 C Figure 52. Viscosity versus time for different material systems: R-M at7% and 20%wt and C-M. at a testing temperature of 160 C Figure 53. Degree of crystallinity for the normal C-M system and for different amounts of rubber in the R-M system Figure 54. Enthalpy of cold crystallisation for the normal C-M system and for different amounts of rubber in the R-M system Figure 55. Peak temperature of cold crystallisation for the normal C-M system and for different amounts of rubber in the R-M system Figure 56. E-modulus for the normal C-M system and for different amounts of rubber in the R-M system Figure 57. Dry impact resistance for the normal C-M system and for different amounts of rubber in the R-M system Figure 58. Dry and conditioned impact resistance for the normal C-M system and for different amounts of rubber in the R-M system Figure 59. The impact resistance for specimens cured at 30 or 45 min, in dry or conditioned state, for 7 and 10% of rubber in the R-M system Figure 60. The degree of conversion for laminates with 15%wt rubber at different processing temperatures Figure 61. The degree of conversion for laminates manufactured at 180 C, with different amounts of rubber Figure 62. The degree of crystallinity for laminates with 15%wt rubber Figure 63. The degree of crystallinity for laminates manufactured at 180 C, with different amounts of rubber Figure 64. Effect of the fibre sizing of the short-beam strength at different processing temperatures Figure 65. Comparison of the tensile strength of different fibre types Figure 66. Al supporting plate (for both winding and infusion) Figure 67. Dry filament wound fibres around Al supporting plate iv

7 Figure 68. Tacky-tape applied and in- and outlet connected to the mould set-up Figure 69. Al covering plates placed on the fibres for better surface quality Figure 70. Fully vacuum bagged mould ready for infusion Figure 71. A fully cured UD laminate Figure 72. Average maximum transverse tensile strength of UD laminates for different infusion parameters Figure 73. Average maximum transverse tensile strength of.the upper and lower part of the obtained UD laminate after infusion List of tables Table 1. Test matrix for the effect of the residence time and temperature on the viscosity of the reactive resin in a buffer vessel... 6 Table 2. Physical properties at different infusion temperatures Table 3. The effect of the degassing procedure on the composite properties Table 4. Thickness (t) and inter laminar shear strength (ILSS) near the inlet and outlet of the composites, and the infusion times at different infusion and post-fill pressures Table 5. Infusion settings for the infusion of long laminates Table 6. Infusion times for the infusion of long laminates Table 7. Comparison of heating rate of different heating methods Table 8. Infusion window for different material systems: C-M and R-M at different amount of R at different testing temperatures Table 9. Comparison of tensile properties of glass fibre rovings Table 10. Infusion parameters for the UD infusion process, for different type of tests v

8 1 Introduction Nowadays, the development of production processes for large composite structures is of great importance, especially for wind energy companies and their production of wind turbine blades with lengths exceeding 60 meters. The life expectancy of 20 years of the wind turbine blades, combined with the fact that wind energy is one of the fastest growing and most promising energy source on the renewable energy market, also leads to the necessity of the development of a recyclable material for the wind turbine blades. Since thermoplastic composites can be fully recyclable and may lead to lower production times compared to their thermoset composite counterparts, increased interest is shown in the development of these composites. Other advantages of thermoplastic composites are the easy assembly of parts through welding and their high toughness. Thermoplastic composite materials are traditionally manufactured through melt processing using a hot press or an autoclave. The melt processing technique however limits the size and thickness of the obtainable parts. From the requirement of making large blades, the use of a different process is needed. For thermoset large size wind turbine blades, a vacuum infusion process is currently used. The vacuum infusion process uses a pressure difference between resin inlet and resin outlet, as a driving force for infusing the resin into the dry fibre preform. The vacuum infusion process has an atmospheric inlet pressure (p 1 bar) and an outlet pressure lower than 1 bar. Because the vacuum infusion process is a closed process (on one side a mould is present and the other one a vacuum bag is used to seal the whole set-up), there are low volatile emissions. The low weight combined with the low volatile emissions and low production costs compared to other fabrications processes, makes the vacuum infusion process a valuable alternative fabrication process technique for manufacturing of large composite structures [1]. However, several limitations for liquid moulding of thermoplastics can be observed: (i) a high resin viscosity which impairs the wet-out of the reinforcement; (ii) the high temperature processing requirement and (iii) a small processing window [2, 3]. The vacuum infusion process can only be used with thermoplastic resins, when their poor impregnation characteristics can be overcome by lowering the viscosity of the resin. The viscosity of thermoplastic resins is typically Pa.s whereas infusion can only take place at a viscosity of 1 Pa.s or less. A trade-off between the possible thermoplastic resins based on availability, costs and suitability for the production process, resulted in the selection of anionic polyamide-6 (APA-6) as the most suitable candidate, as shown in the work of K. van Rijswijk [4,p.35]. Due to the low viscosity at the beginning of the reaction, followed by a fast reaction rate and the high conversion that can be obtained, anionic polyamide-6 is very suitable for the vacuum infusion process. This report is based on the research work that continued after van Rijswijk s thesis on the PhD@Sea project. O C N - + O C N O C * NH O C O C N N - NH C * O O C O C N N - NH C * + O O C NH O C O C N NH C O NH * + O C N - O C N O C NH NH C * O n Figure 1. Anionic polymerisation of caprolactam into polyamide-6 The reactive processing used combines all the manufacturing stages into one single production step: after vacuum infusion of the reactive mixture into the fibre pre-form, a delayed in-situ polymerisation of the monomer to the high molecular weight thermoplastic resin takes place. The high molecular weight APA-6 is obtained through ring-opening polymerisation, as shown in Figure 1. When demoulding the composite, one hour after infusion, the cured, fully recyclable thermoplastic composite part is obtained. 1

9 In the above-mentioned previous work, the resin constitution needed for the production of glass fibre reinforced APA-6 composites was established and the most important parameters in the process were identified. It was shown that APA-6 composites have the potential to be used for wind turbine applications. The requirements, which can be set for the vacuum infusion of large, thermoplastic wind turbine blades, are: The manufacturing process must be able to produce homogeneous large composite parts. The composite needs to be uniform in mechanical and resin properties, both in the flow and through-the-thickness directions. The manufacturing process needs to be suitable for production of complex parts, with complex shapes and thickness variations. The two previous requirements have to be met in combination with good mechanical properties (static and dynamic) and optimum durability. The previous research also posed a number of questions to be solved before these requirements can be met with the APA-6 system: Fluctuations in the composite properties were observed, both in the flow and through-thethickness directions, and also between laminates manufactured following the same infusion strategy. A different infusion strategy and different heating methods will be tested to overcome these problems. For the optimization of the dynamic properties and the durability of the composite, the fibre-to-matrix adhesion was identified as the most critical parameter. The effect of the process parameters, as well as the effect of different fibre sizings, on the optimization of the fibre-to-matrix bond needs to be investigated. The research carried out up till now was focused on small laminates (with a length of 25 cm). More research is needed on bigger panels to investigate the effect of the flow length on the properties of the laminates Voids generated during the infusion and curing process have a detrimental effect on the mechanical properties. Two different lines of research will be described in this report to reduce the void content in the APA-6 composites: a further investigation on the infusion strategy as well as a reduction in cure shrinkage by altering the resin constitution. Chapter 2 of this report presents the optimization of the process for thin laminates. This optimization is based on a number of factors: temperature, pot life, infusion strategy, up-scaling, curing and post-processing. In Chapter 3, different heating methods and their potential for heating thick-walled large composites are discussed. A different resin constitution is investigated to minimize the voids and shrinkage of the composites and the results shown in Chapter 4. Finally, the optimization of the fibre-to-matrix bond by testing different fibre sizings is presented in Chapter 5. In Chapter 6, the conclusions and recommendations will be given. 2

10 2 Thin laminates: process optimization 2.1 Introduction The previous work of van Rijswijk et al. [4] showed that the processing temperature is the most important parameter in the infusion of APA-6 composites. The processing temperature results from the sum of the external heat (heating device) and the internal heat (exothermic reaction of the resin). As the effect of the temperature is thought to be significant for every stage in the infusion process, all the steps will be investigated separately. Likewise, other relevant parameters within every stage will be evaluated resulting in the following scheme: 1. Investigation of the effect of the pot life (time and temperature) on the reaction/reactivity of the resin and the increment of its viscosity 2. Investigation of the impact of different infusion strategies, characterized by time, temperature, pressure and velocity during infusion, on the homogeneity of the composite 3. Investigation of increased flow lengths 4. Investigation of the effect of post-curing or post-processing (different times and temperatures) 2.2 Experimental Materials Resin material For the vacuum infusion of the anionic polyamide-6 thermoplastic laminates, a reactive resin mixture is used, which consists of three components: monomer, activator and initiator. The anionic polymerisation grade ε-caprolactam monomer (APCaprolactam) was supplied by Brüggemann Chemicals, Germany. As the monomer is extremely sensitive to moisture, it was stored at room temperature in PE sealed tanks with bags of silica gel.. The combination of the activator and the initiator and the specific doses relative to the monomer content were researched and determined by van Rijswijk et al. [5]. The reaction of this activator/initiator combination is slow initially, which provides a low viscous resin mixture at the beginning of the reaction and guarantees sufficient time for the full wet-out of the fibre preform. Two different activators were used for the studies in this section, activator C and activator T. They were both used in a 1.2% mol concentration in combination with 1.2% mol of the initiator M. Both the activator and the initiator were supplied by Brüggemann Chemicals, Germany and were stored under a nitrogen atmosphere in sealed aluminium lined polyethylene bags. Glass fibres E-glass fibre mats were used to reinforce the APA-6 composites. The 8-harness satin weave E- glass fabrics (SS , weave style 7781, 300 gram/m 2 ) were supplied by Ten Cate Advanced Composites bv., Nijverdal, The Netherlands. The chemical structure of the aminosilane sized glass fibres is presented in Figure 2. Figure 2. Aminosilane sized E-glass fibres Processing methods Preparation of the resin mixture A special designed lab-scale mixing unit (Mini Mixing Unit 'MMU-TUDelft', Bronk Industrial b.v., The Netherlands), see Figure 3, is used to prepare, under a nitrogen atmosphere and at a temperature of 110 C, the resin mixture for infusion. After dispensing the volume needed for the infusion of a laminate into a buffer vessel, the resin is degassed for 5 minutes to remove any N 2 3

11 dissolved in the mixture during its storage in the MMU tanks. This buffer vessel is inserted in a specially designed aluminium cylinder, which is placed on a heated plate to keep the vessel and resin mixture at a temperature of 110 C during infusion. For the research described in sections 2.3 and 2.4.2, the buffer vessel was also set at a temperature of 130 C. Figure 3. Vacuum infusion set-up (from left to right: Mini Mixing Unit, buffer vessel, laminate and heating device, resin trap and vacuum pump). Figure 4. Infrared panel as heating device for the laminate. Production of test tubes Glass test tubes with a diameter of 20 mm and a length of 150 mm (Schott-Fiolax) were used to manufacture pure resin samples. The test tubes were preheated and flushed with nitrogen before being filled with 50 ml of resin. They were then sealed to prevent moisture uptake and placed in an insulated oil bath at 180 C. After the cure cycle, they were quenched in water at room temperature. Infusion strategy 1 A balanced, symmetric 12-ply laminate (25x29cm 2 ) is prepared for vacuum infusion and is placed in the heating device for the production of the laminates, either in between a hot platen press or two infrared panels. The dry fibre preform is then heated to the chosen processing temperature, which is also the cure temperature; a temperature between 150 C and 190 C. After degassing, the resin is allowed to flow through the inlet injection line into the laminate. The resin injection and fibre compaction are achieved under a vacuum of 250 mbar. When the resin reaches the outlet, the resin inlet tube is closed and the part is ready to be cured at the chosen cure temperature. The vacuum is maintained during the cure cycle of one hour. Infusion strategy 2 A balanced, symmetric 12-ply laminate (25x29cm 2 ) is prepared for vacuum infusion and is placed between two infrared panels, which is the heating device used, as it can be seen in Figure 4. Prior to infusion, the dry fibre preform is heated to the same temperature as the buffer vessel and the MMU, ensuring an isothermal infusion. After degassing, the resin is allowed to flow through the inlet injection line into the laminate. The resin injection and fibre compaction are achieved under a vacuum of 3 or 250 mbar. When the resin reaches the line-injection outlet, the resin inlet tube is closed and the infused laminate is heated to the required processing temperature of 180 C. The vacuum is maintained during the cure cyc le of one hour Analysis methods Viscosity measurements The viscosity of the resin was measured by means of a concentric cylinder viscometer Bohlin V88 (speed 6, spindle type 6). A special mould was designed to measure the viscosity versus time of the reactive resin. 4

12 Degree of conversion To determine the degree of conversion (DOC) of the polymer, samples of the composite plate were taken at 5 cm from the resin inlet and outlet. After drying the samples at 50 C in a vacuum oven for at least 48 hours, they were weighed (m tot ) and refluxed overnight in demineralised water. After this process, the residue was dried and weighed again. As the monomer dissolves in the water, the mass loss can be attributed to the monomer (m mon ). Additionally, the weight of the fibres (m fib ) was determined by burning off these residues in a carbolite oven at 565 C for one hour; according to ASTM D The degree of conversion can then be calculated as follows: mmon DOC = 1 100% m tot m (1) fib It should be taken into account that low molecular oligomers susceptible to be formed during polymerisation also dissolve during the reflux process [6]. As these oligomers, as well as the monomer, do not have any load carrying capabilities, Eq. 1 quantifies the amount of non-load carrying substances in the composite but does not provide the exact degree of conversion. Degree of crystallinity The degree of crystallinity (X c ) and the melting temperature (T m ) of the polymer were determined by means of a Perkin Elmer Differential Scanning Calorimeter (DSC). Disc-shaped samples of approximately 5 mg (m sp ) were taken from the laminate at the same location as the degree of conversion samples and were dried at 50 C in a vacuum oven for at least 48 hours. In the DSC, the test sample was held at 25 C for two minutes an d then heated up to 240 C at a rate of 10 C per minute. The results had to be corrected in account of the fibre content of the sample (m fib ), that was determined by burning off the resin after the DSC test. The degree of crystallinity was calculated as follows: H m m sp 1 Xc = 100% H m m DOC (2) 100 sp fib 100 In which H m is the melting enthalpy of the specimen [J/g] and H 100 is the melting enthalpy of a fully crystalline polyamide-6 ( H 100 = 190 J/g) [7]. The third term in Eq. 2 is a correction factor for the degree of conversion of the sample. Ultrasonic inspection To check the homogeneity of the laminate, a high frequency (10 MHz) ultrasonic C-scan was used. A single through-transmission mode was used to analyse differences in quality within the material, with water as the contact medium. Scans were performed at a speed of 352 mm/s to obtain images with a grid length of 0.5 mm and a width of 1 mm that were analyzed using ALIS software. Short-beam strength The short-beam strength tests were conducted on a Zwick Roell 20 KN testing machine according to the ASTM D2344 standard. The test specimens for the short-beam strength test were rectangular shaped and with the following nominal dimensions: 16.2 x 5.4 x 2.7 mm 3. From the maximum force (P max ), the interlaminar shear strength (ILSS) was be calculated as follows: Pmax ILSS = 0.75 (3) b t In which b and t are the width and thickness of the specimen, respectively. A minimum of 5 specimens per condition was tested. According to the reference standard, the short-beam strength test is suited for comparative testing of composite materials, provided that failures occur consistently, i.e. in the same mode. Temperature measurements Both the polymerisation and crystallization process involve an exothermal reaction [8]. To observe the temperature behaviour of the laminate during infusion and cure, a temperature 5

13 logging test was carried out. Three type K thermocouples were positioned at several locations between the layers 6 and 7 of the glass fabric laminate. 2.3 Pot life The effect of the residence time and the temperature of the reactive resin in the buffer vessel (before infusion) on the resin properties and the resin viscosity were investigated. The viscosity measurements were done according to the test method described above. The mould used for the viscosity measurements was set at a curing temperature of 180 C. The temperature of the buffer vessel and the residence time of the reactive resin inside were varied within the test according to the test matrix in Table 1. The viscosity versus time of a mixture poured into the viscosity mould immediately after mixing was used as a baseline for this experiment. The time limit of 13.5 min was set by the requirement that the resin needs to be pourable into the viscosity mould after the preset residence time in the buffer vessel. At the critical buffer vessel temperature of 130 C, this requirement was met at approximately 13.5 min. Table 1. Test matrix for the effect of the residence time and temperature on the viscosity of the reactive resin in a buffer vessel Temperature of buffer vessel [ C]/ 0 5 min 13.5 min Residence time [min] min 110 C X X 130 C X X No buffer vessel (baseline) X From the test results, it could be concluded that the longer the residence time, the higher the initial viscosity, especially significant for higher temperatures in the buffer vessel as it can be seen in Figure 5. For the long residence time and the buffer vessel temperature of 130 C it was seen that the initial viscosity does not drop to the baseline viscosity after heating up the resin to the mould temperature, showing that a reaction already took place during the residence time in this vessel. For this testing condition, it was also seen that the sharp increase in viscosity until 1 Pa.s (which is the viscosity limit for vacuum infusion of composites) occurs earlier than in the baseline, thus reducing the infusion time. Viscosity [Pa.s] 1,1 1 0,9 0,8 0,7 0,6 0,5 0,4 0,3 0,2 0,1 0 * , Time [min] Figure 5. Viscosity in time of a reactive mixture with an initial buffer vessel residence temperature (110 or 130 C) and time (0, 5 or 13.5 min) Therefore, it can be concluded that the longer the residence time and the higher the initial buffer vessel temperature, the higher the initial viscosity and the faster the reaction. Since the infusion of large products, like large wind turbine blades, requires long residence times, a deeper insight in the effect of this critical parameter is needed to find optimal infusion strategies that do not compromise the resin properties. It is believed that a deeper knowledge on the reaction kinetics of the resin will help understanding the reaction and the effect of the residence time and the temperature. With this purpose, initial tests were performed and the definition of the semiempirical equations for the simulation of the resin reaction has been recently started. 6

14 2.4 Infusion strategy Some research was focused on the infusion strategy pursuing ways to increase the homogeneity of the composites. Firstly, different infusion strategies were investigated (see 2.2.2), as in the work of van Rijswijk et al. And secondly, the research was focused on the effect of the different infusion parameters (temperature, time, pressure) on the laminate properties and its homogeneity Optimization of infusion temperature for APA-6 composites While manufacturing glass fibre reinforced anionic polyamide-6 composites, it was observed that the usual high temperature vacuum infusion process (infusion strategy 1, in section 2.2.2) induces thermal and chemical gradients in the flow direction. A change in the production process was tested to try and solve this problem: both the resin and the dry fibre pre-form were kept at the same temperature during infusion, after which they were heated to the required curing temperature. In this investigation, initiator M and activator T were used. The effect of the infusion temperature was investigated and it was ascertained that at all temperatures a certain gradient was present, although it was less pronounced with a lower injection temperature. However, the mechanical properties did not necessarily give the best results at these low infusion temperatures due to an increased heating time to the required curing temperature. It is thought that the heating rate is of high importance for the polymerisation reaction. If the heating rate is too low, the polymerisation reaction will start at a low temperature, while the crystallisation reaction occurs simultaneously [4]. Therefore the reactive groups will get trapped inside a crystal and the polymerisation reaction will be hindered. Therefore, a trade-off had to be made between the reduction in gradient at low infusion temperatures and the heating time to the cure temperature. Temperature gradient As mentioned before, the current infusion strategy for the production of the laminates induces a gradient of properties in the flow direction. During the impregnation of the laminate, the resin from the buffer vessel is at a temperature of 110 C, whe reas the fibre bed itself is at the required processing/curing temperature, which ranges from C, see infusion strategy 1, section When infused, the resin at the outlet is at a higher temperature than at the inlet, since it must pass through the fibre bed, which is at the processing temperature. Due to this temperature gradient in the flow direction, a difference in properties between the inlet and the outlet is expected. To get an overview of these differences, the degree of conversion and crystallinity were measured for several processing temperatures. Likewise, the temperature was measured by placing themocouples at several locations within the laminate. Degree of crystallinity [%] APA-6 composite (sized/inlet) 25 APA-6 composite (sized/outlet) Mould temperature [ C] Figure 6. Degree of crystallinity of the inlet and the outlet of a vacuum-infused APA-6 composite Temperature [ C] Tmould = 160 C (inlet) Tmould = 160 C (center) Tmould = 160 C (outlet) Time [min] Figure 7. Temperature logging of the infusion (t=0sec) and reaction of a laminate at 160 C at three locations: resin inlet, middle and resin outlet 7

15 The lower temperature at the inlet is translated into a higher crystallinity of the resin in that part of the laminate, as it can be observed in Figure 6. This can be explained taking into account that the thermodynamic driving force to crystallize is higher for decreasing temperatures [6]. Therefore, the rate of crystal growth and, as a result, the degree of crystallinity, is higher at the inlet. It should be noted that the results in this section on the temperature gradient were obtained with a different activator, namely activator C in the resin mixture. The degree of crystallinity results for the laminates in which the activator T, initiator M was used, show the same behaviour as the results in Figure 6. Because of the higher temperatures at the resin outlet, the polymerisation as well as the crystallisation process starts at the outlet of the laminate. The reaction will then move towards the inlet opposite to the flow direction, as it can be seen in Figure 7, in which the peaks represent the polymerisation exothermal reaction. This also induces a gradient in the composite properties as well as the occurrence of voids near the inlet of the composite, which is explained in more detail in ref. 15. Depletion of activator in flow direction The same temperature gradient in the direction of the flow was observed when investigating the T activator. Previous research on APA-6 mainly focused on the initiator M and activator C combination for the reactive resin mixture [4-5, 10-11]. Both activators are blocked isocyanates which can unblock above their de-blocking temperature [12]. This is an equilibrium chemical reaction, as shown in Figure 8, step 1. Activator T has a lower unblocking temperature (125 C) compared to activator C, which de-blocks at a temperature of 160 C. When the activator is in the unblocked state, it has the ability to form branches in a similar way as in the neat resin [10]. It can also chemically react with the aminosilane sizing on the glass fibres by the formation of urea links, as it can be seen in Figure 8, step 2. The difunctional nature of the activator causes the polymer chains to grow from the fibre surface, while still having one side of the chemical group that is able to polymerise, as can seen in Figure 1. These chains can then form an interpenetrating network with the polymer chains in the bulk matrix, enhancing the fibre-to-matrix bond significantly. Due to the presence of unblocked activators, it was shown that the M-T resin system has better bonding properties. It is however thought that activator T already attaches to the fibres during infusion, leading to a depletion of the activator concentration in the flow direction and creating a difference in the resin constitution between the inlet and the outlet. The difference between the inlet and outlet properties was more present in the non-isothermal infusion when activator T was used. Figure 8. De-blocking of carbamoylcaprolactam (step 1) and the subsequent formation of fibre-to-matrix bond (step 2) 8

16 Optimization of the infusion temperature To eliminate the temperature gradient and the filtering effect of the activator through the laminate, the following change in the processing procedure was investigated: The dry fibre pre-form was not heated to the required processing temperature prior to infusion but to the same (lower) temperature as the resin in the buffer vessel. As a result, the temperature gradient was eliminated in the production process. The lower infusion temperature also leads to less unblocked activators, and thus a lower filtering of the activator is expected. After infusion, the laminate was heated to the required processing temperature of 180 C. In th is part of the study, activator T was used. The quality of the glass-fibre reinforced APA-6 composites not only depends on the chemical composition of the resin but also on the processing temperature. In previous research on the neat APA-6 resin by van Rijswijk et al. [10], it was shown that the polymerisation temperature is the most important parameter for the resin properties. If the infusion temperature is below the de-blocking temperature of the activator, but above the initiation temperature for the polymerisation, the blocked activators will take part in the polymerisation reaction, as shown in Figure 1. Hence, when the temperature rises above the deblocking temperature, during the heating phase of the production process, less activator groups will be available for the de-blocking reaction, resulting in a lower fibre-to-matrix bond. However, when the infusion temperature is above the de-blocking temperature of the activator, it can be in an unblocked state during infusion and hence react with the fibres during infusion, leading to the previously observed filtering (see section on filtering of the chemical ingredients). To investigate the effect of the infusion temperature, glass fibre reinforced APA-6 composites were infused at temperatures between 110 and 150 C with an interval of 10 C. They were then heated to a cure temperature of 180 C, with a heati ng rate of 16 C/min. For every infusion temperature, two or more laminates were produced and tested. The properties of the resulting composites were investigated by means of a non-destructive C-scan test (overview of the composite s quality), a degree of conversion test and degree of crystallinity test (assessment of the resin properties in the flow direction) and a short-beam strength test (mechanical properties). Results C-scan 120 C 130 C 140 C 150 C Figure 9. C-Scan images of APA-6 laminates at different infusion temperatures ( C) with the inlet at the bottom side of the picture As it can be seen in the C-scan pictures in Figure 9, there is a transition in the damping of the ultrasonic signal in the flow direction at all infusion temperatures. It can therefore be concluded that filtering is still taking place, resulting in a change in the composite quality between the inlet and outlet areas of the laminate. In the vacuum infusion process, three spacers were used to be able to control the fibre compaction and infusion rate, one at the inlet of the laminate and two near the outlet. Their influence on the resin flow during infusion is clearly visible and hence the transition is not as uniform as expected. 9

17 The best quality was found at the inlet areas of the laminates manufactured with infusion temperatures of 130 and 140 C. The C-scan results f or the infusion temperatures of 120 and 150 C are similar to each other, although the lower quality of the laminates has to be addressed differently for both cases, as explained in the paragraph on the optimisation of the infusion temperature. The result of 120 C is thought to be c aused by the long heating time required to reach the cure temperature, whereas at 150 C the lo w quality can be ascribed to a high filtering of the activator and therefore a less reactive resin system. For all the infusion temperatures, it is expected that voids arise when a low reaction rate has been established. Physical properties of the composites The results for the degree of conversion are shown in Figure 10. As it can be seen in this graph, there is a difference between the inlet and outlet areas of the laminate at all infusion temperatures. Since the quality of the laminates decreases severely towards the outlet, the degree of conversion at that area shows a bigger scatter. For the results of the inlet, however, there is a parabolic trend with a clear optimum, as indicated in the graph. A maximum is found at 130 C, which is believed to indicate an optimum equ ilibrium between the blocked and unblocked state of the activator DOC [%] Temperature [ C] inlet panel 1 inlet panel 2 inlet panel 3 outlet panel 1 outlet panel 2 outlet panel 3 Figure 10. Degree of conversion at different infusion temperatures At infusion temperatures lower than 130 C, the equi librium for the activator is mostly at the blocked side during the infusion, thus promoting polymer growth. The heating rate to the required curing temperature is not sufficiently high to shift the de-blocking equilibrium to the unblocked side (Figure 8, step 1) before the first initiator anions react with the activator, see Figure 1. This eliminates the possibility of linking to the aminosilane sizing groups on the fibre surface, creating a poor fibre-to-matrix bond, as it will be shown when discussing the mechanical properties (see Figure 11). As almost no activator is depleted, a too high activator concentration will be present in the bulk matrix, thus resulting in extensive oligomer formation [10], which causes the degree of conversion to drop as explained in section For infusion temperatures above 130 C, more filtering of the activator will take place. Therefore resulting in a lower reaction rate and hence a lower degree of conversion. For all the infusion temperatures considered in this research, however, an enhancement of the polymer growth compared to the non-isothermal infusion was observed (see Table 2). The difference between the inlet and outlet is also present in the degree of crystallinity (see Table 2), with a higher crystallinity at the inlet area of the laminate. Research on the interface of APA-6 composites showed that a good fibre-to-matrix bond resulted in a lower degree of crystallinity [13]. However, it will be shown that does not correspond with a poor fibre-to-matrix bond compared with the outlet properties, as will be shown in the next paragraph. The degree of crystallinity at the inlet found for the isothermal infusion was higher than for the non-isothermal infusion process. 10

18 Table 2. Physical properties at different infusion temperatures Degree of Degree of Infusion crystallinity [%] conversion [%] temperature [ C] inlet outlet inlet outlet 110 C C C C C Nonisothermal infusion Mechanical properties The results for the short-beam strength are shown in Figure 11. The difference between the inlet and outlet areas of the laminate, as seen in the C-scan results, is visible in the mechanical properties as well, except for the infusion temperature of 110 C. At this temperature, the filtering of the activator throughout the laminate is expected to be at its lowest level, as it is represented in the mechanical behaviour of the composites. The mechanical properties at this temperature are however low compared to other infusion temperatures, due to a poor fibre-to-matrix bond. 90 outlet inlet 80 Short-beam strength [MPa] Infusion temperature [ C] Figure 11. Average short-beam strength values at different infusion temperatures In general, the mechanical properties at the inlet are higher than at the outlet, which is related to the higher degree of crystallinity at the inlet. The mechanical properties of the laminates are however not only dependent on the physical properties of the resin but also on the ability to create a fibre-to-matrix bond. The best mechanical properties are found at an infusion temperature of 130 C. It can therefore be concluded that at that temperature an optimum equilibrium is found between the unblocked state of the activator, which provides the good fibreto-matrix bond, and its blocked state, which ensures a good polymerisation for the current heating rate of 16 C/min. With the non-isothermal infusion process, the difference in mechanical properties in the flow direction was around 25%. With the isothermal infusion at the optimal temperature of 130 C a difference of 15% was observed, confirming a lower filtering throughout the laminate, while still maintaining good physical properties. If the heating rate of the composites could be increased, this optimum might be found at a lower infusion temperature, with even lower filtering as a result. Conclusions To overcome thermal and chemical composition gradients, a new infusion strategy was proposed for the high temperature vacuum infusion process of glass fibre reinforced APA-6 composites: Both the resin and the dry fibre pre-form were at the same temperature during infusion after which the infused preform was heated to the required curing temperature. Research was done on 11

19 the infusion temperature to study its effect on the composite properties and their homogeneity in the flow direction. Some filtering could be observed at all the infusion temperatures, which could be mainly attributed to the de-blocking equilibrium reaction and subsequent filtering of the activator. The lowest filtering was observed at the lowest infusion temperature tested (110 C). However, the mechanical properties at this temperature were low compared to other infusion temperatures. An optimum for the mechanical properties was found at 130 C for the current heating rate, indicating an optimum equilibrium between the blocked and de-blocked state of the activator. If the heating rate of the composites could be increased, this optimum might be found at a lower infusion temperature, with lower filtering as a result and thus more homogeneous properties in the flow direction. Increasing the heating rate proved difficult for the heating method currently in use Process optimization for vacuum infused APA-6 composites In order to apply fibre-reinforced thermoplastic anionic polyamide-6 composites to large structures, such as wind turbine blades, an optimisation of the infusion process for obtaining more homogeneous properties is needed. Previous research on the isothermal infusion of the APA-6 composites (section and [14]) focused on studying the effect of the infusion temperature on the composite properties and their homogeneity in the flow direction. The isothermal infusion strategy was of interest because a non-isothermal infusion was found to induce thermal and chemical composition gradients in the flow direction and hence a gradient in the composite properties (section and [15]). This gradient, also present for the isothermal infusion but to a much lesser extent, could be mainly attributed to the filtering of the activator, due to its deblocking reaction, and also to void formation in the flow direction. An optimum for the mechanical properties was found at 130 C for the current heati ng rate, which results from an optimum equilibrium between the blocked and unblocked state of the activator. Therefore, an isothermal infusion at 130 C (infusion strategy 2 in see section 2.2.2) was used for the research presented in this section, which main goals are: Improving the degassing procedure in order to minimise the void content in the composite Minimising the flow induced gradient by optimizing the process parameters during infusion and post-filling Finding the optimal flow conditions via different infusion velocities and pressures Degassing Procedure The presence of voids in composites is detrimental for several mechanical properties, especially for matrix-dominated properties, such as interlaminar shear strength, compressive strength and fatigue life [16]. Judd et al. [17] found that each 1% increase in void content induces a 7% reduction in the interlaminar shear strength of the composite. This drop of the interlaminar shear strength is approximately linear up to a total void content of 4%. A void content of 1% is generally taken as the maximum acceptable value for structural applications [18]. A void content of about 8% was observed for the APA-6 composites and is, therefore, one of the major issues to address. In previous research, degassing of the MMU tanks for 15 minutes at 100 mbar was proven sufficient for infusing a void free neat polymer panel [10]. However, Tabel 3 shows that the same degassing procedure leads to a void content of approximately 8% when applied to composites (infusion according to infusion strategy 1). As a consequence, an alternative procedure was investigated, consisting of degassing in the resin buffer for 5 minutes at 3 mbar. This way the void content was reduced to about an average 5%. The resulting improvement in the composite properties is clearly visible in the ILSS data presented in Figure

20 Table 3. The effect of the degassing procedure on the composite properties. Degassed in buffer vessel (3 mbar, 5 minutes) Degassed in MMU (100 mbar, 15 minutes) T mould [ C] Conversion [%] Crystallinity [%] Void content [%] Conversion [%] Crystallinity [%] Void content [%] ILSS [MPa] Degassed in buffer vessel 10 Degassed in MMU mould temperature [ C] Figure 12. The effect of the degassing procedure on the inter laminar shear strength (ILSS) of the composites. Flow Conditions To investigate the flow behaviour and identify the important parameters that influence the flow and the quality of the composites, research was focused on the cross-sectional area of inlet tube and on the infusion and post-fill (curing) pressures. The infusion and curing stages were analysed separately. The infusion with activator T as the activator was done by infusion strategy 2, with an infusion temperature of 130 C and a cure temperature of 180 C. For all the laminates manufactured, the flow front position was monitored by means of a video camera. The camera recorded the flow of the resin in flow direction along the right side of the laminate. From this information, the flow rate could be calculated. The overall quality of the laminates was assessed by non-destructive C-scan inspection, after which they were tested to obtain their physical and mechanical properties. The baseline composites used in previous research (for both infusion strategy 1&2) were infused with an inlet tube cross-section of 20mm 2 and the infusion pressure and post-fill pressure was set to 250 mbar, to obtain the same fibre volume fraction as melt processed PA-6 composites. In this research, a full-factorial design was established in which the following values were used: Inlet tube cross-section: 4 and 20 mm 2 Infusion pressure: 3 and 250 mbar Curing pressure: 3 and 250 mbar Infusion stage The infusion time was observed to be mainly dependent on the infusion pressure and not as much on the inlet cross-section. For an infusion pressure of 250 mbar the infusion time was 55±11s while for an infusion pressure of 3 mbar the infusion time was only 35±4s. The flow front position and the flow rate through the fibre preform are shown in Figure 13. As it can be seen in this graph, the flow is quite stabilised and mostly pressure driven from approx. 7.5 cm in the flow direction. From the results displayed in Table 4, it could also be concluded that the infusion pressure determines the thickness of the obtained laminate, regardless the post-fill vacuum pressure. A thickness gradient, usually reported when using vacuum infusion [19], was not present in these experiments. 13

21 Table 4. Thickness (t) and inter laminar shear strength (ILSS) near the inlet and outlet of the composites, and the infusion times at different infusion and post-fill pressures. Infusion pressure [mbar] Post-fill pressure [mbar] t (inlet) [mm] t (outlet) [mm] ILSS (inlet) [MPa] ILSS (outlet) [MPa] Difference ILSS [%] Infusion time [s] 3 3 2,69 2, , ,71 2, ,75 2, , ,76 2, ,5 Curing stage After infusion the inlet tube was closed, the temperature was raised to the required curing temperature and the pressure was set to the post-fill pressure. The difference in mechanical properties between the inlet and the outlet was found to be significant: a drop between 10 and 31% in the interlaminar shear strength over a length of 25cm was observed, as shown in Table 4. To assess the cause of this gradient, the laminates were C-scanned (Figure 14) and crosssections in the flow direction were investigated by means of a field emission scanning electron microscope. Flow rate [cm 3 /s] mm2; 3 mbar 4 mm2; 3 mbar 4 mm2; 250 mbar Flow front position [cm] Figure 13. Flow front position during infusion and their corresponding flow rate at different inlet tube cross-sections and infusion pressures. The micrographs showed macro voids between the fibre bundles, mostly located in the upper part when looking through the thickness of the laminates, to which the loss in mechanical properties could be attributed. The C-scan images (Figure 14) showed that the highest infusion vacuum pressure (3 mbar) resulted in a more uniform damping but a lower overall quality, while the lowest infusion vacuum pressure resulted in a clear transition of the composite quality in the flow direction. Likewise, the ultrasonic attenuation of the laminates near the inlet had a good correspondence with the results of the ILSS tests. 3mbar-3mbar 3mbar-250mbar 250mbar-3mbar 250mbar-250mbar Figure 14. C-scan results for different infusion pressures - post-fill pressures respectively, with an inlet tube cross-section of 20mm 2 14

22 Barraza et al. [20] reported that applying a post-fill pressure in a resin transfer moulding (RTM) process reduced the void content significantly, independently of the infusion velocity, as it was observed within this research. The higher post-fill pressure (3 mbar) led to an increase in mechanical properties (tensile strength and stiffness) of at least 13%, which can be attributed to a lower void content. When looking at the results for the ILSS in Table 4, it can be seen that the post-fill pressure influences the mechanical properties, especially at the outlet as well as reduces the gradient in flow direction. Due to the low de-blocking temperature of activator T (125 C), the infusion temperature of 130 C used in this research might result in filtering of the activator, inducing, as a result, a chemical composition gradient in the flow direction. Therefore, the degree of conversion in the panels with the highest difference in quality was measured and plotted in Figure 15. It can be seen that the degree of conversion near the outlet is higher than that near the inlet in all the studied cases. Likewise, it has approximately the same value for all infusion settings, which indicates that an optimum resin constitution was found at the outlet. Near the inlet, more resin passed through the fibre bed ensuring a proper wet-out and a better fibre-to-matrix bond, therefore reducing the polymerisation, as explained in and [14]. It could be concluded that mainly the amount of resin that flowed through the cross-section influences the degree of conversion. Therefore, the author suggests that flushing the laminate with extra resin after full wet-out could provide with a better fibre-to-matrix bond in the whole laminate as well as more homogeneous properties in the flow direction. However, such a procedure could lead to a lower overall degree of conversion. Degree of conversion [%] inlet outlet Figure 15. Degree of conversion near the inlet and outlet of the laminates at different inlet tube cross-sections [mm 2 ] - infusion pressures [mbar] - cure pressures [mbar] Conclusions In order to increase the homogeneity of the APA-6 glass fibre reinforced composites, an alternative degassing method was investigated: after mixing the reactive resin, it was degassed in the buffer vessel. This decreased the void content significantly and therefore resulted in better mechanical and physical properties. To minimise the flow induced gradient, the effect of several process parameters during infusion and curing, namely the inlet tube cross section, the infusion and post-fill pressures, was investigated. Although none of the considered combinations of parameters led to an outstanding increase in the homogeneity of the composites, basic understanding was gained on the effect of these settings in the overall quality of the laminates. A fast infusion, corresponding to a high infusion vacuum pressure (3 mbar), was found to result in a less visible flow gradient but also in a lower overall quality of the laminates. Therefore, research on even lower infusion pressures (500 mbar) is highly recommended. It was also found that a high post-fill pressure (3 mbar) led to a reduction in the void content, therefore resulting in an increase in the mechanical properties of the laminates as well as more uniform properties in flow direction. As for the physical properties of the laminates, a flow induced gradient was also visible with a lower degree of conversion near the inlet. By flushing the laminate with extra resin after the wet-out of the fibre preform, this gradient might be reduced. 15

23 2.5 Upscaling Research aiming at upscaling the length of the laminates was based on an isothermal infusion. For this study, activator C was used because the de-blocking temperature of this activator is 160 C instead of 125 C for activator T. It is thought that by using activator C, the gradient in flow direction will be less significant than for activator T. The only drawback of using activator C is the fact that its faster exothermal reaction once a certain temperature, at which the resin starts reacting, has been reached within the system. This activator C will thus give more infusion challenges for infusing long panels than activator T. Isothermal infusion was already investigated in section 2.4. Results showed an improvement in the uniformity of the laminate properties at 130 C although some gradient in the flow direction still remained, as shown in Figure 9. To check whether this lack of uniformity is just a local phenomenon in the vicinity of the outlet or, on the contrary, it is induced by the flow itself, the infusion length was increased from 25 to 45 cm, which is the maximum laminate length allowed by the 50x50cm 2 area of the infrared heating panels. Research was focused on the properties of the resin and the composite as well as on the infusion strategy of these longer laminates. The different infusions performed with this purpose are presented in Table 5. Test nr. Table 5. Infusion settings for the infusion of long laminates Temperature of Temperature Infusion resin in the buffer of fibres during pressure vessel [ C] infusion [ C] [mbar] Inlet clamped after resin reached outlet Yes No No Yes Yes Within this study, the infusion temperature that was chosen from the standpoint of the reduction of the flow gradient of the laminates was 110 C. For this temperature, the effect of the infusion pressure was tested. This was done to investigate whether the effect that was seen in section (lower visible flow gradient but lower overall quality for higher infusion pressures) was also valid for the activator C system. For the lower infusion pressure (250mbar, test 2), the inlet was not closed when the resin reached the outlet but was left open until the resin stopped flowing. The laminate was heated up to the cure temperature from the moment that the resin reached the outlet. As an infusion temperature of 130 C showed the best mechanical properties in combination with an improvement in uniformity, it was chosen for the study on activator C as well. For this infusion temperature, the effect of closing the inlet when the resin reached the outlet (test 3) or flushing the laminate with extra resin during the heating phase (test 4) was investigated. During the isothermal infusion, heating to the required temperature is started when the resin reaches the outlet. Normally, the inlet is clamped just before the heating process starts. It is however thought that flushing the laminate with extra resin during the heating phase might improve the properties significantly since the product would be filled with extra resin if needed, diminishing by that the number of voids. To see whether it was possible to infuse long panels with the non-isothermal infusion (infusion strategy 1), a high cure temperature (180 C) was chosen as this is the most critical test case. At this high temperature, the cure of the resin will start fast and the exothermic reaction will increase the viscosity of the resin quickly, therefore hindering the resin flow in the long panel. For all the tests, the infusion time that was needed for the resin to reach the outlet was noted and C-scan images were made. 16

24 Tests 1 and 2 were done to check whether the effect of the infusion pressure was also valid for the activator C system at an infusion temperature of 110 C. It was found that increasing the vacuum infusion pressure to 3 mbar slightly reduced the flow induced gradient while it did not affect the overall quality of the laminates, even when the inlet was clamped after the resin reached the outlet. The optimization of the infusion temperature as described in section was performed for the resin system with activator T. The results of tests 2 and 3 provided information about the effect of the infusion temperature for the activator C system (see Figure 16 and Figure 17 respectively for the results of the C-scan inspection). As it can be seen in these figures, the quality of the products and the flow gradient is comparable for both infusion temperatures, although the best result in reducing the flow gradient while maintaining a good overall quality was obtained with the infusion temperature of 130 C. This result does not entirely agree with those in section 2.4.1, where a clear improvement in flow gradient was observed for the mechanical properties at low infusion temperatures, 110 C, but a better quality at the higher infusion temperatures. It can be concluded then that due to the high de-blocking temperature of the activator C, the flow gradient and quality of the composite is less dependent on the infusion temperature. Figure 16. C-scan results for infusion temperature of 110 C when inlet is not clamped during heating (test 2) Figure 17. C-scan results for infusion temperature of 130 C when inlet is not clamped during heating (test 3) Figure 18. C-scan results for inlet clamping after infusion (test 4) Flushing the laminate with extra resin was performed in tests 2 and 3 by leaving the inlet open during the heating process. In test 3 and 4, the only parameters that was changed, was the clamping of the inlet right after the infusion. The C-scan inspections corresponding to tests 3 and 4 are displayed in Figure 17 and Figure 18 respectively. It can be seen that the flow induced gradient is significantly reduced and that the overall quality of the laminates improved significantly when the resin is flushed during heating. One of the panels manufactured with the parameters of test 4 did not even show any flow gradient, which is the one of the goal of the optimization of the process in combination with a good quality of the laminate. Table 6. Infusion times for the infusion of long laminates Test nr. Temperature of resin in the buffer Temperature of fibres during Infusion time [sec] vessel [ C] infusion [ C] The infusion times for all the tests are also shown in Table 6. It can be seen in this table that increasing the infusion temperature, decreases the infusion time. Due to the high heat 17

25 conductivity of the fibres compared to the resin, the resin that is infused into the 180 C fibre preform will quickly heat up, lowering the viscosity of the resin and facilitating the infusion. However, due to this high temperature, the resin will start its reaction faster after infusion. An optimum for the time available for the infusion process is therefore expected to be found at lower infusion temperatures. It was also observed that a lower infusion time resulted in a better quality of the composite. When the infusion time of the 180 C laminate (144 sec) is compared with that corresponding to a shorter flow length of 25cm (75 sec, [4, p.148]) a linear increase of the infusion time with the flow length can be ascertained. The laminates that were infused at C while letting the resin flow through the laminate while heating up to the cure temperature of 180 C, showed a good quality. This result was achieved without increasing the infusion time into the laminate. It can therefore be concluded that the up-scaling of the laminates is feasible. 2.6 Cure Research on post-processing of anionic polyamide-6 composites was performed, consisting of two main areas of interest: the cure time and the anneal cycle, discussed in section 2.6 and 2.7 respectively. The cure time is the time that the product is at the cure temperature. The anneal cycle is the heat treatment that is given to the laminate once it is cured and cooled down. This annealing cycle can be done at different temperatures and for different times. The cure time and annealing temperature proved to be the crucial parameters to increase the degree of crystallinity, minimise the residual stresses and hence improve the mechanical properties. The effect of the cure time is assessed on both neat resin and composites. For the neat resin, a broad spectrum of cure times was tested ranging from 15 to 120 minutes, with a 15 min interval with an extra data point at 180 min. Glass fibre reinforced APA-6 were cured at min, firstly to determine if the trends in material properties correlate to the neat resin samples and secondly to determine optimum cure time. Pure resin For the pure resin, it was seen that with increasing cure time, the degree of crystallinity rose, until a plateau was reached at approximately 100 minutes of curing, as can be seen in Figure 19. The cure time as researched by van Rijswijk et al. [10], had little effect on the degree of conversion between 30 and 45min, over a range of cure temperatures. The results from this research show a clear agreement in an approximately constant DOC of 92% between the 30 to 75min range. Figure 20 also shows an optimum in degree of conversion at the same cure time as for the high degree of crystallinity. The almost 7% change in conversion observed over a cure time range of 75min to 120min clearly indicates that further conversion of monomer, at a cure temperature of 180 C, can take place for neat resin. After that, t he degree of conversion decreases due to degradation of the polymer or due to a shift in the equilibrium reaction between the monomer and polymer. Degree of crystallinity [%] Cure time [min] Figure 19. Evolution of degree of crystallinity of pure resin in time. Degree of conversion [%] Cure time [min] Figure 20. Evolution of degree of conversion of pure resin in time. 18

26 Laminates For the production of the laminates, infusion strategy 1 was used. The effect of cure time on the laminate is greatly less significant for the degree of conversion, which proofed to be almost constant at a value of 98.5%±0.3%. For the effect of cure time on the degree of crystallinity, Figure 21 shows a linear increase to a peak value of 36.5% at 90min and an almost constant peak crystallinity of 36% even at 120min. Figure 22 shows that the maximum short-beam strength is 77.4MPa, coinciding with the cure time of 90min for peak crystallinity. Degree of crystallinity [%] Cure time [min] Figure 21. Degree of crystallinity for different cure times Short-beam strength [MPa] Cure Time [min] Figure 22. Short-beam strength for different cure times The increase in crystallinity with cure time is in the same order of magnitude as for the pure resin. The increased crystallinity for both the pure resin and the laminate can be explained by the greater extent of primary crystallization that occurs during curing over the secondary crystallization that only initiates from subsequent cooling. It is known [10] that at higher processing temperature branching may occur in the resin. Branching causes a crystal type transition from α to γ structure [21-22], which are two types of crystals with different levels of perfection. With an increased primary crystallization, a greater amount of the more perfect α crystals are formed, while γ or the less perfect α crystals formed during secondary crystallization make up a lower composition of the overall crystal structure. An inherently greater extent and perfection of the crystal structure in the polymer clearly would increase the strength of the polymer due to increased covalent bonds present. It was observed that an optimal degree of crystallinity could be found and that this cure time corresponded to optimal mechanical properties (short-beam strength). Literature states that all semi-crystalline polymers have a direct relationship between the short-beam strength and the degree of crystallinity [23]. This direct relation further accounts for the results showing an increase in the short-beam strength of the laminate due to the increased crystallinity with cure time. Conclusions An increase in degree of crystallinity was found with increasing cure time until an equilibrium value was reached. The results were comparable for the pure and the fibre-reinforced APA-6, with a max increase of 15% after approx. 90 minutes of cure. The mechanical properties followed the same trend, having a peak value (77.4MPa) at the cure time with the maximum degree of crystallinity. 2.7 Post processing of APA-6 composites In previous research on pure APA-6 [10], it was shown that the polymerisation process was completed within 30 minutes. At the end of polymerisation, however, the crystallisation is still in progress. As the properties of the resin are mainly driven by the level of crystallinity [10], the cure and/or annealing cycle may influence the composite properties significantly. By selecting the right curing and post-processing parameters, it is expected that the demoulding time can be optimised, residual stresses minimised and hence the neat resin and composite properties significantly improved. The influence of annealing parameters will be studied. 19

27 All the samples in this study were manufactured at a processing temperature of 180 C, infused by strategy 1 (see section 2.2.2). At this temperature, polymerisation precedes crystallisation. When compared to lower processing temperatures, this processing temperature yields a low degree of crystallinity and has high residual stresses, due to thermal shrinkage [24] after a normal cure cycle. Therefore, it is thought that the post-cure cycle will have the largest effect when the samples are manufactured at this processing temperature. Annealing conditions Initial cure times of 30 and 60 min were chosen for the research on annealing of APA-6 composites. As seen in the last paragraph, full crystallisation has not been reached yet for those cure times. It is expected that the crystallization behaviour will improve and the residual stresses will diminish during the annealing cycle. Within this study, the cured laminate is cooled down to room temperature and removed from the mould, after which it is placed in an oven for the annealing cycle. Design of Experiments (DoE) was used for this research. DoE is a method that is used to determine the relationship between the different parameters affecting the process (annealing time, annealing temperature, rate of cooling after annealing process) and the results of that process (change in crystallinity and in mechanical properties during the annealing process. Three important temperatures can be identified for annealing: the glass transition temperature Tg ( C), the crystallization temperature Tc ( C) and the melting temperature Tm ( C). Semi-crystalline materials are known to undergo additional crystallization when held at a temperature between its glass and melting temperature. Conventional annealing is done at temperatures at or above Tc, leading to increased crystallinity and reduced toughness [25]. If embrittlement is undesirable, a lower temperature annealing can be effective. A lower annealing temperature reduces the heat energy requirements without extending annealing time and thus reducing the costs. The following temperatures were selected for this study: C. The laminates were held at the annealing temperature for 2, 4 or 6 hours. Temperatures at or above 200 C will not be explored in this study as thermal and thermal-oxidative degradation is considerably high [26] and cannot be used practically especially with significant material discoloration. After the annealing cycle, the laminate is subjected to a controlled cooling of either fast or slow rate. Fast cooling, is performed by removing the laminate from the oven and allowing it to cool naturally to room temperature. Slow cooling, is carried out by switching off the oven and allowing the laminate to cool inside the oven, with the oven ventilator opened. Annealing results Degree of conversion For reinforced APA-6, the effect of reheating the polymer for a period of time is expected to increase the degree of conversion, improve the crystallinity by inducing favourable crystal type transition (from γ-crystals to α-crystals) [27] and remove imperfections in crystal structure. The degree of conversion, however, seemed to be constant for all laminates, complying with the results of van Rijswijk et al. [10], stating that the conversion of the APA-6 resin is completed or stops within 30 minutes. Degree of crystallinity For every laminate, crystallinity samples were taken before and after annealing. The relative change in crystallinity [%] in relation to the annealing temperature is presented in Figure 23. The results presented here are the averages of all the annealing times and rates of cooling at a certain annealing temperature. Hereby, the main effects due to a change in parameter in the process can be seen. The mean crystallinity for un-annealed samples is 32.13% after 30 minutes of cure and 29.6% after 60 min. It can be seen that for both curing times, the crystallinity increases significantly with increasing annealing temperatures, whereby the biggest increase is found around the crystallization temperature of the laminate. It is known that crystalline polymers 20

28 heated to temperatures slightly under melting temperature results in increased lamellar crystal thickness due to sufficient thermal energy given to the polymer that enables molecular motion [28]. Change in crystallinity [% min cure 30 min cure Change in crystallinity [%] min cure 30 min cure -10 Annealing temperature [ C] Annealing time [hr] Figure 23. Change in crystallinity at different annealing temperatures (averaging the results for different annealing times and cooling rates at the different annealing temperatures). Figure 24. Change in crystallinity at different annealing times (averaging the results for different annealing temperatures and cooling rates at the different annealing times). Gogolewski [29] outlines the fact that recrystallization of the polymer proceeds at a faster rate at higher temperatures due to the increase in the number of dislocations in the crystal structure, which forms very favourable nucleation sites. Furthermore, the release of topological constraint and stresses in the amorphous phase at higher temperature promotes crystallization [27]. The thermal energy seems crucial in the annealing process, as temperatures close to the Tg do not show an increase in crystallinity. Gogolewski et al. [29] observed that having a lower annealing temperature reduces the post-processing effectiveness on the nylon-6. For the annealing at 100 C in this study, even a decrease in crystallinity was seen, see Figure 23. This result can not be explained at this moment. In Figure 24, the change in crystallinity with respect to the annealing time is shown (averaging the results for different annealing temperatures and cooling rates at the different annealing times). As can be seen in this graph, no significant effect could be identified. For the annealing study on melt-processed PA-6 [29], experiments were performed at hrs at various temperatures up to 215 C. An incubation period of 1 0hrs was noted at temperature of 215 C to see a change in the long spacing of SAXS experiments. An even longer incubation time (20hrs) was noted at 200 C after which material properties started to improve significantly. It can therefore be concluded that the results in this study, for the lower annealing temperatures, do not give any significant results at this point, but if the annealing time is increased it might show an increase in polymer properties. A reduction in the cooling rate after annealing, keeps the polymer chains mobile for a longer duration, which allows more time to form crystals and thus increase the crystallinity of the polymer or increase the level of perfection of the crystals. The cooling rates of the laminates are non-linear, unlike the heating rates to the annealing temperature. As can be seen in Figure 25, the cooling rates show the same trend when plotted against their annealing temperature. In this annealing study, the cooling rate appears to be less significant however, as can be seen in Figure 26. As the polymer chains have sufficient thermal energy for crystal formation during the annealing phase, less can be gained from a slow cooling rate at the end. 21

29 Cool down rate [ C/min] slow cooling fast cooling Anneal temperature [ C] Figure 25. Cooling rates at different annealing temperatures Change in crystallinity [%] Slow 15 Fast Cooling rate Figure 26. Change in crystallinity for different cooling rates As can be concluded from these results, the annealing temperature is the key factor for an increase in crystallinity for both curing times. Short-beam strength The results for the short-beam strength with respect to the annealing temperature (see Figure 27) do not show the same trend as for the degree of crystallinity for a cure time of 60 minutes. A peak value of 72.6 MPa was reached for the 60 min cured laminate, compared with the 70.8 MPa obtained with an un-annealed sample. However, for the 30 min cured laminate, the same trend was seen as for the change in crystallinity, going from a baseline of 59.6 MPa to a maximum of MPa. The biggest difference between the two cure times is that the fibre-to-matrix bond, its intertwining with the bulk matrix and the degree of branching in the bulk are less for the 30 min cured laminates. It can thus be concluded that the degree of crystallinity is not the only parameter determining the mechanical properties. By annealing, the structure of the polymer chains is improved, the amount of dislocations reduced and a more entangled network formed, leading to a more distinct increase in mechanical properties after annealing a 30 min. cured laminate when compared to a more perfect laminate (60 min of cure). The same observation was done when looking at the annealing times, see Figure 28. The effect of the annealing time is not significant for the 60 min. cured laminate unlike for the 30 min cured one. For the 30 min. cured samples, the longer the laminate is held at the annealing temperature, the bigger the increase in mechanical properties. After 30 min. of cure, more imperfections will be present in the matrix as well as a lower fibre-to-matrix bond and less branching. The annealing time therefore becomes an important factor in the increase in mechanical properties. The effect of the cooling rate proved to be insignificant for the mechanical properties, which is in agreement with the results for the change in crystallinity. Short-beam strength [MPa] min cure 30 min cure Annealing temperature [ C] Figure 27. Short-beam strength for different annealing temperatures. Short-beam strength [MPa] min cure 30 min cure Annealing time [hr] Figure 28. Short-beam strength for different annealing time s. 22

30 Conclusions The results for the study on the annealing cycle showed that the change in degree of crystallinity depended mostly on the annealing temperature. The biggest improvement was seen near the crystallization temperature of APA-6 (175 C) with an increase in crystallinity of 20-30%. It was also observed that the mechanical properties only followed the same trend as the crystallinity behaviour for the laminates cured for 30 minutes, leading to a significant improvement in shortbeam strength (+17%), as big improvements could be reached with respect to fibre-to-matrix bond and branching of the matrix. As these properties were already attained during curing for a 60 min laminate, their mechanical properties did not show the same trend as its crystallinity after annealing, showing only a slight increase (+3%) in short-beam strength It could be concluded that the selection of the long cure time is most effective for obtaining good mechanical properties in combination with a high degree of crystallinity. The same properties could be obtained with an annealing cycle, although requiring more time in a heated oven and thus more heating energy. 2.8 Conclusions In this chapter, the process optimization of thin glass fibre reinforced APA-6 composites was discussed. First of all, the effect of the pot life (time and temperature) on the reaction/reactivity of the resin as well as on the increase of the viscosity was investigated. It could be concluded that the longer the residence time and the higher the initial buffer vessel temperature, the higher the initial viscosity and the faster the reaction respectively, hence, complicating the infusion of large products. Investigating the effect of this residence time or pot life more fundamentally by means of reaction kinetics is seen as a critical part in the investigation on the infusion of large wind turbine blades. Secondly, the effect of the infusion itself was tested. Different infusion strategies were tested in order to increase the homogeneity. To overcome thermal and chemical composition gradients, an isothermal infusion strategy was implemented for the high temperature vacuum infusion process. Research was done on the infusion temperature to study its effect on the composite properties and their homogeneity in flow direction. It was seen that at all infusion temperatures, there was still filtering present. A trade-off needed to be made between low filtering (low infusion temperatures) and high mechanical properties (higher infusion temperatures). With the current heating rate, the optimum was found at an infusion temperature of 130 C. As the current heating methods did not allow a faster heating rate between the infusion and cure temperature, research was focused on minimizing the flow induced gradient and finding the optimal flow conditions The effect of different process parameters, like the degassing procedure, the infusion pressure (infusion velocity) and cure pressure (post-filling) were studied. Degassing the resin in a buffer vessel after mixing decreased the void content significantly and therefore resulted in better mechanical and physical properties. A high infusion velocity caused a lower gradient in flow direction, but resulted in a lower overall quality compared to the lower infusion velocity. Applying a post-fill pressure after infusion proofed to reduce the void content and therefore increasing the mechanical properties and reducing the flow induced gradient. The effect of increasing the flow length of the infusion was tested. All laminates were fully infused without changing the processing parameters and resulted in good quality laminates. It can therefore be concluded that the up-scaling of the laminates is feasible. It was seen that there was a linear increase of infusion time with flow length. Flushing the laminate with extra resin during the heating up phase in the infusion process resulted in a laminate with a significant reduction, or even disappearance of the flow gradient. And finally, the effect of a post-cure or post-processing was investigated. For the research on the cure parameters, it was seen that the optimum in mechanical properties was found at a cure time of 90 minutes which corresponded to the maximum in degree of crystallinity. The results for the study on the annealing cycle showed that the change in degree of crystallinity depended mostly 23

31 on the annealing temperature. The biggest improvement was seen near the crystallization temperature of APA-6 (175 C) with an increase in crystallinity of 20-30%. It was also observed that the mechanical properties followed the same trend as the crystallinity behaviour for the laminates cured for 30 minutes, leading to a significant improvement in short-beam strength. It could be concluded that the selection of the long cure time is most effective for obtaining good mechanical properties. The same properties could be obtained with an annealing cycle, although requiring more time in a heated oven and thus more heating energy. 24

32 3 Thick laminates: Heating methods 3.1 Introduction The isothermal infusion strategy, as described in section , showed a lot of potential for the infusion of large APA-6 composites. Due to the low infusion temperature of both the resin and the fibres a long infusion window is obtained while obtaining good mechanical and more uniform properties. The study on the isothermal infusion strategy was focused on thin laminates. However, if this strategy is wanted for the infusion of thick laminates, it needs to be researched whether the heating of the whole product from the infusion temperature to the cure temperature is feasible, especially with the low heat conductivity of the resin material. The heat inside the product needs to reach the required polymerization temperature before the resin starts its exothermal reaction. Otherwise, the reaction will start at a lower temperature than the required temperature for best quality; good resin and mechanical properties. It is expected that with the current heating method a temperature gradient will arise through the thickness of the product, as the inside of the laminate is heated by conduction. In this chapter, different heating methods will be tested and will be compared to each other. The heat distribution through the thickness of the laminates and the heating rate from the infusion temperature to the curing temperature will be considered. 3.2 Experimental Materials Resin material The resin materials used are described in section In this research, the activator C is used for the infusion of the thick-walled composites. Glass fibres E-glass fibre mats are used to reinforce the APA-6 composites. In the research for the thickwalled composites are coarser weave is used than compared to the 8HS glass fabric used in the study of the thin laminates. The coarse 5 mm wide 12k tow plain woven glass fabric (RP0840) was supplied by Ten Cate Advanced Composites with the same fibre sizing as on the 8HS weave. The coarse weave had a high permeability resulting in a fast infusion. Mould material For the infrared heating, which is one of the heating methods investigated in this chapter, it is important to test the influence of the surface finish of the aluminium mould used in the set-up on the absorption of the infrared radiation. Therefore, moulds were prepared with different surface treatments and tested, as will be discussed in 3.3. A 1.5 mm thick mould with a black paint finish was used for all the moulds used in this research after this investigation Processing methods To investigate the heating rate through the thickness of the laminates, a laminate of considerable thickness is tested. All heating methods were tested on a balanced, symmetric 40-ply laminate (22x22cm 2 ), see Figure 29, resulting in a thickness of 24 mm once consolidated. The vacuum bag is made as described in section Infusion strategy The infusion strategy is the infusion strategy 2 as described in section Before the infusion takes place, the thermocouples are connected to Keithley Temperature Logging Computer to record the temperatures during heating. All tests are done with an infusion temperature of 110 C and the laminates are cured at a temperature of 180 C. Heating from 110 C to 180 C is considered to be the maximum temperature range possibly needed for the production of APA-6 25

33 laminates and in this 70 C heating step, the capaci ty differences between the heating techniques will become clear. Figure 29. Vacuum infusion bag for thick laminates with the coarse fabric weave Figure 30. (Kapton ) Thermofoil Heater Heating devices Infrared panels Two 5000 watts infrared emitters of 50x50cm 2 are used, as can be seen in Figure 4. They are provided by Watlow USA. Two PID controller are used to control the top and bottom panels. A thermocouple is placed on the surface of the kapton foil and on the mould to be used as the control input. Before infusion, the power controller was tuned for optimal heating of the dry laminate. Heated press The heated plate consists in two aluminum blocks heated by tubular heating elements. The power output is 3000 Watt over an area of 25x25cm (4.8W/cm 2 ). Three elements are inserted in each block. Thermofoil Another contact heating method was tested, namely a thermofoil. This Polyimide (Kapton ) Thermofoil Heater is sold by Minco Products inc., USA. The kapton foil protects the heating system from the chemicals in the process. A 10x10cm 2 sized flexible foil with a thickness of 0.3 mm was purchased and used in this research. The thermofoil was used on the mould side where the fibres are placed. The power of the thermofoil is 1.8W/m 2. To avoid heat loss through the aluminium plate, isolation material was installed. The heating device for the top of the vacuum bag was the infrared panel as described above. Resistive mesh As the heating rates for the surface heating turned out to be not sufficient, a heating source from the inside out needed to be implemented. A metal mesh will be used as a heating device in the middle of the fibre lay-up to provide resistive heat from the inside out while heating up the laminate from infusion to curing temperature. The resistivie mesh consists of RVS steel wires (thickness: 200µm, width: 858µm) provided by Dinxperlo Wire Weaving Co Ltd. This heating source is placed in the middle of a laminate, see Figure 31. Since the temperature of the metal mesh had to be controlled to the infusion and polymerization temperature, the power controller was adjusted manually to the desired temperature by manual adjustment by varying either the applied voltage or amperes. First, a high power input is given for the maximum heating rate to the required temperature and when this temperature was reached, the power was slowly reduced until a constant temperature. The laminate was produced in combination with the infrared heating from the outside. 26

34 Carbon fabric As carbon fibres exhibit a high thermal conductivity, it was also tested as a heating source in the middle of the laminate. It can be heated up in two ways by induction or by direct voltage. Direct voltage will be used in this study. Although it might be advantegeous from a hardware point of view to test the induction heating in the future. A power supply was connected to a plain weave carbon fabric. The carbon fabric was clamped between two copper plates to connect the fabric to the wire of the power supply for good connectivity and good heat distribution, see Figure 32. The wire was soldered to one of the copper plates and connected to the power supply. The power controller was also adjusted manually to the desired temperature by manual adjustment by varying either the applied voltage or amperes. First, a high power input is given for the maximum heating rate to the required temperature and when this temperature was reached, the power was slowly reduced until a constant temperature. The laminate was produced in combination with the infrared heating from the outside. Figure 31. Position of the resistive mesh in the middle of the fabric lay-up and its connection to the power supply Figure 32. Connection of the carbon fabric via two copper plate to the copper wire of the power supply Analysis methods Temperature measurements Through the thickness of the laminate, K-type thermocouples were placed to record the temperature during the infusion and the cure. The position of the thermocouples can be seen in Figure 33. Heating experiments are performed on the dry fibre lay-up prior to infusion to check the heat distribution through the thickness without the exothermic reaction of the resin. During infusion and heating to the required curing temperature, the temperatures are also logged. Figure 33. Position of the thermocouples through the thickness of the lay-up. Viscosity measurements Viscosity measurements were done, as described in section , to select the required heating rate. Reaction kinetics Reaction kinetics tests were performed to investigate the temperature build-up. To simulate the adiabatic conditions of the polymerization in thick laminates, a wide 600 ml beaker (VWR ) was used as a mould. For ensuring a nitrogen protective environment during the whole polymerization, a cover lid was designed. The lid has two small holes for rigid thermocouples, one in the middle, for adiabatic conditions, (inner thermocouple) and the second one on the side 27

35 (outer thermocouple). On the left side of Figure 34, there is a connection for nitrogen. Also, the design of a supporting aluminium plate to hold one beaker was necessary. Figure 34.Glass beaker with cover lid and supporting plate The polymerization mixture was prepared as described in section The glass beaker with the reactive mixture was transferred into the first oil bath, which was set to the same temperature as mixture in the MMU, for simulation of the pot life and of the time to infuse the thick laminates. After 5 minutes, the beaker was placed into the second oil bath set to a higher temperature for the polymerization reaction to take place, simulating the heating phase of thick laminates by contact heating. The cover lid was then placed over the beaker and the temperature measurements, as described above, were started. The thermocouples were lubricated by grease for easier removing from the product after reaction. 3.3 Heat distribution In this section, the heating profiles through the thickness of the laminates and the heat distribution will be discussed. In the next section (3.4) a comparison will be made between all the heating method comparing the heating rate of the slowest and the fastest layer in the laminate. From the viscosity measurements and adiabatic temperature build-up could be estimated. As can be seen from Figure 35, the viscosity increase for the resin mixture used in this research is highly dependent on the processing temperatures. 1,2 1 Viscosity [Pa.s] 0,8 0,6 0,4 0,2 0-0, Time [min] Figure 35.Viscosity measurements at different mould temperatures These viscosity measurements were done with a mould that was already at the curing temperature of the product. For a cure temperature of 180 C, an infusion window of three minutes is seen, while for a cure temperature of 130 C, the reaction starts after 12 minutes (initial viscosity build-up) reaching an infusion window of 15 minutes. During the heating phase from 115 to 180 C, the average temperature of the resin over the heating time is 145 C. If a linear relationship with mould temperature is assumed for the viscosity build-up, the reaction at 145 C would start at approx 8.5 minutes. 28

36 The adiabatic reaction kinetics test is more representative for the production process of the thickwalled laminates. An infusion time of 5 minutes is assumed during the test at an infusion temperature of 115 C after which the product is heated by contact heating (oil) to the required curing temperature. It can be seen (Figure 36) that for a cure temperature of 160 C, the reaction starts to build up exothermal heat fast after 10 minutes. It is also clear from this figure that the reaction already starts before the cure temperature is reached. The start of the temperature build-up obtained from the adiabatic measurements could be considered as the maximum heating time preferred for the infused laminate. From the results at 140, 150 and 160 C, it could be seen that a linear relationship between the start of the temperature build-up and the cure temperature is a safe way of extrapolating the results to a higher temperature. The start of the reaction at 170 C would be extrapolated to 8.3 minutes and at 180 C about 6.6 minutes. It could be concluded that the required heating time from 115 to 180 C would be around minutes. Temperature [C] (2) Time [sec] Figure 36. Adiabatic temperature measurements with an initial pot life of 5 minutes at 115 C, cured at different temperatures Heated press As can be seen from the heating profiles through the thickness of the laminate in Figure 37, the layer that heats up the slowest is the middle layer of the laminate. For the middle layer, 20 minutes were needed to heat the resin from 110 to 180 C, which is too slow for a good reaction. Furthermore, it could be concluded that there is a good heat uniformiuty around the centre. However, the heated platen is considered to be a non suitable heating method for curing thick laminates. Heated press temperature transition (1) Temperature (C) 160 Top Middle Bottom 110 Under mold Time (min) Figure 37. Heating profiles through the thickness of the laminate during heating from 110 to 180 C with the heated platen 29

37 3.3.2 Infrared The influence of the surface finish of the aluminium mould on the absorption of the infrared radiation was tested prior to the heating experiments with the infrared heating. Therefore, moulds with different surface treatments were tested: an untreated mould, a sanded mould, a mould with a black paint finish and an anodized mould. Also different thickness were tested; 2 and 6 mm. Figure 38. Heating profiles of different mould plates using the infrared panels as heating device It can be seen from Figure 38 that the reduction in thickness leads to a significantly faster heating, as expected. The difference between the untreated and black painted surface can be clearly seen for the 2 mm thick mould plate in which the advantage of using a black paint finish is proven. Also compared to the other surface finishes, the black paint seems to increase the infrared absorption the most. Figure 39. Heating profiles through the thickness of the laminate during heating from 110 to 180 C with the infrared panels The results for the heating profiles using the infrared panels as a heating device are shown in Figure 39. The middle of the laminate and the 15 th layer show a similar temperature behaviour, showing a non-symmetric heating behaviour throught the thickness of the laminate. 12 minutes were needed to heat up the whole laminate to 180 C, which is considerably faster than the 30

38 heated platen press. However, from a polymerisation reaction point of view, this heating rate is still too slow Thermofoil Temperature (ºC) top kapton Under mold Time (minutes) Figure 40. Heating profiles through the thickness of the laminate during heating from 110 to 180 C with a thermofoil on the mould side and an infrared panel at the top For the test with the thermofoil as a heating device, the top part of the vacuum bag was heated with the top infrared panel. A non-symmetric heating distribution can be observed in Figure 40. To have a better look to the performance of the thermofoil, these results are compared with the double IR panel experiment, see Figure 41. The top layer shows the same temperature behaviour for both heating methods as can be expected for the same heating device. The biggest difference can be found for the bottom layer. It can clearly be seen that the thermofoil is more powerful. However, this difference can not be seen in the slowest heated layer, the middle. Here, the heating curves look similar again. Although the chemical reaction of the laminate heated by the double infrared panels starts earlier Temp. (C) Time (min) Thermofoil+IR 40 Thermofoil+IR 20 Thermofoil+IR 0 Double IR 40 Double IR 20 Double IR 0 Figure 41. Comparison of heating profiles of laminates heated by double infrared, or by thermofoil in combination with the top infrared panel. The thermofoil has a good heating rate. Its small thickness and the fact that it is a resistive heating element provide a very efficient solution with minimal hardware. Due to the low thermal conductivity of the resin, that heat however is not conducted to the laminate. Hence the heating rate is still not enough for heating the laminate to the required curing temperature. A crosssection of the cured laminate can be seen in Figure

39 Figure 42. Cross-section of a cured 24 mm thick laminate heated by thermofoil in combination with the top infrared panel Resistive mesh Both the infrared panels and the resistive heating mesh were used in this experiment, in which the temperature of the resistive mesh is controlled manually to the required temperature. 45 amperes were applied to the mesh. The layer that is heating up the fastest is now the middle layer, see Figure 43, where the resistive mesh is located. It reached the cure temperature at 2 min 40 seconds. A cross-section of a 24 mm thick laminate that was heated with a resistive mesh can be seen in Figure Temp. ( C) Top 30 layers 25 layers Middle 15 layers 10 layers Bottom Time (min) Figure 43. Heating profiles through the thickness of the laminate during heating from 110 to 180 C with a resistive mesh in the middle of the fibre lay-up and the infrared heating as surface heating It could be concluded that using a resistive mesh is a suitable solution for heating the laminates from the inside out. The interaction of the mesh with the galss fibres and the resin needs to be investigated, as well as the application of the mesh as a heating method in case of heating up a large wind turbine blade. Metal mesh Carbon fibres Figure 44. Cross-section of laminates heated by a metal mesh or a carbon fabric in combination with the intrared panels Carbon fabric When the carbon fabric is used as the inside heating method in combination with the infrared panels to heat up the laminate, it can be seen the middle layer heats up the fastest to the curing 32

40 temperature (in 1.5 minutes). This result, see Figure 45, is similar to the result obtained with the resistive mesh, as described in section A cross-section of a 24 mm thick laminate that was heated with a carbon fabric in the middle can be seen in Figure Temperature (ºC) Time (minutes) Top 30 layers 25 layers Middle 15 layers 10 layers Bottom Figure 45. Heating profiles through the thickness of the laminate during heating from 110 to 180 C with a resistive carbon fabric in the middle of the fibre lay-up and the infrared heating as surface heating The slowest layer is the 10 th layer in the lay-up and reaches the curing temperature after 8 minutes, which is an acceptable heating time for the reaction of the APA-6 resin. To check the heat distribution through the thickness, two graphs were made. The first graph, Figure 46 shows the heating profiles of the heat input layers, while in Figure 47 the heat distribution throughout the other layers is shown. It can be seen that the infrared heating profiles are similar to each other and that the heat distribution is symmetric throughout the thickness of the laminate Top Middle Bottom Time (minutes) layers 10 layers 25 layers 15 layers Time (minutes) Figure 46. Heating profiles of the heat input layers in the carbon fabric experiment. Figure 47. Heating distribution through the thickness in the carbon fabric experiment. The carbon fabric as a heating source offers a good heating performance. Also the implementation of the heating technique is easy. Moreover, many layers of carbon fabric can be integrated inside the product,resulting in a hybrid composite Combination of carbon fibre and thermofoil After the tests with all the heating methods described above, it was decided to test a combination of the fastest and most uniform heating methods. This resulted in the following heating method: thermofoil at the mould side of the laminate, the carbon fabric as inside heating and the top infrared panel as the top surface heating. A heating cycle with the dry fibres showed excellent heat symmetry. As can be seen from the heating profiles during the heating to cure temperature, see Figure 48, the thermofoil and the carbon fabric heat up faster that the top IR panel. The layer 33

41 slowest in heating up is therefore the 30 th layer. To reach 180 C, it took approximately 7 minutes. It can be seen that after those 7 minutes, all the layers in which no heat was supplied showed a similar exothermal behaviour, resulting in a uniform temperature distribution, which was the objective of this study Temperature (C) Top Middle Bottom Time (min) Figure 48. Heating profiles through the thickness of the laminate during heating from 110 to 180 C with a resistive carbon fabric in the middle of the fibre lay-up and the infrared heating as surface heating 3.4 Comparison of heating methods Heating rate The heating rates that were obtained during the experiments for the slowest heated layers are summarized in Table 7. It can be seen that a few heating methods that were tested comply to the required heating rate needed for the isothermal infusion strategy being researched; namely the thermofoil with the top infrared panel, the carbon fabric in combination with the infrared panels and the final combination of thermofoil, carbon fabric and the top infrared panel. Table 7. Comparison of heating rate of different heating methods Heating method Slowest layer Max heating rate( C/min.) To 180(min.) Heated platen Middle 11 13,5 IR top + bottom Middle 6 12,0 Thermofoil + top IR Middle 11 7,0 Resistive ,0 Carbon fabric ,0 Thermofoil + Carbon fabric + top IR , Heat distribution Three possible heating methods were selected from the heating rate of the whole laminate. However, not only the heating rate is of importance, also the heat distribution through the thickness. Therefore, the temperature results through the thickness of the laminate that were obtained right before the polymerzation reaction started are shown in Figure 49. The heating method with the thermofoil, carbon fabric and infrared panel proofed to have the most unifrom 34

42 heat distribution. The heating method with the carbon fabric also shows a heat distribution that is within the limits of what is needed for the process (±5 C) Layer Thermofoil+Carbon+IR Carbon+IR Resistive+IR Thermofoil+Top IR Platen press Temperature (ºC) Figure 49. Heat distribution for different heating methods, before start of the chemical reaction 3.5 Conclusions For the isothermal infusion of thick laminates, different heating methods were tested. It was shown that for obtaining a sufficiently high heating rate (from 110 to 180 C in 8-10 minutes) only one type of surface heating of the mould gave adequate results, namely thermofoil, which has a high heat transfer in combination with infrared. The other possible heating methods had a heating source in the middle of the laminate: namely a carbon fabric that was used as a resistive heating source. This technique was used in combination with the double infrared panels or with the thermofoil and infrared panel. As a uniform heat distribution is wanted as well for the uniformity of the laminate properties through the thickness of the laminate, it was clear that that goal was only reached if heating from the inside was applied. However, these resistive heating sources, as researched in this chapter, need connections for the power supply which are not easily adapted for the heating of large wind turbine blades. It is therefore advised to investigate the potential of induction heating. Induction heating is the heating process of a metal or conductive object by electromagnetic induction, where eddy currents are generated within the metal resulting in resistance heating of the metal. Because no contact is needed to heat a piece of metal, this would be an ideal technique to heat metal objects inside a laminate, thus heating from the inside. The basic components that are needed are an AC power supply, induction coil and the material to be heated. Another heating source that might be worth researching is microwave heating. This microwave technology is also an inside out heating method. From a mould point of view, several conclusions could be drawn when heating with infrared. Painting the Al mould plate with a mat black paint increases the radiation absorption. An aluminium mould is preferred over a Steel one as its thermal diffusivity is higher. And the thickness of the mould plate is also important, a thinner plate needs less energy to heat up itself and can therefore transport heat quicker. However, the stiffness of the mould is also important for the final product. 35

43 4 Reduction of cure shrinkage induced voids 4.1 Introduction As could be concluded from the work described in section , voids have a great influence on the mechanical properties of the composite. Three types of voids can be distinguished: Mechanical air entrapment during flow in fibre reinforcement Gas diffusion into reactive resin mixture Resin shrinkage upon solidification In section , the reduction of voids due to gas diffusion and mechanical air entrapment during the flow were described. However, the pure APA-6 resin also exhibits high levels of shrinkage (up to 9% in volume) and therefore shrinkage induced voids are one of the issues for the composites. To counteract this shrinkage, a rubber-toughened activator (R) is suggested. In [30], the material requirements for wind turbine blades are described as followed: the material needs to exhibit a high material stiffness, a low density and a long fatigue life, also the material fracture toughness is of importance. The addition of rubber to the material system will soften the matrix material, and will therefore render the material more fatigue resistant which is one of the key requirements for large-scale wind turbines. From literature, it can be expected that the use of activator R will improve the impact properties due to the rubber content. A trade-off between a loss in stiffness and improvement in fatigue properties needs to be made. Therefore, an optimization of the amount of rubber that is required in the composites needs to be performed as well as a comparison between the system that currently has been researched (activator C). 4.2 Experimental Materials For the research on the rubber toughened activator, the baseline material had the same constitution as described in section For the rubber toughened activator, a liquid activator, activator R was used in different concentrations, as will be described in this chapter. This activator was used in combination with initiator M, in the same concentration as previous research namely 1.2% mol. All the components were supplied by Brüggemann Chemicals, Germany Processing methods The processing methods for the preparation of the resin and the infusion of the laminates are the same as described in section Infusion of pure resin panels The manufacturing of the 3 mm thick pure resin panel was done at Bruggemann Chemicals using an injection machine. The mould temperature was set to 160 C while the MMU was used at 110 C. After 45 min of cure, the panel was removed from the mould and was cooled down to room temperature Analysis methods The viscosity measurements and the test for the determination of the degree of conversion are the same as described in section Reactivity tests Glass test tubes with a diameter of 20mm and a length of 150 mm (Schott-Fiolax) were used to make pure resin samples. The first one, tube A, is filled with a solution of activator/monomer and 36

44 the second one, tube B, with a solution of initiator/monomer. The test tubes are then put into a preheated silicon oil bath at 160 C. The resin in t he test tubes is stirred regularly and its temperature is monitored. When both solutions reach 120 ºC, solution B is poured down into the tube A. The mixture is then stirred and the tube is placed back in the silicon bath. From then on, three times are recorded. The first (t1) is when the solution becomes viscous, which is a measure for the infusion window of the resin, the second (t2) when the solution becomes turbid, which is a measure for the start of the polymerization. And finally, the time at which the polymer formed comes loose from the test tube wall (t3), which is a measure for the time to crystallisation. Any special features observed will also be reported. Degree of crystallinity The degree of crystallinity (X c ), the melting temperature (T m ) and the cold crystallisation temperature (Tcc) of the polymer are determined by means of a Perkin Elmer Differential Scanning Calorimeter (DSC). Disc-shaped samples (m sp ) of approximately 5 mg are taken from the laminate at the same location as the degree of conversion samples and are dried overnight. In the DSC, the test sample is held at 25 C for two minutes and is then heated up to 240 C at a rate of 10 C per minute, where they are held for tw o minutes. After that, they are cooled down to room temperature at a rate of -10 C per minute. The results had to be corrected for the fibre and rubber content of the sample. The degree of crystallinity can now be calculated as follows: X c H m mtotal = 100 H m m m m 100 total fibres rubber mon (4) Density ASTM Standard D792-00, Test Method A, was used to calculate the density of the material. To determine the practical density of the specimen, the specimen was first weighed in air (m sp ) on the AG204 DeltaRange balance. Next, the specimen had to be fully submersed in water. The temperature of the water was measured, from which the density of the water could be calculated and a wire assembly was connected to the balance and suspended in water. After that, the specimen was attached to the wire, suspended in water and measured (m sp,water ). The density of the specimen could then be calculated as follows: (5) ρ pract = ρ water m sp ( msp msp, water ) Tensile testing The tensile properties of the pure resin were measured by means of the ISO Standard, with dogbone shaped specimens that had a thickness of 3 mm. The test specimens were milled out of the flat plate and dried before testing for at least 48hours at 50C under full vacuum. At least five samples were tested per panel on a Zwick 20kN static test machine. Impact resistance The impact resistance of pure resin samples were tested by means of an Izod test, ASTM D256-05a. The notched specimens were milled according to the standard Test Method A and had a thickness of 3 mm. The type of failure was noted and the energy to break was measured for each specimen and from that, the impact resistance of the specimens could be calculated. At least, five specimens were tested per panel. The samples that were tested were either dried at least 48 hours at 50 C under full vacuum, or conditioned up to the equilibrium water uptake in a climate chamber at 70 C and 62%RH, according to ISO1110 Sta ndard. Moisture uptake Samples were conditioned in a climate chamber at 70 C and 62%RH until the equilibrium moisture uptake was reached. Then the samples were mechanically tested. 37

45 4.3 Resin constitution The potential advantages of activator R for the APA-6 composites need to be investigated. Therefore, the amount of the rubber content and the initiator needs to be found by means of reactivity tests and viscosity measurements. First of all, it needs to be checked whether the resin reacts well, gives enough time to infuse the composites, and see the effect on shrinkage time. Therefore, reactivity tests were performed at different levels of initiator and activator. A rubber content of 7% and a 1.2%mol concentration of the initiator gives the best compromise for the vacuum infusion system, (as can be seen in Figure 50): Biggest infusion window Highest time to polymerize (enough time to heat laminate before reaction starts) Average time to shrink. After polymerization, the polymer chains needs to react with the fibre sizing as well, therefore enough time between polymerization and crystallization is needed. Time [min] T1 T2 T3 C-M system Amount of rubber [%wt] Figure 50. Reactivity of the resin with different amounts of rubber. T1 stands for the infusion window, T2 for the start of the polymerisation and T3 for the start of the crystallisation. The system has been compared to the reaction times of the normal system. t3 t2 t1 With increasing amount of activator R, the mixture becomes more reactive, therefore, the time to polymerize decreases. As the rubber content complicates the crystallization the time to crystallize increases with increasing amount of rubber. When compared to the activator C system, it can be concluded that the time difference between t1 and t2 is significantly higher. It is therefore needed to check the reaction rate of the system by means of viscosity measurements. 1,200 1,000 viscosity (Pa.s) 0,800 0,600 0,400 0,200 0,000 0:00 1:45 3:30 5:15 7:00 8:45 10:3 12:1 12:5 13:2 14:0 14:3 15:1 15:4 16:2 16:5 17:3 18:0 18:4 19:1 19:5 C-M 180 C-M 170 C-M 160 R 7% 160 R 7% 170 R 7% 180 time (min) Figure 51. Viscosity versus time for different material systems: C-M and R-M at different testing temperatures: and 180 C As can be seen in Figure 51, the effect of the testing temperature is the same for activator C and R. The higher the testing temperature, the faster the reaction starts and the higher the reaction rate. It can also be seen that for the optimal resin constitution of 7%wt of activator R, the reaction starts significantly later than when compared with the activator C. Also, the reaction rate is lower. Using the activator R slows down the start of the reaction, as the amount of reactive parts in the 38

46 mixture is lower that with the mixture with activator C. The equivalent amount of activator R needed to have the same amount of reactive groups is 24.15%. Viscosity measurements were done for activator R concentrations of %wt at three testing temperatures: C. In Figure 52, a summary of the results is shown for a testing temperature at 160 C. By increasing the rubber content in the resin, the reaction rate and initiation of the reaction increases significantly due to addition of the reactive groups. Although, the reaction rate is still lower than with the activator C system. It can also be seen that the initial viscosity of the resin, at the beginning of the test, is higher than with the normal system. This might be disadvantageous for the infusion, as a low viscosity is wanted. 1,200 viscosity (Pa.s) 1,000 0,800 0,600 0,400 C-M 160 R 20% 160 R 7% 160 0,200 0,000 0:00 1:45 3:30 5:15 7:00 8:45 10:30 12:15 12:50 13:25 14:00 14:35 15:10 15:45 16:20 16:55 17:30 18:05 18:40 19:15 19:50 time (min) Figure 52. Viscosity versus time for different material systems: R-M at7% and 20%wt and C-M. at a testing temperature of 160 C The value of 1 Pa.s is seen as the maximum viscosity for vacuum infusion of composites. If a look is given to Table x, it can be seen with low rubber contents in the resin, the infusion window is long. Increasing the rubber content does not seem to reduce the infusion window as significantly from a rubber content of 15% on. Also the reaction rate is comparable for the tests done for a rubber content of 15, 20 and 24.15%wt. Table 8. Infusion window for different material systems: C-M and R-M at different amount of R at different testing temperatures Time to 1 Pa.s (min) C 7% 10% 15% 20% 24,15% 160 C 12,35 20, ,32 15, C 10,22 15,8 10,83 9,3 9,76 9, C 5,2 12,44 9,9 7,87 8,43 8,25 It can be concluded from these results that the activator R seems a suitable candidate for infusion of large composites, due to its long infusion window and its acceptable reaction rate. 4.4 Pure resin To investigate the behaviour of the pure resin, 3mm thick panels were manufactured with different amounts of rubber: %wt activator R. Two panels per activator concentration were infused. Also, two baseline panels with the C-M formulation were made for comparison Those panels were tested for degree of conversion, degree of crystallinity, cold crystallization, tensile properties, water uptake and impact properties. It was seen that the degree of conversion of the panels with activator are comparable with the normal system and that there is no difference in degree of conversion when the rubber content is increased. Therefore, it can be concluded that the degree of conversion is not a significant factor in the interpretation of the results. 39

47 Xc 50,00 48,00 46,00 44,00 42,00 40,00 38,00 36,00 34,00 32,00 30,00 Cx S-7 S-10 1 S-15 S-20 Figure 53. Degree of crystallinity for the normal C-M system and for different amounts of rubber in the R-M system. For the degree of crystallinity, it can be seen that the standard deviation is high for all the test samples, see Figure 53. Although no trend can be seen with increasing rubber content, a significant increase in degree of crystallinity can be seen when the activator R system is compared to the normal system. The increase in crystallinity with activator R can be explained due to the fact that more mobile chains align more easily to form crystals. To verify this, the enthalpy during the cold crystallization (crystallization from the melt) was studied in the DSC. With activator C, more energy was generated for the formation of the crystals. A clear decrease in crystallization enthalpy (Figure 54) and cold crystallization temperature (Figure 55) can be seen with increasing rubber content. Up to 15%, the polymer with activator R crystallises faster, more easily, at higher temperature than with activator C. The increase of rubber above 15%wt shows that a more difficult crystallization behaviour, as a more cross-linked polymer system is obtained. Therefore it becomes more difficult to crystallize and more supercooling is needed to form crystals in the structure, hence a lower T cc. Enthalpy (J/g) percentage of activator R Figure 54. Enthalpy of cold crystallisation for the normal C-M system and for different amounts of rubber in the R-M system. Peak temperature ( C) Percentage of activator R Figure 55. Peak temperature of cold crystallisation for the normal C-M system and for different amounts of rubber in the R-M system. For the mechanical properties, the tensile and impact properties were measured. For which the results are shown in Figure 56 and Figure 57 respectively. Activator C clearly exhibits better tensile properties due to the higher stiffness of the chains. The more activator R is added to the system, the lower the stiffness. As the crystallinity increases by using the rubber-toughened activator, a higher E-modulus is expected when compared to the normal system. However, by using rubber, more flexible parts are added in the polymer chains and therefore the E-modulus will decrease, as can be seen in the graph. A lower impact resistance is seen for the activator R panels when compared to the activator C system. The higher amount of crystallinity of the P5 polymer might render it more brittle on impact. All the specimens exhibited a complete break, showing a brittle -like behaviour. Increasing the amount of activator R, lowers the impact resistance up to 15%. Increasing the rubber content to 20% increases the impact resistance again. The decrease in impact resistance can be attributed 40

48 to a more cross-linked system. Again, at around 15% of rubber, a transition point from a reaction point of view is seen. E modulus [MPa] C Figure 56. E-modulus for the normal C-M system and for different amounts of rubber in the R-M system. Impact resistance [J/m] DRY 0 C Figure 57. Dry impact resistance for the normal C-M system and for different amounts of rubber in the R-M system. While in the dry state, the activator C outperforms the R system in impact resistance, the rubber toughened system takes over in a conditioned state, see Figure 58. The impact resistance results are increased with a minimum factor of 8. The high increases with 7 and 10%wt of rubber can be attributed to a incomplete break (increase with a factor of min. 31) which can be attributed by their higher equilibrium water uptake level. The transition in the material is seen at 15% from incomplete break to a hinge break. It was also seen that the cure time of the samples has a significant effect on the impact resistance see Figure 59. For a shorter cure time, a higher impact resistance was found due to a less entangled system upon curing, therefore lowering the impact resistance. Impact resistance [J/m] DRY WET C Impact resistance [J/m] min dry 45 min dry 30 min wet 45 min wet Figure 58. Dry and conditioned impact resistance for the normal C-M system and for different amounts of rubber in the R-M system. Figure 59. The impact resistance for specimens cured at 30 or 45 min, in dry or conditioned state, for 7 and 10% of rubber in the R-M system. As could be concluded from the research on the pure resin, the rubber toughened system shows a lot of potential for the use in production of laminates. The resin reacts nicely at all rubber contents tested and shows a higher crystallinity than the normal system. A loss of stiffness was found, while the impact resistance and material toughness increases when the samples are conditioned. 4.5 Laminates Now that the behaviour of the pure resin is known, laminates can be produced and tested. After that, the properties like degree of conversion, degree of crystallinity can be tested to see whether the same trends as for the pure resin can be found. A comparison can be made with the normal system. Only preliminary results can be shown in this section. Panels with different amounts of 41

49 activator R (10, 15 and 20%wt) were infused with infusion strategy 1. Also different cure temperatures were tested for the rubber content of 15%. All panels reacted as expected, but due to a change in the chemicals and infusion behaviour, the results that are shown are not yet repetitive. The laminates have therefore not been tested mechanically yet. DOC (%) 96,0% 95,5% 95,0% 94,5% 94,0% 93,5% 93,0% 92,5% 92,0% Temperature ( C) DOC (%) 98% 97% 96% 95% 94% 93% 92% 91% concentration of activator R (%) Figure 60. The degree of conversion for laminates with 15%wt rubber at different processing temperatures. Figure 61. The degree of conversion for laminates manufactured at 180 C, with different amounts of rubber. As can be seen from both Figure 60 and Figure 61, the values for the degree of conversion of the laminates are lower than for the pure resin. The addition of the fibres is known to affect the conversion negatively, as seen in the work of van Rijswijk. As was seen from this work, the decrease in degree of conversion (Figure 60) and in degree of crystallinity (Figure 62) with increasing processing temperature is as expected. The value of the degree of conversion at 170 C shows a high standard deviation due to a bad quality of the laminate, which was a production error. What can be seen from Figure 61 is that the degree of conversion increases with the increase of concentration of activator R. Increasing the amount of rubber at a processing temperature of 180C, does not show any trend for the degree of crystallinity, due to the high standard deviation. All the values for the degree of conversion are still in the acceptable range for the laminates. The selection of the best quantity of rubber to be used will thus be dependent on the mechanical properties of the laminates, which is not part of this work. Xc Temperature ( C) Figure 62. The degree of crystallinity for laminates with 15%wt rubber. Xc 44,00 42,00 40,00 38,00 36,00 34,00 32,00 30,00 0% 5% 10% 15% 20% 25% Concentration of activator R Figure 63. The degree of crystallinity for laminates manufactured at 180 C, with different amounts of rubber. 4.6 Conclusions It can be concluded from the results on reactivity that the activator R seems a suitable candidate for infusion of large composites, due to its long infusion window and its acceptable reaction rate. As could be concluded from the research on the pure resin, the rubber toughened system shows a lot of potential for the use in production of laminates. The resin reacts nicely at all rubber contents tested and shows a higher crystallinity than the normal system. A loss of stiffness was found, while the impact resistance and material toughness increases when the samples are conditioned. All the values for the resin properties are in the acceptable range for the laminates. The selection of the best quantity of rubber to be used will thus be dependent on the mechanical properties of the laminates, which is not part of this work. 42

50 5 Optimisation of fibre-to-matrix bond 5.1 Introduction It could be concluded from the work of van Rijswijk [23] that the fibre-to-matrix bond in combination with good resin properties is the key parameter to good mechanical properties. An improvement in fibre-to-matrix bond will result in an increase in fatigue properties and in a better resistance to moisture absorption. It was also seen that the void content is closely related to the interface properties; the better the interface, the lower the final void content. The fibre-to-matrix bond between the APA-6 resin and the glass fibres used to reinforce the laminates can be improved by using a chemical sizing on the glass fibres that is compatible with the resin system. It is proven to be difficult to find a compatible sizing for the APA-6 system. To be able to test the fibre-to-matrix bond strength of different sizing, uni-directional (UD) transverse tensile tests have proven to give good quantitative results. For this purpose, a method to infuse UD laminates needs to be developed. In this chapter the effect of the fibre sizing on the mechanical properties will be described, the UD-infusion set-up and the optimization thereof. All the fibre rovings that will be tested in this research are classified as E-glass fibres. They have the same density and their nominal E-moduli should be around 75GPa. To exclude the differences between the fibres when testing the fibre-tomatrix bond, the E-modulus of the different fibres will be tested, and afterwards the fibre-to-matrix bond of the UD laminates will be tested. 5.2 Experimental Materials The same resin materials were used as described in section Glass fibres Different types of glass fibres were tested in this research, all with different sizings and properties: P4510, J058, J858, and J870. All the glass fibres that were tested are E-glass fibres. For the optimization of the infusion process of the UD laminates, the 4510 fibre was used Processing methods The same processing methods were used as described in section For the infusion strategy of the UD-laminates, an optimization of the infusion strategy 1 was done Analysis methods Tensile tests Tensile tests were done on the dry glass fibre rovings and on the UD laminates. The tensile test of the roving was performed according to the ISO 3341 Standard. The roving was clamped with a radius clamp to avoid slippage during the test. The gauge length of the fibre with the radius clamps was set at 300mm. An extensometer was used to measure the elongation of the specimen. The two parts of the extensometer were placed at a distance of 200 mm to measure the E-modulus. At least five tests were done with each fibre roving for the determination of the E-modulus. A tangent method was used to determine the E-modulus. The tests were performed on a 20kN Zwick static test machine. The transverse tensile test was used to test the fibre-to-matrix bond in which the specimens are tested in transverse fibre direction, therefore testing only the fibre-to-matirx bond. The tensile specimens with a thickness of 2 mm were prepared and tested according to ASTM Standard D3039. The width and the length of the specimens were 25 and 175 mm respectively. No gripping 43

51 tabs were applied on specimens. The 20kN Zwick testing machine was used for the tensile test. The testing speed applied was 2mm/min. Degree of conversion The degree of conversion of the UD laminates were measured by means of the same test as described in section Effect of fibre sizing The effect of the fibre-to-matrix bond was previously tested by means of the short-beam strength test. This mechanical test is however also dependent on the resin properties, the mechanical locking due to resin shrinkage and the chemical bonding. It is difficult to get a good overview of the real fibre-to-matrix bond via this test method. Therefore, the UD laminates and the transverse tensile test needs to be tested for its suitability to test the fibre-to-matrix bond. 80 Interlaminar shear strength [MPa] APA-6 composite (unsized/outlet) APA-6 composite (sized/outlet) Figure 64. Effect of the fibre sizing of the short-beam strength at different processing temperatures. From the previous research, it was seen that a drop in short-beam strength could be observed when using the unsized glass fibres. The drop in short-beam strength can be explained by a poor fibre-to-matrix bond. With unsized fibres, the chemical bonding of the polymer chains to the glass fibre surface is not present, which leads to a significant drop in short-beam strength. It can therefore be concluded that the usage of a fibre sizing as well as unblocked activators (which are responsible for the adhesion to the fibre sizing) are essential for chemical bonding and good mechanical properties. 5.4 Characterisation of glass rovings In Figure 65, the test results of the four types of fibre rovings are shown. As said previously, the nominal E-modulus for the E glass fibre has to be 75 GPa. To determine the best properties between fibres tested, there are some characteristics that can be studied, which can be found in Table 9. The E-modulus is the most important factor, corresponding to a material with a high strength with minimal deformation. Force (N) Mould temperature [ C] P4510 J058 J858H J ,5 1 1,5 2 Strain (mm) Figure 65. Comparison of the tensile strength of different fibre types 44

52 From Table 9, it can be concluded that the J870 fibre has the lowest E-modulus due to a different, rougher texture, as a result of the manipulation in the test set-up. The other three types of fibres are considered good enough to be used in laminates. In Figure 65, it can be seen that there is no difference between the other three fibre types when it comes to strain. The J058 fibre shows the highest maximum force but with a high standard deviation. Table 9. Comparison of tensile properties of glass fibre rovings Fibres P4510 J058 J858 J870 E-Modulus (MPa) Stdev Max_value (N) Stdev 80, ,2 49,0 Elongation (mm) 0,874 0,835 0,802 1,78 stdev 0,150 0,207 0,108 0, Vacuum infusion set-up for UD laminates For the production of UD laminates, the dry fibre rovings are first uni-directionally filament wound around an Al mould. The mould used is a supporting Al plate (45x45cm 2 with a thickness of 2 mm) to which two levers can be connected for the filament winding process, see Figure 66. The fibres are wound around the mould in four layers, see Figure 67, which will give a final thickness of the laminate of 2 mm after cure, as required by the test method. The width of the laminate will be 20 cm. Figure 66. Al supporting plate (for both winding and infusion) Figure 67. Dry filament wound fibres around Al supporting plate After the winding process, the vacuum bagging can begin and is done as with the normal laminates, as described in section The tacky tape is placed around the mould on both sides of the mould. Then, the inlet and outlet springs and tubes are connected, as can be seen in Figure 68. The type, size and placement of the inlet and outlet were changed in the optimization process. It turned out that when the inlet or outlet spring were connected to an aluminium tube which is placed on the edge of aluminium supporting plate gives the best connection. Thin Al cover plates with a thickness of 2 mm were used to improve the surface quality (Figure 69) and a more uniform thickness. 45

53 Figure 68. Tacky-tape applied and in- and outlet connected to the mould set-up Figure 69. Al covering plates placed on the fibres for better surface quality Finally, the bagging could be finished by applying the polyimide foil as a vacuum bag and is now ready to be infused (see Figure 70). The process of the vacuum infusion of UD laminnate has been already described as infusion strategy 1 in Section The values of the infusion parameters are summarized in Table 10. However, because of complexity of UD laminates vacuum infusion, the process parameters like polymerization temperature, temperature in MMU and in the resin buffer, the infusion pressure and polymerization time were variable. Figure 70. Fully vacuum bagged mould ready for infusion Figure 71. A fully cured UD laminate. As it turned out that it was difficult to fully impregnate the fibres, the infusion parameters stated above were optimized. When a glass fibre fabric is used in the infusion process, the infusion of the resin takes place uniformly due to the fibres that are not only orientated in flow direction but also in transverse direction. Therefore the flow is restricted during infusion but the resin is distributed over the complete surface. For the infusion of UD laminates, there are only fibres in flow direction. The resin will always flow along the easiest path, for example where the density of winded fibres is lower. Therefore, the infusion process will be more susceptible to nonuniformities resulting in non-impregnated areas in the final product. Table 10. Infusion parameters for the UD infusion process, for different type of tests MMU & buffer Preheated Cure Infusion Flow length vessel glass fibres temperature pressure Start values 110 C 180 C 180 C 250 mbar 45 cm Test C 180 C 180 C mbar 45 cm Test C 160 C 180 C 100 mbar 45 cm Test C 180 C 180 C mbar 30 cm 46